Full Guideline PDF 2176838029

NCC-WCH
Version 2.0
Preterm labour and birth

Full guideline
NICE Guideline 25
Methods, evidence and recommendations
November 2015, updated June 2022
Final
Commissioned by the National Institute for
Health and Care Excellence
Error! No text of specified style in document.
Disclaimer
The recommendations in this guideline represent the view of NICE, arrived at after careful
consideration of the evidence available. When exercising their judgement, professionals are
expected to take this guideline fully into account, alongside the individual needs, preferences
and values of their patients or service users. The recommendations in this guideline are not
mandatory and the guideline does not override the responsibility of healthcare professionals
to make decisions appropriate to the circumstances of the individual patient, in consultation
with the patient and/or their carer or guardian.
Local commissioners and/or providers have a responsibility to enable the guideline to be
applied when individual health professionals and their patients or service users wish to use it.
They should do so in the context of local and national priorities for funding and developing
services, and in light of their duties to have due regard to the need to eliminate unlawful
discrimination, to advance equality of opportunity and to reduce health inequalities. Nothing
in this guideline should be interpreted in a way that would be inconsistent with compliance
with those duties.
NICE guidelines cover health and care in England. Decisions on how they apply in other UK
countries are made by ministers in the Welsh Government, Scottish Government, and
Northern Ireland Executive. All NICE guidance is subject to regular review and may be
updated or withdrawn.
Copyright
© 2015 National Collaborating Centre for Women’s and Children’s Health
Update information
In June 2022 this document was updated to redact some content that was now out of date as
a result of the 2022 evidence review on the use of repeat courses of maternal
corticosteroids. See the NICE website for the current recommendations at
https://www.nice.org.uk/guidance/ng25.
Funding
Registered charity no. 213280
Contents
Contents
1 Guideline summary .................................................................................................... 15
1.1 Guideline Committee membership, NCC-WCH staff and acknowledgements...... 15
1.1.1 Guideline Committee membership ........................................................... 15
1.1.2 Acknowledgements .................................................................................. 15
1.1.3 NCC-WCH staff........................................................................................ 15
1.2 Care algorithm ..................................................................................................... 17
1.3 Recommendations .............................................................................................. 19
1.4 Key research recommendations .......................................................................... 20
1.4.1 Prophylactic cervical cerclage compared with prophylactic vaginal
progesterone for preventing preterm birth ................................................ 20
1.4.2 Identifying infection in women with preterm prelabour rupture of
membranes (P-PROM) ............................................................................ 20
1.4.3 Effectiveness of ‘rescue’ cerclage ............................................................ 20
1.4.4 Magnesium sulfate for neuroprotection: bolus plus infusion compared
with bolus alone ....................................................................................... 21
1.5 Research recommendations ................................................................................ 21
1.6 Other versions of the guideline ............................................................................ 21
1.7 Schedule for updating the guideline..................................................................... 21
2 Guideline development methodology ....................................................................... 22
2.1 Development of the guideline .............................................................................. 22
2.1.1 What is a NICE guideline? ....................................................................... 22
2.1.2 Remit ....................................................................................................... 23
2.1.3 Who developed this guideline? ................................................................ 23
2.1.4 What this guideline covers ....................................................................... 23
2.1.5 What this guideline does not cover........................................................... 24
2.1.6 Relationships between this guideline and other NICE guidance ............... 25
2.2 Methods .............................................................................................................. 26
2.2.1 Developing the review questions and outcomes....................................... 26
2.2.2 Searching for evidence ............................................................................ 33
2.2.3 Evidence of effectiveness ........................................................................ 34
2.2.4 Methods of combining clinical studies ...................................................... 35
2.2.5 Type of studies......................................................................................... 38
2.2.6 Appraising the quality of evidence by outcomes ....................................... 39
2.2.7 Quality assessment of NMA ..................................................................... 44
2.2.8 Assessing clinical importance .................................................................. 45
2.2.9 Evidence statements ................................................................................ 45
2.3 Evidence of cost effectiveness ............................................................................ 45
2.3.1 Literature review ...................................................................................... 45
4
Contents
2.3.2 Undertaking new health economic analysis.............................................. 47

2.3.3 In the absence of economic evidence ...................................................... 48
2.4 Developing recommendations ............................................................................. 48
2.4.1 Research recommendations .................................................................... 49
2.4.2 Validation process.................................................................................... 49
2.4.3 Updating the guideline ............................................................................. 49
2.4.4 Disclaimer ................................................................................................ 50
2.4.5 Funding .................................................................................................... 50
3 Information and support ............................................................................................ 51
3.1 Introduction ......................................................................................................... 51
3.2 Review question .................................................................................................. 51
3.3 Description of included studies ............................................................................ 51
3.4 Evidence profile ................................................................................................... 54
3.5 Evidence statements ........................................................................................... 70
3.6 Health economics profile ..................................................................................... 71
3.7 Evidence to recommendations ............................................................................ 71
3.7.1 Relative value places on the outcomes considered .................................. 71
3.7.2 Consideration of clinical benefits and harms ............................................ 71
3.7.3 Consideration of health benefits and resource uses ................................. 73
3.7.4 Quality of evidence .................................................................................. 73
3.7.5 Other considerations ................................................................................ 73
4 Prevention of preterm birth ....................................................................................... 74
4.1 Introduction ......................................................................................................... 74
4.2 Prophylactic progesterone ................................................................................... 74
4.3 Prophylactic cervical cerclage ............................................................................. 74
4.3.1 Introduction .............................................................................................. 74
4.3.2 Review question....................................................................................... 74
4.3.3 Description of included studies................................................................. 75
4.3.4 Evidence profile ....................................................................................... 75
4.3.5 Evidence statements ................................................................................ 87
4.3.6 Health economics profile .......................................................................... 88
4.3.7 Evidence to recommendations ................................................................. 88
4.3.8 Key conclusions ....................................................................................... 91
4.5 Resarch recommendations .................................................................................. 91
5 Diagnosing preterm prelabour rupture of membranes (P-PROM)........................... 93
5.1 Introduction ......................................................................................................... 93
5.2 Review question .................................................................................................. 93
5
Contents

5.5 Evidence statements ........................................................................................... 96
5.6 Health economics profile ..................................................................................... 96
5.7 Evidence to recommendations ............................................................................ 96
5.7.1 Consideration of clinical benefits and harms ............................................ 96
5.7.2 Relative value placed on the outcomes considered .................................. 96
5.7.3 Consideration of health benefits and resource uses ................................. 97
5.7.4 Quality of evidence .................................................................................. 97
5.7.5 Other considerations ................................................................................ 97
6 Antenatal prophylactic antibiotics for women with P-PROM .................................. 98
6.1 Introduction ......................................................................................................... 98
6.2 Review question .................................................................................................. 98
6.5 Evidence statements ......................................................................................... 109
6.5.1 Any antibiotic therapy compared with placebo ....................................... 109
6.5.2 Antibiotic therapy compared with either placebo or no antibiotic therapy 109
6.5.3 Antibiotic therapy compared with no antibiotic therapy (childhood
outcomes at 7-year follow-up) ................................................................ 110
6.6 Health economics profile ................................................................................... 110
6.7 Evidence recommendations .............................................................................. 110
6.7.1 Relative value placed on the outcomes considered ................................ 110
6.7.2 Consideration of clinical benefits and harms .......................................... 111
6.7.3 Consideration of health benefits and resource uses ............................... 112
6.7.4 Quality of evidence ................................................................................ 112
6.7.5 Other considerations .............................................................................. 113
6.8 Recommendations ............................................................................................ 113
7 Identifying infection in women with P-PROM ......................................................... 114
7.1 Introduction ....................................................................................................... 114
7.2 Review question ................................................................................................ 114
7.3 Description of included studies .......................................................................... 114
7.4 Evidence profile ................................................................................................. 115
7.4.1 Additional data ....................................................................................... 124
7.7 Evidence to recommendations .......................................................................... 126
6
Contents

7.7.6 Key conclusions ..................................................................................... 127
7.9 Research recommendations .............................................................................. 127
8 ‘Rescue’ cervical cerclage ....................................................................................... 129
8.1 Introduction ....................................................................................................... 129
8.2 Review question ................................................................................................ 129
8.7.6 Key conclusions ..................................................................................... 135
9 Diagnosing preterm labour in women with intact membranes ............................. 137
9.1 Introduction ....................................................................................................... 137
9.2 Review question ................................................................................................ 137
9.3.1 Diagnosis using clinical examination ...................................................... 138
9.3.2 Diagnosis using biochemical tests.......................................................... 138
9.3.3 Diagnosis using transvaginal ultrasonography to measure cervical
length ..................................................................................................... 139
9.5.1 Diagnosis using clinical examination ...................................................... 161
9.5.2 Diagnosis using biochemical tests.......................................................... 161
9.5.3 Diagnosis using ultrasound features ...................................................... 162
9.5.4 Diagnosis using combination of tests ..................................................... 163
7
Contents

9.7.6 Key conclusions ..................................................................................... 171
10 Tocolysis................................................................................................................... 171
10.1 Introduction ....................................................................................................... 171
10.2 Review question ................................................................................................ 174
10.3 Description of included evidence ....................................................................... 174
10.3.1 Summary of included studies ................................................................. 175
10.4 Introduction to the new network meta-analysis .................................................. 176
10.4.1 Methods ................................................................................................. 177
10.4.2 Limitations in the data ............................................................................ 178
10.5 Introduction to pair-wise meta-analysis.............................................................. 178
10.6.1 Neonatal mortality .................................................................................. 179
10.6.2 Perinatal mortality .................................................................................. 184
10.6.3 Delay of birth by more than 48 hours ..................................................... 190
10.6.4 Neonatal sepsis ..................................................................................... 196
10.6.5 Intraventricular haemorrhage ................................................................. 201
10.6.6 Discontinuation of treatment due to maternal adverse events ................ 207
10.6.7 Estimated gestational age at birth .......................................................... 212
10.6.8 Respiratory distress syndrome (RDS) .................................................... 217
10.7 Evidence profiles for the pairwise comparisons ................................................. 223
10.8 Evidence statements on NMA results ................................................................ 231
10.9 Evidence statements on pair-wise comparisons ................................................ 233
10.11.1 Relative value placed on the outcomes considered.............................. 234
10.11.2 Consideration of clinical benefits and harms ........................................ 234
10.11.3 Consideration of economic benefits and harms ................................... 236
10.11.4 Quality of evidence .............................................................................. 237
10.11.5 Other considerations............................................................................ 238
10.11.6 Key conclusions ................................................................................... 238
11 Maternal corticosteroids .......................................................................................... 238
11.1 Introduction ....................................................................................................... 238
11.2 Single course of maternal corticosteroids at different gestations ....................... 239
11.2.1 Review question..................................................................................... 239
11.2.2 Description of included studies............................................................... 239
11.2.3 Evidence profile ..................................................................................... 240
8
Contents
11.2.4 Evidence statements .............................................................................. 252

11.2.5 Health economics profile ........................................................................ 254
11.2.6 Evidence to recommendations ............................................................... 254
11.2.10 Quality of evidence .............................................................................. 256
11.2.11 Other considerations............................................................................ 256
11.2.12 Key conclusions ................................................................................... 257
11.2.13 Recommendations ............................................................................... 257
11.3 Repeat course corticosteroids for fetal lung maturation ..................................... 257
12 Magnesium sulfate for neuroprotection ................................................................. 257
12.1 Introduction ....................................................................................................... 257
12.2 Review question ................................................................................................ 258
13 Fetal monitoring ....................................................................................................... 272
13.1 Introduction ....................................................................................................... 272
13.2 Interpretation of the fetal heart rate (FHR) pattern ............................................. 272
13.2.1 Introduction ............................................................................................ 272
13.2.2 Review question..................................................................................... 272
13.2.8 Recommendations ................................................................................. 291
13.3 Monitoring options: cardiotocography and intermittent auscultation ................... 291
13.3.1 Introduction ............................................................................................ 291
9
Contents
13.3.2 Review question..................................................................................... 291

13.4 Fetal scalp electrode ......................................................................................... 299
13.4.1 Introduction ............................................................................................ 299
13.4.2 Review question..................................................................................... 299
13.4.6 Health Economics profile ....................................................................... 299
13.5 Fetal blood sampling ......................................................................................... 301
13.5.1 Introduction ............................................................................................ 301
13.5.2 Review question..................................................................................... 301
13.5.6 Health Economics profile ....................................................................... 302
14 Mode of birth............................................................................................................. 305
14.1 Introduction ....................................................................................................... 305
14.2 Review question ................................................................................................ 305
14.5.1 Neonatal outcomes ................................................................................ 311
14.5.2 Maternal outcomes ................................................................................ 311
10
Contents

14.7.6 Key conclusions ..................................................................................... 314
15 Timing of cord clamping for preterm babies .......................................................... 314
15.1 Introduction ....................................................................................................... 314
15.2 Review question ................................................................................................ 314
15.7.1 Relative value places of the outcomes considered ................................. 328
15.7.3 Consideration of health benefits and resource use................................. 329
15.7.6 Key conclusions ..................................................................................... 330
16 Health economics ..................................................................................................... 331
16.1 What is the clinical effectiveness of prophylactic progesterone (vaginal or oral)
in preventing preterm labour in pregnant women considered to be at risk of
preterm labour and birth? .................................................................................. 331
16.1.1 Review of the literature .......................................................................... 331
16.1.2 Introduction ............................................................................................ 332
16.1.3 Methods ................................................................................................. 333
16.1.4 Diagnostic strategies, test accuracy and prevalence of actual preterm
labour ..................................................................................................... 335
16.1.5 Costs ..................................................................................................... 335
16.1.6 Baseline risk and treatment effectiveness .............................................. 336
16.1.7 Quality adjusted life years (QALY) ......................................................... 336
16.1.8 Results ................................................................................................... 337
16.1.9 Discussion ............................................................................................. 354
16.1.10 Conclusion ........................................................................................... 357
16.2 What is the clinical and cost effectiveness of magnesium sulfate given to
women at high risk of giving birth preterm(defined as those suspected to be in
preterm labour or diagnosed as being in preterm labour and those having
planned preterm birth) for preventing cerebral palsy and other neurological
disorders in babies born at different preterm gestations? .................................. 357
16.2.1 Introduction ............................................................................................ 357
16.2.2 Methods ................................................................................................. 358
16.2.3 Model probabilities and treatment effect size ......................................... 359
11
Contents
16.2.4 Costing and resource use ...................................................................... 360

16.2.5 QALYs ................................................................................................... 361
16.2.6 Sensitivity analysis ................................................................................. 362
16.2.7 Results ................................................................................................... 362
16.2.8 Sensitivity analysis ................................................................................. 366
16.2.9 Discussion ............................................................................................. 376
16.2.10 Conclusion ........................................................................................... 377
16.3 What is the clinical and cost effectiveness of tocolyticsgiven to women with
suspected or diagnosed preterm labour to improve outcomes: .......................... 377
16.3.1 Introduction ............................................................................................ 377
16.3.2 Methods ................................................................................................. 378
16.3.3 Discussion ............................................................................................. 402
16.3.4 Conclusion ............................................................................................. 404
17 Glossary and abbreviations ..................................................................................... 405
17.1 Key terms .......................................................................................................... 405
17.2 Glossary ............................................................................................................ 405
17.3 Abbreviations .................................................................................................... 419
18 References ................................................................................................................ 421
Appendices ...................................................................................................................... 435
12
Introduction
Preterm birth is the single biggest cause of neonatal mortality and morbidity in the UK,
affecting over 52,000 babies (around 7.3% of live births) in England and Wales in 2012.
There has been no decline in the UK preterm birth rate over the last 10 years.
Babies born preterm – that is, before 37+0 weeks of pregnancy – have high rates of early, late
and postneonatal mortality, with the risk of mortality being inversely proportional to
gestational age at birth. Babies who survive have increased rates of disability compared with
babies who are not born preterm. Recent UK studies comparing cohorts born in 1995 and
2006 have shown improved rates of survival (from 40% to 53%) for extreme preterm births
(born between 22 and 26 weeks). Rates of disability among survivors have hardly changed
over this time period, with rates of bronchopulmonary dysplasia, major cerebral scan
abnormality, weight and head circumference (less than 2 standard deviations) being
maintained at 68%, 13%, 44% and 23% respectively, although there has been an increase in
the proportion treated for retinopathy of prematurity from 13% to 22% (Costeloe 2012).
The major long-term consequence of prematurity is neurodevelopmental disability. This can
range from severe motor abnormalities, such as cerebral palsy, through to less severe
cognitive abnormalities (MacKay 2010). Although the risk for the individual child is greatest
for those born at the earliest gestations, the global burden of neurodevelopmental disabilities
is driven by the number of babies born at each of these gestations, and is therefore greatest
for babies born between 32 and 36 weeks, less for those born between 28 and 31 weeks,
and least for those born at less than 28 weeks’ gestation (Blencowe 2013) .
Around 75% of women delivering preterm do so after preterm labour. In the majority of
women with preterm labour, a ‘cause’ is not found, although it is known that a significant
proportion of preterm labours are associated with infection. Preterm labour may or may not
be preceded by preterm prelabour membrane rupture (P-PROM). The remaining women
delivering preterm have undergone elective or iatrogenic preterm delivery when this is
thought to be in the fetal or maternal interest (for example because of extreme growth
restriction in the baby or because of maternal conditions such as pre-eclampsia).
‘Treatments’ for preterm labour include strategies to reduce the risk in women who are at
high risk of preterm labour, tocolytics to delay preterm delivery and extra antenatal strategies
(in addition to normal care) to improve outcomes for babies who will be born preterm.
Risk reduction strategies include the use of progesterone prophylaxis (in which there has
been an increase in interest since the early part of this century) and cervical cerclage.
Tocolytics are used to stop uterine contractions. However, there is considerable variation in
practice and there is little agreement about whether an attempt to delay delivery will improve
outcomes for the baby.
Antenatal strategies to improve outcomes for babies who will be born preterm include the
common practice of using prenatal steroids to improve lung maturation. A more recent
development is the use of magnesium sulfate administered to the mother for neuroprotection
of the baby. Again, there is considerable variation of practice around this latter agent and
little consensus about the subgroup of babies who might benefit.
This guideline reviews the evidence for the care of women who present with signs and
symptoms of preterm labour and those who are scheduled to have a preterm birth. It also
reviews how preterm birth can be optimally diagnosed in symptomatic women, given that
many women thought to be in preterm labour when clinically assessed will not deliver
preterm. Optimal diagnosis can facilitate transfer to a place where appropriate neonatal
intensive care can be provided, a strategy known to improve rates of survival for the baby.
13
Additional areas that will be covered by the guidance (such as information needs for women
who presents with signs and symptoms of preterm labour) are outlined in the guideline
scope.
This guideline does not cover who should and should not have medically indicated preterm
delivery, nor diagnostic or predictive tests in asymptomatic women. These issues have been
reviewed in other NICE guidelines (Diabetes in Pregnancy, Hypertensive Disease in
Pregnancy and Antenatal Care) or will be covered in the NICE guideline on High Risk
Intrapartum Care.
14
Guideline summary
1 Guideline summary
1.1 Guideline Committee membership, NCC-WCH staff
and acknowledgements
1.1.1 Guideline Committee membership
Name Role
Judi Barratt Clinical midwife specialist, Worcester Royal Hospital
Paul Eunson Consultant Paediatric Neurologist & Honorary Senior Lecturer, Royal
Hospital for Sick Children, Edinburgh
Jane Hawdon Consultant Neonatologist, Barts Health NHS Trust
Jane Norman (Chair) Professor of Maternal and Fetal Health, Director of the Tommy's Centre
for Maternal and Fetal Health, University of Edinburgh MRC Centre for
Reproductive Health Queen's Medical Research Institute
Philip Owen Consultant Obstetrician and Gynaecologist, North Glasgow NHS Trust
Jane Plumb Lay member
Farrah Pradhan Lay member
Marianne Rowntree Midwife, Plymouth Hospitals NHS Trust
Meekai To Consultant in Fetal Medicine and Obstetrics, Kings College Hospital
Martin Ward Platt Consultant Paediatrician (neonatal medicine), The Newcastle upon
Tyne Hospitals
Louise Weaver-Lowe Neonatal Nurse, Central Manchester University Hospitals NHS Trust
1.1.2 Acknowledgements
Additional support was received from Ed Peston, Wahab Bello, Sarah Bailey, Sofia Dias,
Norin Ahmed and Karen Packham.
1.1.3 NCC-WCH staff
Name Role
Ebenezer Ademisoye Health Economist (from February 2015)
Zosia Beckles Information Scientist (from October 2014)
Liz Bickerdike Research Assistant (until September 2013)
Shona Burman-Roy Senior Research Fellow
Anne Carty Project Manager (from March 2015)
Melanie Davies Clinical Director for Women’s Health (from December 2014)
Maryam Gholitabar Research Fellow
Paul Jacklin Senior Research Fellow – Health Economist
15
Guideline summary
David James Clinical Director for Women’s Health (until November 2014)
Juliet Kenny Project Manager (until March 2015)
Rosalind Lai Information Scientist (until October 2014)
Hugo Pedder Statistician (from September 2014)
Grammati Sarri Senior Research Fellow – Guideline Lead (from October 2014)
Roz Ullman Senior Research Fellow and Clinical Lead for Midwifery (until May 2014)
Amy Wang Research Fellow (from January 2015 to June 2015)
16
Guideline summary
1.2 Care algorithm
The preventive care section was updated and replaced in 2019. Please see the NICE
website for the updated guideline.
17
Guideline summary
18
Guideline summary
1.3 Recommendations
This section was updated and replaced in 2022. Please see the NICE website for the
updated guideline.
19
Guideline summary
1.4 Key research recommendations

1.4.1 Prophylactic cervical cerclage compared with prophylactic vaginal
progesterone for preventing preterm birth
What is the clinical effectiveness of prophylactic cervical cerclage alone compared with
prophylactic vaginal progesterone alone and with both strategies together for preventing
preterm birth in women with a short cervix and a history of spontaneous preterm birth?
Why is this important
Preterm birth causes significant neonatal morbidity and mortality, as well as long-term
disability. Therefore strategies for preventing preterm birth are important. There are
recognised risk factors for preterm birth, and so interventions can be offered to women with
these risk factors. Both prophylactic cervical cerclage and prophylactic vaginal progesterone
are effective in preventing preterm birth in women with a short cervix and a history of preterm
birth, but there is limited evidence on which is more effective, and the relative risks and
benefits (including costs) of each. More randomised research is needed to compare the
relative effectiveness of prophylactic cervical cerclage and prophylactic vaginal progesterone
in improving both neonatal and maternal outcomes. This will help women and healthcare
professionals to make an informed decision about which is the most effective prophylactic
option.
1.4.2 Identifying infection in women with preterm prelabour rupture of

membranes (P-PROM)
What is the diagnostic accuracy of serial C-reactive protein testing to identify
chorioamnionitis in women with P-PROM?
Identifying infection in women with P-PROM is needed to provide best practice care. Early
diagnosis of infection allows consideration of therapeutic strategies (including antibiotics
and/or early birth). Effective treatment of infection is particularly important given that sepsis is
a common direct cause of maternal death. There is currently limited evidence that serial C-
reactive protein testing might be useful, but the Committee is aware that this strategy is in
common practice. Evidence from diagnostic studies is needed about the accuracy of serial
C-reactive protein testing for identifying chorioamnionitis, which is one of the most common
and serious infective complications of P-PROM.
1.4.3 Effectiveness of ‘rescue’ cerclage

What is the clinical effectiveness of ‘rescue’ cerclage in improving outcomes for women at
risk of preterm birth?
There is some evidence from randomised studies that ‘rescue’ cerclage might be effective in
improving neonatal outcomes in women with a dilated cervix and exposed, unruptured fetal
membranes. However, there is uncertainty about the magnitude of this effect. The full
consequences of this strategy and the subgroups of women at risk of preterm labour who
might particularly benefit are not known. A randomised controlled trial would best address
this question, but a national registry of the most critical outcomes (neonatal mortality and
morbidity, maternal morbidity) could also be considered for women who did not want to
participate in a randomised trial but who opted for ‘rescue’ cerclage.
20
Guideline summary
1.4.4 Magnesium sulfate for neuroprotection: bolus plus infusion

compared with bolus alone
What is the clinical effectiveness of a bolus plus infusion of magnesium sulfate compared
with a bolus alone for preventing neurodevelopmental injury in babies born preterm?
There is evidence from randomised studies that magnesium sulfate has neuroprotective
properties for the baby when given to women who will deliver preterm up to 34+0 weeks of
pregnancy. However, there is uncertainty about the best method of administering magnesium
sulfate for this purpose, with different studies using different strategies. There are significant
advantages for the woman and for reducing healthcare costs if a bolus is as effective as a
bolus plus infusion, because magnesium sulfate has side effects for the woman, and more
monitoring is needed for infusion, with additional associated healthcare costs. A randomised
controlled trial would best address this question by assessing the effects of each method on
neonatal and maternal outcomes.
1.5 Research recommendations

1. What is the clinical effectiveness of prophylactic cervical cerclage alone compared with
prophylactic vaginal progesterone alone and with both strategies together for preventing
preterm birth in women with a short cervix and a history of spontaneous preterm birth?
2. What is the diagnostic accuracy of serial C-reactive protein testing to identify
chorioamnionitis in women with P-PROM?
3. What is the clinical effectiveness of ‘rescue’ cerclage in improving outcomes for women at
risk of preterm birth?
4. What is the clinical effectiveness of a bolus plus infusion of magnesium sulfate compared
with a bolus alone for preventing neurodevelopmental injury in babies born preterm?
5. Is intermittent auscultation or electronic fetal monitoring effective in the preterm fetus?
6. Is there any advantage to preterm babies from delayed versus early cord clamping, or
cord milking?
1.6 Other versions of the guideline

NICE produce a number of versions of this guideline:
• The ‘short guideline’ lists the recommendations, context and recommendations for
research.
• ‘Information for the public’ is written using suitable language for people without specialist
medical knowledge.
• NICE Pathways brings together all connected NICE guidance.
1.7 Schedule for updating the guideline

For the most up-to-date information about guideline reviews, please see the latest version of
the NICE guidelines manual available from the NICE website.
21
Guideline development methodology
2 Guideline development methodology

2.1 Development of the guideline
2.1.1 What is a NICE guideline?
National Institute for Health and Care Excellence (NICE) clinical guidelines are
recommendations for the care of individuals in specific clinical conditions or circumstances
within the NHS – from prevention and self-care through primary and secondary care to more
specialised services. We base our clinical guidelines on the best available research
evidence, with the aim of improving the quality of healthcare. We use predetermined and
systematic methods to identify and evaluate the evidence relating to specific review
questions.
NICE clinical guidelines can:
• provide recommendations for the treatment and care of people by healthcare
professionals
• be used to develop standards to assess the clinical practice of individual healthcare
professionals
• be used in the education and training of healthcare professionals
• help patients to make informed decisions
• improve communication between patients and healthcare professionals.
While guidelines assist the practice of healthcare professionals, they do not replace
their knowledge and skills. We produce our guidelines using the following steps:
• The guideline topic is referred to NICE from the Department of Health.
• Stakeholders register an interest in the guideline and are consulted throughout the
development process.
• The scope is prepared by the National Collaborating Centre for Women’s and
Children’s Health (NCC-WCH).
• The NCC-WCH establishes a Guideline Committee.
• A draft guideline is produced after the committee assesses the available evidence
and makes recommendations.
• There is a consultation on the draft guideline.
• The final guideline is produced.
The NCC-WCH and NICE produce a number of versions of this guideline:
• The ‘full guideline’ contains all the recommendations, together with details of the
methods used and the underpinning evidence.
• The ‘short guideline’ lists the recommendations.
• ‘Information for the public’ is written using suitable language for people without
specialist medical knowledge.
• NICE Pathways brings together all connected NICE guidance.
22
2.1.2 Remit
NICE received the remit for this guideline from the Department of Health. It commissioned
the NCC-WCH to produce the guideline.
The remit for this guideline is to develop a clinical guideline on preterm labour and birth.
2.1.3 Who developed this guideline?

A multidisciplinary Guideline Committee comprising healthcare professionals, researchers
and lay members developed this guideline (see Section 1.1.1).
NICE funds the NCC-WCH and thus supported the development of this guideline. The
committee was convened by the NCC-WCH and chaired by Professor Jane Norman in
accordance with guidance from NICE.
The committee met every 4 to 6 weeks during the development of the guideline. At the start
of the guideline development process all committee members declared interests including
consultancies, fee-paid work, shareholdings, fellowships and support from the healthcare
industry. At all subsequent committee meetings, members declared arising conflicts of
interest.
Members were either required to withdraw completely or for part of the discussion if their
declared interest made it appropriate. The details of declared interests and the actions taken
are shown in Appendix C.
Staff from the NCC-WCH provided methodological support and guidance for the
development process. The team working on the guideline included a project manager,
systematic reviewers, health economists and information scientists. They undertook
systematic searches of the literature, appraised the evidence, conducted meta-analysis and
cost effectiveness analysis where appropriate and drafted the guideline in collaboration with
the committee.
2.1.4 What this guideline covers
2.1.4.1 Groups that will be covered

This guideline covers the following groups:
• pregnant women who are considered to be at risk of preterm labour and birth
because they have a history of:
o spontaneous preterm birth
o preterm prelabour rupture of membranes
o mid-trimester loss
o cervical trauma (including surgery – for example, previous cone biopsy
[cold knife or laser], large loop excision of the transformation zone [LLETZ
– any number] and radical diathermy)
• pregnant women who are considered to be at risk of preterm labour and birth
because they have a short cervix that has been identified on ultrasound scan
and/or bulging membranes in the current pregnancy.
• pregnant women with preterm prelabour rupture of membranes
• pregnant women clinically suspected to be in preterm labour
23
• women diagnosed to be in spontaneous preterm labour

• women having a planned preterm birth.
2.1.4.2 Key clinical issues that will be covered

The following clinical issues are covered in this guideline:
• prophylactic use of vaginal progesterone for women considered to be at risk of

preterm labour and birth because they have any of the factors listed in 2.1.4.1
• prophylactic use of cervical cerclage for women considered to be at risk of
preterm labour and birth because they have any of the factors listed in 2.1.4.1
• non-prophylactic ('rescue') cervical cerclage for women in suspected preterm
labour
• diagnosis of preterm prelabour rupture of membranes using biochemical tests
• diagnosis of preterm labour by clinical assessment, biochemical testing and
cervical ultrasound (alone or in combination)
• routine surveillance (temperature monitoring and cardiotocography) of women
with suspected or diagnosed preterm prelabour rupture of membranes
• antenatal antibiotic prophylaxis for women diagnosed with preterm prelabour
rupture of membranes
• use of progesterone/progestogens for women with suspected or diagnosed
preterm labour to improve the outcomes of preterm labour
• use of tocolytic agents (beta-sympathomimetics, oxytocin receptor antagonists,
calcium channel blockers, cyclo-oxygenase enzyme inhibitors, non-steroidal anti-
inflammatory drugs, nitroglycerin, magnesium sulfate) for women with suspected
or diagnosed preterm labour to improve the outcomes of preterm labour
• pharmacological interventions to improve neonatal outcomes including:
o maternal corticosteroids for fetal lung maturation
o magnesium sulfate for preterm neonatal neuroprotection
• information and support for women at risk of preterm labour, or who are
suspected or diagnosed to be in preterm labour, and women having a planned
preterm birth
• fetal monitoring for women suspected to be in or diagnosed to be in preterm
labour
• mode of birth for women suspected to be in or diagnosed to be in spontaneous
preterm labour
• timing of cord clamping.
For further details please refer to the scope in Appendix A and review questions in
Appendix D.
2.1.5 What this guideline does not cover
2.1.5.1 Groups that will not be covered

This guideline does not cover:
• women in labour at term
• women with a multiple pregnancy.
24
2.1.5.2 Clinical issues that will not be covered

This guideline does not cover:
• routine screening for preterm labour in all pregnant women, including fibronectin
testing
• risk factors for preterm labour
• laparoscopic cerclage
• indications for planned preterm birth
• methods of induction of preterm labour
• mode of birth where this is planned antenatally (for women not in suspected or
diagnosed preterm labour)
• use of intrapartum analgesia
• care of preterm neonates including resuscitation
• additional care that is specific to women with co-existing conditions such as
hypertension, diabetes or blood-borne viruses.
2.1.6 Relationships between this guideline and other NICE guidance
2.1.6.1 Related NICE guidance

Diabetes in pregnancy (2015) NICE clinical guideline NG3
Antibiotics for early-onset neonatal infection (2012) NICE clinical guideline 149
Drainage, irrigation and fibrinolytic therapy (DRIFT) for post-haemorrhagic
hydrocephalus in preterm infants (2011) NICE interventional procedure guidance 412
Multiple pregnancy (2011) NICE clinical guideline 129
Quitting smoking in pregnancy and following childbirth (2010) NICE public health
guidance 26
Pregnancy and complex social factors (2010) NICE clinical guideline 110
Hypertension in pregnancy (2010) NICE clinical guideline 107
Neonatal jaundice (2010) NICE clinical guideline 98 Induction of labour (2008) NICE
clinical guideline 70
Antenatal care (2008) NICE clinical guideline 62
Antenatal and postnatal mental health (2007) NICE clinical guideline 45
Laparoscopic cerclage for prevention of recurrent pregnancy loss due to cervical
incompetence (2007) NICE interventional procedure guidance 228
Postnatal care (2014) NICE clinical guideline 37
Endovascular closure of patent ductus arteriosus (2004) NICE interventional
procedure guidance 97
Cerebral Palsy update (under development) NICE clinical guideline
Vision Amniotic Leak Detector to assess unexplained vaginal wetness in pregnancy
(2013) NICE medical technology guidance MTG15
25
2.2 Methods
This chapter sets out in detail the methods used to review the evidence and to generate
the recommendations that are presented in subsequent chapters. This guidance was
developedin accordance with the methods outlined in the NICE guidelines manual 2012
for the stages up to and including guideline development and then in accordance with
the updated NICE guidelines manual 2014 from the consultation stage.
2.2.1 Developing the review questions and outcomes

Review questions were framed according to the type of question:
• intervention – PICO (patient, intervention, comparison and outcome)
• dianostic test accuracy – population, index tests, reference standard and target condition
for reviews of diagnostic test accuracy
• qualitative – population, area of interest, outcomes.
These frameworks guided the literature searching process, critical appraisal and
synthesis ofevidence and facilitated the development of recommendations by the
committee. The review questions were drafted by the NCC-WCH technical team and
were refined and validated by the committee. The questions were based on the key
clinical areas identified in the scope (Appendix A).
A total of 18 review questions were identified.
Full literature searches, critical appraisals and evidence reviews were completed for all the
specified review questions.
Table 1: Review questions

Chapte Type of review Review questions Outcomes
r
3 Qualitative What additional Maternal/family outcomes
information and support • psychological outcomes
should be given to
women (antenatally or • satisfaction/views of care
during labour) and their • knowing choices
families where the • experience of childbirth
woman is at increased • established
risk of preterm labour, breastfeeding/breastfeeding
or is suspected or ondischarge from hospital
diagnosed to be in
• breastfeeding in longer term
preterm labour, or hasa
(asdefined by research
planned preterm birth?
authors)
• ‘bonding’ with the baby
• plans to have/not to have any
morechildren
• impact on family/siblings –
anyreported
26

r
Neonatal outcomes
any longer term outcomes
4.2 Interventional What is the clinical Maternal outcomes
effectiveness of • maternal mortality
prophylactic
progesterone (vaginalor • side effects/adverse effects
oral) in preventing • emotional/psychological impact/effect
preterm labour in Neonatal outcomes
pregnant women • all mortality
considered to be at risk
• number/proportion of babies
of preterm labour and
bornpreterm
birth?
• time from intervention to birth (delay
to birth was selected as a
surrogate)
• bronchopulmonary
dysplasia/chroniclung disease
• neonatal sepsis
• neurodevelopmental disability
congenital abnormalities
4.3 Interventional What is the clinical Maternal outcomes
prophylactic cervical
cerclage in preventing • maternal adverse effects
preterm labour in includinginfection requiring
women considered to intervention, cervical trauma
be at risk of preterm requiring repair
labour and birth? • maternal
emotional/psychologicalimpact
Neonatal outcomes
• mortality up to 1 year
• interval between procedure
anddelivery
• preterm birth
• serious neonatal morbidity
• sepsis
• chronic lung disease/
bronchopulmonary
dysplasia
• long-term infant
neurodevelopmental
outcomes/neurodevelopmen
tal disability
27

r
5 Diagnostic What is the • sensitivity / specificity
diagnosticaccuracy • positive / negative likelihood ratio
of the following tests
to identify preterm
prelabour rupture of
membranes:
• placental
alpha-
microglobulin-
1
• nitrazine (pH)
• insulin-like
growthfactor
binding protein-
1
• fetal fibronectin
• panty liner
6 Interventional What is the clinical Maternal
antenatal
• any infection
prophylactic
including
antibiotics given to
chorioamnionitis
women with
diagnosedpreterm • major adverse drug reaction
prelabour rupture of Neonatal
membranes to • infant/neonatal/perinatal mortality
improve outcomes of • number/proportion of babies
preterm labour? bornpreterm
• interval between intervention and
delivery (delay to birth was
selectedas a surrogate)
• brain injury
• necrotising enterocolitis
• any neonatal infection
(including.neonatal sepsis)
• cerebral palsy (CP) or other
neurodevelopmental
disability
• any composite neurological outcomes
7 Diagnosti What is the • sensitivity / specificity
c diagnostic value of • positive / negative likelihood ratio
accuracy temperature, pulse,
white cell count,C-
reactive protein and
cardiotocography
(CTG) to identify
infection in women
with preterm
prelabourrupture of
membranes (P-
PROM)?
28

r
8 Interventional What is the clinical Maternal
non-prophylactic
• maternal adverse effects
'rescue' cervical
cerclage in • maternal
preventing preterm emotional/psychologicalimpact
birth for women in Neonatal
suspected preterm • mortality up to 1 year
labour? • interval between procedure and
delivery (delay to birth was
selectedas a surrogate)
• preterm birth
• serious neonatal morbidity
• sepsis
bronchopulmonary
dysplasia
• long-term infant
neurodevelopmentaloutcomes
9 Diagnosti What is the • sensitivity / specificity
c diagnostic accuracy • positive / negative likelihood ratio
accuracy of the following
(alone or in
combination) in
women with intact
membranes to
identifypreterm
labour leadingto
preterm birth:
• clinical assessment
• biochemical testing
for markers for
preterm labour
namely
cervicovaginal fetal
fibronectin and
IGF- BP1 insulin-
like growth factor
bindingprotein-1
• cervical
ultrasound
features (such as
cervical length
andfunnelling)?
29

r
10 Interventional What is the clinical Maternal outcomes
andcost effectiveness • maternal mortality
of tocolytics given to
• adverse events – discontinuation
women with
oftreatment
suspectedor
diagnosed preterm • maternal infection
labour to improve Neonatal outcomes
outcomes: • perinatal mortality
• progesterone/prog • neonatal mortality
estogens • time from administration to birth
• beta- (delayof birth by 48 hours or more
sympathomimeti was selected as a surrogate)
cs • mean gestational age at birth
• oxytocin • respiratory distress syndrome
receptor
• chronic lung
antagonists
disease/bronchopulmonary
• calcium dysplasia
channel
• intraventricular haemorrhage
blockers
• white matter
• cyclo-
injury/periventricular
oxygenase
leucomalacia
enzyme
inhibitors • neonatal infection/sepsis
• non-steroidal • neurodevelopmental disability
anti- inflammatory • quality of life
drugs
• nitric oxide donors
• magnesium
sulfate
effectiveness of a • maternal mortality
single course of
• adverse events – all
maternal
corticosteroids for Neonatal outcomes
fetallung maturation • all deaths up to 1 year
given at different • need for mechanical ventilation
gestations in • bronchopulmonary
improving preterm dysplasia/chroniclung disease
neonatal outcomes?
• neonatal sepsis
What is the clinical
effectiveness of • neurodevelopmental disability
repeatcourses of
maternal
corticosteroids for
fetallung maturation
in improving preterm
neonatal outcomes?
30

r
andcost effectiveness • mortality
of magnesium sulfate
• side effects
given to women at
high risk of giving Neonatal outcomes
birth preterm (defined • stillbirth
as those suspected • neonatal/perinatal mortality
to be in preterm • apgar score <7 at 5 minutes
labour or diagnosed
• need for transfer to neonatal
as being in preterm
intensivecare unit (NICU)
labour and those
having planned • need for mechanical ventilation
preterm birth) for • infant feeding at 1 and 6 weeks
preventing cerebral (breastfeeding or exclusive
palsy and other formula)
neurological • longer term outcomes (record any
disorders in babies thatare reported)
born at different • major neonatal morbidity
preterm gestations? (anyreported)
13.2 Predictiv What are the criteria Maternal outcomes
e for best interpreting • mortality
accurac the preterm fetal
• mode of birth (and indication if
y heartrate trace at
operative delivery, and type of
different gestational
caesarian section [CS]
ages for unborn
incision)
babies whose
mothers are in Neonatal outcomes
suspected or • mortality (all death up to 1 year –
diagnosed preterm includes stillbirth, perinatal
labour? mortality,neonatal mortality and
infant death)
• trauma/injury to infant (specify type)
• intraventricular
haemorrhage/periventricula
r
leucomalacia (PVL)/white matter
injury (and sub-group analysis
neededwhere reported separately)
• neonatal sepsis
• need for mechanical ventilation
• length of stay in neonatal
intensivecare unit or neonatal unit
cord blood gas values at birth
31

r
13.3 Diagnosti What is the clinical Maternal
c effectiveness of • mortality
accuracy electronic fetal
• mode of birth (and indication if
monitoring compared
operative delivery, and type of
with intermittent
CSincision)
auscultation at
differentgestational Neonatal
ages for unborn • mortality (all death up to 1 year –
babies whose includes stillbirth, perinatal
mothers are in mortality,neonatal mortality and
suspected or infant death)
diagnosed preterm • trauma/injury to infant (specify type)
labour? • intracranial or interventricular
haemorrhage/periventricular
leucomalacia (PVL)/white matter
injury(and sub-group analysis needed
where reported separately)
• respiratory distress syndrome
• neonatal sepsis
• cord blood gas values at birth
13.3 Interventional What is the utility of Maternal
fetal blood sampling • mortality
(FBS) as an adjunct
• mode of birth
tointrapartum fetal
heartrate monitoring Neonatal
at different • mortality
gestational ages? • trauma/injury to infant
• neonatal sepsis
• cord blood gas values at birth
0 Interventional For women who Maternal outcomes
present in • maternal mortality
suspected or
• hysterectomy/postpartu
diagnosed preterm
mhaemorrhage
labour (who have not
planned antenatally • infection
to give birth by • sepsis
caesarean section Neonatal outcomes
[CS] and for whom • mortality up to 1 year
there are no other
known indications for
bronchopulmonary
CS birth), what is the
dysplasia
clinical effectiveness
ofdeciding to carry • respiratory distress syndrome
out a CS compared • intracranial haemorrhage
32

r
withdeciding to allow • long-term infant neurodevelopmental
vaginal birth? outcomes
15 Interventional In preterm birth, Maternal outcomes

does later or delayed • mortality
cord clamping in
• primary postpartum haemorrhage
active management
of third stage • retained placenta
improve maternal Neonatal outcomes
and neonatal • neonatal or infant mortality
outcomes compared • anaemia requiring transfusion
with earlier or
• respiratory distress
immediate cord
clamping? • brain injury
• treatment for hyperbilirubinaemia
withblood exchange transfusion
• blood counts at 6 and 12 hours
• Apgar score
2.2.2 Searching for evidence
2.2.2.1 Clinical literature search

During the scoping stage, a search was conducted for guidelines and reports on websites
of organisations relevant to the topic. Searching for grey literature or unpublished
literature was not undertaken. Searches for electronic, ahead of print or ‘online early’
publications were notroutinely undertaken.
Systematic literature searches were undertaken to identify all published clinical
evidencerelevant to the review questions. Searches were undertaken according to the
parametersstipulated within the NICE guidelines manual 2012.
Databases were searched using relevant medical subject headings, free-text terms and
study type filters where appropriate. Studies published in languages other than English
werenot reviewed. Where possible, searches were restricted to retrieve only articles
published in English. All searches were conducted in Medline, Embase and The
Cochrane Library. All searches were updated in March 2015 with the exception of the
search for the review question that included the Network Meta-Analysis (NMA) which
was last updated in January2015. Any studies added to the databases after this date
(even those published prior to this date) were not included unless specifically stated in
33
the text.
Search strategies were quality assured by cross checking reference lists of highly
relevantpapers, analysing search strategies in systematic reviews (SRs) and asking
the committeemembers to highlight any additional studies. The questions, the study
types applied, the databases searched and the years covered can be found in
Appendix E.
The titles and abstracts of records retrieved by the searches were sifted for relevance,
withpotentially significant publications obtained in full text. These were assessed against
the inclusion criteria specified in the protocols (Appendix D).
2.2.2.2 Health economic literature search

Systematic literature searches were also undertaken to identify health economic evidence
within published literature relevant to the review questions. The evidence was identified by
conducting a broad search relating to preterm labour in the NHS Economic Evaluation
Database (NHS EED), the Health Economic Evaluations Database (HEED) and Health
Technology Assessment (HTA) databases with no date restrictions. Additionally, the search
was run on Medline and Embase, using a specific economic filter, from 2011 to ensure
recent publications that had not yet been indexed by the economic databases were
identified. This was supplemented by additional searches that looked for economic papers
specifically relating to gas exchange management as this was an area identified for original
economic modelling. Studies published in languages other than English were not reviewed.
Where possible, searches were restricted to articles published in English.
The search strategies for the health economic literature search are included in Appendix E.
All searches were updated in March 2015. No papers published after this date were
considered.
2.2.3 Evidence of effectiveness

The evidence was reviewed following these steps:
• Potentially relevant studies were identified for each review question from the relevant
search results by reviewing titles and abstracts. Full papers were then obtained.
• Full papers were reviewed against pre-specified inclusion and exclusion criteria to identify
studies that addressed the review question in the appropriate population and reported on
outcomes of interest (review protocols are included in Appendix D).
• Relevant studies were critically appraised using the appropriate checklist as specified in
the NICE guidelines manual 2012. For diagnostic questions the Quality Assessment of
Diagnostic Accuracy Studies (QUADAS‐2) checklist was followed.
• Key information was extracted on the study’s methods, PICO factors and results. These
were presented in summary tables in each chapter and evidence tables (in Appendix H).
• Summaries of evidence were generated by outcome and were presented in committee
meetings:
o randomised studies – data were meta-analysed where appropriate and reported in the
Grading of Recommendations Assessment, Development and Evaluation (GRADE)
profiles (for interventional reviews)
34
o diagnostic/predictive accuracy studies – presented as measures of

diagnostic/predictive test accuracy (sensitivity, specificity, positive and negative
predictive value); a meta-analysis was only conducted when the included studies were
not heterogeneous
o qualitative studies – the themes of the studies were organised in a modified version ofa
GRADE profile, where possible, along with quality assessment otherwise presentedin a
narrative form.
• Of all data extracted, 50% was quality assured by a second reviewer and 50% of the
GRADE quality assessment was quality assured by a second reviewer to minimise any
potential risk of reviewer bias or error.
2.2.3.1 Inclusion and exclusion criteria

The inclusion and exclusion of studies was based on the review protocols, which can be
found in Appendix D. Excluded studies by review question (with the reasons for their
exclusion) are listed in Appendix G. The committee was consulted about any uncertainty
regarding inclusion or exclusion.
2.2.4 Methods of combining clinical studies
2.2.4.1 Data synthesis for intervention reviews

Where possible, meta-analyses were conducted to combine the results of studies for each
review question using Cochrane Review Manager (RevMan5) software or STATA. Fixed-
effects (Mantel–Haenszel) techniques were used to calculate risk ratios (relative risk) for the
binary outcomes.
For the continuous outcomes, measures of central tendency (mean) and variation (standard
deviation) were required for meta‐analysis. A generic inverse variance option in RevMan5
was used if any studies reported solely the summary statistics and 95% confidence interval
(95% CI) or standard error; this included any hazard ratios reported. However, in cases
where standard deviations were not reported per intervention group, the standard error (SE)
for the mean difference was calculated from other reported statistics (probability [p] values or
95% CIs) if available: meta‐analysis was then undertaken for the mean difference and SE
using the generic inverse variance method in RevMan5. When the only evidence was based
on studies that summarised results by presenting medians (and interquartile ranges), or only
p values were given, this information was assessed in terms of the study’s sample size and
was included in the GRADE tables without calculating the relative or absolute effects or as a
narrative summary. Consequently, aspects of quality assessment such as imprecision of
effect could not be assessed for this evidence and this has been recorded in the footnotes of
the GRADE tables. When more than 2 studies reported a continuous outcome, the
presentation of mean (SD) per comparison group was taken by averaging the means of
included studies.
In instances where multiple scales were reported for a single outcome, mean differences
were standardised (divided by their SD) before pooling, giving meta-analysed results that
were reported as standardised mean differences (SMD), with a standard deviation of 1.
35
Where reported, time-to-event data were presented as a hazard ratio or results from a Cox
hazard proportion model were given as a result from a multivariate analysis.
Statistical heterogeneity was assessed by visually examining the forest plots and by
considering the chi-squared test for significance at p less than 0.1 or an I-squared
inconsistency statistic (with an I-squared value of 50–74.99% indicating serious
inconsistency and I-squared value of over 75% indicating very serious inconsistency). If the
heterogeneity still remained, a random effects (DerSimonian and Laird) model was employed
to provide a more conservative estimate of the effect. Where considerable heterogeneity was
present, we set out to perform predefined subgroup analyses based on the following factors:
• different gestational age of fetus
• inclusion of studies with mixed populations of women with single and multiple
pregnancies
• different groups of women at high risk of preterm labour.
2.2.4.2 Data synthesis for diagnostic test accuracy review

For diagnostic test accuracy studies, the outcomes reported were sensitivity, specificity,
positive likelihood ratio and negative likelihood ratio.
The assessment of usefulness of the diagnostic or predictive accuracy of tests followed the
terms and thresholds below:
Sensitivity and specificity:

• high – 90% and above
• moderate – 75% to 89.9%
• low – 74.9% or below.
Positive likelihood ratio:
• very useful – more than 10
• moderately useful – 5 to 10
• not useful – less than 5.
Negative likelihood ratio:
• very useful – 0 to 0.1
• moderately useful – more than 0.1 to 0.5
• not useful – more than 0.5.
2.2.4.3 Data synthesis for qualitative review

For the qualitative review in the guideline, results were presented in 2 ways:
• NICE checklists on assessing qualitative studies were used to assess the quality
assessment of individual studies.
• Results were reported narratively by individual study when appropriate
2.2.4.4 Data synthesis using a network-meta analysis

A network meta-analysis (NMA) was formulated to synthesise direct and indirect
evidence oftreatments’ efficacy to determine which treatments are most effective at
delaying preterm birth to improve the outcomes for the baby with least harm to, and
36
least adverse effects for, the woman while preserving randomisation within primary
studies for the outcomes of:
• neonatal mortality
• perinatal mortality
• respiratory distress syndrome (RDS)
• intraventricular haemorrhage (IVH)
• adverse events requiring discontinuation of treatment
• delay of birth by at least 48 hours
• neonatal sepsis
• gestational age at birth.
Hierarchical Bayesian network meta-analyses (NMAs) were performed using the
softwareWinBUGS version 1.4. These models were based on original work from the
University of Bristol (https://www.bris.ac.uk/cobm/research/mpes/mtc.html).
A class effect model was adopted for the new NMA because it was hypothesised that
treatments within class would borrow similar clinical characteristics and mechanisms
of effect. In other words, results for one member of the class in relation to efficacy and
side effects were considered to be generalisable to other members of that same class.
Since there was no evidence of within-class variability for any of the outcomes
considered, all theresults presented assume that all treatments in a class have the
same relative effect (see Appendix J).
Trials with non UK licensed interventions were included in the NMA to allow the
maximum use of available evidence and borrow strength of loops in the network only if
there was at least 1 trial that included licensed (for preterm labour or for other
conditions) interventions forthe same class. Some other considerations in the design of
the NMA were:
• The committee discussed that although dosage, mode of administration and
timing of treatment may influence the effectiveness of different tocolytics
interventions, it was considered unlikely for this factor to change the
direction of relative effect for the differentinterventions tested in the analysis.
The committee therefore decided not to consider anyconfounding effect of
these factors in the NMA.
• Some of the included studies examined drugs that are not licensed as
tocolytics for use inpregnancy (including nylidrin and barusiban). These
drugs were included in the NMA to increase the size of the network and
because it is not uncommon for drugs that are not licenced for pregnancy
indications to be prescribed for use in this context.
• The committee chose to have separate classes for alcohol/ethanol and
combinationtreatments (classed as ‘other’) in the new NMA.
Standard deviations (SDs) were imputed where they were not reported for 5 studies
assessing estimated gestational age. Imputed values were based on the median SD for each
of these treatments from other included studies. A sensitivity analysis using the upper
quartile of the reported SD was carried out. Apart from increased uncertainty in estimates the
main results were not affected.
37
2.2.4.5 Assessment of consistency

Consistency was assessed by checking the agreement of direct and indirect evidence using
a node-split model.Bayesian p values for agreement between direct and indirect evidence
were calculated. When these were lower than 0.05, included trials were inspected to help
determine reasons for the potential inconsistency, bearing in mind that multiple probabilities
of disagreement are being calculated and there is the potential to find spurious results.
Consistency was considered as part of the quality appraisal of the evidence for the NMA (see
below).
2.2.4.6 Model evaluation

For all the networks set up in the NMA, models for fixed and random effects were developed
and then these were compared based on residual deviance and deviance information criteria
(DIC). The model with the smallest DIC is estimated to be the model that would best predict
a replicate dataset which has the same structure as that currently observed. A small
difference in DIC between the fixed and random effects models (3–5 points) implies that the
better fit obtained by adding random effects does not justify the additional complexity.
However, if the difference in DIC between a fixed and random effect model was less than 5
points, and the models make very similar inferences, then we would report the results from a
fixed effects model as it does not make as many assumptions as the random effect model
and contains fewer parameters, and it is easier for clinical interpretation than the random
effects model.
2.2.5 Type of studies

Systematic reviews (SRs) with or without meta-analyses were considered the highest quality
evidence to be selected for inclusion. Individual patient data (IPD) meta-analyses are
considered the gold standard type of meta-analysis and were prioritised for inclusion in the
evidence base of this guideline when appropriate.
Randomised trials and observational studies (including diagnostic or prognostic studies)
were included in the evidence reviews as appropriate.
Literature reviews, posters, letters, editorials, comment articles, conference abstracts,
unpublished studies and studies not in English were excluded.
For intervention reviews in this guideline, randomised controlled trials (RCTs) were included
because they are considered the most robust study design for unbiased estimation of
intervention effects.
If the committee believed RCT data were not appropriate or there was limited evidence from
RCTs, well-conducted non‐randomised comparative studies were included. For diagnostic
reviews, cross‐sectional, retrospective studies and case series were included. Please refer to
Appendix D for full details on the study design of studies selected for each review question.
The committee defined primary outcomes as women’s and babies’ mortality and birth within
48 hours and within 7 days; and secondary outcomes as long-term infant
neurodevelopmental outcomes, birth events (mode of birth, complications of birth, perineal
trauma), newborn events (condition at birth, birth injuries, admission to neonatal units) and
women’s assessment of birth experience. The committee considered other outcomes when
they were relevant to specific questions.
38
2.2.6 Appraising the quality of evidence by outcomes

The evidence for outcomes from the included RCTs and, where appropriate, observational
studies was evaluated and presented using an adaptation of the GRADE toolbox developed
by the international GRADE working group. The software developed by the GRADE working
group (GRADEpro) was used to assess the quality of each outcome, taking into account
individual study quality factors and the meta-analysis results and clinical evidence profile
tables were generated. The clinical evidence profile table includes details of the quality
assessment and pooled outcome data, where appropriate, an absolute measure of
intervention effect and the summary of quality of evidence for that outcome. In this table, the
columns for intervention and control indicate summary measures and measures of dispersion
(such as mean and standard deviation or median and range) for continuous outcomes and
frequency of events (n/N: the sum across studies of the number of patients with events
divided by sum of the number of completers) for binary outcomes.
The evidence for each outcome was examined separately for the quality elements listed and
defined in Table 2. Each element was graded using the quality levels listed in Table 3.
The main criteria considered in the rating of these elements are discussed below. Footnotes
were used to describe reasons for grading a quality element as having serious or very
serious limitations. The ratings for each component were summed to obtain an overall
assessment for each outcome (Table 2).
The GRADE toolbox is currently designed only for randomised trials and observational
studies but we adapted the quality assessment elements and outcome presentation for
diagnostic accuracy and prognostic studies subject to data availability.
Table 2: Description of quality elements in GRADE for intervention studies

Description of quality elements in GRADE for interventionstudies
Quality element
Risk of bias (study Limitations in the study design and implementation may bias the
limitations) estimates of the treatment effect. High risk of bias for the majority ofthe
evidence decreases confidence in the estimate of the effect.
Inconsistency Inconsistency refers to an unexplained heterogeneity of results.
Indirectness Indirectness refers to differences in study population, intervention,
comparator and outcomes between the available evidence and the
review question, or recommendation made, such that the effect
estimate is changed.
Imprecision Results are imprecise when studies include relatively few patients and
few events and thus have wide confidence intervals around the
estimate of the effect. Imprecision results if the confidence interval
includes the clinically important threshold.
Publication bias Publication bias is a systematic underestimate or an overestimate ofthe
underlying beneficial or harmful effect due to the selective publication
of studies.
39
Table 3: Levels of quality elements in GRADE Level

Levels of quality
elementsin GRADE Description
level
None There are no serious issues with the evidence.
Serious The issues are serious enough to downgrade the outcome evidence by
1 level.
Very serious The issues are serious enough to downgrade the outcome evidence by
2 levels.
Table 4: Overall quality of outcome evidence in GRADE Level

Overall quality of
outcome evidence in
GRADE level Description
High Further research is very unlikely to change our confidence in the
estimate of effect.
Moderate Further research is likely to have an important impact on our confidence
in the estimate of effect and may change the estimate.
Low Further research is very likely to have an important impact on our
confidence in the estimate of effect and is likely to change the
estimate.
Very low Any estimate of effect is very uncertain.
2.2.6.1 Grading the quality of clinical evidence

After results were pooled, the overall quality of evidence for each outcome was considered.
The following procedure was adopted when using GRADE:
• A quality rating was assigned based on the study design. RCTs start as high,
observational studies as moderate and uncontrolled case series as low or very
low.
• The rating was then downgraded for the specified criteria: risk of bias (study
limitations); inconsistency; indirectness; imprecision; and publication bias. These
criteria are detailed below. Evidence from observational studies (which had not
previously been downgraded) was upgraded if there was a large magnitude of
effect or a dose-response gradient, and if all plausible confounding would reduce
a demonstrated effect or suggest a spurious effect when results showed no effect.
Each quality element considered to have ‘serious’ or ‘very serious’ risk of bias
was rated down by 1 or 2 points respectively.
• The downgraded/upgraded ratings were then summed and the overall quality
rating was revised. For example, all RCTs started as high and the overall quality
became moderate, low or very low if 1, 2 or 3 points were deducted respectively.
• The reasons or criteria used for downgrading were specified in the footnotes.
40
The details of the criteria used for each of the main quality elements are discussed further in
Sections 2.2.6.2 to 2.2.6.6.
2.2.6.2 Risk of bias

Bias can be defined as anything that causes a consistent deviation from the truth. Bias can
be perceived as a systematic error; for example if a study was carried out several times and
there was a consistently wrong answer, the results would be inaccurate.
The risk of bias for a given study and outcome is associated with the risk of over‐ or
underestimation of the true effect.
The risks of bias are listed in Table 5.
A study with a poor methodological design does not automatically imply high risk of bias; the
bias is considered individually for each outcome and it is assessed whether this poor design
will impact on the estimation of the intervention effect.
Table 5: Risk of bias in randomised controlled trials

Risk of bias Explanation
Allocation concealment Those enrolling patients are aware of the group to which the next
enrolled patient will be allocated (this is a major problem in
‘pseudo’ or ‘quasi’ randomised trials with allocation by, for
example, day of week, birth date, chart number).
Lack of blinding Patient, caregivers, those recording outcomes, those
adjudicating outcomes or data analysts are aware of the arm to
which patients are allocated. As mortality is the most critical
outcome for this guideline and its effect is not biased by lack of
blinding, unblinded studies were not automatically downgraded
for this outcome.
Incomplete accounting of Missing data not accounted for and failure of the trialists to
patients and outcome events adhere to the intention to treat principle when indicated.
Selective outcome reporting Reporting of some outcomes and not others on the basis of the
results.
Other risks of bias For example:
• stopping early for benefit observed in randomised trials, in
particular in the absence of adequate stopping rules
• use of unvalidated patient-reported outcome
• recruitment bias in cluster randomised trials.
2.2.6.3 Diagnostic studies

For diagnostic accuracy studies, the QUADAS version 2 (QUADAS‐2) checklist was used
(see Appendix F in The guidelines manual 2012). Risk of bias and applicability in primary
diagnostic accuracy studies in QUADAS‐2 consists of 4 domains (see Figure 1):
• patient selection
• index test
• reference standard
• flow and timing.
41
Figure 1: Description of QUADAS-2 domains
2.2.6.4 Inconsistency
Inconsistency refers to an unexplained heterogeneity of results. When estimates of the
treatment effect across studies differ widely (that is, when there is heterogeneity or variability
in results), this suggests true differences in underlying treatment effect.
Heterogeneity in meta‐analyses was examined and sensitivity and subgroup analyses
performed as pre‐specified in the protocols (Appendix D).
When heterogeneity existed (chi-squared p less than 0.1, I-squared inconsistency statistic of
between 50% and 74.99% or I-squared greater than 50% or evidence from examining forest
42
plots) but no plausible explanation was found (for example duration of intervention or
different follow-up periods) the quality of evidence was downgraded by 1 or 2 levels,
depending on the extent of uncertainty to the results contributed by the inconsistency in the
results. In addition to the I-squared and chi-squared values, the decision for downgrading
was also dependent on factors such as whether the intervention is associated with benefit in
all other outcomes or whether the uncertainty about the magnitude of benefit (or harm) of the
outcome showing heterogeneity would influence the overall judgment about net benefit or
harm (across all outcomes).
When outcomes are derived from a single trial, inconsistency is not an issue for downgrading
the quality of evidence. However, ‘no inconsistency’ is nevertheless used to reflect the
decision not to downgrade the evidence for this quality assessment domain.
2.2.6.5 Indirectness
Directness refers to the extent to which the populations, intervention, comparisons and
outcome measures are similar to those defined in the inclusion criteria for the reviews.
Indirectness is important when these differences are expected to contribute to a difference in
effect size or may affect the balance of harms and benefits considered for an intervention.
2.2.6.6 Imprecision
Imprecision in guidelines concerns whether the uncertainty (confidence interval) around the
effect estimate means that it is not clear whether there is a clinically important difference
between interventions or not. Therefore, imprecision differs from the other aspects of
evidence quality in that it is not really concerned with whether the point estimate is accurate
or correct (has internal or external validity) but instead is concerned with the uncertainty
about what the point estimate is. This uncertainty is reflected in the width of the confidence
interval.
The 95% confidence interval (95% CI) is defined as the range of values that contain the
population value with 95% probability. The larger the trial, the smaller the 95% CI and the
more certain the effect estimate.
Imprecision in the evidence reviews was assessed by considering whether the width of the
95% CI of the effect estimate was relevant to decision‐making, considering each outcome in
isolation.
When the confidence interval of the effect estimate is wholly contained in 1 of the 3 zones
(clinically important benefit, clinically important harm, no clinically important benefit or harm)
we are not uncertain about the size and direction of effect (whether there is a clinically
important benefit, or the effect is not clinically important, or there is a clinically important
harm), so there is no imprecision (Figure 2).
When a wide confidence interval lies partly in each of 2 zones, it is uncertain in which zone
the true value of effect estimate lies and therefore there is uncertainty over which decision to
make (based on this outcome alone). The confidence interval is consistent with 2 decisions
and so this is considered to be imprecise in the GRADE analysis and the evidence is
downgraded by 1 level (‘serious imprecision’).
If the confidence interval of the effect estimate crosses into 3 zones, this is considered to be
very imprecise evidence because the confidence interval is consistent with 3 clinical
43
decisions and there is a considerable lack of confidence in the results. The evidence is
therefore downgraded by 2 levels in the GRADE analysis (‘very serious imprecision’).
Implicitly, assessing whether the confidence interval is in, or partially in, a clinically important
zone, requires the committee to estimate a minimally important difference (MID) or to say
whether they would make different decisions for the 2 confidence limits.
The committee considered it clinically acceptable to use the GRADE default MID to assess
imprecision: a 25% relative risk reduction or relative risk increase was used, which
corresponds to clinically important thresholds for a risk ratio of 0.75 and 1.25 respectively.
This default MID was used for all the dichotomous outcomes in the interventions evidence
reviews. For continuous outcomes, a MID was calculated by adding or subtracting 0.5 times
standard deviations. For outcomes that were meta-analysed using the standardised mean
difference approach (SMD), the MID was calculated by adding or subtracting 0.5 (given SD
equals 1).
For the diagnostic questions, we assessed imprecision on the outcome of positive or
negative likelihood ratio because these were prioritised by the committee as the most
important diagnostic outcomes for their decision- making. The assessment of imprecision for
the results on positive or negative likelihood ratio followed the same concept as used in
interventional reviews. For example, if the 95% CI of the positive likelihood ratio crossed
2 zones (from moderately useful [5 to 10] to very useful [greater than10]) then imprecision
was downgraded by 1, or if crossed 3 zones (not useful [less than 5], moderately useful [5 to
10] and very useful [greater than 10] then imprecision was downgraded by 2.
Figure 2: Illustration of imprecision assessment
2.2.7 Quality assessment of NMA

The quality of evidence from NMA was assessed using a modified GRADE appraisal
process.
Risk of bias was assessed using the quality assessment undertaken by Haas 2012 and for
all additional studies using the checklist developed by the Technical Support Unit (TSU)
commissioned by NICE.
Indirectness was assessed using information about the study population and imprecision
based on credible intervals in line with standard GRADE methodology. Inconsistency was
44
assessed by comparing estimates based on direct and indirect data included in the network.
Where there was evidence of inconsistency (see Appendix J) then quality was downgraded.
Imprecision was assessed based on the credible interval within each comparison. Data were
downgraded if a credible interval crossed the two default MIDs or majority of the
comparisons.
2.2.8 Assessing clinical importance

The committee assessed the evidence by outcome in order to determine if there was, or
potentially was, a clinically important benefit, a clinically important harm or no clinically
important difference between interventions. To facilitate this, binary outcomes were
converted into absolute risk differences (ARDs) using GRADEpro software: the median
control group risk across studies was used to calculate the ARD and its 95% CI from the
pooled risk ratio.
2.2.9 Evidence statements

Evidence statements are summary statements that are presented after the GRADE profiles,
summarising the key features of the clinical evidence presented. The wording of the
evidence statements reflects the certainty or uncertainty in the estimate of effect. The
evidence statements are presented by comparison (for intervention reviews) or by outcome
and encompass the following key features of the evidence:
• the number of studies and the number of participants for a particular outcome
• a brief description of the participants
• an indication of the direction of effect (if a particular treatment is beneficial or harmful
compared with the other, or whether there is no difference between the 2 tested
treatments)
• a description of the overall quality of evidence (GRADE overall quality).
2.3 Evidence of cost effectiveness

The committee is required to make decisions based on the best available evidence of both
clinical and cost effectiveness. Guideline recommendations should be based on the expected
costs of the different options in relation to their expected health benefits (that is, their ‘cost
effectiveness’) rather than the total implementation cost. Thus, if the evidence suggests that
a strategy provides significant health benefits at an acceptable cost per patient treated, it
should be recommended even if it would be expensive to implement across the whole
population.
Evidence on cost effectiveness related to the key clinical issues being addressed in the
guideline was sought, a systematic review of the published economic literature was
undertaken and a new cost effectiveness analysis was conducted in priority areas.
2.3.1 Literature review

The health economist:
45
• identified potentially relevant studies for each review question from the economic search
results by reviewing titles and abstracts and full papers were then obtained
• reviewed full papers against pre-specified inclusion/exclusion criteria to identify relevant
studies (see Section 2.3.1.1 for details)
• critically appraised relevant studies using the economic evaluations checklist as specified
in the guidelines manual
• extracted key information about study methods and results into evidence tables (included
in Appendix H)
• generated summaries of the evidence in NICE economic evidence profiles (included in
the relevant chapter for each review question) – see Section 2.3.1.2 for details.
2.3.1.1 Inclusion and exclusion criteria

Full economic evaluations (studies comparing costs and health consequences of alternative
courses of action: cost–utility, cost effectiveness, cost–benefit and cost–consequence
analyses) and comparative costing studies that addressed the review question in the relevant
population were considered for inclusion as economic evidence.
2.3.1.2 NICE economic evidence profiles

The NICE economic evidence profile has been used to summarise cost and cost
effectiveness estimates. The economic evidence profile shows, for each economic study, an
assessment of applicability and methodological quality for each economic evaluation. These
assessments were made by the health economist using the economic evaluation checklist
from The Guidelines Manual 2012. It also shows the incremental costs, incremental effects
(for example quality-adjusted life years [QALYs]) and the incremental cost effectiveness ratio
for the base-case analysis in the evaluation, as well as information about the assessment of
uncertainty in the analysis. See Table 6 for more details.
Table 6: Content of NICE economic evidence profile

Content of NICE economic evidenceprofile item
Description
Study First author name, reference, date of study publication
and countryperspective.
Applicability An assessment of applicability of the study to the clinical
guideline, thecurrent NHS situation and NICE decision-
makinga:
• Directly applicable – the study meets all applicability
criteria, or failsto meet 1 or more applicability criteria
but this is unlikely to change the conclusions about
cost effectiveness.
• Partially applicable – the study fails to meet 1 or
more applicabilitycriteria and this could change the
conclusions about cost effectiveness.
• Not applicable – the study fails to meet 1 or more
applicability criteriaand this is likely to change the
conclusions about cost effectiveness. Such studies
would usually be excluded from the review.
46
Content of NICE economic evidenceprofile item
Description
Limitations An assessment of methodological quality of the studya:
• Minor limitations – the study meets all quality criteria,
or fails to meet 1 or more quality criteria, but this is
unlikely to change the conclusions about cost
effectiveness.
• Potentially serious limitations – the study fails to
meet 1 or more quality criteria and this could
change the conclusion about cost effectiveness.
• Very serious limitations – the study fails to meet 1 or
more quality criteria and this is highly likely to
change the conclusions about cost effectiveness.
Such studies would usually be excluded from the
review.
Other comments Particular issues that should be considered when
interpreting the study.
Incremental cost The mean cost associated with 1 strategy minus the
mean cost of a comparator strategy.
Incremental effects The mean QALYs (or other selected measure of health
outcome) associated with 1 strategy minus the mean
QALYs of a comparator
strategy.
Cost effectiveness Incremental cost effectiveness ratio (ICER): the
incremental cost divided by the incremental effects.
Uncertainty A summary of the extent of uncertainty about the ICER
reflecting the results of deterministic or probabilistic
sensitivity analyses, or stochastic analyses of trial data,
as appropriate.
a. Applicability and limitations were assessed using the economic evaluation
checklist from the guidelinesmanual.
2.3.2 Undertaking new health economic analysis

As well as reviewing the published economic literature for each most review questions, as
described above, new economic analysis was undertaken by the health economist in
selected areas. Priority areas for new health economic analysis were agreed by the
committee after formation of the review questions and consideration of the available health
economic evidence.
2.3.2.1 Cost effectiveness criteria

NICE’s report Social value judgements: principles for the development of NICE guidance
sets out the principles that committees should consider when judging whether an intervention
offers good value for money. In general, an intervention was considered to be cost effective if
either of the following criteria applied (given that the estimate was considered plausible):
47
• The intervention dominated other relevant strategies (that is, it was both less costly in
terms of resource use and more clinically effective compared with all the other relevant
alternative strategies).
• The intervention cost less than £20,000 per QALY gained compared with the next best
strategy.
If the committee recommended an intervention that was estimated to cost more than £20,000
per QALY gained or did not recommend one that was estimated to cost less than £20,000
per QALY gained, the reasons for this decision are discussed explicitly in the
‘Recommendations and link to evidence’ section of the relevant chapter with reference to
issues regarding the plausibility of the estimate or to the factors set out in Social value
judgements: principles for the development of NICE guidance’ guidance.
If a study reported the cost per life year gained but not QALYs, the cost per QALY gained
was estimated by multiplying by an appropriate utility estimate to aid interpretation. The
estimated cost per QALY gained is reported in the economic evidence profile with a footnote
detailing the life years gained and the utility value used. When QALYs or life years gained
are not used in the analysis, results are difficult to interpret unless a particular strategy
dominates the others with respect to every relevant health outcome and cost.
2.3.3 In the absence of economic evidence

When no relevant published studies were found and a new analysis was not prioritised, the
committee made a qualitative judgement about cost effectiveness by considering expected
differences in resource use between options and relevant UK NHS unit costs alongside the
results of the clinical review of effectiveness evidence.
2.4 Developing recommendations

Over the course of the guideline development process, the committee was presented with:
• evidence tables of the clinical and economic evidence reviewed from the literature (all
evidence tables are in Appendix H)
• summary of clinical and economic evidence and quality assessment (as presented in
Chapters 3 to 15)
• forest plots (Appendix I)
• a description of the methods and results of the cost effectiveness analysis undertaken for
the guideline (Chapter 16).
Recommendations were drafted on the basis of the committee’s interpretation of the
available evidence, taking into account the balance of benefits, harms and costs between
different courses of action. Firstly, the net benefit over harm (clinical effectiveness) was
considered, focusing on the prioritised outcomes and taking into account the clinical benefits
and harms when one intervention was compared with another. The assessment of net
benefit was moderated by the importance placed on the outcomes (the committee’s values
and preferences) and the confidence the committee had in the evidence (evidence quality).
Secondly, it was assessed whether the net benefit justified any differences in costs.
In areas where no substantial clinical research evidence was identified, the committee
considered other NICE relevant guidelines and consensus statements or used their collective
experience to identify good practice. The health economics justification in areas of the
guideline where the use of NHS resources (interventions) was considered was based on
48
committee consensus in relation to the likely cost effectiveness implications of the

recommendations. The committee also identified areas where evidence to answer their
review questions was lacking and used this information to formulate recommendations for
future research. When clinical and economic evidence was of poor quality, conflicting or
absent, the committee members drafted recommendations based on their expert opinion.
The considerations for making consensus-based recommendations include the balance
between potential harms and benefits, the economic costs or implications compared with the
economic benefits, current practices, recommendations made in other relevant guidelines,
patient preferences and equality issues.
The wording of recommendations was agreed by the committee and focused on the
followingfactors:
• the actions healthcare professionals need to take
• the information readers need to know
• the strength of the recommendation (for example the word ‘offer’ was used for
strongrecommendations and ‘consider’ for weak recommendations)
• the involvement of patients (and their carers if needed) in decisions on treatment and
care
• consistency with NICE’s standard advice on recommendations about drugs, waiting
timesand ineffective interventions.
The main considerations specific to each recommendation are outlined in the
‘Evidence torecommendations’ sections within each chapter.
2.4.1 Research recommendations

When areas were identified for which good evidence was lacking, the committee considered
making recommendations for future research. Decisions about inclusion were based on
factors such as:
• the importance to patients or the population

• national priorities
• potential impact on the NHS and future NICE guidance
• ethical and technical feasibility.
2.4.2 Validation process

This guidance is subject to a 6-week public consultation and feedback as part of the quality
assurance and peer review of the document. All comments received from registered
stakeholders are responded to in turn and posted on the NICE website when the pre-
publication check of the full guideline occurs.
2.4.3 Updating the guideline

Following publication, and in accordance with the NICE guidelines manual, NICE will
undertake a review of whether the evidence base has progressed significantly to alter the
guideline recommendations and warrant an update.
49
2.4.4 Disclaimer
Healthcare providers need to use clinical judgement, knowledge and expertise when
deciding whether it is appropriate to apply guidelines. The recommendations cited here are a
guide and may not be appropriate for use in all situations. The decision to adopt any of the
recommendations cited here must be made by practitioners in light of individual patient
circumstances, the wishes of the patient, clinical expertise and resources.
The National Collaborating Centre for Women and Children’s Health disclaims any
responsibility for damages arising out of the use or non-use of these guidelines and the
literature used in support of these guidelines.
2.4.5 Funding
The National Collaborating Centre for Women and Children’s Health (NCC-WCH) was
commissioned by NICE to undertake the work on this guideline.
50
Information and support
3 Information and support

3.1 Introduction
Although many preterm babies arrive early without warning, some pregnancies are known to
be at high risk of ending in preterm birth: when this is the case, there may be opportunities
for steps to be taken to try to reduce that risk. The earlier a baby is born, the more severe his
or her health problems are likely to be. Some babies born preterm require special care and
may spend weeks or months hospitalised in neonatal intensive care units.
At such a vulnerable time in their lives, families need to be given information and support to
meet their needs, without causing unnecessary anxiety. This section considers the evidence
for providing information and support to pregnant women and their families whose babies are
at high risk of arriving early.
3.2 Review question

What additional information and support should be given to women (antenatally or during
labour) and their families where the woman is at increased risk of preterm labour, or is
suspected or diagnosed to be in preterm labour, or has a planned preterm birth?
This review question has two sections. The first section aims to identify themes of additional
information given prior to birth that would be considered as important for women at increased
risk, or suspected or diagnosed as being in preterm labour or has planned a preterm birth.
The second section examines whether interventions or packages of care designed to provide
additional information prior to or during preterm birth compared with usual care could result in
better maternal, family and/or neonatal outcomes.
3.3 Description of included studies

Seven studies were included in this review. Five qualitative studies (Gauche 2011, Gupton
1994, Sawyer 2013, Young 2012, Griffin 1997) were included in the first section of the
review, which aims to identify the key areas of antenatal and intrapartum information and
support needs of women and their families at increased risk for preterm labour. Two
randomised controlled trials (RCTs) (Oakley 1990, Villar 1992) were included in the second
section of the review, which aimed to assess the effectiveness of interventions or packages
of care with regard to maternal, family and/or neonatal outcomes. The women in the included
studies had pregnancies between 15 to 36 weeks at the time.
More information on the study characteristics and population characteristics are given in
Table 7.
Table 7: Study and population characteristics

Type of study/methods/
Included comparison groups (if Population
studies applicable) characteristics Outcomes
Section 1 of review; key areas of information for families at risk for preterm birth
Gaucher 2011 Qualitative study/ face-to- 5 mothers Explored the women’s
face interviews using Range of gestational concerns regarding
grounded theory age: 26 to 30 2/7 weeks possible preterm labour
3/5 had full term and their expectations of
pregnancies after
51
Type of study/methods/
Included comparison groups (if Population
studies applicable) characteristics Outcomes
hospital discharge the prenatal consultation
and of the neonatologist
Gupton 1994 Qualitative study using • Convenience Ranking ordering of
Preterm Birth Learning sample of 34 priorities for learning
Needs Questionnaire women needs of hospitalised
(PBLNQ) • Range of women at risk of preterm
gestational age: birth
26–36 weeks
• All high
risk
pregnanci
es
Sawyer 2013 Qualitative study/interview • 25 mothers Experiences and
with open-ended questions and 7 couples satisfaction with care
• Range of during preterm birth
gestational age:
24–32
Young 2012 Qualitative study/ face-to- • Preterm labour Exploring the areas of
face semi-structured between 23 and importance for
interviews using 26 weeks counselling for extreme
ethnography methods • 10 families prematurity
• 80% high
risk
pregnancie
s
Griffin 1997 Qualitative study/ face-to- • Convenience Evaluating the experience
face interviews with open sample of 13 of a prenatal tour of the
ended questions parents (10 neonatal intensive care
mothers) unit during high risk
• All high pregnancy
risk
pregnanci
es
Section 2 of review: effectiveness of interventions or packages of antenatal care for
pregnancies at risk of preterm birth
Oakley 1990 RCT/intervention: a • Intervention: Postnatal depression
minimum of 3 home visits 255 control:
from a midwife at 14, 20, 254
and 28 weeks’ gestation, • At risk pregnancies
plus 2 telephone contacts
Villar 1992 RCT/intervention: a • Intervention: Satisfaction with
minimum of 4 home visits 1115 control: antenatal care
from specially trained 1120
female social workers or • At risk pregnancies
obstetrical nurses and had
access to a special 'drop in'
support office at each study
hospital
52
Section 1 – themes of additional information given prior to birth

All qualitative studies employed appropriate study methods.
Preterm Birth Learning Needs Questionnaire (PBLNQ) was included in 1 study to elicit both
qualitative and quantitative data (Gupton 1994), grounded theory in another (Gaucher 2011),
qualitative ethnography using semi-structured interviews was employed in another study
(Young 2012) and the last study used semi-structured interviews (Sawyer 2013).
Two qualitative studies examined cases of extreme prematurity (Sawyer 2013, Young 2012).
The first study assessed parents’ positive and negative experiences and satisfaction with
care during very preterm births (less than 32 gestational weeks), identifying determinants of
their experience of care (Sawyer 2013). The second study interviewed parents with babies
born at 23–26 gestational weeks to ascertain retrospectively how pre-delivery counselling
could be improved (Young 2012).
One study (Griffin 1990) qualitatively examined a specific intervention (prenatal tour of the
neonatal intensive care unit during high-risk pregnancy) to elicit parents’ experiences
regarding this.
Section 2 – interventions or packages of care designed to provide additional
information prior to or during birth compared with usual care
Two RCTs (Oakley 1990, Villar 1992) included in a Cochrane review (Hodnett 2010) that
compared routine care provision with provision of additional support to reduce the likelihood
of preterm birth or low birth weight in pregnant women matched the second part of this
review protocol. One RCT (Oakley 1990) was conducted in the UK and usual antenatal care
was compared with the addition of a social support intervention consisting of, at a minimum,
3 home visits from a midwife (at 14, 20 and 28 weeks’ gestation) plus 2 telephone contacts
or brief home visits between these times. The midwife was also on call to provide support to
mothers if necessary for 24 hours a day.
The second RCT (Villar 1992) was an international multicentre study that aimed at increasing
social support and reducing stress and anxiety in pregnancy. Women in the control group
received standard antenatal care whereas women in the intervention group received a
minimum of 4 home visits from specially trained female social workers or obstetrical nurses
and had access to a special 'drop in' support office at each study hospital. The purpose of
both interventions was to strengthen the woman’s social network and provide direct
emotional support and health education.
The committee anticipated that information and support needs might vary for women in
different clinical scenarios and hence that specific recommendations might be needed for
women:
• at increased risk of preterm labour (the risk could be either known prior to conception,
early in pregnancy or later in pregnancy) and who may be having a planned preterm birth
• who are suspected or diagnosed to be in preterm labour (where preterm birth had not
been expected).
However, the information included in the selected studies did not allow for further stratified
analysis based on these different clinical scenarios.
53
3.4 Evidence profile

The search strategies for this chapter can be found in Appendix E, the excluded studies in
Appendix G, the evidence tables in Appendix H and the forest plots in Appendix I.
Section 1
Information is presented in the following tables:
• Table 8: Quality assessment of qualitative studies

• Table 9: Information needs of hospitalised women at high risk of preterm birth (Gupton
1994, n=34)
• Table 10: Aspects of hospitalised women’s stressful experience of their possible preterm
labour (Gaucher 2011, n=5)
• Table 11: The expectations of women hospitalised for preterm labour regarding the
prenatal consultation (Gaucher 2011, n=5)
• Table 12: The expectations of women hospitalised for preterm labour regarding the
neonatologist (Gaucher 2011, n=5)
• Table 13: Parents’ views of a prenatal tour of a neonatal intensive care unit (NICU)
during a high-risk pregnancy (Griffin 1997, n=13)
• Table 14: Information and support needs of women during the birth of their preterm baby
(Sawyer 2013, n=39)
• Table 15: Pre-delivery counselling experiences and information and support needs of
parents with babies born between 23 and 26 gestational weeks (Young 2012, n=10)
Section 2
Information is presented in the following table:
• Table 16: GRADE findings for the comparison of antenatal information/support
intervention with routine care in women with a high risk of preterm birth
Full description of the characteristics and results of the included studies can be found in the
evidence tables in Appendix H. A summary quality assessment for each qualitative study is
given in Table 8.
Table 8: Quality assessment of qualitative studies

Results
transferrab
le to the
population
specified in
Populatio Analysi the
Study n Methods s Relevance to guideline population protocol
Gauche Well Well Well Canada. Very
r 2011 reported1 reported reported Women aged between 24 and 36 unlikely
years, gestational age of 26–30+2
weeks, admitted for preterm labour
with diverse
reasons for hospitalisation and
diverse social backgrounds
Griffin Well Well Well USA. Unlikely
1997 reported reported reported Women aged between 20 and 42
years, with high-risk pregnancies and
with the
54
Results
transferrab
le to the
population
specified in
Populatio Analysi the
Study n Methods s Relevance to guideline population protocol
majority married and completed high
school education.
Gupton Well Well Well Canada. Unlikely
1994 reported2 reported reported The majority of women were white,
married and had completed high
school education. The mean
gestational age was 31+3 weeks
(range 26–36 weeks). The majority of
women were hospitalised for
spontaneous premature rupture of
membranes (35%), twin pregnancy
with
cervical dilation and/or contractions
(18%) or antepartum haemorrhage
(12%).
Sawyer Well Well Well UK. Likely to be
2013 reported reported reported Women whose babies were born at transferrabl
24 to 32 gestational weeks. The e
majority of women were white
European, married, had completed
secondary education, were employed
and had 1 previous birth.
Young Well Poorly Well Canada. Likely to be
2012 reported reported reported Women aged between 22 and 37 transferrabl
3,4 years, with high-risk pregnancies of e
24–26 gestational weeks, educated to
college or university level.
1. 5 of 7 women who agreed to participate were interviewed. Women were enrolled until no additional
themeswere identified. 3 of the women went on to have a term birth.
2. A convenience sample was used.
3. Interviews were conducted a long time after the birth (recall bias). All but 1 were conducted within 4
years, andthe mean was 3.2 years after the birth.
4. Interviews are stated to be semi-structured but no further details of the questions asked are presented.
The evidence from the qualitative studies that explored the areas of additional information
and support needs for women at increased risk are presented in Table 9 to Table 15. As the
nature of this review was explorative, details of the main themes identified in each of these
studies are given in the following tables along with direct quotations from studies’ participants
when necessary. Given that each of these qualitative studies explored different aspects of
information needs, results are presented separately by study.
In Table 9 to Table 15 content in italics represents direct quotations of women or fathers, with
the rest of the content (non-italics) representing field-workers’ reporting of women’s words.
Table 9: Information needs of hospitalised women at high risk of preterm birth

(Gupton 1994, n=34)
What information is the most important for a mother who is at risk for preterm birth to know?
The Preterm Birth Learning Needs Questionnaire consists of 18 topics commonly included in
55
educational programs for women at high risk of preterm birth. Participants scored each topic
applying a score range of 0–20. The 5 most important topics (estimated using mean score
[SD])were:
• the consequences of prematurity for the baby (mean 19.38 [1.65])
• problems of the newborn associated with preterm birth (mean 19.29 [1.66])
• how premature babies are cared for at home (mean 19.21 [1.82])
• how premature babies grow and develop (mean 18.71 [3.40])
• the signs and symptoms of preterm labour (mean 18.53
[2.60])In the study:
• 22/34 (67%) women indicated a need to know the possible risks or complications to the baby
andthe baby's chance of survival if premature birth should become a reality.
• 11/34 (32%) women indicated a need for reassurance – to be told that "the baby will be OK";
"forthe staff to be supportive of the mother" – and assistance in coping – to know "how to
prepare oneself psychologically and physically to face the stress, fear, etc."
• 9/34 (27%) women indicated that it was most important for them to know how a premature
birthcould be prevented
• 6/34 (18%) women indicated that they wanted ongoing information on the condition of their
babyas their pregnancy progressed.
3/34 (9%) women indicated that they wanted information on how to care for a preterm baby
What concerns do you have about being considered at high risk for preterm birth?
• 31/34 (91%) women indicated concern regarding the baby's survival chances, possible
complications or permanent disabilities associated with prematurity and fetal
development,especially lung maturation
Additional concerns:
future care of the baby, how long the baby might be in hospital, whether it would be possible to
breastfeed a premature baby, the uncertainty of the situation – “so many unknowns, so many ‘ifs’
cause fear”
Are there things that mothers at high risk of preterm birth do not need to know or should be
taught?
All those responding to this question expressed a desire to be told "everything":
"I like to know exactly what is going on and get all the facts straight, so I can prepare myself both
physically and psychologically"
"The more knowledge that I have the more positive I feel. Not knowing the possibilities is
frightening"
"...if you are prepared for the worst and it doesn't happen, it feels great. If it does, I think that being
totally unprepared could cause serious problems - both personally and in your family"
3/34 (9%) women indicated the need for honesty:
"Up front honesty is the best way to go. This is enough of a surprise; you don't need any more
surprises because you weren't told something"
"I prefer to know as much as possible and appreciate honesty in my doctors, coupled with human
compassion"
Several women included the need for advice for those who communicate information to women
athigh risk of preterm birth:
"Give information gradually so mother has time to absorb and accept at her own pace"
"Don't tell them something they may have done or not done has increased the risk. It adds to
theguilt"
"The use of alarming-sounding medical terms that when defined aren't life-threatening [is
frightening] – not talking down to a mother but make sure she's familiar with the phases and
terminology you're using – don't assume someone else has already explained – don't get overly
technical – quoting statistics doesn't reassure – you want to know how your baby is doing"
56
What would you tell someone (a friend or relative) to help them cope with being at high risk
for preterm birth?
6/34 (18%) women indicated to tell other women to rest and relax
6/34 (18%) women indicated to tell other women to trust in the healthcare system:
"I would try to remind them how advanced medicine is and the chances for survival are high"
"Reassure them that absolute care is taken when handling preterm labour – competent doctors
andnurses, modern technology"
"Make sure you know what is happening at all times. Listen closely to what you are told and
obeythe medical staff"
4/34 (12%) women indicated the importance of keeping informed:
"Inform yourself – talk to others who have gone through it"
"To seek professional help and information and not to listen to those who know little or
nothing" "Ask as many questions as they can regarding effects of preterm labour on baby and
mother andread articles/books on preterm births"
Advice to maintain a positive attitude was also given:
"Don't go on a guilt trip"
"Keep an optimistic and positive attitude no matter
what""Hope for the best, prepare for the worst"
"Positive imagery and relaxation help"
Table 10: Aspects of hospitalised women’s stressful experience of their possible

preterm labour (Gaucher 2011, n=5)
Mourning
Having faced bad news regarding several aspects of their health or pregnancy, women tried to
adapt quickly from having a health pregnancy to preparing for the challenges of prematurity
andfound this to be difficult; the roles they had been preparing to play as parents changed.
Some women at risk of a hysterectomy faced the possibility of no longer being able to bear children.
Perceptions of prematurity
All women had negative views about prematurity; several of them compared it with “horror stories”
or “hell”. All women wished to avoid delivering prematurely.
Isolation
Women felt isolated from their usual support systems: 4 had been transferred from another hospital
and their families lived far from the institution used for the study. They expected their hospitalisation
and bed rest to become prolonged, which was perceived as another difficult challenge to overcome.
Furthermore, participants believed that they had lost their intimacy or privacy during their
hospitalisation experience.
Powerlessness
Women expressed a strong feeling of powerlessness and loss of control. They believed that they
had to accept all treatments offered to them to obtain the best possible outcome for themselves
andfor their baby:
"There is nothing we can do. We’re a little powerless in all this. So we let ourselves go. We let
goand we let them do anything to us." (Mother 5)
They were overwhelmed by the number of events experienced in a short period of time; the
uncertainty of these events added insecurity and stress:
"Uncertainty, it’s like vertigo or a precipice. And there is a lot of uncertainty. We don’t know when I
will deliver. We don’t know how I will deliver. We don’t know how it will go for the baby. We don’t
57
know what awaits the baby after. And we can get surprises, good or bad, for months after that. So
it’s a lot of uncertainty for a long time." (Mother 3)
Main concerns
The baby’s health and outcome were the main concerns for most women. One was most
worried about her own medical condition. Another had been born prematurely herself, and
focused on potential attachment difficulties as a parent and on a prolonged separation from her
other children.All participants expressed some concerns about organising their families’ lives
around a prolongedhospital stay:
"Yesterday, I was preparing my children’s things, but I didn’t know what to prepare. I had to give
them extra everything because I didn’t know when I would be back. One of my children goes to
school, one goes to daycare and the third one stays at home (…) and he’s having his first birthday
tomorrow. Now they are staying in 2 different households. One child is at my mother’s house and 2
children are at my mother-in-law’s." (Mother 2)
Consultation as a stressor
Women were generally informed by the obstetrical team in charge of their medical care that
theywould meet with a neonatologist. However, 1 woman had not been told this and found
out only when approached about participating in the present study; she asked to partake in
the study andwas, therefore, included after she met with the team responsible for her care.
Similar to other participants, she perceived the consultation as an additional source of stress:
"Simply knowing that we’ll meet the neonatologist is a stressor in itself. It’s something really big
(…)The fact that I am being offered to meet the neonatologist before anything else makes me
realise that, in my case, it is highly probable that I will deliver prematurely." (Mother 5)
However, all of the participants looked forward to the consultation so that their questions would be
answered; they also hoped that the neonatologist could somehow reassure them, although the
information they sought was not perceived as reassuring in itself:
"I think that the more the neonatologist will tell me, the more stressed I will be. But I don’t like
(…)not knowing the answers." (Mother 1)
"I am looking forward to meeting them so that they can reassure us. Well, maybe not so that they
can reassure us, but so that they can tell us the truth." (Mother 2)
Table 11: The expectations of women hospitalised for preterm labour regarding the
prenatal consultation (Gaucher 2011, n=5)
Expectations from the consultation – Reassurance
Being reassured was the most important objective of the prenatal consultation. Women realised
thatthey might receive worrisome information about possible complications related to prematurity.
They hoped that the neonatologist would find ways to reassure them:
"Being reassured and just knowing what to expect. Because right now, I don’t really know what to
expect. So it’s those 2 aspects, I think. (…) And what I can do as a mother to make sure, really
make sure, that my baby is healthy and happy. Because that’s really what I want." (Mother 4)
Expectations from the consultation – Information and content

All women expected to receive clear, precise details and statistics about short-term and long-
term complications of prematurity specific to their medical condition and related to gestational
age. Someanticipated themes were respiratory distress, neurological complications, sepsis,
feeding difficulties and length of hospitalisation. They hoped the neonatologist would describe
some of the technology in the NICU. They reported having learned about prematurity and its
complications from friends working in healthcare, from the media or from their own physicians.
Only 2 of the participants underwent active follow-up for high-risk pregnancies before their
enrolment in the present study.
One woman suggested that parents visit the NICU before delivery and believed that written
documentation or pictures could be helpful.
58
Expectations from the consultation – Parental roles and responsibilities
Women expected the neonatologist to explain what their responsibilities would be and what would
be expected of them. They wanted help organising their professional and family lives so they could
be available for their baby. They wanted to know how they would be allowed to touch or hold their
babies, and wanted to discuss breastfeeding and feeding strategies.
Some wanted to know how they might participate in decision-making processes regarding
their baby’s treatment plans. One woman expressed concern about excessive care and had
preparedquestions to ask the neonatologist about her legal rights:
"I’m not sure the neonatologists would make the same decisions that I would and I am worried they
might impose their decisions on us." (Mother 3)
Expectations from the consultation – Consistency of information
Women expected all of the different medical teams involved in their care to communicate amongone
another to hold consistent discourses about their situation. They reported inconsistency between
healthcare providers’ messages as an added source of stress.
NICU neonatal intensive care unit
Table 12: The expectations of women hospitalised for preterm labour regarding the
neonatologist (Gaucher 2011, n=5)
Expectations from the neonatologist – Structure of the consultation
Women who were interviewed believed that the best time to meet the neonatology team was
beforelabour and delivery. They hoped their spouse would be present. They believed that the
neonatologists should explain their role first, and then volunteer information about prematurity
and its possible complications. One woman suggested that they sit down during the consultation.
They all expected the neonatologists to be open to listening to their concerns and to provide time
to answer their questions:
"Sometimes, I find it goes fast, that we don’t have time to ask our questions. (…) It would only take
the doctor an extra minute or 2, but it would save us from being anxious and having unanswered
questions." (Mother 3)
Expectations from the neonatologist – Trust

It was very important that the neonatologist instils a feeling of trust. Women wanted to know
thatthey were in the best place for their baby and themselves to receive optimal care:
"We are handing over our lives and our baby’s life into the hands of people we’ve never met before.
So, if there’s no trust, it’s impossible." (Mother 3)
Expectations from the neonatologist – Support and strategies

Most women expected the neonatologist to offer support and help them develop strategies to cope
with their situation:
"It’s very important to have a good doctor who can answer your questions and reassure you. (…) I
mean, at least they’re there to answer your questions and be supportive." (Mother 4)
Some also thought that neonatologists should refer them to other members of the healthcare team
to explore various aspects of the problem. One woman, who had undergone in vitro fertilisation and
fetal reduction, would have preferred to be referred to her own obstetrician for additional information
and support.
59
Table 13: Parents’ views of a prenatal tour of a neonatal intensive care unit (NICU)
during a high-risk pregnancy (Griffin 1997, n=13)
Benefits of the tour
Parents described benefits of the tour, including that it
• decreased their fears
• inspired hope for their baby’s prognosis
• provided reassurance about care in the NICU
• prepared them for their baby’s NICU hospitalisation
All parents described at least one of these benefits, including 5 mothers who said the tour was
overwhelming or difficult because of the appearance of newborns.
‘Well, it’s just hard when you see something like that. They were so young and so precious and
fighting for their lives…. But you are more put at ease by seeing the care that they do receive andthe
attention that you get. But it’s still frightening to see babies that small’
Decreased their fears
Parents reported that because the tour was informative, it decreased their fears about the NICU and
the type of care that their newborn might require:
‘Because it’s so difficult to handle when you don’t know. I know it’s scary at times and I think
themore education that you can receive about it, the better prepared you are to handle it
should it happen”
Parents stated that just knowing that the NICU existed was helpful.
“Just to know that it was there. And I think it put my wife more relaxed and at ease the fact that
theyhad a facility there that was nearby. We didn’t have to worry about going to another hospital
because they didn’t have a special care nursery. Just the fact that it was there, we could see it,
we know that it looked like and so if we were faced with that problem we were at least familiar
with it.”
The tour gave mothers information about the NICU they needed to share with other family
members. One mother indicated that she had gained an understanding of the unit and was
better prepared to talk to her child about the NICU. Three of 4 mothers who were not who were
not accompanied on the tour by the fathers reported that they had shared information about
the NICUwith the fathers, which was comforting to them. One of these mothers described her
husband’s reaction to their infant’s admission to the NICU:
“My husband was calm because I had already told him what to expect.”
Inspired hope for newborn’s prognosis

For several mothers, the tour inspired hope for their newborn’s prognosis, especially when the
mothers saw very premature infants who were said to be progressing well. One mother said:
“The tour gave me hope that he was going to be fine. Seeing babies younger than him
thrive……and then seeing the babies approximately his age survive thriving and doing well.”
Another mother said
“It showed me that there is a lot more hope, and I thought about a few years ago or even 10 years
ago, babies like this wouldn’t have made it.”
One mother said that after the tour, she was determined to take better care of herself and adhere to
her prescription for bed rest to decrease the chance that her infant would be born prematurely.
Provided reassurance about care in the NICU

Parents reported that the tour was comforting and reassuring because it gave them an
opportunityto observe the type and quality of care that the infants received. One mother said:
“I was a lot more comfortable now seeing how they are giving the care and just seeing the
environment they are in.”
Parents felt encouraged when they observed the way that nurses cared for the infants. One
mothersaid:
“I saw the love, compassion and empathy that they showed for each of the babies there. So I
60
knewhe was going to be treated well.”

Another mother commented:
“Knowing they do care about them and they do realise that they are human and not
machines….You could feel that they really cared and worried.”
It was especially helpful for the parents to see so many nurses and physicians in the NICU;
hearingspecific information about primary nursing also helped some mothers to feel more
comfortable.
Those mothers explained that it was reassuring to know that their questions could be answered
because the primary nurse would know their infant.
Prepared parents for their newborn’s NICU hospitalisation

All parents whose infants subsequently were cared for in the NICU reported that the tour prepared
them for the experience. These parents explained that it helped to acquaint them with the NICU
before delivery. One father said:
“…..we didn’t have to worry and wonder. It (the tour) made us understand how it all worked so
that we were familiar with it when we did go there. And we didn’t worry about what was going to
be donebecause they explained everything beforehand. So, we pretty much knew exactly what
their procedures were and how everything was dealt with instead of finding out as they did it….
The tour pretty much prepared us for what we were going to see when we went up there.”
One mother speculated on how her reaction to her infant’s hospitalisation in the NICU would have
been had she not toured the NICU while she was pregnant. She said:
“I think it would have been a much more negative experience had I not toured and when there
andsaw the tubes in my baby’s throat and the tape and everything. I don’t know if I would have
been able to take that….”
For one mother, the tour’s importance became evident after her infant was born:
“Well I didn’t really think much of it until she was born. I thought, well this is an interesting place
andall that, but after she was actually born and brought here I kept thinking to myself, I’m glad I
came and saw the place before she was born. It kind of helped ease knowing where she was
going to be.It made it a lot easier.”
Finally, a mother who initially was overwhelmed after the tour expressed how it prepared her for
hernewborn’s admission to the NICU. She said:
“I knew what to expect once I was there. So, I relaxed, and it wasn’t overwhelming after I had him
and he went to the (NICU).”
Evaluation of arrangement and conducting of the tour

Parents evaluated and provided suggestions on the way the tour was arranged and conducted and
offered advice to other parents. In general, all parents recommended that parents in similar
circumstances should be offered a prenatal tour of the NICU. One father said:
“I think you should go to the hospital and should try to get a tour of it…. You shouldn’t be
intimidatedby the hospital and all the goings on in a nursery…. you have to get over the fear and
ask the right questions and be familiar with that.”
Parents advised that more healthcare providers suggest tours to parents diagnosed with a high-
riskpregnancy. Two mothers also recommended that other perinatal healthcare providers should
tour the NICU so that they can be supportive to parents. One mother perceived that her need to
tour theNICU was not supported by the staff on the antepartum unit. She said:
“So, I think some of them should be a little bit more realistic and help the patient prepare for
their early delivery much more, rather than saying ‘Oh, I don’t think they should have taken her
there’ or‘it’s too much for her’…. If they can just empathise with the patient and be a little more
positive, I think the whole stay there would be a lot better as a result.”
Parents also evaluated and gave specific advice in a number of areas, including:
• tour arrangements
• type of information provided on the tour
• the behaviours and knowledge of the tour conductor.
Evaluation of the tour – Arrangement of the tour
61
Parent’s recommendations for timing of the tour varied. However, several recommended that
parents tour the NICU soon after their pregnancies are identified as high-risk. One mother
recommended that to minimise anxiety parents take the tour soon after deciding to do so.
Parents who toured with their partners commented that having each other as a support person was
helpful. They recommended that the tour be scheduled so that the partner or other support person
could accompany the parent. One mother said:
“Now that’s the part I wish I could have changed. I wished my husband or somebody had been with
me. But nobody was with me at the time.”
One couple also recommended that the tour should be scheduled around other appointments to
avoid an additional trip to the hospital.
Evaluation of the tour – Type of information given on the tour

Parents reported that it was important to receive detailed information on the following:
• newborns who had a diagnosis or gestational age similar to what was anticipated for
theirnewborn
• a description of equipment for their newborns
• roles of staff members
• a description of the parental role in the NICU, including the visitation
policy.A mother said:
“Just by introducing me to people and explaining the various ages of and their survival and the
babies that make it there. That was very comforting.”
A parent suggested that parents meet with the neonatologist before the tour. It was important for
parents to hear about the parental role. One mother said:
“They said if your baby was there, you could come up at any time, if you were the parent…..
youcould come in and they do encourage bonding with the baby, you can feed the baby, that
type ofthing. That did put me at ease.”
However, not all parents perceived that they received adequate information on the parental role. A
mother said:
“The parental role during the tour could have been more explicit because I was sure of my role
during the tour, what would be expected of me or what I could do as far as caring for my baby.”
The need for more specific information became apparent to parents after their infants were cared
forin the NICU. These parents indicted that they wanted more information on expectations for
their rolein the NICU, breastfeeding, sibling visitation and the potential for the baby to be
transferred from theNICU to another unit before discharge. Two parents suggested that hand-
outs would supplement orreinforce information that was given during the tour and assist parents
to inform family and friends about the NICU.
Parents reported that the tour should be individualised to meet the specific needs of parents.
Parents perceived the tour as individualised when they went as a couple or an individual rather
thanin a group, had an opportunity to ask questions and saw newborns who had a diagnosis or
gestational age similar to that expected for their newborn. Therefore, it was critical for the nurse
conducting the tour to know the parents’ maternal–fetal diagnosis. Several parents made
additional suggestions, such as having an opportunity to go on a second tour or changing the
order in which the NICU patient care areas are shown; these demonstrate the parents’ individual
needs.
Evaluation of the tour – Behaviour/knowledge of the tour conductor

Most parents reported that the nurses who conducted the tours were knowledgeable and
comforting. These nurses were describe as compassionate, concerned, helpful and considerate
ofthe time parents needed to understand the information and ask questions.
One mother said:
“She was a warm lady… putting her hand on my arm, and just somebody touching me made me
feel like (I was) relaxed…”
One father stated that the nurse who conducted the tour “knew what was going on and knew the
staff, and the staff apparently thought a lot of her…”
62
Table 14: Information and support needs of women during the birth of their preterm
baby (Sawyer 2013, n=39)
Overall satisfaction with care
“Overall, how satisfied would you say you were with the care that you received during the birth?”
Extremely satisfied with care and nothing could be improved = 31/39 (80%) parents
Generally satisfied with care but certain things could have been improved (such as provision
ofinformation) = 7/39 (18%)
Dissatisfied with her care = 1/39 (2%)
Staff professionalism
Staff professionalism – Information and explanation

33/39 parents (4 fathers and 6 mothers in a couple) mentioned this theme
Provision of information was really important and was mentioned by 33 participants (85%). They
wanted to be told what would happen during the birth (particularly if they were having a caesarean
section), what type of anaesthetic would be administered and what was going to happen to their
baby when he or she was born.
The anaesthetist was someone who stood out in participants' minds in terms of providing detailed
information and explanations.
“So we actually go down into the operating theatre and again the anaesthetist was there and
talkingto [us] as she said ‘I will stay with you the whole time’ and she talked us through
everything that
was happening and for both of us that was just outstanding, absolutely.” (Mother 1,
caesareansection [C/S]).
It was perceived that someone taking the time to explain what was happening helped them cope
with the situation and made the experience less ‘traumatic’:
“It was a traumatic experience. I think, if it hadn’t been explained to us exactly step by step it
wouldhave been more traumatic It was just so much easier, because they did go out of
their way and
they explained absolutely everything to you.” (Father 2, C/S).
Participants also wanted information to be explained in a way they could easily understand:
“They told you everything that was going on, what was happening. They make sure you understood,
make sure he [father] understood what was going on.” (Mother 7, C/S).
One mother wanted more information than she was given during the birth. She had some
medicalknowledge, and would have liked to know about what was happening throughout her
operation inmore detail:
“So you feel prodding, and I wasn’t told much. I felt I wasn’t told much when I was actually in
there and hadn’t, I didn’t know when they’d started to open me up, cut me open...So I didn’t
know what they were doing, water’s, broken my waters None of that was ever communicated to
me.” (Mother
8, C/S).
Six participants (15%) commented that the different members of staff introduced themselves and
told them what they would be doing. This helped them feel less like they were in a room with
peoplethey did not know:
“I mean they were all very, I remember there being people in the room and they were all introducing
themselves and what they did.” (Mother 6, C/S).
Staff professionalism – Staff calm in a crisis

11/39 parents (7 mothers, 1 father and 3 mothers in a couple) mentioned this theme. Nineteen
participants (49%) described feeling frightened of what was going to happen during the birth and
forthe outcome of their baby. However, the calm attitude of the staff helped them feel more
comfortable and at ease:
63
“You’re not as frightened. It’s daunting going in a room when you’ve never been in. All your bits
aregoing to be on show. And you’re worried about your children. Are they gonna survive? Are
they gonna be born stillborn? You know. they were so relaxed, they made me feel so
comfortable” (4
Mother, C/S).
“I think it was them staying relaxed. Even though it was a rush, it was a stressful time, you could see
that, but they were very good at staying calm. But I suppose that’s their job in a way, but they were
actually very good at it.” (Mother 19, C/S).
Staff professionalism – Confident and in control

8/39 parents (8 mothers, no couples) mentioned this theme. The confidence displayed by staff was
important to participants as it demonstrated capability and control. One woman described that the
surgeon in charge of her operation portrayed total confidence:
“And the way he mastered the team, I got the absolute… he had an air of confidence and in
control of the entire team. He knew what every person was doing. And he was very commanding
as well” (5Mother, vaginal delivery [V]).
Having confidence in the staff seemed to make it easier to hand over control to them. One woman
described that she did not feel that she needed to be in control. She trusted the staff and was
happy for them to take control of the situation:
“Absolute confidence in the staff. I didn’t feel like I needed to know every step of the way. I was
ableto just step back, realise that control was not mine. The control was where it should be, with
professionals, and they would take good care of them [the babies]” (5 Mother, V).
Four mothers (10%) described the doctors as being firm with them, but said this was exactly
whatthey needed. They wanted the staff to take control of the situation and tell them what to
do:
“It was very very quick, very shouty: ‘you have to do this, you have to do this now’. It was made very
clear to me if I didn’t push he wouldn’t survive. Erm, which was absolutely fantastic, which was what
needed to be done” (3 Mother, V).
Staff professionalism – Staff not listening to the woman

8/39 parents (6 mothers, 1 father and 1 mother in a couple) mentioned this theme. This area
contributed to a negative experience of care for participants. Seven mothers (18%) expressed
disappointment that the staff did not always listen to what they had to say. These women
describedtelling staff that they felt they were in labour and close to giving birth, and often the staff
did not believe or trust what they were saying, which left women feeling ignored and frustrated.
“And then when I started to get pains, I started to tell the midwives, or the nurses that were
there. And felt that they didn’t actually believe me, because they put me on monitors. And
where my waters had gone, the monitors don’t pick up the contractions as well. So they were
just saying ‘no,no, no, the contractions are not real basically [you] can’t be feeling this
amount of pain” (Mother
19, C/S).
One woman described how she tried to tell the midwife that she was about to have her baby,
butwas not listened to, and as a result no staff were present at the birth:
“The only kind of downside to it, was I kept saying to her, all my family have very quick labours…..I
kept saying to her I need to push I need to push and she said I’ve only checked you half an hour
ago, you’re only 3cm ..... and she went I’m just popping out the room ..... and at that point I just
pushed and her head popped out, and no one was in the room apart from me and my partner” (23
Mother, V).
Staff empathy
21/39 parents (15 mothers, 1 father and 5 mothers in a couple) mentioned this theme. Participants'
experiences of their care during the birth were also influenced by the interpersonal interactions
with
care providers, in particular by caring and emotional support, and encouragement and reassurance.
Staff empathy – Caring and emotional support
64
Twenty-one participants (54%) spoke about the ‘warm and friendly’ attitude of the staff. In terms
of satisfaction with their experience, it was important that they were treated in a pleasant
manner. Twovery different quotes illustrate the importance of the staff treating them as an
individual and receiving personalised care:
“I just found our experience very good, it was very I suppose personal in a sense. I wasn’t, I
didn’tfeel like a piece of meat. I felt like a human. and people were caring.” (3 Mother,
V).
“But the midwives that should have shown me compassion in the beginning didn’t. They were
justnot bothered.” (30 Mother, V).
Mothers spoke about the importance of a member of staff always being with them, and
thisgenerally referred to the presence of a midwife:
“One of the nurses just steps out the way, holds your hand, and talks to you ..... So it’s just nice to
have someone there, talking to you and holding your hand and sort of walking you through
everything instead of everyone buzzing around.” (2 Mother, C/S).
One mother whose baby was born with many complications and died less than 24 hours after
thebirth described how the caring and supportive attitude of one midwife made her experience
of thebirth less traumatic than it could have been:
“The midwives were incredible, so during the birth,. we had this amazingly lovely kind of West
African midwife who was, oh just love, like lovely, so nice so, supportive and caring and empathetic
and everything that you could possibly want and just really supportive and, so the birth process itself
actually, in the scheme of things was relatively easy thing then to go to because I felt very
supportive … and she was so lovely.” (32 Mother, V).
Staff empathy – Encouragement and reassurance

23/39 parents (16 mothers, 3 fathers and 4 mothers in a couple) mentioned this theme. Twenty-
three participants (59%) mentioned wanting encouragement and reassurance from the staff.
They understood that the staff have to be realistic about the situation and the prognosis for their
baby, butfound it really helpful and encouraging if the staff were able to reassure them in some
way:
“Obviously so they can’t lie... but just kind of being positive I think really really helps um ‘cause
youknow, it’s it’s quite terrifying not having had an operation before and um you know you don’t
quite know what to expect and things so just people you know just reassuring you, saying nice
things” (14, C/S Mother).
“And that’s what you want is reassurance, that time, and so yeah, it was very good.” (1 Father,
C/S).Encouragement from the staff also influenced their experience with care at birth. One
woman who was feeling scared and tired described how a midwife encouraged her to continue:
“Yeah we were whisked upstairs and at that point I couldn’t feel the hand moving so I really freaked
out. One of the midwives was there and she could feel a pulse, calm down, gave me cuddles,
reallycalmed me down and said ‘you’re ok, you’ve got to do this, you’ll get through it.’ Really sort
of geedme up and gave me that extra bit of strength really.” (3 Mother, V).
Another mother described how praise from a midwife contributed positively to her experience:
“You know she was constantly praising ‘you, you’re doing really well, just breathe through it’, you
know and things like that whereas you get some midwives who just aren’t the nicest, so um, the fact
that she was as nice as she was.” (23 Mother, V).
Involvement of the father

16/39 parents (7 mothers, 5 fathers and 4 mothers in a couple) mentioned this theme. It was
important to the mothers that the baby's father was involved in the birth, and the extent to
whichstaff involved them contributed to a positive or negative experience with care. For
example, 2 women (5%) described how the staff tried to delay the caesarean section so the
father could getthere for the birth.
Three women (8%) also discussed that they had planned their partner's involvement in the birth,and
therefore appreciated any effort the staff made to make them feel more involved:
“He got there really quick. But they involved him, once they brought him [to the operating theatre],
they told him everything while he was getting changed, what to expect.” (2 Mother, C/S).
65
“I found it reassuring that they were very happy with [husband] to be sort of looking over
theirshoulders and sticking his nose in and whatever, so there was no ‘stand over there
dad'.” (12Mother, C/S).
Four women (10%) talked of regret that the baby's father was not able to participate more and
wasnot encouraged to feel more involved in the birth by the staff:
“Erm he found it very awkward...When they were being born he just sat out there, wasn’t really
ableto participate...So he felt like a spare part. when we were rushed to the surgical unit… there
were
so many people in the room, he felt he didn’t know where to stand. He didn’t want to get in the
way.He knew he needed to get there… let everyone get on with their job. But he felt in the way.”
(5 Mother, V).
“I don’t think anyone even really spoke to [the father], I mean I I’m reflecting on it now, I don’t think
anyone did, how was he involved, he wasn’t involved at all, so yeah ‘how are you feeling?’, ‘is
thereanything I can do?’, yeah.” (31 Mother, V).
It was also important to fathers that they were encouraged to feel involved in the birth. One of
thefathers interviewed described how fathers are not normally made to feel involved in the
birth, but that this time he was involved from the start:
“Because normally they don’t talk to you. To a woman, they say ‘right we’ve got to do this, got to
dothat’ so the lady knows exactly what’s happening to her and why. For the bloke.. ‘Stay down
the pub and we’ll give you a ring when it’s all done and you can come up when it’s all nice and
clean, ina blanket.’ But with [name of hospital], it was completely different.” (2 Father, C/S).
Birth environment
17/39 parents (11 mothers, 3 fathers and 3 mothers in a couple) mentioned this theme.
Participantsdiscussed features of the delivery suite and operating theatre that contributed to their
positive experience at the birth. Five participants (13%) described that the radio was playing during
the birth,which made the environment seem less frightening:
“You know they didn’t make it scary in any way at all, they were all quite happy, I think the radio
was playing, which was good, you know things like that. The environment didn’t seem scary.” (1
Mother, C/S).
Three women (8%) also commented on the views from the windows of the operating theatre. It
helped them feel ‘connected’ with the outside world and help take their mind off things:
“It can take your mind off it a bit rather than just sort of grey walls um so yea so I mean that’s very
much what we remember actually and often sort of comment on it you know to people.” (14
Mother, C/S).
C/S caesarean section, V vaginal delivery
Table 15: Pre-delivery counselling experiences and information and support needs of
parents with babies born between 23 and 26 gestational weeks (Young 2012,
n=10)
Content
Content – Theme: Knowledge

None of the families had any previous knowledge regarding preterm birth. (Family 1 did have 2
children who were extremely preterm births at 24 and 26 weeks’ gestational age, but they
responded in reference to their first child.) Before being counselled, most parents had assumed
thatwith extreme preterm labour there was no chance of survival:
“[He] told me all the issues…I didn’t even think that… it was an option to even have a [baby at] 26
weeks…. We were, in all honesty and bluntness, prepared to have a burial for this child. We didn’t
know what to expect, or severe abnormalities, and we talked about it…through the night.” (Family
3)
All parents wanted information that was clearly stated regarding the likelihood of survival and
what to expect at delivery. All parents desired to be fully informed of the immediate risks for their
child:
“…what we needed would to be told that [they] would administer steroids, his best chances are
66
thatyou last another 48 hours there could be complications if he doesn’t, um, vis-à-vis, breathing…
moment by moment until his birth happens and then [they’ll] let you know what you have to face.”
(Family 4)
One set of parents recounted the experience of having multiple members of the neonatal team
counsel them about various aspects of the NICU including ongoing research projects. They
believed that this manner of counselling lacked compassion and would have preferred fewer
counsellorsfocusing on information of immediate relevance such as survival and prognosis:
“…it would almost be a bit more compassionate to tell people we’ll deal with it once the baby comes
then, you know, we’ll see what problems arise, there could be some, but going into the great detail
before added a lot of stress to the fact that we were early and all of those things just kept going
through our head.” (Family 4)
Content – Theme: Resuscitation wishes

Most families did not recall explicitly being asked about their resuscitation wishes:
“We want to focus on just the baby and then if that happens, then we’ll deal with it at that time.
Butwe never had that opportunity, other than just between ourselves…they should bring it up
and they should discuss it with the parents and then the parents have that opportunity to say,
‘no, we don’t want to talk about it’…” (Family 8)
In retrospect, 3 couples (Families 3, 5 and 9) may not have chosen resuscitation, had they
known all of the potential complications of prematurity. One couple who had twins, of whom one
died, (Family 9) believed the other twin suffered to such an extent while in the NICU that they
would not have proceeded with resuscitation had they known ‘what was in store’. One mother
was counselledalone in the middle of the night and believed her awareness was affected by
medication:
“But, to be honest, if somebody would have told me that this is what my life would be like, I don’t
think that I would have chosen resuscitation. I might have chosen to hold (twin A) for the 7
minutesthat he cried and let him die.” (Family 5)
Even parents who had deferred the ultimate decision to the team indicated that parents should have
clear opportunities to express their wishes.
Content – Theme: Additional resources
All parents suggested that, in addition to verbal counselling, written information would have helped
them feel informed and supported. The parents who were provided with pictures of NICU found that
they enhanced their understanding (Family 1). One mother suggested having a video or a virtual
tour of the NICU (Family 10) to help prepare for this experience.
Process
Process – Theme: Timing of counselling during pregnancy

Most of the families were seeing high-risk obstetricians during the pregnancy. They wished that
theyhad received counselling about prematurity when the pregnancy was first deemed high-risk.
Three couples believed they were falsely reassured by their doctor about the risks of preterm
delivery (Families 3, 4 and 9). One mother, who finally conceived via in vitro fertilisation after
having multiplemiscarriages due to an incompetent cervix, recalled:
“They were just saying don’t worry about it though, so I said OK. But I knew when I got pregnant
itwas pretty iffy all the way.” (Family 4)
One couple (Family 1) did suggest that early information regarding prematurity would cause
needless worry; this couple was one of 2 who did not need to see a high-risk specialist before
delivery. Two couples (Families 3 and 6) commented that while the risks for conditions such as
Down’s syndrome are discussed antenatally, there is no information routinely given about
prematurity even though it is common. They suggested that written pamphlets be available at
obstetricians’ or family physicians’ offices.
Process – Theme: Timing of counselling during maternal hospitalisation
Seven families waited in hospital more than 24 hours and even couples requiring emergent
management waited a few hours before birth occurred. One mother (Family 5) recalled being
67
admitted twice with spotting at 24 and 25 weeks before going into labour at 26 weeks. She was not
counselled until the third admission in the middle of the night. By then she was anaemic and on
medications that affected her awareness, and fell asleep during the conversation.
Process – Theme: Ongoing counselling

After the initial emergency counselling, parents wanted the opportunity to hear the news again,
together, if there was time (that is, if delivery was not imminent). The mother who was admitted
forweeks after the initial counselling, due to an incompetent cervix, and her partner did not see
the team until after the birth:
“…if they’d have come in even one or 2 at a time instead of 6 at a time, and spaced it out and then
revisit a day later, just to even pop their head in to say ‘hi, how are you doing? Oh, I’m OK’….that
would have made the just before the birth thing a whole lot easier…” (Family 4)
Although parents acknowledged that physicians are busy and cannot always cater to parents’
schedules, they believed that a follow-up visit after parents have had a chance to digest information
and formulate questions would improve the communication process.
Process – Theme: Impact of counsellors’ attitude

Parents indicated that counsellors’ messages regarding the survival and prognosis of their
extremely premature neonate should be performed in a compassionate manner and that hope
should be conveyed after the decision to resuscitate had been made:
“I don’t know what the legalities are, but my feeling at the time was that oh, we needed a lot of
positive reinforcement at that moment and what we got was the exact opposite.” (Family 4)
Parents believed that some counsellors were unnecessarily negative. One mother recalls a
physician who simply stated that the team would not proceed with resuscitation:
“He said to me, ‘OK, if the baby is born today, what we are going to do is just wrap it up, we
won’tdo any heroics, we’ll just wrap him up you can hold him for a little bit and then he’ll
probably just go.’” (Family 1)
This mother recalled being devastated by this mental imagery and described how she subsequently
avoided this particular physician throughout the child’s course in the NICU.
The second part of this evidence review, which aimed to test the effectiveness of
interventions or packages of care for women at risk for preterm labour, included the results of
2 RCTs and a qualitative study of a convenient sample to evaluate parents’ views of a
prenatal tour of a neonatal intensive care unit (NICU). The qualitative study is considered as
indirect evidence for this part of the evidence review, given the non-comparative nature of its
study design and the limitation on testing the role of intervention. However, it gives insight of
parents’ care experience, which was considered as complimentary to the results of 2 RCTs.
68
Table 16: GRADE findings for the comparison of antenatal information/support intervention with routine care in women with a high risk
of preterm birth
Number of
Quality assessment women/babies Effect
Other
Inconsistenc consideratio Interventio Routine Relative Absolute
Number of studies Design Risk of bias y Indirectness Imprecision ns n care (95% CI) (95% CI) Quality
Postnatal depression
1 study Randomised Serious1 No serious No serious Serious2 None 92/230 107/228 RR 0.85 70 fewer Low
(Oakley 1990) trial inconsistency indirectness (0.69 to per 1000
1.05) (from 145
fewer to
23 more)
Less than very satisfied with antenatal care
1 study Randomised No serious No serious No serious Serious2 none 51/945 45/942 RR 1.13 6 more Moderate
(Villar 1992) trial risk of bias inconsistency indirectness (0.76 to per 1000
1.67) (from 11
fewer to
32 more)
CI confidence interval, MID minimally important difference, RR relative risk
1. Oakley 1990: Blinding of participants, clinicians and outcome assessors was unclear
2. Evidence was downgraded by 1 due to serious imprecision as 95% confidence interval crossed one default MID.
69
3.5 Evidence statements

Section 1 – themes of additional information given prior to birth
One descriptive study (n=34) found that women hospitalised for being at high risk of
pretermbirth rated all 18 teaching topics on a questionnaire as being important to know.
Teaching topics relating to the consequences of prematurity for the baby and problems for
the baby associated with preterm birth were the most important for women and their
families and partners. The responses to open-ended questions confirmed that the women’s
and families’overriding concerns were for the wellbeing of their unborn babies.
One qualitative study (n=5) of women who had a baby born before 32 gestational weeks
reported that the majority of women were very satisfied with their care. Four main
themesemerged as important determinants of positive or negative experiences of care
during preterm birth: staff professionalism; staff empathy; involvement of fathers; and
the birth environment.
One qualitative study (n=39) found that women admitted to hospital for preterm labour
reported that their experience was stressful and that they felt a sense of isolation and
powerlessness. Although the prenatal consultation with a neonatologist was also regarded
asstressful, they expected that it would empower them by providing reassurance and clear
information regarding the consequences of prematurity for their baby and regarding their
parental role. Consistency of information from medical teams was important. They expected
to have trust in their neonatologist who would provide support and strategies to help cope
with their situation.
One qualitative study (n=10) of pre-delivery counselling for extremely premature babies
identified the content and process of the counselling as the 2 main themes concerning
parents. Parents wanted clearly stated information regarding likelihood of survival and what
to expect at delivery, and wanted to be informed of the immediate risks to their baby.
Parentsreported that additional written information would be helpful. Only 40% recalled
explicit questions regarding their resuscitation wishes but believed that parents should have
the opportunity to state explicitly their resuscitation wishes. Parents would have liked to
receive counselling when the pregnancy was diagnosed as being at high risk of preterm
delivery.
The timing of counselling when mothers were admitted was not always optimal and
theystated that ongoing counselling would be helpful, as well as counselling for fathers.
Counselling should be performed in a compassionate manner.
One qualitative study of a prenatal NICU tour during a high-risk pregnancy (n=13) identified
that the tour was of benefit to parents in that it decreased their fears, inspired hope for their
baby’s prognosis, provided reassurance about NICU care and prepared parents for their
baby’s NICU stay. Parents commented on the timing of the tour and their wish to have their
partner present. They believed it was important to have detailed information regarding
babieswith a diagnosis or gestational age similar to that anticipated for their baby and
descriptions of equipment for their babies, the roles of staff members and the parental role
in the NICU, including the visitation policy and feeding policy.
Section 2 – interventions or packages of care designed to provide additional

information prior to or during birth compared with usual care
70
Two RCTs of 458 and 1887 women reported that there were no significant differences in
postnatal depression (low quality) or satisfaction with care (moderate quality) when
women with a high risk of preterm birth who received additional antenatal support were
compared with those who received routine care.
3.6 Health economics profile

No search for health economic evidence was undertaken for the question on what
additional information and support should be given to women (antenatally or during labour)
and their families where the woman is at increased risk of preterm labour, or is suspected
or diagnosed to be in preterm labour, or has a planned preterm birth. It was anticipated that
the opportunity cost of any recommendations arising from this review would be negligible
and, reflecting that this is usually not an issue with important cost implications, it is very
unusual tofind published health economic evidence assessing the provision of information
and support in this area. Furthermore, the committee was keen that existing NICE
guidelines on patient experience in adult NHS services be followed.
Therefore this question was not identified as a priority for health economic analysis.
3.7 Evidence to recommendations

3.7.1 Relative value places on the outcomes considered
The effectiveness review found low to moderate quality evidence that the additional care
provided in the packages examined did not improve postnatal depression or satisfaction with
care. However, the committee believed that the qualitative review did elicit themes that
should underpin recommendations for the delivery of information and support to women at
high risk of preterm delivery and to their partners and families.
3.7.2 Consideration of clinical benefits and harms

As a first principle, the committee agreed that kindness, compassion and empathy were
crucial principles in all interactions between clinicians and the pregnant woman and her
partner because otherwise the content of any information imparted becomes redundant.
The effectiveness review found evidence that the additional care provided in the packages
examined did not improve the outcomes of postnatal depression or satisfaction with care.
However, the committee supported the recommendations on the themes revealed by the
qualitative review for the delivery of information and support to women at high risk of preterm
delivery and to their partners and families.
The committee noted that women might prioritise information about the risks and
consequences of preterm birth for their babies over information regarding their own
outcomes of preterm birth. However, they acknowledged that, due to time pressures during
preterm delivery, midwives and antenatal staff often have to fine-tune information given
before the preterm birth to provide the most important details regarding the delivery. They
therefore recognised the value of providing information and support as early as possible in
the antenatal period, taking into account the level and nature of the risk(s) and the
imminence of delivery for women with suspected, diagnosed or established preterm labour.
The committee recognised that not all women would be aware of the symptoms and signs of
preterm labour or have prior knowledge of the care that would be offered to them and their
babies if they delivered preterm and considered this to be an important component of
information provision. The committee considered that women should receive information
71
from healthcare professionals regarding the immediate and long-term consequences to the
baby of preterm birth.
Reassurance and trust in healthcare professionals were recurring themes in the qualitative
evidence. The committee has interpreted this as requiring the provision of honest and
realistic information about a woman’s individual situation. Honesty about the level of certainty
regarding possible outcomes was considered valuable among members of the committee,
although it was recognised that healthcare professionals would need to achieve a balance
between preventing further anxiety and not withholding information. Although the committee
recognised that not all women would want statistical predictions of future events, if statistics
regarding risk were to be provided, then this should be done in line with the NICE guideline
on patient experience in adult NHS services. It was recommended that natural frequencies of
outcomes might be more acceptable in providing information to women and their partners or
families.
The committee agreed that verbal information should be supplemented with written
information and recognised that it could be helpful to provide guidance on where to access
further information, including information in other formats or on other media, and how to
contact support organisations.
In each case healthcare providers should revisit the provision of information during
pregnancy and labour, for example because of a changing clinical situation, or because
information provided verbally may not be absorbed at the time it is given, and the woman and
her partner may find that further questions may come to mind subsequently.
Consistency of information was also considered a key consideration. Inconsistencies in
information provided to the woman by different members of the healthcare team as well as
differences in the information given to each parent were acknowledged as a source of
anxiety that led to a reduction in trust. The committee believed that improved communication
between members of the healthcare team could help mitigate against this.
Other themes of information provision that the committee considered particularly relevant
related to mother’s attachment to their newborn, the stress of hospitalisation necessitating
separation from older children and disruption to daily life, and the importance of joint
decision-making with healthcare professionals. The role of neonatal staff in listening to
women was considered critical, and women should be offered the opportunity to speak with a
neonatologist within 24 hours if they wish to do so.
It was noted that providing women at high risk of preterm labour with a tour of the NICU as
soon as possible after the risk is identified might increase their confidence, knowledge and
reduce anxiety. Meeting staff and seeing equipment seems to reassure parents that if their
baby was born prematurely there were mechanisms and people in place to help their baby.
The committee considered it important that these tours were individualised (for example, the
tour might include some rooms and not others, depending on level of risk) and that partners
were encouraged to participate in this tour where possible. It was felt that the tour needs to
be ‘real’, not virtual, because part of the reassurance will be derived from contact with the
healthcare professionals during the tour. In cases where women are not clinically able to visit
the unit, a virtual tour and discussion with staff from the neonatal unit should be facilitated
Finally, the committee recognised that information regarding resuscitation and withdrawal of
care was important to inform discussions that women at risk of preterm or very preterm
labour may wish to have antenatally. It was acknowledged that the woman’s or parents’
wishes might change over time, following reflection on their present situation or as situations
changed, and hence ongoing opportunities for dialogue regarding resuscitation and
treatment should be made available.
72
3.7.3 Consideration of health benefits and resource uses

The recommendations made by the committee have negligible resource implications and aim
to follow NICE guidelines on patient experience in adult NHS services. They reflect best
current practice where information and support to women will always be an important
component of the package of care that women receive. The recommendations stress key
elements of information and support that women should be made aware of, but do not imply
that alternative forms of information and support should not be used.
3.7.4 Quality of evidence

All of the qualitative studies used appropriate research methods and were well reported.
However, only 2 were considered to have results that were directly transferable to women
receiving NHS care in England and most of the included populations in the studies were
restricted to women with medium or higher educational status. The committee considered the
likelihood of recall bias in 1 study which decreased their confidence in its findings, although
similar themes were expressed in a second study of a similar population of very preterm
babies.
3.7.5 Other considerations

The recommendations in this section were based on both the interpretation of qualitative
evidence reviewed and on committee members’ expert opinions. The committee was aware
of several reports that aligned with the findings of this review and with their own experience;
for example the NICE quality standard on specialist neonatal care (QS4) and a report from
the Bliss Organisation. The latter’s POPPY report recommended that a tour of a neonatal
unit be provided for parents at high risk of preterm labour to provide information and prepare
them for what to expect. This is reassuring in terms of the existing guidance which includes
delivery of information and support and which implements available evidence. Furthermore
the NICE guideline on patient experience in adult NHS services makes recommendations,
the principles of which should be used for women who are at high risk of or who are in
preterm labour.
This section was identified as a priority area for equality issues in the scoping stage of
guideline’s development. The committee followed the suggestions from the stakeholders in
the scoping stage and recommendations on additional information and support were made
for women at increased risk of preterm labour with suspected, diagnosed or established
preterm labour or having a planned preterm birth. The recommendations supported the
provision of information in different formats (for example both oral and written information)
and followed the principles in the NICE guideline on patient experience in adult NHS services
to address any equality issues.
3.8 Recommendations
updated guideline.
73
Prevention of preterm birth
4 Prevention of preterm birth

4.1 Introduction
Prevention of preterm birth would reduce the number of babies suffering from prematurity
and its complications, and would reduce the burden of disability that results. Effective
prevention depends upon identifying women at high risk of preterm delivery, either because
they are at risk of spontaneous preterm labour or have medical complications which make
them likely to require planned preterm birth. This chapter addresses care for women with
recognised risk factors for spontaneous preterm labour; that is, a history of spontaneous
preterm birth or mid-trimester loss, a history of preterm premature rupture of membranes (P-
PROM) in any previous pregnancies or a history of cervical surgery (for example knife cone
biopsy).
Prevention also depends upon availability of effective interventions which can be offered to
high-risk women. This chapter reviews the evidence for the use of prophylactic vaginal
progesterone and prophylactic cervical cerclage.
Appendix G, the evidence tables in Appendix H and the forest plots in Appendix I
4.2 Prophylactic progesterone

updated guideline.
4.3 Prophylactic cervical cerclage

4.3.1 Introduction
Cervical cerclage, also known as a cervical stitch, is a treatment used to prevent the cervix
opening too early during a pregnancy causing either a late miscarriage or preterm birth.
Indications for prophylactic cerclage include a history that increases the risk of spontaneous
second-trimester loss or preterm delivery, and/or cervical shortening seen on ultrasound.
Preventing or delaying preterm birth may have significant benefit in terms of reducing the
severity of diseases of prematurity and associated complications. However, there is
uncertainty about which women are most likely to benefit from this intervention.
4.3.2 Review question

What is the clinical effectiveness of prophylactic cervical cerclage in preventing preterm
labour in women considered to be at risk of preterm labour and birth?
Women with the following are considered at risk of preterm labour and birth:
• a history of spontaneous preterm birth

• a history of preterm prelabour rupture of membranes
• a history of mid-trimester loss
• a history of cervical trauma (including surgery)
• a short cervix that has been identified by scan.
74
4.3.3 Description of included studies

Two SRs and meta-analyses (Alfirevic 2012, Berghella 2011) were included to answer this
review question, 1 of which is an IPD meta-analysis (Berghella 2011) and the other a
Cochrane SR and meta-analysis (Alfire 2012).
The population included in both SRs was women at risk of preterm labour; namely those
with previous spontaneous preterm birth or midtrimester loss (singleton gestation) or short
cervix (less than 25 mm).
Only RCTs were included in both SRs. The IPD meta-analysis (Berghella 2011) of 5 RCTs
(sample size of 504) looked only at the role of cerclage to prevent preterm birth and perinatal
mortality and morbidity when compared with no treatment. The SR and meta-analysis of 12
RCTs by Alfirevic (2012; total sample size of 3328) expanded its focus to include
comparisons between prophylactic cerclage and alternatives (either no treatment or any
alternative intervention, such as progesterone), or included comparisons between different
cerclage protocols (history, ultrasound, physical exam-indicated cerclage). Another
difference between the 2 meta-analyses was that Alfirevic (2012) included RCTs (n=9) which
compared cervical cerclage with no treatment for women at risk of preterm delivery as
diagnosed with either previous history alone (n=4) or cervical length ultrasound examination
(single or serial; n=5) whereas Berghella 2011 restricted its inclusion criteria to those trials
(n=5) including women who were assessed as high risk by having both a relevant history and
ultrasound testing. More specifically, inclusion criteria for the IPD meta-analysis was
restricted to short cervical length (less than 25 mm in 4 trials and 15 mm or less in 1 trial)
identified on either single or serial ultrasound scans before 24 weeks’ gestation (4 trials) or
before 27+0 weeks’ gestation (1 trial). All the studies reported by Berghella (2011) were also
included in Alfirevic (2012).
In the 2 trials included in the meta-analysis by Alfirevic (2012) that compared a policy of
history-indicated cerclage with a policy of ultrasound-indicated cerclage, all women had a
history of spontaneous preterm birth (or midtrimester loss) suggesting a high risk of preterm
birth in the current pregnancy. In both trials women allocated to the ultrasound arms had a
cerclage when cervical length was found to be 20 mm or less (54% and 33% respectively). In
1 trial, all women in the history-only indicated arm had a cerclage; however, in the second
trial, for women randomised to the history-only indicated cerclage arm, cerclage was
performed only if the obstetrician considered it necessary (20% cerclage).
Multiple pregnancies were excluded or results were presented separately from single
pregnancies in both SRs.
The included RCTs came from a variety of locations: USA (4 trials), UK (1 trial), The
Netherlands (1 trial), Iran (1 trial), Nigeria (1 trial), France (1 trial), South Africa (1 trial),
multicentre (2 trials). Individual patient data (IPD) from some of the trials were used by the
authors of the SR.
The gestational age of women in the included studies was well reported, ranging from 10+0
weeks to 33+6 weeks at each study’s recruitment, but the gestational age when the cerclage
was performed was not reported. There was a wide variation on the type of suture used
across the included studies; in the majority (n=8) the McDonald suture was used, whereas in
1 study a suture “similar” to the McDonald suture was used and in another the Shirodkar
suture was included. Two studies did not provide any information on the type of suture used.
4.3.4 Evidence profile

75
Evidence is presented in the following GRADE profiles. Subgroup analyses were performed
on the different methods of assessment of high risk status for preterm delivery (whether only
based on history taking or ultrasound [single or serial] or based on both methods).
• Table 21: GRADE profile for comparison of prophylactic cervical cerclage versus no cerclage
• Table 22: GRADE profile for comparison of prophylactic cervical cerclage versus progesterone
(17OHP-C)
• Table 23: GRADE profile for comparison of policy of prophylactic history-indicated cerclage versus
policy of cerclage indicated by serial ultrasound scanning in women with a previous preterm birth
Full descriptions of the characteristics and results of the included studies can be found in the evidence
tables in Appendix H.
Serious neonatal morbidity was defined in the Cochrane review by Alfirevic (2012) as any of the
following:
• necrotising entercolitis or sepsis
• mechanical ventilation
• major adverse outcome before hospital discharge
• bronchopulmonary dysplasia
• retinopathy of prematurity
• positive fetal blood culture
• other life-threatening morbidity.
76
Table 17: GRADE profile for comparison of prophylactic cervical cerclage versus no cerclage
Quality assessment Number of women Effect
Number of Other No Relative Absolute
studies Design Risk of bias Inconsistency Indirectness Imprecision considerations Cerclage cerclage (95% CI) (95% CI) Quality
Perinatal death (women considered at risk of preterm birth by any indication)
1 meta- Randomised No serious risk No serious No serious Serious1 None 100/1196 128/1195 RR 0.78 24 fewer Moderate
analysis of trials of bias inconsistency indirectness (8.4%) (10.7%) (0.61 to per 1000
8 studies 1) (from 42
(Alfirevic
fewer to 0
2012)
more)
Perinatal death (subgroup analysis only for those women considered at high risk of preterm labour due to their previous history alone)
1 meta- Randomised Serious2 No serious No serious Serious1 None 62/770 77/769 RR 0.8 20 fewer Low
analysis of trials inconsistency indirectness (8.1%) (10%) (0.58 to per 1000
3 studies 1.1) (from 42
(Alfirevic fewer to 10
2012) more)
Perinatal death (subgroup analysis only for those women considered at high risk of preterm labour due to both their previous history and identification of a short cervix in the current
pregnancy by one-off ultrasound scan)
1 study Randomised No serious risk No serious No serious Very serious3 None 2/26 (7.7%) 3/30 (10%) RR 0.77 23 fewer Low
(Alfirevic trials of bias inconsistency indirectness (0.14 to per 1000
2012) 4.25) (from 86
fewer to
325 more)
Perinatal death (subgroup analysis only for those women considered at high risk of preterm labour due to both their previous history and identification of a short cervix in the current
pregnancy by serial ultrasound scan)
analysis of trials of bias inconsistency indirectness4 (9.5%) (14.5%) (0.41 to per 1000
2012) more)
Perinatal death (subgroup analysis only for those women considered at high risk of preterm labour due to their history of previous preterm birth and identification of a short cervix in
the current pregnancy by either one-off or serial ultrasound scan)
(Berghella fewer to 10
2011) more)
77

Perinatal death (subgroup analysis only for those women considered at low or unspecified risk of preterm labour due to their previous history but with a short cervix in the current
pregnancy identified by one-off ultrasound scan)
1 meta- Randomised No serious risk No serious No serious Very serious3 None 12/147 11/140 RR 1.01 1 more per Low
analysis of trials of bias inconsistency indirectness4 (8.2%) (7.9%) (0.46 to 1000
(Alfirevic fewer to
2012) 96 more)
Serious neonatal morbidity (all women considered at risk of preterm birth for any indication)
1 meta- Randomised No serious risk No serious No serious Very serious3 None 39/407 42/411 RR 0.95 5 fewer Low
(Alfirevic fewer to
2012) 44 more)
Serious neonatal morbidity (subgroup analysis only for those women considered at high risk of preterm labour due to both their previous history and identification of a short cervix in
the current pregnancy by one-off ultrasound scan)
1 study Randomised No serious risk No serious No serious Very serious3 None 2/26 (7.7%) 3/30 (10%) RR 0.77 23 fewer Low
(Alfirevic trials of bias inconsistency indirectness (0.14 to per 1000
2012) 4.25) (from 86
fewer to
325 more)
Serious neonatal morbidity (subgroup analysis only for those women considered at high risk of preterm labour due to both their previous history and identification of a short cervix in
the current pregnancy by serial ultrasound scan)
1 meta- Randomised No serious risk No serious No serious Very serious3 None 25/234 30/241 RR 0.84 20 fewer Low
(Alfirevic fewer to
2012) 46 more)
Serious neonatal morbidity (subgroup analysis only for those women considered at high risk of preterm labour due to their history of previous preterm birth and identification of a
short cervix in the current pregnancy by either one-off or serial ultrasound scan)
(Berghella
fewer to 9
2011)
more)
Serious neonatal morbidity (subgroup analysis only for those women considered at low or unspecified risk of preterm labour due to their previous history but with a short cervix in the
current pregnancy identified by one-off ultrasound scan)
78

1 meta- Randomised No serious risk No serious No serious Very serious3 None 12/147 9/140 RR 1.4 26 more Low
(Alfirevic fewer to
2012) 143 more)
Preterm birth before 37+0 weeks ( all women considered at risk of preterm birth for any indication)
1 meta- Randomised Serious5 Serious6 No serious Serious1 None 389/1464 480/1434 RR 0.8 67 fewer Very Low
analysis of trials indirectness7 (26.6%) (33.5%) (0.69 to per 1000
(Alfirevic fewer to
2012) 104 fewer)
Preterm birth before 37+0 weeks (subgroup analysis only for those women considered at high risk of preterm labour due to their previous history alone)
1 meta- Randomised Serious8 Serious6 No serious Serious3 None 215/1038 249/1007 RR 0.86 35 fewer Very Low
analysis of trials indirectness9 (20.7%) (24.7%) (0.59 to per 1000
(Alfirevic fewer to
2012) 67 more)
Preterm birth before 37+0 weeks (subgroup analysis only for those women considered at high risk of preterm labour due to both their previous history and identification of a short
cervix together in the current pregnancy by one-off ultrasound scan)
1 study Randomised No serious risk No serious No serious Serious1 None 9/26 (34.6%) 19/30 RR 0.55 285 fewer Moderate
(Alfirevic trials of bias inconsistency indirectness (63.3%) (0.3 to per 1000
2012) 0.99) (from 443
fewer to 6
fewer)
cervix in the current pregnancy by serial ultrasound scan)
1 meta- Randomised No serious risk Serious6 No serious Serious1 None 110/253 144/257 RR 0.78 123 fewer Low
analysis of trials of bias indirectness4 (43.5%) (56%) (0.6 to per 1000
(Alfirevic fewer to
2012) 11 more)
Preterm birth before 37+0 weeks (subgroup analysis only for those women considered at high risk of preterm labour due to their history of previous preterm birth and identification of a
analysis of trials of bias inconsistency indirectness (42%) (60.6%) (0.58 to per 1000
5 studies 0.83)
79

(Berghella (from 103
2011) fewer to
255 fewer)
Preterm birth before 37+0 weeks (subgroup analysis only for those women considered at low or unspecified risk of preterm labour due to their previous history but with a short cervix
in the current pregnancy identified by one-off ultrasound scan)
1 meta- Randomised No serious risk Serious6 No serious Serious1 None 55/147 68/140 RR 0.8 97 fewer Low
analysis of trials of bias indirectness4 (37.4%) (48.6%) (0.55 to per 1000
(Alfirevic fewer to
2012) 78 more)
(Berghella fewer to
2011) 186 fewer)
Preterm birth before 34+0 weeks (in all women considered at risk of preterm birth for any indication)
1 meta- Randomised No serious risk No serious No serious Serious1 None 210/1196(17 277/1196 RR 0.79 49 fewer Moderate
analysis of trials of bias inconsistency indirectness4 .6%) (23.2%) (0.68 to per 1000
(Alfirevic fewer to
2012) 16 fewer)
1 meta- Randomised Serious2 Serious6 No serious Very serious3 None 106/770 138/769 RR 0.76 43 fewer Very Low
analysis of trials indirectness (13.8%) (17.9%) (0.4 to per 1000
(Alfirevic fewer to
2012) 83 more)
cervix in the current pregnancy by one-off ultrasound scan)
1 study Randomised No serious risk No serious No serious Very serious3 None 6/26 (23.1%) 11/30 RR 0.63 136 fewer Low
2012) 1.46) (from 268
fewer to
169 more)
80

analysis of trials of bias inconsistency indirectness4 (25.7%) (35%) (0.55 to per 1000
(Alfirevic fewer to
2012) 35 more)
(Alfirevic fewer to
2012) 60 more)
Preterm birth before 32+0 weeks subgroup analysis in women considered at high risk of preterm labour due to their history of previous preterm birth and identification of a short cervix
in the current pregnancy by either one-off or serial ultrasound scan
(Berghella fewer to
2011) 154 fewer)
Preterm birth before 28+0 weeks (all women considered at risk of preterm birth for any indication)
8 studies 1) (from 45
2012) more)
1 meta- Randomised Serious2 No serious No serious Serious1 None 60/770 73/769 RR 0.82 17 fewer Low
analysis of trials inconsistency indirectness (7.8%) (9.5%) (0.59 to per 1000
(Alfirevic fewer to
2012) 12 more)
Preterm birth before 28+0 weeks (subgroup analysis only for those considered at high risk of preterm labour due to both their previous history and identification of a short cervix in the
current pregnancy by one-off ultrasound scan)
81

1 study Randomised No serious risk No serious No serious Very serious3 None 3/26 (11.5%) 5/30 RR 0.69 52 fewer Low
2012) 2.62) (from 137
fewer to
270 more)
Preterm birth before 28+0 weeks (subgroup analysis only for those women considered at high risk of preterm labour due to both their previous history )and identification of a short
1 study Randomised No serious risk No serious No serious Serious1 None 36/253 52/257 RR 0.71 59 fewer Moderate
(Alfirevic trials of bias inconsistency indirectness4 (14.2%) (20.2%) (0.48 to per 1000
2012) 1.04) (from 105
fewer to 8
more)
(Berghella fewer to
2011) 114 fewer)
1 meta- Randomised No serious risk No serious No serious Very serious3 None 19/147 18/140 RR 1.01 1 more per Low
analysis of trials of bias inconsistency indirectness (12.9%) (12.9%) (0.55 to 1000
(Alfirevic fewer to
2012) 107 more)
analysis of trials of bias inconsistency indirectness (5.2%) (11%) (0.26 to per 1000
(Berghella fewer to
2011) 82 fewer)
Maternal side effects (vaginal discharge, bleeding, pyrexia Not requiring antibiotics) (in all women considered at risk of preterm birth for any indication)
1 meta- Randomised Serious10 Serious11 No serious Serious1 None 83/491 49/462 RR 2.25 133 more Very Low
analysis of trials indirectness (16.9%) (10.6%) (0.89 to per 1000
3 studies 5.69)
82

(Alfirevic (from 12
2012) fewer to
497 more)
Maternal side effects (vaginal discharge, bleeding, pyrexia not requiring antibiotics) (subgroup analysis only for those women considered at high risk of preterm labour due to their
previous history alone)
1 meta- Randomised Serious10 Serious11 Serious9 Serious1 None 71/364 47/336 RR 1.57 80 more Very Low
analysis of trials (19.5%) (14%) (0.76 to per 1000
(Alfirevic fewer to
2012) 313 more)
Maternal side effects (vaginal discharge, bleeding, pyrexia not requiring antibiotics) (subgroup analysis only for those women considered at low or unspecified risk of preterm labour
due to their previous history but with a short cervix in the current pregnancy identified by one-off ultrasound scan
1 study Randomised No serious risk No serious No serious No serious None 12/127 2/126 RR 5.95 79 more High
(Alfirevic trials of bias inconsistency indirectness imprecision (9.4%) (1.6%) (1.36 to per 1000
2012) 26.06) (from 6
more to
398 more)
Pyrexia (fever of 38°C or more) in all women considered at risk of preterm birth (any indication)
1 meta- Randomised Serious12 No serious No serious No serious None 38/630 (6%) 15/615 RR 2.39 34 more Moderate
analysis of trials inconsistency indirectness imprecision (2.4%) (1.35 to per 1000
(Alfirevic
more to 79
2012)
more)
Pyrexia (fever of 38°C or more) (subgroup analysis only for those women considered at high risk of preterm labour due to their previous history alone)
1 meta- Randomised Serious12 No serious No serious Serious1 None 34/503 15/489 RR 2.22 37 more Low
(Alfirevic more to 92
2012) more)
Pyrexia (fever of 38°C or more) (subgroup analysis only for those women considered at high risk of preterm labour due to both their previous history and identification of a short cervix
in the current pregnancy by one-off ultrasound scan)
1 study Randomised No serious risk No serious No serious Very serious3 None 1/26 (3.8%) 0/30 (0%) RR 3.44 NC Low
(Alfirevic trials of bias inconsistency indirectness (0.15 to
2012) 81.09)
Pyrexia (fever of 38°C or more) (subgroup analysis only for those women considered at low or unspecified risk of preterm labour due to their previous history but with a short cervix in
the current pregnancy identified by one-off ultrasound scan)
83

1 study Randomised No serious risk No serious No serious Very serious3 None 3/101 (3%) 0/96 (0%) RR 6.66 NC Low
(Alfirevic trials of bias inconsistency indirectness (0.35 to
2012) 127.2)
CI confidence interval, MID minimally important difference, NC not calculable, PTL preterm labour, RR relative risk
1. Evidence was downgraded by 1 due to serious imprecision as 95% confidence interval crossed one default MID
2. Method of randomisation not clearly reported in 3 trials; allocation concealment not clearly reported in 2 trials
3. Evidence was downgraded by 2 due to very serious imprecision as 95% confidence interval crossed 2 default MIDs
4. One study did not exclude women with advanced cervical dilatation or exposed fetal membranes (numbers not reported). 7% of women in the control arm of that study
received cerclage and it is not clear whether intention-to-treat analysis was performed
5. Method of randomisation not clearly reported in 6/9 trials; allocation concealment not clearly reported in 4/9 trials
6. Evidence was downgraded by 1 due to serious heterogeneity (chi-squared p<0.1, I-squared inconsistency statistic of 50%-74.99%) and no plausible explanation was found
with subgroup analysis
7. In 2 studies women in the control arm received cerclage and it is unclear whether intention-to-treat analysis performed in those trials
8. Method of randomisation not clearly reported in 4 trials; allocation concealment not clearly reported in 3 trials
9. 11% of women in the control arm of one study received cerclage and it is unclear whether intention-to-treat analysis was performed
10. Unclear method of randomisation and allocation concealment in 2 trials
11. Evidence was downgraded by 1 due to serious heterogeneity (chi-squared p<0.1, I-squared inconsistency statistic of 50%-74.99%) and no plausible explanation was found
12. Method of randomisation not clearly reported in 2 trials; allocation concealment not clearly reported in 1 trial
Table 18: GRADE profile for comparison of prophylactic cervical cerclage versus progesterone (17OHP-C)
Number Other Relative Absolute

of studies Design Risk of bias Inconsistency Indirectness Imprecision considerations Cerclage Progesterone (95% CI) (95% CI) Quality
Perinatal death
1 study Randomised trials No serious risk No serious No serious Very serious1 None 14/42 11/37 (29.7%) RR 1.12 36 more per Low
(Alfirevic of bias inconsistency indirectness (33.3%) (0.58 to 1000
2012) 2.16) (from 125
fewer to 345
more)
Serious neonatal morbidity (respiratory distress syndrome requiring mechanical ventilation >24 hours, intraventricular haemorrhage, neonatal sepsis or necrotising enterocolitis)
2012) 2.74)
84
Number Other Relative Absolute

of studies Design Risk of bias Inconsistency Indirectness Imprecision considerations Cerclage Progesterone (95% CI) (95% CI) Quality
(from 100
fewer to 329
more)
Preterm birth before 37 completed weeks
1 study Randomised trials No serious risk No serious No serious Very serious1 None 22/42 22/37 (59.5%) RR 0.88 71 fewer Low
(Alfirevic of bias inconsistency indirectness (52.4%) (0.6 to per 1000
2012) 1.3) (from 238
fewer to 178
more)
2012) 2.97) (from 89
fewer to 373
more)
17OHP-C 17 α-hydroxyprogesterone caproate, CI confidence interval, MID minimally important difference, RR relative risk
Table 19: GRADE profile for comparison of policy of prophylactic history-indicated cerclage versus policy of cerclage indicated by serial
ultrasound scanning in women with a previous preterm birth
Number History-
of Risk of Other indicated Serial Relative
studies Design bias Inconsistency Indirectness Imprecision considerations cerclage scanning (95% CI) Absolute (95% CI) Quality
Perinatal death
1 study Randomised Serious1 No serious No serious Very serious2 None 14/125 (11.2%) 10/122 RR 1.37 30 more per 1000 Very
(Alfirevic trials inconsistency indirectness (8.2%) 0.63 to 2.96) (from 30 fewer to 161 low
2012) more)
Serious neonatal morbidity (composite measure of morbidity not adequately described)
1 study Randomised Serious1 No serious No serious Very serious2 None 7/125 (5.6%) 4/122 (3.3%) RR 1.71 23 more per 1000 Very
(Alfirevic trials inconsistency indirectness (0.51 to (from 16 fewer to 154 low
2012) 5.69) more)
85
Number History-
of Risk of Other indicated Serial Relative
studies Design bias Inconsistency Indirectness Imprecision considerations cerclage scanning (95% CI) Absolute (95% CI) Quality
1 study Randomised Serious3 No serious Serious4 Very serious2 None 5/45 (11.1%) 8/52 (15.4%) RR 0.72 43 fewer per 1000 Very
(Alfirevic trials inconsistency (0.25 to (from 115 fewer to low
2012) 2.05) 162 more)
1 study Randomised Serious1 No serious No serious Very serious2 None 19/125 (15.2%) 18/122 RR 1.03 4 more per 1000 Very
(Alfirevic trials inconsistency indirectness (14.8%) (0.57 to (from 63 fewer to 128 low
2012) 1.87) more)
Maternal infection requiring intervention (antibiotics or delivery)
1 study Randomised Serious1 No serious No serious Very serious2 None 0/125 (0%) 1/122 RR 0.33 5 fewer per 1000 Very
2012) 7.91) more)
Maternal side effects (vaginal discharge, bleeding, pyrexia not requiring antibiotics)
1 study Randomised Serious1 No serious No serious Very serious2 None 6/122 (4.9%) 11/121 RR 0.54 42 fewer per 1000 Very
2012) 1.42) more)
1. Women in the serial scanning group received significantly more progesterone than women in the history-indicated cerclage group (39% vs 25%)
3. Unclear method of randomisation and allocation concealment
4. 54% of women in the control arm received cerclage and it is unclear whether intention-to-treat analysis performed
86

Prophylactic cerclage compared with no cerclage
Moderate to very low quality evidence from 2 meta-analyses (one was IPD) of 500 to over
3000 women found that, although fewer perinatal deaths and preterm births, and lower rates
of neonatal morbidity, occurred in women who had cerclage compared with women who did
not have cerclage, this difference was only significant for the outcomes of preterm birth
before 37+0 weeks, 34+0 weeks and 28+0 weeks. Further sub-analysis based on the way
the high risk status for preterm birth was assessed showed that this significant difference
only remained for the subgroup of women assessed by the combination of risk due to
previous history (preterm labour or midtrimester loss) and short cervix by serial or one-off
ultrasound testing.
High to very low quality evidence from the aggregated meta-analysis (from several hundred
women) showed that although a higher proportion of women in the cerclage group
experienced maternal adverse events and pyrexia compared to the group who received no
treatment, this difference was significant only for the outcome of pyrexia when all women
who were considered at risk of preterm birth were included (independently of indication).
Subgroup analyses of high quality evidence showed that a significantly higher proportion of
women who were considered at low or uncertain risk of preterm delivery based on their
previous history but with a short cervix in current pregnancy (identified by one-off ultrasound
scan) experienced maternal adverse events in the cerclage group compared with those in
the group who received no treatment.
Further subgroup analyses for the outcome of pyrexia showed that the difference in the
proportion of women in the cerclage arm who experienced pyrexia compared with the ‘no
treatment’ arm remained significantly high only for those women who were considered at
high risk based on their previous history alone. There was no significant difference in risk of
pyrexia between subgroups classified as at high risk of preterm labour due to both their
previous history and identification of a short cervix in the current pregnancy by one-off
ultrasound scan and those women considered at low or unspecified risk of preterm labour
due to their previous history but with a short cervix in the current pregnancy identified by one-
off ultrasound scan (low quality evidence).
Prophylactic cerclage compared with prophylactic progesterone
Low quality evidence from the single trial (of 79 women) which reported this comparison
showed no significant differences in the outcomes of perinatal deaths, serious neonatal
morbidity or preterm births before 37+0 weeks of pregnancy or before 28+0 weeks between
women who received prophylactic cerclage and women who received prophylactic
progesterone.
Policy of prophylactic history-indicated cerclage compared with policy of cerclage
indicated by serial ultrasound scanning
Very low quality evidence from a single trial (of over 200 women) comparing a policy of
prophylactic history-indicated cerclage with a policy indicated by serial ultrasound scanning
found no significant difference in perinatal deaths, serious neonatal morbidity, preterm births
before 37+0 weeks or 34+0 weeks of pregnancy, maternal infection requiring intervention
and maternal side effects (vaginal discharge, bleeding and pyrexia not requiring antibiotics),
although only 20% of women in the ‘history indicated’ group were actually treated with
cerclage. Results should be interpreted with caution as the obstetrician in this trial decided
that women in the group with a policy of prophylactic history-indicated cerclage did not need
87
one (there were no strict criteria for this decision) when 100% should have had it in this
group if it were truly an RCT comparing prophylactic history-indicated cerclage.
4.3.6 Health economics profile

A single search was undertaken for health economic evidence on prophylactic cervical
cerclage to prevent preterm labour in women considered to be at risk of preterm labour and
birth and rescue cervical cerclage in preventing preterm birth in women in suspected preterm
labour. A total of 60 articles were identified by the search. After reviewing titles and abstracts,
3 papers were obtained. These studies were all excluded because they were not economic
evaluations or were published conference abstracts. Therefore, no relevant economic
evidence was identified for this question.
This question was identified as a priority for health economic analysis as current practice is
varied and there is a lack of consensus on many aspects of care. The committee therefore
thought it would be important for recommendations to be supported by cost effectiveness
evidence, especially as there are potentially large cost savings from preventing preterm birth.
However, no new economic analysis was undertaken due to a lack of evidence of difference
on ‘functional outcomes’ (such as perintal death and serious neonatal morbidity) in the
clinical review, with most studies not powered to detect any differences, making it difficult to
assess treatment effectiveness.
4.3.7 Evidence to recommendations
4.3.7.1 Relative value placed on the outcomes considered

The committee considered that reductions in serious neonatal outcomes and longer term
morbidity were the most important outcomes for this review question. They considered a
reduction in incidence of preterm birth to be a useful proxy measure for these outcomes, so
this was prioritised along with the other neonatal outcomes. The committee agreed that it
would also be informative to report any long-term infant neurodevelopmental outcomes or
neurodevelopmental disabilities as a single outcome for comparing the effectiveness of
prophylactic cerclage with the other options.
In terms of maternal outcomes, the committee prioritised mortality and adverse effects,
including maternal infection requiring further intervention and cervical trauma that can require
future repair because prophylactic cerclage is an invasive procedure. The committee also
discussed the importance of the emotional or psychological impact on women undergoing
this type of prophylactic intervention.
The committee agreed at the protocol stage that outcomes would be assessed according to
the way the risk of preterm labour is assessed in the studies, whether there is a previous
experience of preterm delivery (or mid-trimester loss) or there are risk factors associated with
the current pregnancy, such as a short cervix.
4.3.7.2 Consideration of clinical benefits and harms

The evidence from the 2 SRs and meta-analyses (1 aggregated and 1 IPD) showed that
among the neonatal outcomes reported in the studies (perinatal death, serious neonatal
morbidity, preterm delivery), only delivering preterm before 37+0, 34+0 and 28+0 weeks of
pregnancy was significantly different between the cerclage and no treatment groups, with the
cerclage group favoured for this outcome.
The evidence base also suggested that there may be an increase in maternal adverse
effects in women who received prophylactic cerclage compared with those who did not. The
committee did note, however, that it was not possible to distinguish the nature of the
88
individual adverse effect and thus it was hard to determine the clinical significance of this
result. However, they discussed in depth the associated risks for the pregnancy from this
technique, such as uterine contractions, bleeding or infection which may lead to miscarriage
or preterm labour. These risks were balanced against the benefit from mechanical support to
the cervix.
The only available data on specific adverse events was for pyrexia which was analysed
separately. The results did show a significant increase in the risk of experiencing pyrexia in
the group that received prophylactic cerclage compared with the group that received no
treatment. However, there was still some uncertainty as to the clinical significance of this
result given that none of the trials specified whether the women who had pyrexia had also
received antibiotics.
The IPD meta-analysis reported outcomes specific to women with a history of previous
preterm birth and a short cervix in the current pregnancy identified by ultrasound scan. No
further evidence was identified that provided information about women with other historical
indications, for example a history of cervical trauma (including surgery). However, enough
information was available to perform subgroup analysis distinguishing between those women
assessed as at high risk for preterm labour only from history taking or from investigating the
cervical length (with serial or one-off ultrasound testing). Sub-group analyses were
performed to look at the outcomes according to the different risk factors for preterm birth that
had been used as indicators for the use of prophylactic cerclage in the trials. Analyses were
conducted for sub-groups of women who had been identified as being at high risk due to
their history alone, or at high risk due to their history and the presence of a short cervix in the
current pregnancy identified by one-off or serial ultrasound scan. No differences were found
between groups for the outcomes of perinatal death or serious neonatal morbidity. The
committee felt the contrast between the findings for perinatal death in these sub-groups and
the findings in the overall analysis could potentially be attributed to the smaller sample sizes
included in the sub-groups and were therefore reluctant to draw any firm conclusions from
this.
For women with a history of previous preterm birth who were also found to have a short
cervix on either a single ultrasound scan or serial ultrasound scans, there was evidence of a
significant reduction in preterm birth before 37+0, 35+0, 32+0, 28+0 and 24+0 weeks of
pregnancy for women who had received prophylactic cerclage compared with those who did
not. This conclusion was in line with the committee members’ clinical experience.
Prophylactic cerclage was also compared with progesterone: no difference in rates of
preterm birth was found between the 2 interventions for any of the neonatal outcomes.
Therefore a recommendation with a choice of either of these prophylactic interventions was
drafted.
The committee noted that in women with a previous preterm birth, the comparison of a policy
of prophylactic cerclage on the basis of clinical history with a policy of cerclage indicated by
serial ultrasound scanning was not very informative because the estimates of effects
between the 2 groups were biased by the design limitations of the trial. For these reasons,
the committee did not place confidence in these results.
In summary, the results of the review were that the benefits of prophylactic cerclage, in terms
of reduction in preterm birth, were more likely to be seen in the sub-group of women who had
both had a previous preterm birth and a short cervix in the current pregnancy. This reflected
the committee members’ clinical experience. Moreover, as women in the overall analysis
included a proportion of women with this particular combination of risk factors, they felt it was
plausible that the benefits seen in the overarching group were likely to be due to the
influence of these women on the overall result. They noted that there was a paucity of
evidence about the emotional and psychological impact of prophylactic cerclage and
89
transvaginal scanning. Hence they concluded that the recommendations should be tailored
to a specific group of women for whom the benefit of this intervention is most certain.
4.3.7.3 Consideration of health benefits and resource uses

The committee felt that prophylactic cerclage was likely to be an expensive intervention due
to the setting in which it is delivered and because it requires the healthcare professionals
providing the care to have clinical expertise.
However, they also acknowledged that the management of preterm birth and the associated
neonatal outcomes are extremely costly. They therefore considered that the overall health
benefits likely to be obtained from offering this prophylactic intervention to selected women
not only justified the resource use but also that the initial costs incurred would be likely to be
offset by large cost savings downstream.
4.3.7.4 Quality of evidence

The committee noted that evidence was available for most of the prioritised outcomes with
the exception of cervical trauma and women’s emotional/psychological impact, but that the
quality varied among the outcomes.
The quality of the majority evidence was moderate to low with risk of bias (due to lack of
information on study design) and imprecision to be the most affected areas of quality
assessment.
The committee also concluded that, although results of subgroup analyses informed their
decision-making, most studies were not sufficiently powered to detect a significant difference
in outcomes (resulting in the evidence being downgraded for very serious imprecision for
these outcomes) between the groups, and therefore effects on subgroups should be
interpreted with caution. This does not apply for the subgroup analyses presented by
Berghella (2011) in the IPD meta-analysis, as the quality of this analysis was superior. IPD
meta-analysis is considered a gold standard in meta-analysis as it uses the ‘raw data’ of
individual patients from included studies instead of the published summary results of studies
in a traditional meta-analysis. Compared with subgroup analyses in a single study or in a
traditional meta-analysis, an IPD offers important potential advantages, such as:
• increased possibilities to perform more complex statistical analyses that better match the
underlying data
• more power compared with single studies and traditional meta-analyses
• higher validity of subgroup analyses by avoiding ecological bias and by taking the
distribution of other patient characteristics into account
• improved flexibility and standardisation of defining subgroups across studies
• opportunities to examine the consistency of subgroup effects across studies.
4.3.7.5 Other considerations

These recommendations were based on both the clinical interpretation of evidence and on
committee members’ expert opinions. Because individual studies included women whose
history of previous preterm birth varied between 16 and 36 weeks of pregnancy and who had
cervical length of less than 25 mm in the current pregnancy, they relied on the characteristics
of women in the majority of studies in both meta-analyses to inform their recommendations.
The committee did not consider the option of giving the combination of progesterone and
cerclage, and was aware that usual clinical practice is to use one or the other.
90
4.3.8 Key conclusions

The committee considered that whether or not cerclage can be said to be beneficial to
improve neonatal outcomes depends on whether the reduction in preterm birth (for which
there is evidence) is translated into a reduction in preterm birth-related neonatal morbidity
(for which there is no evidence). There is evidence of limited harm to the mother in terms of
increased pyrexia, but the clinical significance of this is not clear. The committee concluded
that the evidence of benefit is not high enough to recommend that all women at risk of
preterm birth due to a previous history and/or a short cervix should have prophylactic
cerclage, but neither is there evidence of great enough harm that would justify
recommending against its use.
From their clinical expertise, the committee members felt that some sub-groups of women
may benefit from prophylactic cerclage, but this cannot be fully supported by the evidence. It
was felt that performing cerclage in all women thought to be at risk of preterm birth due to a
previous history and/or a short cervix would result in over-treatment of women who would not
necessarily benefit and may lead to iatrogenic harm. Insertion of a suture is an invasive
procedure and the committee’s experience was that some women will choose not to have the
intervention without clear benefit. They did, however, feel that the evidence of reduced
preterm birth across a range of gestations in women who had a history of preterm birth plus a
short cervix in the current pregnancy identified on ultrasound scan suggested that the
balance of benefits and harms was most likely to be optimised in this group and therefore the
recommendation should be targeted towards these women. However, women at higher risk
of preterm labour, such as those with a history of preterm prelabour rupture of membranes
(P-PROM) in previous pregnancy or a history of cervical surgery, should also be considered
for this prophylactic treatment option. The committee was disappointed with the availability of
appropriate data as they were uncertain of the benefit for women with certain risk criteria,
such as cone biopsy or large loop excision of the transformation zone.
4.4 Recommendations
updated guideline.
4.5 Resarch recommendations

1. What is the clinical effectiveness of prophylactic cervical
cerclage alone compared with prophylactic vaginal
progesterone alone and with both strategies together for
preventing preterm birth in women with a short cervix and a
Research question history of spontaneous preterm birth?
Why this is needed
Importance to ‘patients’ or Preterm birth causes significant neonatal morbidity and mortality, as
the population well as long-term disability. Therefore strategies for preventing
preterm birth are important. There are recognised risk factors for
preterm birth, and so interventions can be offered to women with
these risk factors. Both prophylactic cervical cerclage and
prophylactic vaginal progesterone are effective in preventing preterm
birth in women with a short cervix and a history of preterm birth, but
there is limited evidence on which is more effective, and the relative
risks and benefits (including costs) of each. More randomised
research is needed to compare the relative effectiveness of
prophylactic cervical cerclage and prophylactic vaginal progesterone
91
1. What is the clinical effectiveness of prophylactic cervical

cerclage alone compared with prophylactic vaginal
progesterone alone and with both strategies together for
preventing preterm birth in women with a short cervix and a
Research question history of spontaneous preterm birth?
in improving both neonatal and maternal outcomes. This will help
women and healthcare professionals to make an informed decision
about which is the most effective prophylactic option.
Relevance to NICE The research would affect one of the key recommendations in future
guidance updates of the guideline and would therefore be highly relevant
Relevance to the NHS Preterm birth causes significant neonatal morbidity and mortality, and
survivors may have life-long physical and neurological disabilities.
Thus any intervention that prevents preterm birth will reduce the
requirement for health care and social care resources, with
considerable financial savings.
National priorities NHS Outcomes Framework #1: Preventing people from dying
prematurely
Current evidence base Both prophylactic cervical cerclage and prophylactic vaginal
progesterone were found to be effective in reducing the risk of
preterm birth in women with a short cervix and a history of previous
preterm birth, but there is limited evidence on which is more effective,
and the relative risks and benefits (including costs) of each. More
research is needed to allow women and their caregivers to make an
informed decision.
Equality This group is defined only by gestational age at delivery.
92
Diagnosing preterm prelabour rupture of membrane (P-PROM)
5 Diagnosing preterm prelabour

rupture of membranes (P-PROM)
5.1 Introduction
Preterm prelabour rupture of membranes (P-PROM) is the presenting symptom in around
20% of all women who develop spontaneous preterm labour. Although many women with
preterm rupture of the fetal membranes go into labour fairly quickly thereafter, those who do
not are at risk of infection ascending into the uterine cavity. Such infection can be very
harmful to mother and baby and hence a diagnosis of P-PROM warrants careful clinical
monitoring to facilitate early detection and treatment of in utero infection and
chorioamnionitis. Accurate diagnosis of P-PROM is therefore important.
The aim of this question was to determine the diagnostic accuracy of placental alpha-
microglobulin-1, nitrazine (pH), insulin-like growth factor binding protein-1, fetal fibronectin
and diagnostic panty-liner with polymer-embedded strip to diagnose P-PROM. These index
tests were considered either individually or in combination.
5.2 Review question

What is the diagnostic accuracy of the following tests to identify preterm prelabour rupture of
membranes:
• placental alpha-microglobulin-1
• nitrazine (pH)
• insulin-like growth factor binding protein-1
• fetal fibronectin
• panty-liner with polymer-embedded strip?

Two prospective cohort studies were included in this review which evaluated the diagnostic
accuracy of specific tests for the detection of P-PROM (Jain 1998, Tagore 2010). One study
(Jain 1998) investigated the diagnostic accuracy of insulin-like growth factor binding protein-1
(IGFBP-1) for diagnosing P-PROM, whereas the other (Tagore 2010) tested the use of
IGFBP-1, placental alpha-microglobulin-1 (PAMG-1) and the nitrazine test to diagnose P-
PROM.
One study was conducted in the UK (Jain 1998) and the other in Singapore (Tagore 2010).
The gestations of women in 1 study ranged from 17 to 37 weeks (Tagore 2010) and in the
second ranged from 24 to 36 weeks (Jain 1998). The use of tocolytics was reported only in 1
study which specified that two-thirds of women in the study received steroids with tocolysis
(Tagore 2010).
No studies that examined the diagnostic accuracy of fetal fibronectin or the diagnostic panty-
liner with polymer-embedded strip for preterm prelabour rupture of membranes were
included according to the protocol.
The reference test (gold standard) varied between the 2 studies. In 1 study (Jain 1998)
pooling of the liquor in the posterior fornix in speculum examination and intact amniotic sac at
birth appeared to be a reference test, although no clear definition was provided. In the
second study (Tagore 2010) the standard positive reference test was defined as the
93
presence of 3 or more of the following conditions: pooling of the clear fluid seen during
speculum examination; oligohydraminous identified on ultrasound scan; signs and symptoms
of chorioamnionitis; and preterm birth within a week of presentation along with a convincing
history of leaking liquor.
The Committee considered that it was important to look at the diagnostic accuracy of tests to
assess rupture of membranes (for example leaking of amniotic fluid) in the population of the
guideline – those at risk of preterm labour (below 37 weeks) with P-PROM.
The decision not to use indirect evidence, either from studies on term PROM or from studies
with mixed populations of term and preterm PROM without the majority being on the
population of interest (at least 2/3 of the sample) or with a subgroup analysis on P-PROM
population, was based on both clinical and methodological grounds. More specifically:
• Different composition of amniotic fluid preterm and term: a variety of substances
(including phospholipids lecithin (phosphatidylcholine), phosphatidylglycerol, sphingomyelin,
phosphatidylinositol, and phosphatidylethanolamine), are released into amniotic fluid whilst
the fetus matures, and are therefore likely to be present at term, but not preterm. Indeed,
detection of these substances is the basis of amniocentesis for fetal lung maturity testing,
where it is used.
• Mechanisms that initiate preterm and term birth may differ: infections of the amniotic
fluid are more frequent in preterm rather than term deliveries, particularly amongst
pregnancies that end before 34 weeks of gestational age.

Data is reported in a modified GRADE table showing summary results of diagnostic studies
(Table 24) for the following tests:
• placental alpha-microglobulin-1
• insulin-like growth factor binding protein-1
• nitrazine.
A description of the usefulness of the positive and negative likelihood ratio (following the
thresholds set up in Section 2.2.4) is given along the summary statistics for each test.
evidence tables in Appendix H.
94
Table 20: GRADE profile for predictive accuracy of diagnostic tests for identifying preterm prelabour rupture of membranes (P-PROM)
Measure of diagnostic accuracy (with 95% confidence
Quality assessment intervals)
Number Number Positive Negative
of Risk of Other of likelihood likelihood
studies Design bias Inconsistency Indirectness Imprecision considerations women Sensitivity Specificity ratio ratio Quality
Placental alpha-microglobulin-1
1 study Case Very No serious Serious3 No serious None 100 92.07%(84 99% (98 to 547 (1.11 to 0.07 (0.02 to Very
(Tagore series serious1,2 inconsistency imprecision to 100) 100) >1000) 0.21) low
2010) Very useful Very useful
Insulin-like growth factor binding protein-1
1 study Case Very No serious No serious No serious None 34 97% (82 to 99% (97 to 293 (0.60 to 0.02 (0.001 Very
(Jain series serious1,2,5,6 inconsistency indirectness imprecision 100) 100) >1000) to 11.1) low
1998) Very useful Very useful
1 study Case Very No serious Serious3 No serious None 94 87.5% (77 94.4%(88 15.75 (5.21 0.13 (0.05 to Very
(Tagore series serious1,2 inconsistency imprecision to 97) to 100) to 47.5) 0.30) low
2010) Very useful Moderately
useful
Nitrazine
1 study Case Very No serious Serious3 No serious None 98 85% (73 to 39.7% (27 1.40 (1.10 0.37 (0.16 to Very
(Tagore series serious1,2 inconsistency imprecision 96)* to 52) to 1.80) 0.84) low
2010) Not useful Moderately
useful
1. Unclear if the reference standard results interpreted without knowledge of the results of the index test
2. Unclear how women were selected for the study (a consecutive or random sample)
3. n=6 women had twin pregnancy
4. The very wide confidence interval is due to the way this is calculated for likelihood ratios where there are very few false results and does not represent uncertainty around the
point estimate, therefore the study has not been downgraded for imprecision
5. Unclear if the same reference test was used for all participants
6. Reference test/gold standard not clearly specified. Might have used following observations: Pooling of the liquor in the posterior fornix in speculum examination intact
amniotic sac at birth
95
Antenatal prophylactic antibiotics for women with P-PROM

Placenta alpha-microglobulin-1
One prospective cohort study (n=100) found that placenta alpha-microglobulin-1 is a useful
test in diagnosing P-PROM. Positive and negative likelihood ratios were very useful. The
evidence was of very low quality.
Two prospective cohort studies (n=128) found that insulin-like growth factor binding protein-1
had a very useful and moderately useful positive and negative likelihood ratio for diagnosing
P-PROM. The evidence was of very low quality.
Nitrazine
Evidence from 1 prospective cohort study (n=98) found a not useful positive likelihood ratio
and moderately useful negative likelihood ratio for the nitrazine test in diagnosing P-PROM.
The evidence was of very low quality.

A search was undertaken for health economic evidence on diagnostic tests to identify
preterm prelabour rupture of membranes. A total of 82 articles were identified by the search.
After reviewing titles and abstracts, 14 papers were obtained but these were all excluded.
Therefore, no relevant economic evidence was identified for this question.
This question was identified as a medium priority for health economic analysis but more
important priorities meant that new economic analysis for this guideline was not ultimately
undertaken.

The Guideline Committee agreed that diagnosis of P-PROM is key to successful
management and improved perinatal outcomes for affected women. They considered
identification of true positive and true negative cases to be equally important for clinical
decisions regarding further treatment of women. Failure to identify those women with P-
PROM correctly can result in the failure to implement helpful prophylactic measures.
Conversely, failure to identify women without P-PROM correctly can result in delay in
discharge from hospital or inappropriate intervention, such as hospitalisation or induction of
labour for elective preterm birth, and inappropriate use of antibiotics.
5.7.2 Relative value placed on the outcomes considered

The committee recognised that the included studies showed that 2 tests (placenta alpha-
microglobulin-1 and insulin-like growth factor binding protein-1) were useful for correct
identification of P-PROM, although the very low quality of evidence reduced the committee’s
confidence in the results. Based on positive and negative likelihood ratios, they agreed that
96
these 2 tests appeared to be better than nitrazine testing, that the test results can be trusted
to identify women who do have P-PROM and that women who have P-PROM are unlikely to
be missed when using these tests. The committee also agreed that nitrazine should not be
used as a diagnostic test for P-PROM.
The committee noted the potential clinical harm of the ‘not useful’ positive likelihood ratio of
nitrazine and concluded that the test is not useful for identifying P-PROM (because of the risk
of identification of many false positives). The committee was concerned that high rate of false
positives may be problematic because this can unnecessarily result in a cascade of
interventions, such as induction of labour, elective preterm birth and use of antibiotics. The
committee members were aware of evidence from randomised trials showing that
administration of antibiotics to women in preterm labour with intact membranes is associated
with a significant increase in the risk of cerebral palsy during childhood.
The committee discussed amniotic pooling and concluded that this was an obvious and
confirmed sign of P-PROM. Therefore they recommended that no further test for the
diagnosis of P-PROM be performed when pooling of amniotic fluid is observed and that an
additional diagnostic test is only required when there is uncertainty about diagnosis of P-
PROM.

The committee was convinced of clinical usefulness of insulin-like growth factor binding
protein-1 and placental alpha macroglobulin 1 testing but noted that the cost effectiveness
remains to be tested. They also noted that there is a cost associated with ongoing
surveillance if a test isn’t performed, which can include inpatient care.

The quality of evidence was very low and was limited to results from 2 included studies. The
committee acknowledged that many studies had been excluded but agreed that this
restrictive approach was necessary to ensure that any recommendations made were based
on relevant study populations.
The committee had concerns regarding bias because of the small sample size of the
included studies, as reference standards varied between the studies and as no single
strategy can be used as the reference (gold) standard for diagnosis of P-PROM.
They agreed that although some useful positive and negative likelihood ratios were
demonstrated, it was difficult to have confidence in the findings given the poor quality of the
evidence. Hence the committee made a recommendation that clinicians should not use the
tests alone to decide what care to offer the woman.

committee’s clinical expert opinion.
The committee acknowledged that the importance of ascertaining whether membranes are
ruptured is greater in the preterm context than at term because of the higher risk of
complications for both the baby and the woman if a diagnosis is missed. However, they
believed that prophylactic antibiotics should not be offered if diagnostic testing for P-PROM
was negative and in the absence of amniotic fluid pooling, but rather that the woman should
97
be encouraged to return if any further symptoms suggestive of P-PROM or preterm labour

arose.
The Committee was also aware of the NICE Medical Technologies Guidance on the Vision
Amniotic Leak Detector to assess unexplained vaginal wetness in pregnancy (MTG15),
however the evidence underpinned that guidance was largely coming from women at term
whereas the Committee discussed the important of selecting evidence on preterm
population.
5.8 Recommendations
updated guideline.
6 Antenatal prophylactic antibiotics for

women with P-PROM
6.1 Introduction
Preterm prelabour rupture of membranes (P-PROM) is a major risk factor for intrauterine
infection/chorioamnionitis, which itself can be the cause of maternal sepsis (the leading direct
cause of maternal death) and a major contributor to neonatal morbidity (such as pneumonia)
and neonatal mortality. The objective of this review question is to evaluate the effectiveness
of antibiotic prophylaxis offered to pregnant women whose membranes have ruptured
preterm before labour has started for the prevention of early-onset neonatal infection. The
focus population is women who have a diagnosis of P-PROM with no other indication for
antibiotic therapy, for example in the absence of evidence of infection. We looked at the
prophylactic efficacy of antibiotics compared with no antibiotics (or placebo) for improving
neonatal and maternal outcomes in general and not the performance of any individual
antibiotic. However, further subgroup analysis was planned at the protocol stage to present
results for different antibiotic classes in order to facilitate decision-making.
6.2 Review question

What is the clinical effectiveness of antenatal prophylactic antibiotics given to women with
diagnosed preterm prelabour rupture of membranes to improve outcomes of preterm labour?

Three studies were included in this review (Kenyon 2013, Kenyon 2008, Mercer 2011). Two
were systematic reviews (SRs) of randomised controlled trials (RCTs) (Kenyon 2013, Mercer
2011) and 1 was a follow-up of an earlier trial conducted in 2001 (Kenyon 2008) which was
included in the earlier SRs. Sixteen RCTs from the SR by Kenyon (2013) met our protocol.
Mercer 2011 is a further analysis of 5 of the included studies from Kenyon (2013).
The two included SRs evaluated the immediate and long-term effects of administering
antibiotics to women with P-PROM before 37 weeks of pregnancy on maternal and neonatal
outcomes. The third included study (Kenyon 2008) is a follow-up of UK children at 7 years
98
born to women with mean gestational age of 32 weeks who participated in an earlier clinical
trial that evaluated the use of antibiotics in women presenting with P-PROM.
All included studies in the SRs specified that P-PROM was confirmed either with a speculum
examination alone or in combination with a positive nitrazine test and ’ferning’ of amniotic
fluid.
Women with a diagnosis of infection or antibiotics taken during the previous 7 to 10 days
were excluded from the studies.
The type, route, dose and regimen of antibiotics used varied widely between the included
trials in the SRs: for further details see the GRADE profiles below and the evidence tables in
Appendix H.

The findings of the review are presented in the following GRADE profiles:
• Table 25: GRADE profile for comparison of antibiotic therapy versus placebo
• Table 26: GRADE profile for comparison of antibiotic therapy versus either placebo or
no antibiotic therapy
• Table 27: GRADE profile for comparison of antibiotic therapy versus no antibiotic
therapy (childhood outcomes at 7-year follow-up)
Full descriptions of the characteristics and results of the included studies can be found in the
99
Table 21: GRADE profile for comparison of antibiotic therapy versus placebo
Quality assessment No. of women / babies Effect

No. of Gestation Relative
studies Design Risk of bias Inconsistency Indirectness Imprecision (weeks) Antibiotic Placebo (95% CI) Absolute Quality
Neonatal outcomes
Perinatal death/death before discharge – Any antibiotica
1 meta- Randomised Very No serious Serious8 Serious 9 20 to 34 276/4315 138/1986 RR 0.93 5 fewer Very low
analysis of trials serious2,3,4,5,6,7, inconsistency (6.4%) (6.9%) (0.76 to per 1000
(Kenyon fewer to
2013) 10 more)
Perinatal death/death before discharge – All penicillin (excluding co-amoxiclav)
1 meta- Randomised Serious6 No serious Serious10 Serious 9 20 to 34 7/165 10/167 RR 0.78 13 fewer Very low
analysis of trials inconsistency (4.2%) (6%) (0.31 to per 1000
(Kenyon fewer to
2013) 58 more)
Perinatal death/death before discharge – Beta lactam (including co-amoxiclav)b
1 meta- Randomised Serious2 No serious No serious Serious 9 24 to 36 80/1236 46/644 RR 0.62 27 fewer Low
(Kenyon fewer to
2013) 111 more)
Perinatal death/death before discharge – Macrolide (including erythromycin) c
1 meta- Randomised Serious5 No serious Serious10 Serious 9 20 to 36 84/1354 56/784 RR 0.83 12 fewer Very low
analysis of trials inconsistency (6.2%) (7.1%) (0.43 to per 1000
(Kenyon fewer to
2013) 43 more)
Perinatal death/death before discharge – Other antibiotic d
1 meta- Randomised Serious7, 12 No serious Serious10 Serious 9 24 to 36 28/371 26/391 RR 1.13 9 more Very low
analysis of trials inconsistency (7.5%) (6.6%) (0.68 to per 1000
(Kenyon fewer to
2013) 59 more)
100

Neonatal encephalopathy – Any antibiotice
1 study Randomised Serious13 No serious No serious No serious <36 0/30 (0%) 0/30 (0%) NC NC Moderate
(Kenyon trials inconsistency indirectness imprecision
2013)f
Neonatal necrotising enterocolitis – Any antibioticf
4,5,7,12,13
1 meta- Randomised Serious No serious Serious14 Serious 9
20 to 36 100/4273 58/1956 RR 1.09 3 more Very low
analysis of trials inconsistency (2.3%) (3%) (0.65 to per 1000
(Kenyon fewer to
2013) 25 more)
Neonatal necrotising enterocolitis – All penicillin (excluding co-amoxiclav)g
1 meta- Randomised Serious13 No serious No serious Serious 9
20 to 36 5/124 6/138 RR 0.85 7 fewer Low
analysis of trials inconsistency indirectness (4%) (4.3%) (0.25 to per 1000
(Kenyon fewer to
2013) 86 more)
Neonatal necrotising enterocolitis – Beta lactam (including co-amoxiclav)h
1 meta- Randomised Serious4 No serious Serious 10 No serious 20 to 36 29/1236 3/644 RR 4.72 17 more Low
analysis of trials inconsistency imprecision (2.3%) (0.47%) (1.57 to per 1000
(Kenyon more to
2013) 62 more)
Neonatal necrotising enterocolitis – Macrolide (including erythromycin) i
1 meta- Randomised Serious13 No serious No serious Serious 9
20 to 36 21/1322 19/754 RR 0.88 13 fewer Low
(Kenyon fewer to
2013) 72 more)
Neonatal necrotising enterocolitis – Other antibiotic i
1 meta- Randomised Serious7,12 No serious No serious Serious 9
20 to 36 25/402 30/421 RR 0.89 8 fewer Low
(Kenyon fewer to
2013) 33 more)
Neonatal infection including pneumonia – Any antibiotic j
101

1 meta- Randomised Very serious No serious Serious8 No serious 20 to 36 85/823 141/857 RR 0.67 47 fewer Very low
2,4,11,6,5,12
(Kenyon fewer to
2013) 68 fewer)
Neonatal infection including pneumonia – All penicillin (excluding co-amoxiclav) k
1 meta- Randomised Very serious 6,13 No serious Serious10 No serious 20 to 36 6/258 25/263 RR 0.3 67 fewer Very low
(Kenyon fewer to
2013) 83 fewer)
Neonatal infection including pneumonia – Beta lactam (including co-amoxiclav) l
1 study Randomised Serious4 No serious No serious Serious 9 24 to 29 0/31 1/31 RR 0.33 22 fewer Low
(Kenyon trials inconsistency indirectness (0%) (3.2%) (0.01 to per 1000
2013) 7.88) (from 32
fewer to
222 more)
Neonatal infection including pneumonia – Macrolide (including erythromycin) m
20 to 36 19/163 25/171 RR 0.79 31 fewer Very low
(Kenyon fewer to
2013) 54 more)
Neonatal infection including pneumonia – Other antibiotic n
1 meta- Randomised Serious7,12 No serious Serious10 No serious 20 to 36 60/371 90/392 RR 0.71 67 fewer Very low
analysis of trials inconsistency imprecision (16.2%) (23%) (0.53 to per 1000
(Kenyon fewer to
2013) 108
fewer)
Birth before 37 weeks' gestation – Any antibiotic o
20 to 36 3104/3642 1102/1289 RR 1.00 0 fewer Low
(Kenyon fewer to
2013) 26 more)
Birth within 7 days of randomisation – Any antibiotic p
102

1 meta- Randomised Serious10,11 No serious Serious8 No serious 20 to 36 2388/4145 1221/1820 RR 0.79 141 fewer Low
(Kenyon fewer to
2013) 195
fewer)
Serious childhood disability at approximately 7 years – Any antibiotic q
1 study Randomised No serious risk No serious No serious Serious 10 <37 938/2375 311/796 RR 1.01 4 more Moderate
(Kenyon trials of bias inconsistency indirectness (39.5%) (39.1%) (0.91 to per 1000
2013) 1.12) (from 35
fewer to
47 more)
Maternal outcomes
Maternal death – Any antibiotic r
1 meta- Randomised No serious risk No serious Serious 1 No serious 20 to 34 0/369 0/394 NC NC Moderate
analysis of trials of bias inconsistency imprecision (0%) (0%)
3 studies
(Kenyon
2013)
Maternal death – All penicillin (excluding co-amoxiclav)
1 study Randomised No serious risk No serious No serious No serious 20 to 34 0/40 0/45 NC NC High
(Kenyon trials of bias inconsistency indirectness imprecision (0%) (0%)
2013)
Maternal death – Other antibiotic (not penicillin, beta-lactam or macrolide) s
1 meta- Randomised No serious risk No serious Serious1 No serious 20 to 36 0/329 0/349 NC NC Moderate
analysis of trials of bias inconsistency imprecision (0%) (0%)
2 studies
(Kenyon
2013)
Maternal infection after birth prior to discharge – Any antibiotict
1 meta- Randomised Serious12,13 No serious No serious Serious 14 20 to 36 729/3943 306/1604 RR 0.91 17 fewer Low
(Kenyon fewer to 4
2013) more)
Chorioamnionitis – Any antibioticu
103

1 meta- Randomised Very No serious No serious Serious 14 20 to 36 126/767 196/792 RR 0.66 84 fewer Very low
1,2,5,6,7,11,12
analysis of trials serious inconsistency indirectness (16.4%) (24.7%) (0.46 to per 1000
(Kenyon fewer to
2013) 134
fewer)
Major maternal adverse drug reaction – Any antibiotic v
1 meta- Randomised Serious12,13 No serious No serious Serious 14 20 to 36 0/3913 0/1574 NC NC Low
analysis of trials inconsistency indirectness (0%) (0%)
3 studies
(Kenyon
2013)
CI confidence interval, MID minimally important difference, NC not calculable, RR risk ratio,
a. Interventions in the included studies: Mezlocillin followed by ampicillin or placebo, ampicillin and erythromycin, then amoxacillin and erythromycin or placebo, ampicillin,
pivampicillin, and metronidazole or identical placebo
b. Interventions in the included studies: Mezlocillin followed by ampicillin or matched placebo, ampicillin or matched placebo, penicillin or matched placebo, piperacillin or
placebo
c. Interventions in the included studies: Erythromycin or matched placebo, co-amoxiclav and erythromycin or matched placebo, erythromycin or matched placebo, eythromycin
or matched placebo
d. Interventions in the included studies: Ampicillin and erythromycin then amoxicillin and erythromycin or placebo indamycin and gentamycin or matched placebo, ampicillin,
pivampicillin and metronidazole or identical placebo
e. Interventions in the included study: Ampicillin, pivampicillin and metronidazole or matched placebo.
f. Interventions in the included studies: Benzylpenicillin and penicillin or matched placebo, mezlocillin and ampicillin or matched placebo
ampicillin or matched placebo, mezlocillin and ampicillin or matched placebo, co-amoxiclav and erythromycin or matched placebo, ampicillin or matched placebos, penicillin or
matched placebo, piperacillin or placebo, erythromycin or placebo, erythromycin or matched placebo, clindamycin and gentamycin or matching placebo
g. Interventions in the included studies: Mezlocillin and ampicillin or matched placebo, mezlocillin and ampicillin or matched placebo, ampicillin or matched placebos
h. Interventions in the included studies: benzylpenicillin and penicillin or matched placebo, co-amoxiclav and erythromycin or matched placebo
Interventions in the included studies: co-amoxiclav and erythromycin or matched placebo, penicillin or matched placebo, piperacillin or placebo
i. Interventions in the included studies: Ampicillin or matched placebo, erythromycin or matched placebo, erythromycin or matched placebo, clindamycin and gentamycin or
matching placebo
j. Interventions in the included studies: benzylpenicillin and penicillin or matched placebo, metzlocillin or placebo, mezlocillin and ampicillin or matched placebo, ampicillin,
Pivampicillin and metronidazole or identical placebo, co-amoxiclav or matched placebo, eythromycin or matched placebo, ampicillin or matched placebo, penicillin or matched
placebo, piperacillin or placebo, erythromycin or placebo, erythromycin or matched placebo, clindamycin and gentamycin or matching placebo
k. Interventions in the included studies: Metzlocillin or placebo, mezlocillin and ampicillin or matched placebo, co-amoxiclav or matched placebo, erythromycin or matched
placebo, ampicillin or matched placebo
l. Interventions in the included study: -Benzylpenicillin and penicillin or matched placebo
m. Interventions in the included studies: Ampicillin, pivampicillin and metronidazole or identical placebo, penicillin or matched placebo, piperacillin or placebo
104
n. Interventions in the included studies: Erythromycin or placebo, erythromycin or matched placebo, clindamycin and gentamycin or matched placebo
o. Interventions in the included studies: Co-amoxiclav and erythromycin or matched placebo, penicillin or matched placebo, clindamycin and gentamycin or matching placebo
p. Interventions in the included studies: Mezlocillin and ampicillin or matched placebo, ampicillin or matched placebo, mezlocillin and ampicillin or matched placebo, co-
amoxiclav and erythromycin or matched placebo, ampicillin or matched placebo, piperacillin or placebo, erythromycin or placebo
q. Interventions in the included study: Co-amoxiclav and erythromycin or matched placebo
r. Interventions in the included studies: mezlocillin and ampicillin or placebo, ampicillin and erythromycin or placebo, ampicillin, pivampicillin and metronidazole or identical
placebo
s. Interventions in the included studies: ampicillin and erythromycin or placebo, ampicillin and metronidazole or identical placebo
t. Interventions in the included studies: ampicillin, oral pivampicillin and metronidazole or identical placebo, co-amoxiclav and erythromycin or matched placebo, erythromycin or
placebo, clindamycin and gentamycin or matched placebo
u. Interventions in the included studies: IV metzlocillin or placebo, ampicillin, pivampicillin and metronidazole or identical placebo, ampicillin and ampicillin or matched placebo,
co-amoxiclav or matched placebo, erythromycin or matched placebo, ampicillin and ampicillin or matched placebo, penicillin or matched placebo, piperacillin or placebo,
erythromycin or placebo, clindamycin and gentamycin or matched placebo
v. Interventions in the included studies: Co-amoxiclav and erythromycin or matched placebo, erythromycin or placebo, clindamycin and gentamycin or matching placebo
1. 118/614 women were Group B Strep positive
2. Unclear method of randomisation in 1 study
3.. Unclear allocation concealment in 4 studies
4. Data collected from an abstract in 1 study
5. 15% of loss to follow up in 1 study
6. One study specified that 101 women were randomised but results for 115 women are reported
7. One study specified that trial stopped after intermediate evaluation showed treatment group had better outcome
8. Twin pregnancy included in 3 studies
9. Confidence intervals crossed 1 default MID
10. Twin pregnancy included in 1 study
11. Unclear allocation concealment in 3 studies
12. Data from one study extracted from a PhD thesis
13. Unclear allocation concealment in 1 study
Table 22: GRADE profile for comparison of antibiotic therapy versus either placebo or no antibiotic therapy

No. of Gestation No Relative
studies Design Risk of bias Inconsistency Indirectness Imprecision (weeks) Antibiotic antibiotic (95% CI) Absolute Quality
Neonatal outcomes
Perinatal death/death before dischargea
1 meta- Randomised Very No serious Serious9 No serious 20 to 36 299/4604 172/2268 RR 0.89 8 fewer per Very
analysis trials serious1,2,3,4,5,6,7,8 inconsistency imprecision (6.5%) (7.6%) (0.74 to 1000 (from low
of 18 1.08) 20 fewer to 6
studies more)
105

No. of Gestation No Relative
studies Design Risk of bias Inconsistency Indirectness Imprecision (weeks) Antibiotic antibiotic (95% CI) Absolute Quality
(Kenyon
2013)
Intraventricular haemorrhageb
1 meta- Randomised Very serious12,13 No serious Serious 14 No serious 20 to 36 74/572 105/590 RR 0.73 48 fewer per Very
analysis trials inconsistency imprecision (12.9%) (17.8%) (0.56 to 1000 (from 9 low
of 7 0.95) fewer to 78
studies fewer)
(Mercer,
2011)
Neonatal sepsisc
1 meta- Randomised Serious13 No serious Serious 14 No serious 20 to 36 53/485 82/489 RR 0.67 55 fewer per Low
analysis trials inconsistency imprecision (10.9%) (16.8%) (0.49 to 1000 (from
of 5 0.91) 15 fewer to
studies 86 fewer)
(Mercer,
2011)
Birth delayed ≥7 days after randomisationd
1 meta- Randomised Very serious12,13 No serious Serious 14 No serious 20 to 36 237/515 139/537 RR 1.8 207 more Very
analysis trials inconsistency imprecision (46%) (25.9%) (1.52 to per 1000 low
of 6 2.13) (from 135
studies more to 292
(Mercer, more)
2011)
CI confidence interval, RR risk ratio

a. Interventions in the included studies: Ampicillin, ampicillin, gentamycin and amoxicillin and clavulanic acid, co-amoxiclav or matched placebo, erythromycin or matched
placebo, ampicillin or matched placebos, metzlocillin and ampicillin or matched placebo, co-amoxiclav and erythromycin or matched placebo, penicillin or matched placebo,
piperacillin or placebo, co-amoxiclav or placebo, erythromycin or placebo, erythromycin or matched placebo, a-mpicillin, erythromycin and amoxacillin or matched placebo,
ampicillin, clindamycin and gentamycin or matched placebo, ampicillin or matched placebo, ampicillin, pivampicillin and metronidazole or identical placebo
b. Interventions in the included studies: Ampicillin, ampicillin, gentamycin and amoxicillin and clavulanic acid, metzlocillin or placebo, IV Metzlocillin and ampicillin or matched
placebo, piperacillin or placebo, erythromycin or matched placebo
- Cindamycin and gentamycin or matched placebo
c. Interventions in the included studies: Ampicillin, ampicillin, gentamycin, amoxicillin and clavulanic acid, piperacillin or placebo, ampicillin, erythromycin and amoxicillin or
matched placebo, ampicillin
d. Interventions in the included studies: Ampicillin, ampicillin, gentamycin, amoxicillin and clavulanic acid or placebo, metzlocillin or placebo, metzlocillin and ampicillin or
matched placebo, piperacillin or placebo, ampicillin, erythromycin and amoxicillin or matched placebo
1. No blinding in 6 studies
106
2. Unclear method of randomisation in 3 studies

3. Unclear allocation concealment in 8 studies.
4. Data from extracted from a PhD thesis 1 study
5. 15% of loss to follow up in 1 study.
6. One study specified that 101 women were randomised but results for 115 are reported
7. One study specified that trial stopped after intermediate evaluation showed treatment group had better outcome
8. Data from extracted from an abstract in 1 study
11. Unclear method of randomisation in 1 study
12. Unclear allocation concealment in 2 studies
14. Multiple pregnancy included in 1 study
Table 23: GRADE profile for comparison of antibiotic therapy versus no antibiotic therapy (childhood outcomes at 7-year follow-up)
Quality assessment No of patients Effect

Any
antibiotic
No of Risk of Other versus no Relative
studies Design bias Inconsistency Indirectness Imprecision considerations antibiotics Control (95% CI) Absolute Quality
Total death/stillbirths, death in first year, death after first year – Any erythromycin versus no erythromycin
1 study Randomised No No serious Serious1 No serious None 156/2323 172/2389 RR 0.93 5 fewer per Moderate
(Kenyon trials serious inconsistency imprecision (6.7%) (7.2%) (0.74 to 1000 (from
2008) risk of 1.16) 19 fewer to
bias 11 more)
Total death/death after first year – Any erythromycin versus no erythromycin
1 study Randomised No No serious Serious1 No serious None 7/2323 4/2389 RR 1.79 1 more per Moderate
bias 9 more)
Total death/stillbirths, death in first year, death after first year – Any co-amoxiclav versus no co-amoxiclav
1 study Randomised No No serious Serious1 Serious2 None 163/2336 165/2376 RR 1.01 1 more per Low
(Kenyon trials serious inconsistency (7.0%) (6.9%) (0.80 to 1000 (from
bias 18 more)
Total death/death after first year – Any co-amoxiclav versus no co-amoxiclav
107

Any
antibiotic
No of Risk of Other versus no Relative
studies Design bias Inconsistency Indirectness Imprecision considerations antibiotics Control (95% CI) Absolute Quality
1 study Randomised No No serious Serious1 No serious None 5/2336 6/2376 RR 0.85 0 fewer per Moderate
bias 4 more)
Cerebral palsy – Any erythromycin versus no erythromycin
1 study Randomised No No serious Serious1 No serious None 46/1590 41/1671 RR 1.18 4 more per Moderate
bias 20 more)
Cerebral palsy – Any co-amoxiclav versus no co-amoxiclav
1 study Randomised No No serious Serious1 No serious None 39/1632 48/1629 RR 0.81 24 fewer Moderate
(Kenyon trials serious inconsistency imprecision (2.4%) (2.9%) (0.53 to per 1000
2008) risk of 1.24) (from 14
bias fewer to 7
more)
Any functional impairment at age 7 – Any erythromycin versus no erythromycin
1 study Randomised No No serious Serious1 Serious2 None 594/1551 655/1620 RR 0.91 36 fewer Low
(Kenyon trials serious inconsistency (38.3%) (40.4%) (0.79 to per 1000
2001) risk of 1.05) (from 85
bias fewer to 20
more)
Any functional impairment at age 7 – Any co-amoxiclav versus no co-amoxiclav
1 study Randomised No No serious Serious1 Very serious3 None 645/1587 604/1584 RR 1.11 42 more Very low
(Kenyon trials serious inconsistency (40.6%) (38.1%) (0.96 to per 1000
2001)a risk of 1.25) (from 15
bias fewer to 95
more)
CI confidence interval, MID minimally important difference, RR risk ratio

1. Multiple pregnancy included
2. Confidence intervals crossed 1 default MID
3. Confidence intervals crossed 2 default MIDs
108

6.5.1 Any antibiotic therapy compared with placebo
Neonatal outcomes
Very low quality evidence from meta-analysis of 12 studies with 1600 women with suspected
or diagnosed preterm prelabour rupture of membranes (P-PROM) found that neonatal
infection including pneumonia was significantly lower in the group of women who received
prophylactic antibiotics compared with those allocated to receive placebo.
A meta-analysis of 7 studies (low quality evidence) from almost 6000 women with P-PROM
found that significantly fewer of the women who received prophylactic antibiotics gave birth
within 7 days of randomisation compared with those allocated to receive placebo.
Low quality evidence from 1 meta-analysis of 2 RCTs of over 1800 women with suspected or
diagnosed P-PROM showed a higher incidence of necrotising enterocolitis in babies born to
women who received prophylactic beta lactam antibiotics including co-amoxiclav compared
with those allocated to receive a placebo. Moderate to very low quality evidence from meta-
analysis of 2 to 12 RCTs (number of participants ranged from 61 to 6301) showed no
significant difference for the other outcomes (perinatal death, neonatal encephalopathy,
positive neonatal blood culture, birth before 37 weeks of gestation, serious childhood
disability at approximately 7 years).
Maternal outcomes
High to very low quality evidence from meta-analysis of 2 to 11 RCTs (number of participants
ranged from 85 to 1559) found no significant difference in maternal death, maternal infection
after birth prior to discharge and major maternal adverse drug reaction in women with
preterm prelabour rupture of membranes allocated to receive prophylactic antibiotics
compared with those allocated to receive placebo treatment.
Very low quality evidence from meta-analysis of 11 RCTs (n=1559) indicated that incidence
of chorioamnionitis was lower in women with P-PROM allocated to receive prophylactic
antibiotics compared with those allocated to receive placebo treatment.
6.5.2 Antibiotic therapy compared with either placebo or no antibiotic

therapy
Neonatal outcomes
Very low quality evidence from meta-analysis of 18 RCTs (n=6872) showed no difference in
the incidence of perinatal death/death before discharge in neonates whose mother with
P-PROM was allocated to receive prophylactic antibiotics compared with those allocated to
receive no antibiotic treatment.
Low and very low quality evidence from 5 to 7 RCTs (number of participants ranged from 974
to 1162) found significantly lower incidence of each of intraventricular haemorrhage and
sepsis and significantly higher number of births delayed by at least 7 days after
randomisation in neonates whose mother with P-PROM was allocated to receive prophylactic
antibiotics compared with those allocated to receive no antibiotic treatment.
109
Maternal outcomes
Maternal outcomes relevant to this section were not reported.
6.5.3 Antibiotic therapy compared with no antibiotic therapy (childhood

outcomes at 7-year follow-up)
Moderate to very low quality evidence from 1 RCT (n=4712) showed no significant difference
in the incidence of total death, death after first year, cerebral palsy and functional impairment
at age 7 years in neonates whose mother was diagnosed with P-PROM and allocated to
receive prophylactic antibiotics (either erythromycin or co-amoxiclav) compared with those
allocated to receive no antibiotic treatment.

A search was undertaken for health economic evidence on antenatal prophylactic antibiotics
given to women with diagnosed preterm prelabour rupture of membranes to improve
outcomes of preterm labour. A total of 73 articles were identified by the search. After
reviewing titles and abstracts, 2 full papers were obtained and 1 was included for review.
A UK study (Colbourn 2007) considered a range of prenatal strategies for preventing group B
streptococcus and other serious bacterial infections in early infancy. The study looked at 12
different populations of risk groups including a preterm group with membrane rupture more
than 2 hours before labour onset and its analysis suggested that treatment with intravenous
antibiotics would be cost effective in such a population.
This question was not prioritised for health economic analysis as it was thought by the
committee to be a cheap intervention and something that was part of current clinical practice,
which the committee expected would not be changed by the guideline.
6.7 Evidence recommendations

The committee prioritised the following clinical outcomes:
• maternal outcomes:
o mortality
o maternal infections (such as chorioamnionitis)
o major adverse events
• neonatal outcomes:
o neonatal or perinatal mortality
o number of babies born preterm
o brain injury including intraventricular haemorrhage
o periventricular leucomalacia (PVL)/white matter injury
o necrotising enterocolitis
o any neonatal infection (including neonatal sepsis).
The protocol also included any long-term outcomes in childhood (particularly functional
impairments, behavioural difficulties, cerebral palsy, seizures and wheezing) by taking into
account that the long-term impact may be affected by other influences (not necessarily the
administration of antibiotics before delivery) and for that reason long-term neurological
outcomes were not included.
110
Evidence on neonatal encephalopathy was sought but not found, but the committee did not
consider this as a critical outcome for drafting these recommendations in the context of long-
term follow-up.
When considering the relative value of each outcome, the committee assumed that
outcomes relating to infection in the baby would pertain to early onset neonatal infection.
Although antibiotics given to women with P-PROM appeared to reduce the rate of positive
neonatal blood cultures, the committee placed little additional weight on this outcome, over
and above the other beneficial effects for the baby.

The evidence from the included SRs and meta-analyses showed that prophylactic antibiotics
for women with P-PROM may reduce the incidence of chorioamnionitis but this effect was
not found for other indices of maternal infection.
Low quality evidence was found which showed that prophylactic antibiotics for women with
P-PROM might delay birth for more than 7 days. The committee was uncertain whether birth
in the reviewed studies followed spontaneous (preterm) labour or was due to other factors. In
terms of the neonatal outcomes, there was some clear evidence that neonatal infections,
including pneumonia and sepsis, were reduced by antibiotic prophylaxis in women with P-
PROM.
Although intraventricular haemorrhage appeared to be reduced with the use of prophylactic
antibiotics compared with no antibiotics for women with P-PROM, the majority of evidence
was from babies who were born in the late 1980s. The committee questioned the relevance
of the data to current practice and inferred that it is most likely that these babies may already
have had compromised health status, so that, in general, infection would have just been one
more added problem rather than the primary cause of haemorrhage. Hence the committee
gave little weight to the apparent benefit of antibiotics for P-PROM in preventing
haemorrhage.
In addition, the committee discussed the challenge of interpreting the results on infection,
given that there was no clear indication of the nature of infections in the included evidence
and the term ‘pneumonia’ might be too heterogeneous to draw clinically relevant results.
In summary, although antibiotics given to mothers with P-PROM seem to have little effect on
the long-term health outcomes of children, the short-term advantages (reducing neonatal
infection and delaying birth) are such that the committee decided that antibiotics should be
offered routinely to all women with P-PROM.
Although the evidence base for this section was not robust, the committee concluded that
this recommendation should be strong. Giving antibiotics to women with P-PROM is currently
standard clinical practice in the UK and the review of evidence in this question showed no
reason to change this practice. More specifically, the evidence of no harm for the baby in
terms of cerebral palsy or for the mother in terms of major maternal adverse drug reaction
further confirmed the direction and the strength of the recommendation. The committee
discussed the absence of any major maternal drug side effects and that this can be
explained because history taking in women with P-PROM can identify allergies and can
therefore determine the appropriate class of antibiotics to be offered.
6.7.2.1 Choice of antibiotic

The committee noted that although antibiotics overall had no effect on necrotising
enterocolitis and the beta lactam anitbiotics (including co-amoxiclav) reduced neonatal
111
infection including pneumonia, beta lactams (including co-amoxiclav) significantly increased

the risk of necrotising enterocolitis. Hence the committee decided that beta-lactam antibiotics
should not be selected to improve neonatal outcomes in women with P-PROM.
Regarding other available antibiotics, the committee considered that, in addition to the
benefits of erythromycin shown in the evidence summary above, there are additional
potential benefits of erythromycin as the choice of antibiotic in women with P-PROM. Firstly,
erythromycin is not reported to increase the risk of necrotising enterocolitis. Secondly, it can
be administered orally to target group B streptococcus, other streptococcal and
staphylococcal infections, bacteria relevant to early-onset sepsis, and other micro-organisms
affecting the woman and baby before labour. Thirdly, erythromycin offers a theoretical
advantage (for the woman, rather than the baby) in that it can counteract mycoplasma
infection that is implicated in the early stages of chorioamnionitis – this effect is not seen with
penicillins). Finally, the absorption of erythromycin across the gastrointestinal tract and the
placenta is limited, which suggests a potential benefit in terms of minimising the baby’s
exposure to antibiotics. The committee noted that although there was little evidence of
benefit to the baby, there was no evidence of harm. Although few antibiotics are licensed for
use in pregnancy or in preterm babies, there was a strong consensus within the committee
that healthcare professionals should consider antibiotic prophylaxis using erythromycin for
women with P-PROM.
The committee specified that the recommended dosage of erythromycin in this section
should be 250 mg 4 times per day, as this was the dosage used in the largest study included
in the evidence basis of this recommendation.

Antibiotics are cheap and infection could result in longer stay in hospital. Furthermore,
infection may lead to poor health outcomes, the cost of which can be very high. However,
there are concerns that the overuse of antibiotics can promote antibiotic resistance, which
has potentially large implications for future health benefits and costs. There was some
evidence that antibiotics could delay birth which could have some resource implications,
although the committee noted that women with P-PROM are now more likely to be managed
as outpatients than was historically the case.

The majority of evidence was downgraded in this section due to high risk of bias and
imprecision. Lack of blinding was the most common reason for the quality of evidence to be
downgraded. For the outcome of prolonging delivery for 7 or more days, the committee
discussed the importance of masking treatment allocation in trials on interpretation of results
and clinicians decision-making. However, the study with the highest weight in the meta-
analysis was masked and therefore the committee placed confidence in the estimates of
effect for this outcome.
The committee discussed the importance of accurate diagnosis of P-PROM and highlighted
potential for harm to the baby (for example by increased risk of cerebral palsy) from
inadvertent use of prophylactic antibiotics in women with intact membranes who are
incorrectly diagnosed with P-PROM.
When the evidence was examined, the committee noted that participants in the included
trials would have been giving birth more than 15 years ago, with the majority delivering even
longer ago (in the 1980s). These studies would have included a different profile of babies’
health status, being in worse health generally and therefore resulting in an overestimation of
112
the magnitude of any effect (positive or negative). The population in these studies would not
reflect the population now being treated, thereby limiting the generalisability of its results.

committee members’ clinical expert opinion.
When the committee discussed the use of erythromycin, they noted that the benefit might not
be entirely due to bactericidal properties but also that the reduction in associated
inflammatory damage might influence outcomes.
When the committee discussed the role of antibiotics given to women with P-PROM as
prophylactic measures for improving neonatal or mortality outcomes, there was a lack of
information on this role at different gestational ages.
6.8 Recommendations
updated guideline.
113
Identifying infection in women with P-PROM
7 Identifying infection in women with P-

PROM
7.1 Introduction
Preterm prelabour rupture of membranes (P-PROM) exposes both the fetus and the mother
to risk of infection. Bacterial infection can be life threatening to the mother and to the unborn
fetus or the baby postnatally. Ascending infection with Group B streptococcus is particularly
dangerous. Other infections of the amniotic membranes or umbilical cord also put the fetus at
risk of cytokine-induced white matter damage in the brain, which may result in periventricular
leukomalacia and cerebral palsy. The cytokines produced by the maternal immune response
to infection can also trigger premature labour. Hence there is a need to be able to identify
maternal infection early to institute appropriate treatment with antibiotics and to avoid the
unnecessary use of antibiotics where there is no infection. Indiscriminate use of antibiotics
potentially puts the mother and fetus at risk of side effects and may encourage antibiotic
resistance.
7.2 Review question

What is the diagnostic value of temperature, pulse, white cell count, C-reactive protein and
cardiotocography (CTG) to identify infection in women with preterm prelabour rupture of
membranes (P-PROM)?

Thirteen studies were included in this review. One of the included studies was a randomised
controlled trial (RCT) (Lewis 1999) that randomised women with P-PROM to either a CTG or
biophysical profile. Only data for the CTG arm are relevant to this review. One study
analysed women with P-PROM participating in an RCT of corticosteroids (Garite 1982). Nine
studies were prospective case series, 1 of which was a consecutive case series (Kurki 1990);
in the other 8 studies it was unclear whether women were recruited consecutively (Carroll
1995, Farb 1983, Fisk 1987, Hawrylyshyn 1983, Ismail 1985, Perrone 2012, Romem 1984,
Yoon 1996). Two studies were retrospective case series (Del Valle 1992, Smith 2012).
Nine of the studies (Farb 1983, Fisk 1987, Hawrylyshyn 1983, Ismail 1985, Kurki 1990,
Perrone 2012, Romem 1984, Smith 2012, Yoon 1996) reported values for C-reactive protein
at various thresholds as a predictor of clinical amnionitis or chorioamnionitis, or histological
chorioamnionitis. Four of the studies reported values for white blood cell count at various
thresholds as a predictor of clinical or histological chorioamnionitis (Garite 1982,
Hawrylyshyn 1983, Romem 1984, Yoon 1996) and 1 study reported values for maternal
temperature as a predictor of clinical or histological chorioamnionitis (Ismail 1985). The
prevalence of histological chorioamnionitis ranged from 21% to 63% (6 studies) and of
clinical chorioamnionitis ranged from 14% to 29% (4 studies). None of the included studies
looked at the role of maternal C-reactive protein or maternal white blood cell count in
identifying neonatal sepsis.
Three studies reported values for the CTG as a predictor of neonatal infection (Carroll 1995,
Del Valle 1992, Lewis 1999), 1 study reported values for fetal heart rate as a predictor of
both clinical and histological chorioamnionitis (Ismail 1985) and 1 study reported values for
fetal heart rate as a predictor of clinical chorioamnionitis (Garite 1982).
114
The mean gestational age at rupture of membranes was reported in 6 studies and ranged
from 26.7 weeks (standard deviation [SD] 0.8) to 31.8 weeks (SD 2.6). The duration of
preterm prelabour rupture of membranes was reported in 5 studies and ranged from 3.5 days
(SD 12.1) to 16 days (SD 12).
In the majority of studies, maternal serum samples for C-reactive protein determination
and/or white blood cell count were taken on a daily basis from admission until birth. In the 3
studies where CTG was the index test, the test was performed daily. The timing of the test
results selected for analysis was not clearly reported in the majority of the studies. Two
studies reported diagnostic accuracy values for the last CTG performed immediately before
birth (Del Valle 1992, Lewis 1999), 2 studies reported predictive values for the last recorded
C-reactive protein level taken before birth (Fisk 1987, Smith 2012), 1 study reported results
for C-reactive protein taken at admission (Romem 1984), 1 study looked at the role of
C-reactive protein samples taken at admission and samples taken 24 to 48 hours before birth
(Perrone 2012) and 1 study focused at on white blood cell count and fetal heart rate
measured at admission (Garite 1982).
The majority of studies (9/13) required that ruptured membranes were confirmed both by
visualisation of amniotic fluid and a positive biochemical test. Five studies reported that
maternal antibiotic therapy was not given during the period before birth (Farb 1983, Fisk
1987, Hawrylyshyn 1983, Romen 1984, Yoon 1996), 3 reported women were given
antibiotics on clinical diagnosis of chorioamnionitis (Del Valle 1992, Garite 1982, Smith 2012)
and 2 reported the administration of routine prophylactic antibiotics (Lewis 1999, Perrone
2012). Three studies did not give any information in relation to use of antibiotics.
Two studies included women with a multiple pregnancy (10% of women in Kurki 1990, 8% of
women in Fisk 1987; but only data in singleton pregnancies is included in the review for this
study) and 2 studies stated only singleton pregnancies were included (Perrone 2012, Yoon
1996). In the remaining studies it was unclear if women with a multiple pregnancy were
included.
There were no studies identified on the predictive value of maternal pulse to identify infection
in women with preterm prelabour rupture of membranes.
See the evidence table in Appendix H for further details of included studies.

Data is reported separately for different tests in the following GRADE profiles:
• Table 28: GRADE profile for predictive accuracy of C-reactive protein for identifying
infection
• Table 29: GRADE profile for predictive accuracy of maternal white blood cell count for
identifying infection
• Table 30: GRADE profile for predictive accuracy of fetal heart rate for identifying
infection
• Table 31: GRADE profile for predictive accuracy of maternal temperature for identifying
infection.
The specific tests and the thresholds used (such as C-reactive protein greater than 2 mg/100
ml) are listed in the rows of each GRADE table and the outcomes that they predict are listed
in the ‘definition of outcome’ column.
115
Evidence from prospective case series started at high quality for the purposes of this review
question and was then downgraded if there were any issues identified that would undermine
the trustworthiness of the findings (for example if it was unclear whether consecutive women
were included in the study). Retrospective case series started at moderate quality and were
then downgraded if there were any issues (for example if it was unclear whether consecutive
women were included in the study).
Findings are reported separately for each study since the timing of testing, administration of
antibiotic therapy, definitions of outcome measures and thresholds used vary across studies
or are not clearly reported, thus making pooling of data inappropriate. In order to provide a
synthesis of findings the range of all values for a particular test is given in the first row of the
relevant GRADE profile.
116
Table 24: GRADE profile for predictive accuracy of C-reactive protein for identifying infection
Measures of diagnostic accuracy (95% confidence
Quality assessment interval)
Outcome and
prevalence Positive Negative
Number. of Risk of Inconsist Indirectnes Impreci (type of Prevalen Sensitivit likelihood likelihood
studies Design bias ency s sion infection) ce y Specificity ratio ratio Quality
C-reactive protein – all thresholds measured at a range of time points
Overall Case series Very Very No serious Very Clinical Range: Range: 32% Range: Range: Very low to
summary of serious1,2 serious 4 indirectness serious 5 chorioamnionitis 37% to to 100% 1.13 to 0.12 to high
findings from , histological 94% Low to high 23.0 0.72
9 studies chorioamnionitis Low to Not useful Not useful
or histological high to very to
funisitis 14% to useful moderately
63% useful
C-reactive protein ≥0.7 mg/100 ml measured within 72 hours of birth
1 study Case series Serious1 No No serious Serious 6 Histological 54% 86% (72.75 3.8 (1.46 to 0.53 (0.36 Low
(Yoon 1996) serious indirectness chorioamnionitis (37.78 to to 98.68) 9.89) to 0.79)
inconsiste (56%) 70.79) Moderate Not useful Not useful
ncy Low
C-reactive protein >1.2 mg/100 ml *
1 study (Kurki Case series No serious No No serious Serious 7 Clinical and 33/147 94% 50% (40.82 1.88 (1.53 0.12 (0.03 Moderate
1990) risk of bias serious indirectness histological (85.8 to to 59.18) to 2.30) to 0.47)
inconsiste chorioamnionitis 100) Low Not useful Moderately
ncy (22%) High useful
C-reactive protein >1.2 mg/100 ml measured on admission (admission to birth interval: mean 16 days (SD 12 days)
1 study Case series Serious No No serious Serious6 Histological 24/66 41.7% 83.3% (69.4 2.5 (1.10 to 0.70 (0.49 Low
(Perrone risk of bias2 serious indirectness funisitis (24.5 to to 91.7) 5.71) Not to 1.01)
2012) inconsiste (36%) 61.2) Moderate useful Not useful
ncy Low
C-reactive protein >1.2 mg/100 ml measured 24 to 48 hours before birth
1 study Case series Serious No No serious Serious6 Histological 24/66 75.0% 69.0% (54.0 2.42 (1.46 0.36 (0.18 Low
(Perrone risk of bias2 serious indirectness funisitis (55.1 to to 80.9) to 4.02) to 0.75)
2012) inconsiste (33%) 88.0) Low Not useful Moderately
ncy Moderate useful
C-reactive protein >1.25 mg/100 ml measured at birth or last results obtained during hospital admission if discharged undelivered
1 study Case series Serious 2 No No serious Serious7 Histological 26/52 88% 96% (88.76 23.00 (3.35 0.12 (0.04 Low
(Hawrylyshyn serious indirectness chorioamnionitis (76.18 to to 100) to 157.97) to 0.35)
1983) inconsiste (50%) 100) a High Very useful Moderately
ncy Moderate useful
C-reactive protein ≥2 mg/100 ml measured on admission
117

Outcome and
1 study Case series Serious 2 No No serious Very Clinical 7/51 86% 82% (70.42 4.71 (2.35 0.17 (0.03 Very low
(Romem serious indirectness serious 3 chorioamnionitis (59.79 to to 93.21) to 9.46) to 1.08)
1984) inconsiste (14%) 100) Moderate Not useful Moderately
ncy Moderate useful
C-reactive protein >2 mg/100 ml *
1 study (Farb Case series Serious 2 No No serious Serious 6 Clinical 9/31 56% 73% (54.12 2.04 (0.83 0.61 (0.28 Low
1983) serious indirectness chorioamnionitis (23.09 to to 91.34) to 5.00) to 1.33)
inconsiste (29%) 88.02) Low Not useful Not useful
ncy Low
1 study Case series Serious 2 No No serious Serious 6 Clinical 18/100 82% 55% (44.11 1.85 (1.35 0.30 (0.11 Low
(Ismail 1985) serious indirectness chorioamnionitis (66.12 to to 65.65) to 2.53) to 0.87)b
inconsiste (18%) 100) Low Not useful Moderately
ncy Moderate useful
1 study (Farb Case series Serious 2 No No serious Very Histological 5/24 80% 68% (47.52 2.53 (1.15 0.29 (0.05 Very low
1983) serious indirectness serious 5 chorioamnionitis (44.94 to to 89.32) to 5.60) to 1.73)
inconsiste (21%) 100) Low Not useful Moderately
ncy Moderate useful
1 study Case series Serious 2 No No serious Serious 6 Histological 63/100 67% 81% (68.46 3.52 (1.77 0.41 (0.28 Low
(Ismail 1985) serious indirectness chorioamnionitis (55.03 to to 93.70)a to 7.02) to 0.60)
inconsiste (63%) 78.31) Moderate Not useful Moderately
ncy Low useful
C-reactive protein >2 mg/100 ml measured on admission (mean 16 days (SD 12 days) from admission to birth)
1 study Case series Serious 2 No No serious Serious 6 Histological 24/66 37.5% 90.5% (77.9 3.94 (1.36 0.69 (0.50 Low
(Perrone serious indirectness funisitis (21.2 to to 96.2) to 11.43) to 0.96)
2012) inconsiste (33%) 57.3 Low High Not useful Not useful
ncy
C-reactive protein >2 mg/100 ml taken within 48 hours of birth
1 study (Fisk Case series Serious 2 No No serious Serious 6 Histological 30/51 50% 81% (64.16 2.63 (1.01 0.62 (0.41 Low
1987) serious indirectness chorioamnionitis (32.11 to to 97.25)a to 6.80)b to 0.93)b
inconsiste (59%) 67.89)a Moderate Not useful Not useful
ncy Low
C-reactive protein >2 mg/100 ml measured 24 to 48 hours before birth
1 study Case series Serious 2 No No serious Serious 6 Histological 24/66 54.2% 88.1% (75.0 4.55 (1.85 0.52 (0.33 Low
(Perrone serious indirectness funisitis (35.1 to to 94.8) d to 11.21) b to 0.82) b
2012) inconsiste (33%) 72.1) d Moderate Not useful Not useful
ncy Low
C-reactive protein >3 mg/100 ml measured within 48 hours of birth
1 study (Fisk Case series Serious 2 No No serious Serious 6 Histological 30/51 47% 90% (77.92 4.9 (1.24 to 0.59 (0.41 Low
1987) serious indirectness chorioamnionitis (28.81 to to 100) a 19.33)b Not to 0.85)b
inconsiste (59%) 64.52)a High useful Not useful
ncy Low
118

Outcome and
C-reactive protein >3.5 mg/100 ml taken within 48 hours of birth
1 study (Fisk Case series Serious 2 No No serious Serious 6 Histological 30/51 40% 95% (86.13 8.4 (1.18 to 0.63 (0.46 Low
1987) serious indirectness chorioamnionitis (22.47 to to 100) a 59.77) b to 0.86) b
inconsiste (59%) 57.53) a High Moderately Not useful
ncy Low useful
C-reactive protein >4 mg/100 ml
1 study (Kurki Case series Serious 2 No No serious Serious 6 Clinical 33/147 72% 77% (69.49 3.19 (2.14 0.35 (0.20 Low
1990) serious indirectness chorioamnionitis (57.53 to to 84.90)a to 4.74)b to 0.62)b
inconsiste (22%) 87.92) a Moderate Not useful Moderately
ncy Low useful
C-reactive protein >4 mg/100 ml taken within 48 hours of birth
1 study (Fisk Case series Serious 2 No No serious Serious 6 Histological 30/51 37% 100% (100 to NC 0.63 (0.48 Low
1987) serious indirectness chorioamnionitis (19.42 to 100) a Very useful to 0.83) b
inconsiste 59% 53.91)a High Not useful
ncy Low
C-reactive protein >5 mg/100ml (measurement closest to time of birth reported)*
1 study Randomised Serious 2 No No serious Serious 6 Histological 26/73 76.9% 31.9% (18.59 1.13 (0.85 0.72 (0.32 Low
(Smith 2012) trial serious indirectness chorioamnionitis (60.73 to to 45.24)a to 1.51)b to 1.64)b
inconsiste (36%) 93.12)a Low Not useful Not useful
ncy Moderate
NC not calculable, SD standard deviation
* Timing of measurement not reported/unclear
a. As reported in study, confidence intervals calculated by NCC
b. Calculated by NCC
1. Selection bias – only women who gave birth within 72 hours of amniocentesis were analysed and unclear whether consecutive women were included in the study
2. Possible selection bias - unclear whether consecutive women were included in the study
3. Evidence was downgraded by 2 due to 95% confidence interval for the negative likelihood ratio ranges from not useful (>0.5) to very useful (0–0.1)
4. Evidence was downgraded by 2 due to very serious heterogeneity (chi-squared p<0.1, I-squared inconsistency statistic of >75%). 5 Summary of findings from studies
reporting a wide range of values and confidence intervals
5. Evidence was downgraded by 2 due to 95% confidence interval for the positive likelihood ratio ranges from not useful (<5) to very useful (>10)
6. Evidence was downgraded by 1 due to 95% confidence interval for the negative likelihood ratio ranges from not useful (>0.5) to moderately useful (>0.1–0.5)
7. Evidence was downgraded by 1 due to 95% confidence interval for the negative likelihood ratio ranges from moderately useful (>0.1–0.5) to very useful (0–0.1)
119
Table 25: GRADE profile for predictive accuracy of maternal white blood cell count for identifying infection
Measures of diagnostic
Quality assessment accuracy
Outcome and
Number. of Risk of Inconsist Indirectnes Impreci (type of Prevale likelihood likelihood
studies Design bias ency s sion infection) nce Sensitivity Specificity ratio ratio Quality
White blood cell count - all thresholds measured at a range of time points
Overall Case series Serious 1,2 Very No serious Serious4 Clinical or Range:16. Range: 62% Range: 2.10 Range: Low to
summary of serious3 indirectness histological 7% to to 97.5% to 6.70 0.31 to moderate
findings from chorioamnioniti 80% Low to high Not useful 0.85
4 studies s (14% to 56%) Low to to Not useful
moderate moderately
useful
White blood cell count >12,500 cells/mm3 measured at birth
1 study Case series Serious 1 No No serious Serious Histological 26/52 80% 62% (42.84 2.10 (1.25 0.31 (0.13 Low
(Hawrylyshyn serious indirectness 4
chorioamnioniti (65.62 to to 80.24) a to 3.54)b to 0.73)b
1983) inconsiste s (50%) 95.92)a Low Not useful Moderately
ncy Moderate useful
White blood cell count ≥12,500 cells/mm3*
1 study Case series Serious 1 No No serious Serious Clinical 7/51 43% (6.20 82% (70.42 2.36 (0.82 to 0.70 (0.36 Low
(Romem and serious indirectness 5
chorioamnioniti to 79.52)a to 93.21) a 6.81) b Not to 1.35)b
Artal 1984) inconsiste s (14%) Low Moderate useful Not useful
ncy
White blood cell count ≥13,000 cells/mm3 measured within 72 hours of birth
Yoon 1996 Case series Serious2 No No serious No Histological 35/63 40% 82% (67.96 2.24 (0.92 to 0.73 (0.53 Moderate
serious indirectness serious chorioamnioniti (23.77 to to 96.35) a 5.47) b Not to 1.01) b
inconsiste imprecisi s (56%) 56.23) a Moderate useful Not useful
ncy on Low
White blood cell count ≥16,000 cells/mm3*
1 study Case series Serious 1 No No serious Serious Clinical 7/51 29% (0 to 95% (89.30 6.29 (1.05 to 0.75 (0.47 Low
(Romem and serious indirectness 5
chorioamnioniti 62.04) a to 100) a 37.66) b to 1.20) b
Artal 1984) inconsiste s (14%) Low High Moderately Not useful
ncy useful
White blood cell count >20, 000 cells/mm3 measured on admission or 24–48 hours prior to birth
1 study Case series Serious6 No Serious No Clinical 36/237 16.7% (0 97.5% (91 6.70 (2.16 to 0.85 (0.74 Low
(Garite and serious indirectness serious chorioamnioniti to 29.04)b to 100)b 21.0)b to 0.99)b
Freeman inconsiste 7
imprecisi s (15%) Low High Moderately Not useful
1982) ncy on useful
120
* Timing of measurement not reported/unclea

2. Selection bias – only women who delivered within 72 hours of amniocentesis were analysed and unclear whether consecutive women were included in the study
3. Evidence was downgraded by 2 due to very serious heterogeneity (chi-squared p<0.1, I-squared inconsistency statistic of >75%).
6. Possible selection bias - unclear from report how women were selected for inclusion in original trial
7. A proportion of women reported to be in labour (≤44/251)
Table 26: GRADE profile for predictive accuracy of fetal heart rate for identifying infection
Quality assessment Measures of diagnostic accuracy
Outcome and Positive Negative
Number of Risk of Inconsistenc Indirectnes Imprecisi prevalence Preval Sensitivit likelihood likelihood
studies Design bias y s on (type of infection) ence y Specificity ratio ratio Quality
Fetal heart rate abnormality – all definitions
Overall Case Serious Serious No Serious4 All infectious Range: Range: Range: Range: Very low to
summary of series indirectness morbidity (intra- 8% to 41.33% to 0.85 to 0.44 to moderate
findings from amniotic infection, 60.0% 100% infinity 1.00
5 studies neonatal sepsis and Low Low to high Not useful Not useful
presumed neonatal to very to
sepsis, neonatal useful moderately
pneumonia, clinical useful
chorioamnionitis,
intrauterine infection)
(7% to 63%)
Abnormal antenatal CTG – last test before birth
1 study Case No No serious No serious No serious Total infectious 23/69 39.1% 82.6% 2.25 (1.00 0.74 (0.52 High
(Lewis 1999) series serious inconsistency indirectness imprecisio morbidity (intra- (16.93 to (69.43 to to 5.06)c to 1.00)c
risk of n amniotic infection, 65.08)a 90.57)b Not useful Not useful
bias neonatal sepsis and Low Moderate
presumed neonatal
sepsis) (33%)
1 study (Del Case Serious1 No serious No serious Serious 4 Neonatal infection 5/68 60.0% 90.5% 6.30 (2.22 0.44 (0.15 Low
Valle 1992) series inconsistency indirectness (sepsis and (32.53 to (91.99 to to 18.0)c to 1.00)c
pneumonia) (7%) 84.13)c 100)c Moderately Moderately
Low High useful useful
121

Outcome and Positive Negative
Number of Risk of Inconsistenc Indirectnes Imprecisi prevalence Preval Sensitivit likelihood likelihood
studies Design bias y s on (type of infection) ence y Specificity ratio ratio Quality
1 study (Del Case Serious1 No serious No serious No serious Clinical 10/68 30.0%(1. 89.66% 2.90 (0.86 0.78 (0.52 Moderate
Valle 1992) series inconsistency indirectness imprecisio chorioamnionitis 60 to (81.82 to to 9.75) c to 1.00) c
n (15%) 58.40)c 97.49) c Not useful Not useful
Low Moderate
Non-reactive CTG*
1 study Case Serious1 No serious No serious No serious Intrauterine infection 14/89 50% 41.33% 0.85 (0.48 1.00 (0.67 Moderate
(Caroll series inconsistency indirectness imprecisio (positive fetal blood (23.81 to (30.19 to to 1.49) c to 1.00) c
1995) n culture) (18%) 76.18) c 52.48) c Not useful Not useful
Low Low
Fetal heart rate >160 bpm*
1 study Case Serious1 No serious No serious No serious Clinical 18/100 22% 97% (94.22 9.11 (1.80 0.79 (0.62 Moderate
(Ismail series inconsistency indirectness imprecisio chorioamnionitis (3.02 to to 100)a to 45.99)c to 1.00)c
1985) n (18%) 41.43)a High Moderately Not useful
Low useful
1 study Case Serious1 No serious No serious No serious Histological 63/100 8% (1.26 97% (92.07 2.94 (0.36 0.95 (0.86 Moderate
(Ismail series inconsistency indirectness imprecisio chorioamnionitis to 14.61)a to 100)a to 24.18)c to 1.00)c
1985) n (63%) Low High Not useful Not useful
Fetal heart rate >170 bpm on admission
1 study Case Serious No serious No serious No serious Clinical 36/237 13.9% (0 100% (100 NC/infinity 0.86 (0.75 Moderate
(Garite and series 1
inconsistency indirectness imprecisio chorioamnionitis to 21.04)c to 100)c Very useful to 0.98) Not
Freeman n (15%) Low High useful
1982)
bpm beats per minute, CTG cardiotocography, NC not calculated
* Timing of measurement not reported/unclear
c. Error in reported value, value recalculated and confidence intervals calculated by NCC
2. Possible selection bias - unclear from report how women were selected for inclusion in original trial
3. Evidence was downgraded by 1 due to serious heterogeneity (chi-squared p<0.1, I-squared inconsistency statistic of 50%–74.99%)
122
Table 27: GRADE profile for predictive accuracy of maternal temperature for identifying infection
Outcome and
Number of Inconsiste Imprecisi (type of Sensitivit likelihood likelihood
studies Design Risk of bias ncy Indirectness on infection) y Specificity ratio ratio Quality
1 study (Ismail Case series Serious 1
No serious No serious No Histological 17% (8.09 97% (92.07 6.46 (0.87 to 0.85 (0.75 to Moderate
1985) inconsisten indirectness serious chorioamnionitis to 26.83) to 100) 1.03) 0.96)
cy imprecisi (63%) Low High Moderately Not useful
on useful
123
7.4.1 Additional data

1.1.1.1. Serial C-reactive protein measurements
Four observational studies reported additional data for serial C-reactive protein (CRP)
measurements and the timing of CRP elevation in relation to infection. One observational
study (n=147) reported that a constant rise of 12 mg/100 ml/day or more was observed more
frequently in women with chorioamnionitis than in women without chorioamnionitis, both 2
days before giving birth and 12 hours before giving birth.
One observational study (n=54) found that, when measured serially, CRP levels rose before
white blood cell count (no further details are given). Another observational study (n=51) also
found that elevated CRP appeared to be the earliest sign of chorioamnionitis. The normal
within-day percentage coefficient of variation for CRP values obtained at a 6–8 hour interval
was 4.8±0.9% and the between days coefficient was 12.8±2.2% (values ranged from less
than 0.60 to 1.75 mg/100 ml). An increase of over 30% in the day-to-day coefficient of
variation was defined as abnormal based on data collected. A fourth observational study
(n=55) also concluded that CRP elevation preceded birth or clinical infection by “several
days”. The study reported that 12 of 13 women with consecutive estimations of above 2
mg/100 ml were found to have histological chorioamnionitis.
7.4.1.1 Additive effect of tests

One randomised trial (n=73) reported the predictive value for CRP level, maternal
temperature and white blood cell count, alone and taken together. Using a logistic regression
model to develop a receiver-operator characteristics curve, the area under the curve (AUC)
for CRP alone was calculated as 0.566, the AUC for CRP plus temperature at the onset of
labour was 0.696 and for CRP plus temperature plus white blood cell count AUC was 0.697,
indicating that each is 57%, 70% and 70% predictive of chorioamnionitis, respectively.

7.5.1.1 C-reactive protein
The majority of studies (ranging from 24 to 147 women) found that CRP at cut-offs ranging
0.7 mg/100 ml or above to greater than 5 mg/100 ml is not a useful predictor of either clinical
or histological chorioamnionitis (positive likelihood ratios judged to be not useful and mainly
low sensitivity). One observational study (n=51) found the positive likelihood ratio for CRP
greater than 3.5 mg/100 ml to be moderately useful and another study found the positive
likelihood ratio for CRP greater than 1.25 mg/100 ml to be very useful.
The negative likelihood ratios were either not useful or moderately useful. Negative likelihood
ratios were judged to be moderately useful at cut-offs greater than 1.2 mg/100 ml [(in 2 out of
3 studies (n=213)], greater than 1.25 mg/100 ml (1 study n=52), 2 mg/100 ml or more (in 4
out of 8 studies n=175) and greater than 4 mg/100 ml [(in 1 of 2 studies (n=147)]. Specificity
was found to be mainly low or moderate.
The evidence was generally of low quality.
7.5.1.2 White blood cell count

The evidence from 4 observational studies with over 400 women with P-PROM
predominately found that white blood cell count at cut-offs ranging from greater than 12,500
cells/mm3 to 13,500 cells/mm3 or above is not a useful predictor of either clinical or
124
histological chorioamnionitis (positive likelihood ratios judged to be not useful, moderate or

low sensitivity and specificity). One observational study (n=51) found the positive likelihood
ratio for white blood cell count greater than 16,000 cells/mm3 to be moderately useful. The
same observational study found the negative likelihood ratio for white blood cell count
greater than12,500 cells/mm3 to be moderately useful. One observational study (n=237)
found the positive likelihood ratio for white blood cell count greater than 20,000 cells/mm3 to
be moderately useful, and the negative likelihood ratio not useful, in predicting clinical
chorioamnionitis.
The evidence was of moderate to low quality.
7.5.1.3 Fetal heart rate

Two observational studies (total n=158) found that an abnormal CTG result is not a useful
predictor of neonatal infection (positive and negative likelihood ratios judged to be not useful,
and low to moderate sensitivity and specificity). One observational study (n=68) found a
moderately useful positive likelihood ratio and negative likelihood ratio for an abnormal CTG
result in predicting neonatal infection.
One observational study (n=68) found that an abnormal CTG result is not a useful predictor
of clinical chorioamnionitis (positive and negative likelihood ratios judged to be not useful,
and low to moderate sensitivity and specificity).
One observational study (n=100) found that a fetal heart rate greater than 160 bpm was not a
useful predictor of histological chorioamnionitis. The same study found that a fetal heart rate
greater than 160 bpm for predicting clinical chorioamnionitis had a moderately useful positive
likelihood ratio and a not useful negative likelihood ratio (fetal heart rate greater than 160
bpm had a low sensitivity and high specificity for predicting both clinical and histological
chorioamnionitis). One observational study (n=237) found that a fetal heart rate greater than
170 bpm was not a useful predictor of clinical chorioamnionitis (positive and negative
likelihood ratios not useful).
The evidence was generally of moderate quality.
7.5.1.4 Maternal temperature

Evidence from 1 observational study (n=100) found that a raised maternal temperature had a
moderately useful likelihood ratio for predicting histological chorioamnionitis but a not useful
negative likelihood ratio. The specificity of raised maternal temperature was found to be high
but the sensitivity was low for predicting histological chorioamnionitis. The evidence from this
study was of moderate quality.

A search was undertaken for health economic evidence on the diagnostic value of
temperature, pulse, white cell count, CRP and CTG to identify infection in women with P-
PROM. A total of 34 articles were identified by the search. After reviewing titles and
abstracts, 2 full papers were obtained but they were both excluded as they did not evaluate
the relevant investigations. This question was not prioritised for health economic analysis as
they were thought by the Guideline Committee to be low-cost investigations and with an
expectation that they would not be found to be particularly effective.
125

The committee considered all the properties of diagnostic accuracy measurements for
decision-making in this topic – sensitivity, specificity, and positive and negative likelihood
ratio – and the relative importance of having a high false positive and high false negative
result from identification of different types of infections for women with P-PROM. Because
ascending infection from mother to fetus potentially causes perinatal mortality or severe
neonatal morbidity, the committee would accept a test or series of tests that had some false
positives as long as the false negative rate was very low. Conversely, a high false positive
rate would result in an increase in hospital admissions and use of antibiotics.

The committee discussed the benefits of using different tests (CRP, white blood cell count,
fetal heart rate and maternal temperature) in isolation for identification of different types of
infection for women with P-PROM. The evidence of included studies did not show that any of
these tests were helpful in identification of different types of infections (clinical and historical
chorioamnionitis, funisitis and neonatal infection) for this group of women at risk of preterm
birth that would be relevant in clinical practice.
However, the negative likelihood ratios of these tests (for example CRP of 1.2, 1.5 or 2
mg/100 ml or less measured on admission) were found to be better in ruling out women with
P-PROM with no indications of infections. The committee considered that these tests may
provide useful reassurance against a diagnosis of infection when consistent with the clinical
picture, rather than helping to make a positive diagnosis. Different thresholds were used per
test in each study. In the majority of cases these thresholds were fairly low compared with
the ones used for decision-making in clinical practice and the single high value has not been
investigated in the evidence, so the committee considered these limitations in the
interpretation of results.
The committee discussed that a CRP threshold of 2 mg/100 ml and white blood cell count of
12,500 cells/mm3 would be the most common thresholds used in clinical practice, above
which a result would be considered to be abnormal. The evidence did not support these
thresholds as useful markers of infection. White blood cell counts are usually higher than
normal in pregnant women. Therefore, they concluded that when using these tests they
should be combined with clinical assessment to diagnose infections for women with P-
PROM.

A 2014 British Medical Journal (BMJ) report cited a cost of £1.03 for CRP, although this may
reflect just the laboratory costs and not the costs of obtaining the sample: the Guideline
Committee suggested a cost of around £5 per test. The other investigations are also
relatively inexpensive, requiring a very small amount of healthcare professional time.

The majority of evidence was moderate to low quality as some of the included studies were
small and there was serious risk of bias. Measurements of sensitivity and specificity requires
a clinically relevant threshold to be defined, but the evidence was presented based on the
selected thresholds by the authors that were usually lower than the one used in clinical
practice. There was a high variability between the included studies in the selection of
population, definition of diagnostic tools and the measurements reported, but this is not
126
unusual for this type of study (diagnostic). The committee concluded that the generalisation
of results which reported timing of outcomes measured since birth was less useful because it
meant they did not reflect the real clinical scenarios. The age of included studies and the lack
of representativeness of entire population of women with P-PROM were limitations in the
study design of evidence included in this section. The common use of antibiotics across the
included studies was also a factor that may have distorted the direction of results identified.
The committee also discussed the limitations of applicability of evidence due to the way
outcomes were reported as a single cut-off point instead of change infection rates across
time.

Sepsis was the most common direct cause of maternal death in the most recent Confidential
Enquiry on Maternal and Child Health. Bacterial infection with Group A streptococcal
infection or colifors was the most common cause associated with increased odds of
progression from severe sepsis to septic shock in the MBBRACE 2014 report. These
bacteria can spread very rapidly to the fetus and can cause severe disease in both mother
and fetus or baby which may not be amenable to antibiotic therapy. It may not be clinically
appropriate to observe and wait for test results and urgently needed antibiotics, and delivery
may be the more appropriate action.
These recommendations are based on both the clinical interpretation of evidence and on the
Guideline Committee members’ clinical expert opinions.

No consistent findings were found for the different tests for identifying different types of
infections for women with P-PROM.
7.8 Recommendations
updated guideline.
2. What is the diagnostic accuracy of serial C-reactive protein

Research question testing to identify chorioamnionitis in women with P-PROM?
Why this is needed
Importance to ‘patients’ or Identifying infection in women with P-PROM is needed to provide
the population best practice care. Early diagnosis of infection allows consideration of
therapeutic strategies (including antibiotics and/or early birth).
Effective treatment of infection is particularly important given that
sepsis is a common direct cause of maternal death. There is currently
limited evidence that serial C-reactive protein testing might be useful,
but the Committee is aware that this strategy is in common practice.
Evidence from diagnostic studies is needed about the accuracy of
serial C-reactive protein testing for identifying chorioamnionitis, which
is one of the most common and serious infective complications of P-
PROM.
127
Relevance to NICE Medium; the research would inform future updates of the guideline
guidance and addresses a commonly-encountered clinical scenario of clinical
importance.
Relevance to the NHS The Committee was aware that serial measurement of C-reactive
protein is in common practice for monitoring women with P-PROM.
Whilst not an expensive test, if shown to be unhelpful there would be
cost savings.
prematurely
Current evidence base Although limited evidence showed that serial C-reactive protein
testing might be useful, the Committee was aware that this strategy is
in common practice. Evidence is needed on its effectiveness in
identifying chorioamnionitis, one of the most common and serious
infective complications of P-PROM.
Equality The population is defined by gestational age.
Feasibility The research is feasible and the intervention is low-cost. There are
no ethical issues other than those usually pertaining to perinatal
research.
Other comments None
128
‘Rescue’ cervical cerclage
8 ‘Rescue’ cervical cerclage

8.1 Introduction
Cervical cerclage, also known as a cervical stitch, is a treatment used to prevent the cervix
opening too early and causing either a late miscarriage or preterm birth. Cerclage may be
performed as a prophylactic measure where the woman has a history that increases the risk
of spontaneous second-trimester loss or preterm delivery and/or cervical shortening seen on
ultrasound; this type of cerclage is considered in Chapter 4. It can also be performed as a
salvage measure (‘rescue’ or non-prophylactic cerclage) when a woman presents with
premature cervical dilation, often with exposed fetal membranes, and in some cases where
the membranes have prolapsed into the vagina.
‘Rescue’ cerclage is not a common procedure and is most often performed in the middle
trimester around the time of viability. Preventing or delaying preterm birth at this gestation
might have significant benefit in terms of reducing mid-trimester loss and avoiding the
consequences of extreme prematurity. However, the procedure carries risks and there is
uncertainty about which women are most likely to benefit. This chapter considers only
‘rescue’ or non-prophylactic cervical cerclage.
8.2 Review question

What is the clinical effectiveness of non-prophylactic ‘rescue’ cervical cerclage in preventing
preterm birth for women in suspected preterm labour?

Six studies are included in the review of the role of rescue cervical cerclage (Althuisius 2003,
Aoki 2013, Curti 2012, Daskalakis 2006, Olatunbosun 1995, Stupin 2008). As only 1
randomised controlled trial (RCT) was identified to match the protocol (Althuisius 2003),
comparative cohort studies were also considered; either prospective (Daskalakis 2006,
Olatunbosun 1995) or retrospective (Aoki 2013, Curti 2012, Stupin 2008).
All women included in the studies were at risk of preterm birth. Women who did not have
rescue cervical cerclage were confined to bed rest. The additional use of supportive
treatment with tocolysis, antibiotics, corticosteroids and low molecular weight heparins varied
between the studies.
The lowest gestational age of women included in the studies reviewed was 16+0 weeks while
the highest was 27+6 weeks. Mean gestation (standard deviation [SD]) at the time of
emergency cerclage was reported only in 3 out of the 6 included studies and was around
22 weeks (22.4 [1.7] weeks, 22.4 [2.1] weeks and 22.2 [3.3] weeks).

Data is presented in the following GRADE profile:
• •Table 32: GRADE profile for comparison of ‘rescue’ cervical cerclage versus no cerclage
For quantitative outcomes, such as the pregnancy prolongation data that were presented by
either parametric (mean, SD) or non-parametric (median, range) measures, the GRADE
129
profile includes information only on results from studies with parametric measures that could
be used to calculate the absolute effects.
A full description of the characteristics and results of the included studies can be found in the
130
Table 28: GRADE profile for predictive accuracy of maternal temperature for identifying infection
Other
Number of considera ‘Rescue’ No Relative Absolute
studies Design Risk of bias Inconsistency Indirectness Imprecision tions cerclage cerclage (95% CI) (95% CI) Quality
Perinatal death (including any intrauterine death and neonatal death up to 7 days postpartum)
1 study Cohort study No serious risk No serious Serious 1 Serious 2
None 5/89 13/72 RR 0.31 125 fewer per Very
(Stupin 2008) of bias inconsistency (5.6%) (18.1%) (0.12 to 1000 low
0.83) (from 31 fewer
to 159 fewer)
Neonatal survival
1 study Randomised No serious risk No serious Serious 3 Serious 2
None 9/16 4/14 RR 1.97 277 more per Low
(Althuisius trial of bias inconsistency (56.3%) (28.6%) (0.77 to 1000
2003) 5.01) (from 66 fewer
to 1000 more)
1 study Cohort study Serious4 No serious No serious Serious 2
None 30/37 8/15 RR 1.52 277 more per Very
(Curti 2012) inconsistency indirectness (81.1%) (53.3%) (0.92 to 1000 low
2.5) (from 43 fewer
to 800 more)
1 study Cohort study Serious 5
No serious No serious Serious 2
(Olatunbosun inconsistency indirectness (77.3%) (60%) (0.80 to 1000 low
1995) 2.06) (from 120 fewer
to 636 more)
No serious No serious Serious 2
(Daskalakis inconsistency indirectness (96%) (57.1%) (0.88 to 1000 low
2006) 3.21) (from 69 fewer
to 1000 more)
Serious neonatal morbidity (defined as admission to neonatal intensive care unit and/or neonatal deaths)
None 10/16 14/14 RR 0.64 360 fewer per Low
(Althuisius trial of bias inconsistency (62.5%) (100%) (0.43 to 1000
2003) 0.94) (from 60 fewer
to 570 fewer)
Interval between study entry and delivery days) (better indicated by higher value)
1 study Randomised No serious risk No serious Serious 3 Very serious6 None 54 (SD 47) 20 (SD 28) - MD 34 higher Very
(Althuisius trial of bias inconsistency (3.11 higher to low
2003) 64.89 higher)
No serious No serious No serious None 62 (SD27) 22 (SD18) NC MD 40 higher Low
(Daskalakis inconsistency indirectness imprecision (26.97 higher to
2006) 53.03 higher)
131

Other
Number of considera ‘Rescue’ No Relative Absolute
studies Design Risk of bias Inconsistency Indirectness Imprecision tions cerclage cerclage (95% CI) (95% CI) Quality
Preterm birth (not defined)
1 study Cohort studies Very serious7 No serious No serious No serious None 12/15 20/20 RR 0.8 200 fewer per Very
(Aoki 2013) inconsistency indirectness imprecision (80%) (100%) (0.61 to 1000 (from 390 low
1.05) fewer to 50
more)
Preterm birth between 22+0 and 27+6 weeks
1 study Cohort studies Very serious7 No serious No serious No serious None 3/15 16/20 RR 0.25 600 fewer per Very
(Aoki 2013) inconsistency indirectness imprecision (20%) (80%) (0.09 to 1000 (from 240 low
0.7) fewer to 728
fewer)
Preterm birth before 34 weeks
None 7/13 10/10 RR 0.56 440 fewer per Low
(Althuisius trial of bias inconsistency (53.8%) (100%) (0.34 to 1000 (from 70
2003) 0.93) fewer to 660
fewer)
Preterm birth before 32 weeks
No serious No serious No serious None 9/29 16/17 RR 0.33 631 fewer per Low
(Daskalakis inconsistency indirectness imprecision (31.0%) (94.1%) (0.19 to 1000 (from 405
2006) 0.57) fewer to 762
fewer)
Maternal side effects (include cervical laceration and cervical dystocia due to scar tissue preventing cervical dilation)
No serious No serious Very serious6 None 4/29 0/17 RR 5.4 NC Very
(Daskalakis inconsistency indirectness (8.2%) (0%) (0.31 to low
2006) 94.55)
CI confidence interval, MD mean difference, MID minimally important difference, NC not calculable, RR relative risk
1. Majority of evidence has only 1 indirect aspect of population: 19% of women had a multiple pregnancy (20% in cerclage group, 18% in no cerclage group)
2. Evidence was downgraded by 1 due to serious imprecision as 95% CI crossed one default MID
3. Majority of evidence has only 1 indirect aspect of population: 30% of women had a multiple pregnancy (23% in intervention group, 40% in control group)
4. Study states women were allocated to treatment but it was not clear how this allocation was made, Unclear care protocol for women in the no cerclage group
5. Likely variation in additional treatments (e.g. tocolysis, antibiotics, corticosteroids) between groups
Evidence was downgraded by 2 due to very serious imprecision as 95% confidence interval crossed 2 default MIDs 7. High selection, performance, attrition and detectionbias
132

Low and very low quality evidence from 1 small RCT (n=30) and 3 cohort studies (n=32 to
52) found no significant difference for the outcome of neonatal survival in babies whose
mothers had rescue cerclage compared with those who did not have cerclage.
Low and very low quality evidence from 3 individual cohorts and 1 RCT (n=23 to 161)
showed significantly fewer perinatal deaths and preterm births before 22+0 to 27+6, 32 and
34 weeks in women who had rescue cerclage compared with women who did not. Very low
quality evidence from the small RCT and 1 cohort study showed that the interval between the
study entry and birth was longer in women who had rescue cerclage compared with women
who did not. In addition, low quality evidence from the same RCT showed significantly lower
risk of serious neonatal morbidity (defined as admission to neonatal intensive care unit
and/or neonatal deaths) in babies whose mothers had cerclage compared with women who
did not have cerclage.
The outcomes of neonatal survival and maternal side effects were not found significantly
different between the 2 groups (rescue cerclage and no cerclage) based on low to very low
quality evidence from 1 small RCT and individual cohort studies.

A single search was undertaken for health economic evidence on prophylactic cervical
cerclage to prevent preterm labour in women considered to be at risk of preterm labour and
birth and rescue cervical cerclage in preventing preterm birth in women in suspected preterm
labour. A total of 60 articles were identified by the search. After reviewing titles and abstracts,
3 papers were obtained. These studies were all excluded because they were not economic
evaluations or were published conference abstracts. Therefore, no relevant economic
evidence was identified for this question.
This question was not identified as a priority for health economic analysis as the Guideline
Committee reflected that the intervention would only be relevant for a very small proportion of
the patient population.

In terms of neonatal outcomes, the Guideline Committee considered neonatal mortality as
critical for this review, concluding that any mortality up to 1 year could be reported as a single
outcome. The committee included early neonatal survival because only the first week was
available for analysis.
Given that the purpose of this intervention is to delay birth, preterm birth and the interval
between the procedure and delivery were both prioritised as important outcomes. The
committee included neonatal sepsis, chronic lung disease and bronchopulmonary dysplasia
as potential associated adverse events. The committee also considered long-term infant
neurodevelopmental outcomes, such as neurodevelopmental disability, because these are
common adverse events associated with preterm birth and any reduction in these outcomes
would be a significant indicator of the effectiveness of the intervention. However, given the
absence of available data, the committee agreed that serious neonatal morbidity could be
considered as a surrogate for the neonatal adverse events outcomes selected originally.
133
In terms of maternal outcomes, maternal mortality and maternal adverse effects were
prioritised for this review question, including infection requiring intervention and cervical
trauma requiring repair, because rescue cerclage is a difficult procedure and there is a
potential risk of such events occurring. In addition, the committee felt that evidence regarding
maternal emotional and psychological impact should be also assessed, due to the invasive
nature of the procedure and considering the stressful circumstances under which it might be
conducted (when urgent treatment is required).

The committee recognised that both the randomised and observational evidence supported
rescue cerclage for the outcomes of reducing serious neonatal morbidity, reducing preterm
birth below 27, 32 and 34 weeks and increasing the interval between intervention and
delivery. The committee had serious concerns regarding the lack of data for the outcome of
neonatal sepsis as, based on their clinical experience, this was likely to be a significant
complication associated with the procedure. They concluded that information about the risks
of rescue cerclage, as well as its potential aims and benefits, should be communicated to the
woman and their family members or carers as appropriate.
The committee members’ clinical opinion was that any benefit of performing rescue cerclage
beyond 32 gestational weeks would be limited and may not outweigh the potential harms.
They agreed that the recommendation to perform rescue cerclage for women with a dilated
cervix and exposed, unruptured membranes should reflect both the quality of reviewed
evidence (low to very low) and the gestational age of women included in these studies
(between 16+0 and 27+6 weeks).
They were also aware that rescue cerclage can be a technically difficult procedure to
perform, requiring specialist skills and expertise to mitigate the risks of maternal or neonatal
adverse events.
For these reasons, the committee recommended that the decision for rescue cerclage must
be made only after discussion with a consultant obstetrician. The decision should take
account of the woman’s gestation and her own stated wishes after a full discussion. The
committee members decided that, in their clinical opinion, rescue cerclage would cause harm
to women with signs of infection, active vaginal bleeding or uterine contractions and so they
decided upon a strong recommendation of not offering rescue cerclage to these groups of
women.

The committee felt that rescue cerclage was likely to be an expensive intervention due to the
setting in which it is delivered and the expertise required of the healthcare professionals
providing the care.
They also felt that although the evidence shows a reduction in preterm birth, perinatal death
and neonatal morbidity, it was of low quality and the chance of poor outcomes is quite high
despite emergency cerclage. However, they acknowledged that the management of preterm
birth and the associated outcomes is extremely costly – both financially and in terms of
parental anxiety – and therefore if the intervention delayed birth beyond key gestational
milestones, then the initial costs incurred would be likely to be offset by large cost savings
downstream.
Furthermore, the committee felt that the overall cost impact for health services would be
small because rescue cerclage would only be an appropriate intervention for a small
proportion of the patient population.
134

The majority of the evidence included was of low to very low quality. Only a small RCT of 30
women was included in the evidence base whereas the rest of the evidence came from
cohort studies. Indirectness and imprecision were the main areas affected in the quality
assessment of included studies. More specifically, some of the cohorts included were mixed
populations in terms of use of additional tocolytics (or other adjunctive treatment) and the
inclusion of twin pregnancies. Potential bias in the single RCT arose from the imbalanced
allocation of women with multiple pregnancies to treatment groups, which was of particular
concern for the committee. The sample size of the included studies was relatively small, thus
restricting the confidence in the estimates of effects.

The recommendations were based on both the interpretation of clinical evidence reviewed
and on the committee members’ expert opinions.
The committee noted that this review did not set out to compare the different types of stitches
that could be used and the conclusions of the findings may be limited under these
restrictions. They were also aware that there is an existing RCOG guideline on cervical
cerclage that provides more detail on how to perform this intervention.
The committee also noted that it would be important to take into account the effectiveness of
other possible interventions (such as magnesium sulfate for neuroprotection) when
considering whether to offer rescue cerclage at lower gestational ages.

In light of all their reservations about the evidence on the effectiveness of rescue cerclage,
the committee members decided that a decision should be made with caution regarding its
application.
For these reasons, the committee felt that women should be clearly informed about the
potential risks and benefits of the procedure. It was noted that a specific definition of preterm
labour had been used in some of the included studies (dilated cervix with exposed fetal
membranes) and that this should be reflected in the recommendations in terms of indications
and contraindications for the use of rescue cerclage. They also noted that the gestation at
which rescue cerclage was undertaken was an important consideration, as were the skills
and experience of the practitioner performing the procedure. The importance of the woman
and her partner having confidence and trust in the obstetrician was also highlighted. The
committee concluded that whether or not to insert a rescue cervical suture was a complex
judgement that should be undertaken on an individualised basis with full involvement of the
woman.
The committee also took the view that there should be some obligation to collect data about
outcomes of its use in a national registry.
8.8 Recommendations
updated guideline.
135

3. What is the clinical effectiveness of ‘rescue’ cerclage in
Research question improving outcomes for women at risk of preterm birth?
Why this is needed
Importance to ‘patients’ There is some evidence from randomised studies that ‘rescue’ cerclage
or the population might be effective in improving neonatal outcomes in women with a
dilated cervix and exposed, unruptured fetal membranes. However, there
is uncertainty about the magnitude of this effect. The full consequences
of this strategy and the subgroups of women at risk of preterm labour
who might particularly benefit are not known. A randomised controlled
trial would best address this question, but a national registry of the most
critical outcomes (neonatal mortality and morbidity, maternal morbidity)
could also be considered for women who did not want to participate in a
randomised trial but who opted for ‘rescue’ cerclage.
Relevance to NICE The importance is high because rescue cerclage is widely used yet its
guidance evidence base is of relatively low quality, and it is likely that more high
quality evidence would refine or change current recommendations.
Relevance to the NHS If rescue cerclage is effective in delaying delivery, and this benefit is
found to outweigh any harms, then it will be important to define the
groups of women who may benefit from this treatment. If effective it has
potential to reduce morbidity and mortality, healthcare resources and
costs expended on the care of very preterm babies.
National priorities NHS Outcomes Framework 2014-5, #1: Preventing people from dying
prematurely
Current evidence base The Committee recognised that the current evidence base was of low
quality and many important questions remain unanswered. There is
some evidence that ‘rescue’ cerclage might be effective in improving
outcomes in women with a dilated cervix and exposed, unruptured fetal
membranes. However, there was uncertainty in the magnitude of this
effect and the full consequences and the subgroups of women at risk of
preterm labour who might particularly benefit are unknown.
Equality No
Feasibility There is no reason in principle why further trials should not be carried out
to address current uncertainties. The Committee felt that a randomised
trial would provide the best evidence, but that a registry collection of
outcomes (neonatal mortality and morbidity, maternal morbidity) could be
considered for women who did not want to participate in a randomised
trial but who opted for rescue cerclage.
The ethical issues are not in principle different from those affecting other
perinatal trials.
Other comments Trials would have to be carried out in centres with neonatal facilities
equipped to care for very preterm babies.
136
Diagnosing preterm labour in women with intact membranes
9 Diagnosing preterm labour in women

with intact membranes
9.1 Introduction
It can often be unclear whether symptoms of preterm labour will result in progression to
established labour and birth. Symptoms described by women, such as painful contractions,
may be associated with labour, but others could be non-specific, for example low back pain
or abdominal pain. Sometimes symptoms can occur but then settle, allowing the pregnancy
to continue towards term. Investigations performed clinically (such as digital vaginal
examination of the cervix), biochemically or using ultrasound may help to distinguish women
in preterm labour from those who are not. Women in whom preterm labour is correctly
identified may then benefit from clinical management to try to delay birth and/or improve
neonatal outcomes, whereas those who are not in preterm labour can be reassured and
further intervention is not necessary.
9.2 Review question

What is the diagnostic accuracy of the following (alone or in combination) in women with
intact membranes to identify preterm labour leading to preterm birth:
• clinical assessment (such as symptoms expressed by women, strength and frequency of

contractions, findings on vaginal examination)
• biochemical testing for markers for preterm labour, namely cervicovaginal fetal fibronectin
and IGF-BP1 insulin-like growth factor binding protein-1
• cervical ultrasound features (such as cervical length and funnelling)?

A total of 38 prospective cohort studies were included in this review (Azlin 2010, Bagga
2010, Bartnicki 1996, Benattar 1997, Botsis 2006, Brik 2010, Burwick 2011, Danti 2011,
Demirci 2009, Diaz 2009, Eroglu 2004, Giles 2000, Gomez 2005, Gramellini 2007, Holst
2006, Iams 1995, Kwek 2004, LaShay 2000, Lembet 2002, Lukes 1997, Malak 1996,
McKenna 1999, Palacio 2007, Sakai 2003 Schmitz 2006, Schmitz 2008, Schreyer 1989,
Senden 1996, Skoll 2006, Sotiriadis 2010, Swamy 2005, Tanir 2008, Tanir 2009, Tekesin
2005, Ting 2007, Tsoi 2006, Tsoi 2005, Van Baaren 2014).
All studies included women with signs and symptoms of preterm labour who had singleton
pregnancies and intact membranes, apart from 1 study (McKenna 1999) that did not specify
details for women with multiple pregnancies, 1 study (Benattar 1997) in which 13% of women
had a multiple pregnancy and 1 study (Lembet 2002) where reporting of multiple pregnancies
is unclear and that may have included some women with preterm premature rupture of
membranes (P-PROM).
Originally the Guideline Committee decided to consider only studies with women who had
not received tocolytics as part of their management plan to delay preterm labour. There were
no studies where tocolytics were not used, so the committee decided to expand the inclusion
criteria and include studies where all women had had tocolytics to preserve homogeneity of
interpretation of results. However, due to the limited number of such studies (8 studies:
Benattar 1997, Kwek 2004, Palacio 2007, Senden 1996, Sotiriadis 2010, Swamy 2005,
Tekesin 2005, Ting 2007), it was decided to include studies with a mixed population (that is,
those who had and who had not received tocolytics) and downgrade the quality of this
evidence as indirect to the population of interest.
137
A Health Technology Assessment (HTA) systematic review (SR) was published in 2009
examining different screening techniques to diagnose preterm birth. This review included
randomised controlled trials (RCTs), and prospective and retrospective cohort studies.
Diagnostic accuracy results were reported separately for asymptomatic and symptomatic
women for some of the tests that were also the focus of this review question (digital vaginal
examination, phosphorylated insulin-like growth factor binding protein-1 (pIGFBP-1), fetal
fibronectin and transvaginal ultrasound). Outcomes included preterm birth and time to
delivery endpoints. This HTA review was not included in this evidence review as only
prospective cohort studies were considered for inclusion in the protocol. In addition,
outcomes in the HTA were assessed at 7 to 10 days following presentation rather than within
the 7 days specified in the review protocol of this question. However, the individual studies in
the HTA publication were assessed for relevance to this protocol.
An update of the fetal fibronectin section of the HTA review has since been published (HTA
2013) but was excluded from this review, again due to differences in outcomes. This more
recent HTA report also included retrospective studies and populations with multiple
pregnancies and did not report the outcomes of interest, namely birth within 48 hours or 7
days of presentation.
9.3.1 Diagnosis using clinical examination

The presentation of evidence is presented by diagnostic test. The reference standard was
considered to be progression to labour, with labour defined as progressive dilation of the
cervix over the few hours/days after presentation, leading to birth.
9.3.1.1 Bishop score

There were 3 studies (Schmitz 2008, Schreyer 1989, Senden 1996) included, all of which
looked at the accuracy of this diagnostic test. The studies examined different thresholds of
Bishop scores to diagnose birth within 48 hours and 7 days.
The population across all studies were symptomatic women who presented between 24 and
36 gestational weeks.
9.3.2 Diagnosis using biochemical tests

1.1.1.2. Phosphorylated insulin-like growth factor binding protein-1 (pIGFBP-1)
In total 7 studies were included for this test (Azlin 2010, Brik 2010, Danti 2011, Eroglu 2004,
Kwek 2004, Lembet 2002, Tanir 2009), all of which examined pIGFBP-1 to predict preterm
birth within 7 days.
Three of these studies (Brik 2010, Kwek 2004, Lembet 2002) examined pIGFBP-1 to predict
preterm birth within 48 hours
A subgroup analysis of women who had cervical length estimates of 30 mm or less was
given in 2 studies:
• Danti 2011 – cervical length: less than 20 mm; 20–30 mm; 30 mm or less)
• Azlin 2010 – cervical length less than 25 mm.
Overall, 5 studies included symptomatic women who presented during 22–35 gestational
weeks, although 1 study (Lembet 2002) included women over 20 gestational weeks and 3
studies included women up to 36+6 weeks (Azlin 2010, Lembet 2002, Tanir 2009).
138
Four studies were conducted in resource-rich countries: 1 in Malaysia (Azlin 2010), 1 in

Spain (Brik 2010) and 2 in Singapore (Kwek 2004, Ting 2007). Three were conducted in
Turkey (Eroglu 2004, Lembet 2002, Tanir 2009).
9.3.3 Diagnosis using transvaginal ultrasonography to measure cervical

length
In total 13 studies are included in this section (Azlin 2010, Bagga 2010, Botsis, Danti 2011,
Demirci 2009, Eroglu 2004, Gomez 2005, Gramellini 2007, Holst 2006, Palacio 2007,
Schmitz 2008, Sotiriadis 2010, Tsoi 2005).
Four studies (Bagga 2010, Gomez 2005, Schmitz 2008, Tsoi 2005) examined different
thresholds of cervical length measured by transvaginal ultrasound to diagnose preterm birth
within 48 hours.
All 13 studies examined different thresholds of cervical length measured by transvaginal
ultrasound to diagnose preterm birth within 7 days.
Three studies also investigated the diagnostic accuracy of cervical length measurement
using transvaginal ultrasound in specific subgroups. One study reported the use of different
cervical length thresholds to diagnose birth within 7 days in women according to gestational
age above or below 32 weeks (Palacio 2007). A second study also examined the use of
change in cervical length 24 hours after admission as a diagnostic tool to predict preterm
birth within 7 days (Sotiraides 2010). A third study described different thresholds of cervical
length determined by transvaginal ultrasound to diagnose preterm birth within 48 hours and 7
days in women with a Bishop score between 4 and 7 (Schmitz 2008).
Overall most studies included symptomatic women who presented during 24–36 gestational
weeks, although 3 studies included women with lower gestational ages (Gramellini [2007]
from 20 weeks, Gomez [2005] from 22 weeks and Sotiriadis [2010] from 23 weeks) and 1
study included women up to 37 weeks (Bagga 2010).
Most studies for diagnosis using transvaginal ultrasonography to measure cervical length
were conducted in resource-rich countries.

The search strategies for this chapter can be found in Appendix E, the excluded studies
inAppendix G, the evidence tables in Appendix H and the forest plots in Appendix I.
Data is reported in the following GRADE profiles:
• Single tests:
o Table 33: GRADE profile for evaluation of a Bishop score to diagnose preterm
birth within 48 hours or within 7 days
o Table 34: GRADE profile for evaluation of pIGFBP-1 to diagnose preterm birth
within 48 hours or within 7 days
o Table 35: GRADE profile for evaluation of fetal fibronectin to diagnose preterm
birth within 48 hours or within 7 days
o Table 36: GRADE profile for evaluation of fetal fibronectin >50 nanogram/ml
before and after cervical examination to diagnose preterm birth within 7 days
o Table 37: GRADE profile for evaluation of cervical length measured using
transvaginal ultrasound to diagnose preterm birth within 48 hours or 7 days
• Combinations of tests:
o Table 38: GRADE profile for evaluation of Bishop score plus cervical length
139
measured using transvaginal to diagnose preterm birth within 48 hours or 7 days

o Table 39: GRADE profile for evaluation of a selective test based on cervical
length measured using transvaginal ultrasound plus Bishop score to diagnose
birthwithin 48 hours or 7 days
o Table 40: GRADE profile for evaluation of fetal fibronectin greater than 50
nanograms/ml plus Bishop score to diagnose preterm birth within 7 days
o Table 41: GRADE profile for evaluation of pIGFBP-1 test and cervical length
measured using transvaginal ultrasound to diagnose preterm birth within 7 days
o Table 42: GRADE profile for evaluation of fetal fibronectin greater than 50
nanograms/ml plus cervical length to diagnose preterm birth within 48 hours or 7
days
Full description of the characteristics and results of the included studies can be found in
theevidence tables in Appendix H.
Further information regarding the proportion of women with previous preterm births is
givenin the GRADE profiles.
140
Table 29: GRADE profile for evaluation of a Bishop score to diagnose preterm birth within 48 hours or within 7 days
Other
considerations
Number (percentage with Positive Negative
of Risk of previous preterm Sample likelihood likelihood
studies Design bias Inconsistency Indirectness Imprecision birth) size Sensitivity Specificity ratio ratio Quality
Bishop score to diagnose birth within 48 hours
Bishop score of 4 to 6a
1 Prospective No No serious Very Serious4 NR 70 69.2% 73.7% 2.63 (1.27 0.42 (0.14 Very
(Schreyer cohort serious serious1,2 (41.1 to (67.3 to to 4.09) to 0.87) low
1989) 89.0) 78.2) Not useful Moderately
useful
Bishop score ≥4b
1 Prospective No No serious Very Very serious5 NR 395 94.0% 43.0% 1.66 (1.20 0.14 (0.01 Very
(Schmitz cohort serious serious2,3 (71.0 to (38.0 to to 1.76) to 0.72) low
2008) risk of 100.0) 48.0) Not useful Moderately
bias useful
Bishop score ≥8b
2008) 62.0) 98.0) Very useful Not useful
Bishop score to diagnose birth within 7 days
Bishop score >2c
1 (Senden Prospective No No serious Serious2,6 Very serious5 NR 25 100% 73% 3.10 (0.86 0.17 (0.000 Very
1996) cohort serious to 3.83) to 1.09) low
Not useful Moderately
useful
Bishop score of 4 to 6a
(Schreyer cohort serious serious1,2 (44.6 to (68.8 to to 4.64) to 0.81) low
useful
Bishop score ≥4b
2008) 100.0) 50.0) Not useful Very useful
Bishop score ≥8b

Other
considerations
1 Prospective No No serious Very No serious NR 395 34.0% 98.0% 17.83 (6.87 0.67 (0.55 Low
(Schmitz cohort serious serious1,2 imprecision (19.0 to (96.0 to to 47.57) to 0.81)
NR not reported
a. The Bishop score comprises scoring for cervical length, consistency, position, dilation and station of head.
b. Bishop score was not defined.
c .Bishop score defined according to Myerschough P.R.;”Induction of labour” Chap 20 in Munro Kerr’s Operative Obstetrics 10th edn. 1982, pub, Bailliere Tindall
1. Tocolysis may have been used in some women: authors state that women received no medication aside from prenatal vitamins and iron however women who werereadmitted in
actual labour after discharge at 48 hours are stated as having received tocolytic medication.
2. Women included in the study were a mixed population where some received tocoloysis and some did not
3. The proportion of women who received tocolysis was not reported.
4. Evidence was downgraded by 1 due to 95% confidence interval for the negative likelihood ratio ranges from not useful (>0.5) to moderately useful (>0.1-0.5)
5. Evidence was downgraded by 2 due to 95% confidence interval for the negative likelihood ratio ranges from not useful (>0.5) to very useful (0-0.1)
6. 7/25 (28%) women received ritodrine, 8/25 (32%) women received antibiotic therapy and 19/25 (76%) received corticosteroids. Treatment was according to the estapractice of
administration when considered appropriate
Table 30: GRADE profile for evaluation of pIGFBP-1 to diagnose preterm birth within 48 hours or within 7 days
Other
considerations
pIGFBP-1 test to diagnose birth within 48 hours
1 (Brik Prospective Serious1 No serious Very Serious5 9.4% 276 73.7% 64.9% 2.1 (1.52 to 0.41 (0.19 Very
2010) cohort inconsistency2 serious3,4 2.91) to 0.87) low
useful
1 (Kwek Prospective Serious6 No serious Very serious7 Very serious8 NR 42 66.7% 61.1% 1.71 (0.56 0.54 (0.09 Very
2004) cohort inconsistency2 (25.5 to (54.2 to to 2.73)a to 1.37) a low
93.8)a 65.6)a Not useful Not useful

Other
considerations
1 Prospective No No serious Very Serious12 16% 36 93.3% 81.0% 4.90 (2.12 0.08 (0.004 Very
(Lembet cohort serious inconsistency2 serious3,10,11 (72.3 to (65.9 to to 6.85) a to 0.42) a low
2002) risk of 99.6) a 85.5) a Not useful Very useful
bias9
1 (Ting Prospective No No serious Very serious7 No serious NR 94 100% 74% 3.85 NC Low
2007) cohort serious inconsistency2 imprecision Not useful
risk of
bias9
pIGFBP-1 test to diagnose birth within 7 days
1 (Azlin Prospective Serious7 No serious Serious3, 14 Very serious NR 51 80.0% 93.5% 12.27 (2.83 0.21 (0.01 Very
2010) cohort inconsistency1 15
(32.9 to (88.4 to to 22.16) to 0.76) low
98.9) 95.5) Very useful Moderately
useful
1 (Brik Prospective Serious1 No serious Very Serious5 NR 276 73.1% 66.2% 2.16 (1.60 0.41 (0.21 Very
2010) cohort inconsistency2 serious3,4 to 2.92) to 0.78) low
useful
1 (Eroglu Prospective No No serious Serious3,16 Very 3.9% 51 83.3% 84.4% 5.38 (1.83 0.20 (0.01 Very
2007) cohort serious inconsistency2 serious17 (39.2 to (78.6 to to 7.37) a to 0.77) a low
risk of 99.1) 86.5) a Moderately Moderately
bias9 useful useful
1 (Kwek Prospective Serious7 No serious Very serious8 Very NR 42 83.3% 73.3% 3.12 (1.48 0.23 (0.04 Very
2004) cohort inconsistency2 serious18 (55.6 to (62.2 to to 4.56) a to 0.71) a low
96.9) a 78.8) a Not useful Moderately
useful
1 Prospective No No serious Very Serious12 16% 36 93.8% 85.0% 6.25 (2.43 0.07 (0.004 Very
(Lembet cohort serious inconsistency2 serious3,,10,11 (74.3 to (69.4 to to 9.71) a to 0.37) a low
2002) risk of 99.7) a 87.9) a Moderately Very useful
bias9 useful
1 (Tanir Prospective No No serious Serious3,19 Serious12 NR 68 93.3% 79.2% 4.50 (2.53 0.08 (0.004 Very
2009) cohort serious inconsistency2 (69.6 to (72.5 to to 5.25) a to 0.42 a low
risk of 99.6) a 81.0) a Not useful Very useful
bias9
1 (Ting Prospective No No serious Very serious7 No serious NR 94 69% 78% 3.13a 0.40 a Very
2007) cohort serious inconsistency2 imprecision low
risk of
NC not calculable, NR not reported, pIGFBP-1 phosphorylated insulin-like growth factor binding protein-1
a. Calculated by the NCC-WCH technical team
1. It is unclear whether clinicians were blinded to the results of the index test therefore subsequent clinical management, such as use of tocolytics where decided by the
attending clinician, may have been influenced by index test results and have affected when birth (the reference standard) occurred
2. Single study analysis
3. Women included in the study were a mixed population where some received tocoloysis and some did not
4. Tocolysis (with nifedipine or atosiban) was used in all women who were in established preterm labour but the definition of preterm labour and the proportion of women who
received tocolysis are not reported. Steroids (betamethasone) were administered as appropriate
6. Clinicians were not blinded to the results of the index test therefore subsequent clinical management, such as use of tocolytics may have been influenced by index test
results and have affected when birth (the reference standard) occurred
7. All women received tocolysis and corticosteroids according to existing clinical protocols
9. The primary clinician was blinded to the results of the index test therefore subsequent clinical management, such as use of tocolytics was not influenced by index test results.
10. Women with ruptured membranes were not specified as being excluded from the study and the proportion of women with ruptured membranes is not specified
11. 8/18 (44%) women who tested positive for pIGFBP-1 and 13/18 (72.2%) of women who tested negative for pIGFBP-1 received IV tocolysis (1st line treatment with
magnesium sulfate). This was according to an existing protocol where women with progressive cervical change and regular contractions, despite bed rest and hydration with
500ml Ringer’s lactate solution, received tocolysis
12. Evidence was downgraded by 1 due to 95% confidence interval for the negative likelihood ratio ranges from moderately useful (>0.1-0.5) to very useful (0-0.1)
13. Confidence intervals were not calculable
14. 12/51 (23.53%) women received tocolysis at the discretion of the attending clinician (further details of the tocolytic used or whether corticosteroids were administered were
not reported)
16. 8/14 (66.7%) of women who tested positive for pIGFBP-1 and 8/37 (20.5%) of women who tested negative for pIGFBP-1 received tocolysis (first line treatment with calcium
channel blockers) according to an existing protocol where women with progressive cervical change and persistent regular contractions, despite 2 hours bed rest and hydration
with 500ml Ringer’s lactate solution, received tocolysis. Maternal corticosteroids were given. No tocolytics or maternal steroids were used after 34 weeks gestation. The mean
gestational age at enrolment was 29.5 ± 2.6.
19. Decisions regarding tocolytic and steroid use were made by clinicians. 23/25 (92%) women who tested positive for pIGFBP-1 and 40/43 (93 %) of women who tested
negative for pIGFBP-1 received tocolysis. Symptomatic treatment included IV ritodrine hydrochloride or magnesium sulfate. Betamethasone was given twice daily to enhance
fetal lung maturation where indicated.
Table 31: GRADE profile for evaluation of fetal fibronectin to diagnose preterm birth within 48 hours or within 7 days
Other
considerations
Fetal fibronectin test to diagnose birth within 48 hours
1 study Prospective Very No serious Serious3 Serious4 13% 215 58.8% 78.8% 2.77 (1.48 0.52 (0.25 Very
(Gomez cohort serious1,2 inconsistency (34.4 to (76.7 to to 4.13) a to 0.86) a low
2005) 80.0) a 80.6) a Not useful Not useful
1 study Prospective Very No serious Serious3 No serious NR 118 75% 88% 6.25a 0.28 a Very
(LaShay cohort serious1,2 inconsistency imprecision Moderately Moderately low
2000) useful useful
Fetal fibronectin test to diagnose birth within 7 days
1 study Prospective Serious1 No serious Serious6 No serious NR 112 100% 70.9% 3.44 (2.57 0.00 Low
(Bartnicki cohort inconsistency imprecision (19.29 to (61.4 to to 4.60) a
1995) 100) a 79.2) a Not useful
1 study Prospective Serious1 No serious Serious3,7 No serious NR 114 89% (55 to 90 (55 to 8.9 a 0.12 a Low
(Benattar cohort inconsistency imprecision 100) 100) Moderately Moderately
1996) useful useful
1 study Prospective Serious2 No serious Serious3 No serious 28.9 % 52 66.7% 78.3% 3.0 a 0.43 a Low
(Burwick cohort inconsistency imprecision (53.5 to (66.7 to Not useful Moderately
2011) 79.9) 89.8) useful
1 study Prospective Serious2 No serious Serious3 Serious4 30% 170 75% (52.9 78.2 (70.7 3.44 (2.36 0.32 (0.16 Very
(Diaz cohort inconsistency to 89.4)a to 84.2) a to 5.01) a to 0.64) a low
useful
1 (Eroglu Prospective No serious No serious Serious6,10 Very 3.9% 51 83.3% 80.0% 4.17 (1.50 0.21 (0.01 Very
2007) cohort risk of inconsistency serious11 (38.9 to (74.1 to to 5.54) a to 0.82) a low
bias9 99.1) a 82.1) a Not useful Moderately
useful
1 study Prospective Serious2 No serious Serious3 Serious4 NR 151 68.7% 74.8 (67.5 2.73 (1.75 0.41 (0.20 Very
(Giles cohort inconsistency (46.0 to91.5)a to 82.1) a to 4.23) a to 0.87) a low
useful

Other
considerations
1 study Prospective Very No serious Serious3 Serious4 13% 215 64.3% 81.8 (79.1 3.54 (2.19 0.44 (0.24 Very
(Gomez cohort serious1,2 inconsistency (45.8 to to 84.1) a to 5.03) a to 0.69) a low
2005) 79.8) a Not useful Moderately
useful
1 study Prospective Serious2 No serious Serious3,10 Very 32% 192 93.0% 82.0% 5.17 (3.66 0.09 (0.01 Very
(Iams cohort inconsistency serious11 (66.0 to (75.5 to to 7.30)a to 0.58)a low
1994) 99.0) a 87.3) a Moderately Very useful
useful
1 study Prospective Very No serious Serious3 No serious NR 118 67% (CI NR NR NR Very
(LaShay cohort serious1,2 inconsistency imprecision NR) low
2000)
1 study Prospective No serious No serious Serious6 Serious8 21% 763 86.3% 82.3% 4.89 (3.89 0.17 (0.06 Low
(Lukes cohort risk of bias inconsistency (65.0 to (79.3 to to 6.13) a to 0.47) a
1996) 97.0) a 85.0) a Not useful Moderately
useful
1 study Prospective Serious1 No serious Serious6 Very 19% 141 80% (44.4 90.2% 8.16 (4.2 to 0.22 (0.06 Very
(Malak cohort inconsistency serious12 to 96.9) a (82.7 to 15.9) a to 0.77) a low
1996) 95.2)a Moderately Moderately
useful useful
1 study Prospective No serious No serious Very No serious 9% 185 73.8% 74.2% 2.86 a 0.35 a Low
(Sakai cohort risk of bias inconsistency serious13 imprecision Not useful Moderately
2003) useful
1 study Prospective Serious1 No serious Serious6 Very 57% 192 93.0% 79.0% 3.91 (2.96 0.22 (0.09 Very
(Schmitz cohort inconsistency serious11 (66.0 to (74.0 to to 5.17) a to 0.54) a low
2005) 99.0)a 83.0)a Moderately Very useful
useful
1 study Prospective No serious No serious Very Very NR 25 100% 86% 5.75 (1.34 0.15 (0.000 Very
(Senden cohort risk of bias inconsistency serious6,14 serious11 to 7.67) a to 0.89) a low
1996) Moderately Moderately
useful useful
1 study Prospective Serious1,15 No serious Serious6 Very 16% 149 80.0% 85.1% 5.36 (3.32 0.23 (0.08 Very
(Skoll cohort inconsistency serious11 (51.4 to (77.6 to to 8.63) a to 0.64) a low
2006) 94.7)a 90.4)a Moderately Moderately
useful useful

Other
considerations
1 study Prospective Serious1 No serious Serious6 No serious NR 404 67% 92% 8.37 a 0.35 a Low
(Swamy cohort inconsistency imprecision5 Moderately Moderately
1996) useful useful
1 study Prospective Serious2 No serious Serious6 No serious NR 195 68.6% 84.4% 4.3 (2.1 to 0.3 (0.2 to Low
(Tanir cohort inconsistency imprecision 9.8) 0.5)
useful
1 study Prospective Serious2 No serious Serious13 No serious 26% 170 81.8% 76.7% 3.5 0.24 Low
(Tekesin cohort inconsistency imprecision (48.2 to (69.4 to Not useful Moderately
2005) 97.7) 83.1) useful
1 study Prospective No serious No serious Serious13 No serious NR 94 56% 76% 2.33 a 0.73 a Low
(Ting cohort risk of bias inconsistency imprecision Not useful Not useful
2007)
1 study Prospective Serious1 No serious Serious3 Serious8 NR 195 94.7% 61.9% 2.49 (1.81 0.09 (0.004 Very
(Tsoi cohort inconsistency (73.0 to (59.6 to to 2.66)a to 0.45)a low
2006) 99.7)a 62.5)a Not useful Very useful
CI confidence interval, NR not reported
a. Calculated by the NCC-WCH technical team
1. No definition of symptoms of preterm labour
2. Blinding of clinicians to the index test was not reported
3. Whether women received tocolytic therapy was not reported
5. Confidence intervals were not calculable from the data provided
6. Women included in the study were a mixed population where some received tocolysis and some did not
7. n=15 women with twin pregnancy included
9. The primary clinician was blinded to the results of the index test therefore subsequent clinical management, such as use of tocolytics was not influenced by index test results
10. 8/14 (57.1%) women who tested positive for fFN and 8/37 (21.6%) women who tested negative for fFN received tocolytic therapy. Tocolytic therapy (with calcium channel
blockers) was administered according to an existing protocol where women with progressive cervical change and persistent regular contractions, following 2 hours bed rest and
hydration with 500ml Ringer’s lactate solution, received tocolysis. Maternal corticosteroids were given. No tocolytics or maternal steroids were used after 34 weeks gestation.
The mean gestational age at enrolment was 29.5 weeks ± 2.65
12. Evidence was downgraded by 2 due to 95% confidence interval for the negative likelihood ratio ranges from not useful (>0.5) to very useful (0-0.1
13. All women received tocolysis
14. 7/25 (28%) women received ritodrine, 8/25 (32%) women received antibiotic therapy and 19/25 (76%) received corticosteroids. Treatment was according to the established
practice of administration when considered appropriate
15. Reason for admission for 30/130 women not reported
Table 32: GRADE profile for evaluation of fetal fibronectin >50 nanogram/ml before and after cervical examination to diagnose preterm
birth within 7 days
Other
considerations
(percentage with Positive Negative
Number Risk of previous preterm Sample likelihood likelihood
of studies Design bias Inconsistency Indirectness Imprecision birth) size Sensitivity Specificity ratio ratio Quality
Fetal fibronectin >50 nanogram/ml before cervical examinationa
1 Prospective Serious1 No serious Very Very 30% 50 100% (41.1 73.9% 3.83 (1.32 0.00 (0.00 Very
(McKenna cohort inconsistency2 serious3,4 serious5 to 100.0)c (68.8 to to 3.83)c to 0.85)c low
1999) 73.9)c Not useful Very useful
Fetal fibronectin >50 nanogram/ml after cervical examinationa,b

1 Prospective Serious1 No serious Very Very 30% 50 75.0% 65.2% 2.16 (0.58 0.38 (0.02 Very
(McKenna cohort inconsistency 2 serious3,4 serious5 (22.7 to (60.7 to to 3.02)c to 1.27)c low
1999) 98.7)c 67.3)c Not useful Moderately
useful
a. A positive test result for fetal fibronectin was defined as > 50ng/ml. Cervical examinations were performed within 1 to 3 hours of the initial fetal fibronectin test.
b. Results for changes in fetal fibronectin test results following cervical examination were also provided; 5/34 women who initially tested negative changed to positive after the
second fetal fibronectin test, 2/16 women who initially tested positive changed to negative after the second test.
c. Calculated by the NCC-WCH technical team.
1. The symptoms of preterm labour were not defined
2. Single study analysis
3. The use of tocolytic medications was not reported
4. The inclusion of women with multiple pregnancy is unclear
Table 33: GRADE profile for evaluation of cervical length measured using transvaginal ultrasound to diagnose preterm birth within 48
hours or 7 days
Other
considerations
(percentage Positive Negative
Number Risk of with previous Sample likelihood likelihood
of studies Design bias Inconsistency Indirectness Imprecision preterm birth) size Sensitivity Specificity ratio ratio Quality
Cervical length measured using transvaginal ultrasound to diagnose preterm birth within 48 hours

Other
considerations
Cervical length ≤5 mm
1 (Tsoi Multicentre No No serious Serious3,4 Serious5 NR 510 42.9% 97.8% 19.05 (7.93 0.59 (0.39 Low
2005) observational serious inconsistency2 (24.2 to (96.9 to to 41.84)a to 0.78)a
cohort risk of 61.2)a 98.5)a Very useful Not useful
bias1
cohort risk of 93.6)a 94.2)a Very useful Moderately
bias1 useful
cohort risk of 100.0)a 84.9)a Moderately Very useful
bias1 useful
1 (Gomez Prospective Serious6 No serious Very Serious5 NR 215 64.7% 90.4% 6.74 (3.47 0.39 (0.18 Very
2005) cohort inconsistency2 serious3,7 (40.5 to (88.3 to to 10.55) to 0.67) low
83.9) 92.1) Moderately Moderately
useful useful
1 (Bagga Prospective Serious9 No serious Very Serious5 NR 100 62.5% 89.5% 5.94 (2.75 0.42 (0.25 Very
2010) cohort inconsistency2 serious7,10 (44.6 to (83.8 to to 12.60)a to 0.66)a low
76.6)a 93.9)a Moderately Moderately
useful useful
Cervical length <30 mm
1 (Schmitz Prospective No No serious Very Very NR 395 88.0% 40.0% 1.48 (1.22 0.29 (0.08 Very
2008) cohort serious inconsistency2 serious3,7 serious11 (64.0 to (65.0 to to 1.80) to 1.07) low
risk of 98.0) 46.0) Not useful Moderately
bias1 useful
1 (Gomez Prospective Serious6 No serious Very Very NR 215 88.2% 53.0% 1.88 (1.29 0.22 (0.04 Very
2005) cohort inconsistency2 serious3,7 serious11 (63.2 to (50.9 to to 2.12) to 0.72) low
97.9) 53.9) Not useful Moderately
useful
Cervical length measured using transvaginal ultrasound to diagnose preterm birth within 7 days

Other
considerations
1 (Tsoi Multicentre No No serious Very No serious NR 510 37.2% 99.1% 43.44 0.63 (0.57 Low
2005) observational serious inconsistency2 serious3,4 imprecision (26.7 to (98.2 to (14.65 to to 0.75)a
cohort risk of 43.4)a 99.7)a 149.45)a Not useful
bias1 Very useful
1 (Tsoi Multicentre No No serious Very Serious12 NR 510 65.1% 95.7% 15.21 (9.30 0.36 (0.24 Very
2005) observational serious inconsistency2 serious3,4 (51.5 to (94.5 to to 23.68)a to 0.51)a low
cohort risk of 76.5)a 96.8)a Very useful Moderately
bias1 useful
1 (Tsoi Multicentre No No serious Very Serious13 NR 510 97.7% 88.7% 8.61 (7.04 0.03 (0.00 Very
2005) observational serious inconsistency2 serious3,4 (86.9 to (87.7 to to 8.96)a to 0.15)a low
cohort risk of 99.9)a 88.9)a Moderately Very useful
bias1 useful
1 (Gomez Prospective Serious6 No serious Very Serious12 NR 215 60.7% 93.0% 8.73 (4.58 0.42 (0.26 Very
2005) cohort inconsistency2 serious3,4 (43.6 to (90.5 to to 15.66) to 0.62) low
75.1) 95.2) Moderately Moderately
useful useful
1 (Botsis Prospective No No serious Serious3,14 Very NR 62 81.8% 92.2% 10.43 (3.73 0.20 (0.04 Very
2006) cohort serious inconsistency2 serious15 (52.7 to (85.9 to to 20.60)a to 0.55)a low
risk of 96.5)a 95.3)a Very useful Moderately
bias1 useful
1 (DeMirci Prospective No No serious Very Serious8 NR 209 78.9% 94.2% 13.64 (7.15 0.22 (0.08 Very
2011) cohort serious inconsistency2 serious3,16 (57.0 to (92.0 to to 20.89)a to 0.47)a low
risk of 92.5)a 95.6)a Very useful Moderately
bias1 useful
1 Prospective No No serious Very No serious NR 108 26.3% 95.5% 5.86 (1.46 0.77 (0.61 low
(Gramellini cohort serious inconsistency2 serious3,17 imprecision (11.2 to (92.3 to to 24.29)a to 0.96)a
2007) risk of 39.7)a 98.4)a Moderately Not useful
bias1 useful
1 (Holst Prospective Serious9 No serious Very Serious5 NR 55 72.0% 83.0% 4.32 (1.88 0.34 (0.18 Very
useful

Other
considerations
1 (Palacio Prospective No No serious Serious18 No serious NR 333 28.6% 96.5% 8.10 (2.83 0.74 (0.54 low
2007) cohort serious inconsistency2 imprecision (12.9 to (95.4 to to 20.65)a to 0.91)a
risk of 47.1)a 97.7)a Moderately Not useful
bias1 useful
1 Prospective Very No serious Serious18 Very NR 122 83.3% 95.8% 20.00 (5.77 0.17 (0.01 Very
(Sotiriadis cohort serious9,20 inconsistency2 serious21 (43.7 to (89.8 to to 31.16)a to 0.65)a low
2010) 97.0) 98.4) Very useful Moderately
useful
Cervical length <15 mm in women admitted <32 weeks’ gestation
1 (Palacio Prospective No No serious Serious18 Serious5 NR 116 10.0% 96.9% 3.23 (0.00 0.93 (0.49 low
2007) cohort serious inconsistency2 (0.00 to (96.5 to to 41.52)a to 1.04)a
risk of 51.8)a 98.8)a Not useful Not useful
bias 1
Cervical length <15 mm in women admitted ≥32 weeks’ gestation
1 (Palacio Prospective No No serious Serious18 Serious5 NR 217 35.3% 96.0% 8.82 (2.93 0.67 (0.46 low
risk of 55.2)a 97.7)a Moderately Not useful
bias 1 useful
1 (Eroglu Prospective No No serious Serious22 Very 3.9% 51 66.7% 95.6% 15.00 (2.79 0.35 (0.09 Very
risk of 91.3)a 98.8)a Moderately Moderately
bias1 useful useful
1 (Bagga Prospective Serious9 No serious Very Serious5 NR 100 60.0% 96.9% 19.50 (5.14 0.41 (0.36 Very
64.7)a 99.5)a Very useful Moderately
useful
(Gramellini cohort serious inconsistency2 serious3,17 (45.7 to (74.2 to to 5.00)a to 0.73)a low
2007) risk of 83.3)a 83.3)a Not useful Moderately
bias1 useful

Other
considerations
1 (Palacio Prospective No No serious Serious18 Serious12 NR 333 71.4% 79.2% 3.43 (2.17 0.36 (0.16 Low
risk of 87.6)a 80.3)a Not useful Moderately
bias1 useful
1 Prospective Very No serious Serious18 Very NR 122 83.3% 77.1% 3.64 (1.46 0.22 (0.01 Very
(Sotiriadis cohort serious9,20 inconsistency2 serious15 (43.7 to (67.7 to to 16.61)a to 0.84)a low
useful
1 (Azlin Prospective No No serious Serious23 Very NR 51 80.0% 71.7% 2.83 (0.93 0.28 (0.01 Very
2010) cohort serious inconsistency2 serious11 (31.3 to (66.4 to to 3.78) to 1.03) low
bias1 useful
1 (Schmitz Prospective No No serious Very Very NR 395 94.0% 42.0% 1.63 (1.43 0.15 (0.04 Very
2008) cohort serious inconsistency2 serious3,7 serious11 (79.0 to (37.0 to to 1.84) to 0.57) low
bias1 useful
1 (Gomez Prospective Serious7 No serious Very Very NR 215 89.3% 55.6% 2.01 (1.53 0.19 (0.05 Very
2005) cohort inconsistency2 serious3,7 serious11 (71.8 to (53.0 to to 2.25) to 0.53) low
useful
1 (Danti Prospective No No serious Serious25 Very NR 102 90.0% 42.9% 1.58 (0.64 0.23 (0.00 Very
bias24 useful
1 (Palacio Prospective No No serious Serious18 Very NR 116 75.0% 85.7% 5.25 (1.39 0.29 (0.02 Very
risk of 98.7)a 86.6)a Moderately Moderately
bias 1 useful useful
1 (Palacio Prospective No No serious Serious18 Serious5 NR 217 70.6% 75.5% 2.88 (1.69 0.39 (0.15 Low
bias 1 useful

Other
considerations
Change in cervical length 24 hours after admission measured using transvaginal ultrasound to diagnose preterm birth within 7 days
Change in cervical length >20%
1 Prospective Very No serious Serious18 Serious5 NR 122 50.0% 92.7% 6.86 (1.54 0.54 (0.16 Very
(Sotiriadis cohort serious9,20 inconsistency2 (18.8 to (85.7 to to 16.61)a to 0.94)a low
2010) 81.2) 96.4) Moderately Not useful
useful
Cervical length <15 mm at admission plus change >20% 24 hours later

(Sotiriadis cohort serious9,20 inconsistency2 (18.8 to (94.3 to to to 0.87)a low
2010) 81.2) 99.8) 1171.37)a Not useful
Very useful
Cervical length >15 mm at admission plus change >20% 24 hours later
1 Prospective Very No serious Serious18 Very NR 122 25.0% (0.0 93.0% 3.57 (0.00 0.81 (0.10 Very
(Sotiriadis cohort serious9,20 inconsistency2 serious11 to 90.3)a (92.5 to to 16.17)a to 1.08)a low
2010) 94.4)a Not useful Not useful
Cervical length <25 mm at admission plus change >20% 24 hours later
(Sotiriadis cohort serious9,20 inconsistency2 (18.8 to (92.7 to to 173.72)a to 0.88)a low
NR not reported
a. Calculated by the NCC-WCH technical team.
2. Single study analysis.
3. Women included in the study were a mixed population where some received tocolysis and some did not.
4. 52% of women (265/510) received tocolytic medication.
6. Clinicians were not blinded to the results of the index test therefore subsequent clinical management, such as use of tocolytics where decided by the attending obstetrician,
may have been influenced by index test results and have affected when birth (the reference standard) occurred.
9. Blinding of clinicians to the index test was not reported.
10. Whether women received tocolytic medication was not reported.
12. Evidence was downgraded by 1 due to 95% confidence interval for the negative likelihood ratio ranging from not useful (>0.5) to moderately useful (>0.1-0.5)
14. 45.4% of women who gave birth within 7 days of admission received tocolytic medication; 31.3% of women who did not give birth within 7 days received tocolytic
medication.
15. Evidence was downgraded by 2 due to 95% confidence interval for the negative likelihood ratio ranging from not useful (>0.5) to very useful (0-0.1)
16. 81% (21/26) of women with cervical length < 15mm received tocolytic medication; 52% (96/183) of women with cervical length > 15mm received tocolytic medication.
17. 64.8% (70/108) of women received tocolytic medication.
18. All women received tocolytic medication.
19. Very wide CI LR+
20. No baseline characteristics were reported.
22. 6/51 (11.8%) women had a cervical length <20mm and these women received tocolytic therapy. 45 women had a cervical length >20mm and 10 of these women received
tocolytics. Tocolytic therapy (with calcium channel blockers) was administered according to an existing protocol where women with progressive cervical change and persistent
regular contractions, despite 2 hours bed rest and hydration with 500ml Ringer’s lactate solution, received tocolysis. Maternal corticosteroids were given. No tocolytics or
maternal steroids were used after 34 weeks gestation. The mean gestational age at enrolment was 29.5 ± 2.6.
not reported).
24. Clinicians were not blinded to transvaginal ultrasound results which informed decisions regarding admission to hospital (where cervical length ≤ 30mm). However,
subsequent clinical management, such as use of tocolytics, were also decided by the attending clinician, and may have been influenced by index test results, affecting when
birth (the reference standard) occurred.
25. Tocolytics and corticosteroids were administered at the discretion of the attending clinician. 22/60 (37%) of women with cervical length ≤30mm and 5/42 (12%) women with
cervical length >30mm received tocolysis. 28/60 (47%) of women with cervical length ≤30mm and 4/42 (10%) women with cervical length >30mm received corticosteroids.able
38: GRADE profile for evaluation of Bishop score plus cervical length measured using transvaginal to diagnose preterm birthwithin 48 hours or
7 days
Table 34: GRADE profile for evaluation of Bishop score plus cervical length measured using transvaginal to diagnose preterm birth
within 48 hours or 7 days
Other
consideratio
ns
(percentage
with
Number previous Positive Negative
of Risk of preterm Sample likelihood likelihood
Bishop score and cervical length measured using transvaginal ultrasound to diagnose preterm birth within 48 hours
Bishop score between 4 and 7 and cervical length ≤20 mm

Other
consideratio
ns
(percentage
with
Number previous Positive Negative
of Risk of preterm Sample likelihood likelihood
1 Prospective No No serious Very Serious5 15.5% 213 60.0% 64.0% 1.66 (0.75 0.63 (0.21 to Very
(Schmitz cohort serious inconsistency2 serious3,4 (26.0 to (57.0 to to 2.43)a 1.15)a low
2008) risk of 88.0) 71.0) Not useful Not useful
bias1
1 Prospective No No serious Very Very serious6 15.5% 213 80.0% 46.0% 1.48 (0.81 0.44 (0.08 to Very
bias1 useful
bias1 useful
Bishop score and cervical length measured using transvaginal ultrasound to diagnose preterm birth within 7 days
1 Prospective No No serious Very Serious5 15.5% 213 55.0% 65.0% 1.57 (0.90 0.69 (0.37 to Very
2008) risk of 77.0) 71.0) Not useful Not useful
bias1
bias1 useful
1 Prospective No No serious Very Very serious6 15.5% 213 95.0 (75.0 29.0% 1.34 (1.02 0.17 (0.01 to Very
(Schmitz cohort serious inconsistency2 serious3,4 to 100.0) (22.0 to to 1.80)a 0.94)a low
2008) risk of 36.0) Not useful Moderately
bias1 useful
1. Clinicians were blinded to the results of transvaginal ultrasound but not blinded to Bishop score results.

3. Women included in the study were a mixed population where some received tocoloysis and some did not.
5. Evidence was downgraded by 1 due to 95% confidence interval for the negative likelihood ratio ranging from not useful (>0.5) to moderately useful (>0.1-0.5)
Table 35: GRADE profile for evaluation of a selective test based on cervical length measured using transvaginal ultrasound plus Bishop
score to diagnose birth within 48 hours or 7 days
Other
considerations
Selective test based on cervical length measured using transvaginal ultrasound and Bishop score to diagnose birth within 48 hoursa,b
1 Prospective No No serious Very Very 15.5% 213 88% (64 to 58% (54 to 2.08 (1.74 0.20 (0.06 Very
(Schmitz cohort serious inconsistency2 serious3,4 serious5 99) 64) to 2.63) to 0.75) low
2008) risk of Not useful Moderately
bias1 useful
Selective test based on cervical length measured using transvaginal ultrasound and Bishop score to diagnose birth within 7 daysa,b
1 Prospective No No serious Very Serious6 15.5% 213 94% (79 to 60% (55 to 2.35 (2.01 0.10 (0.03 Very
(Schmitz cohort serious inconsistency2 serious3,4 99) 65) to 2.74) to 0.40) low
2008) risk of Not useful Moderately
bias1 useful
a. A test was considered positive for preterm birth if Bishop score was either ≥8 or between 4 and 7 combined with a cervical length ≤30mm.
b. The clinically selected population in this table only comprises women with a Bishop score of 4 to 7. Women with a Bishop score ≥8 were not included as their test results
were deemed positive without additional data on cervical length from transvaginal ultrasound.
6. Evidence was downgraded by 1 due to 95% confidence interval for the negative likelihood ratio ranging from very useful (0-0.1) to moderately useful (>0.1-0.5)
Table 36: GRADE profile for evaluation of fetal fibronectin greater than 50 nanograms/ml plus Bishop score to diagnose preterm birth
within 7 days
Other
considerati
ons
(percentage
with
previous Positive Negative
Number of Inconsisten Indirectnes preterm Sampl likelihood likelihood
studies Design Risk of bias cy s Imprecision birth) e size Sensitivity Specificity ratio ratio Quality
Fetal fibronectin >50 nanograms/ml and Bishop score >2a to diagnose birth within 7 days
1 (Senden Prospective Serious1 No serious Serious2 Very NR 50 100% 95% 13.42 (2.16 0.13 (0.000 Very low
1996) cohort inconsistenc serious3 to 23.0)b to 0.78) b
y Very useful Very useful
a. Bishop score defined according to Myerscough P.R.; ”Induction of labour” Chap 20 in Munro Kerr’s Operative Obstetrics 10th edn. 1982, pub, Bailliere Tindall
b. Calculated by the NCC-WCH technical team.
not reported)
Table 37: GRADE profile for evaluation of pIGFBP-1 test and cervical length measured using transvaginal ultrasound to diagnose
preterm birth within 7 days
Other
considerations
pIGFBP-1 test cervical length <20mm
1 (Danti Prospective Serious1 No serious Very Serious4 NR 19 33% (1-91) 63% (35- 0.89 (0.16 1.07 (0.44 Low
2011) cohort inconsistency serious2,3 85) to 4.97) to 2.59)
Not useful Not useful
pIGFBP-1 test and cervical length <25mm
1 (Azlin Prospective Serious6 No serious Serious2,6 Very NR 51 80.0% (34.4 97.8% (92.9 36.8 (4.83 0.20 (0.02 Very
2010) cohort inconsistency serious7 to 98.2) a to 99.8) a to 508.35) a to 0.71) a low

Other
considerations
Very useful Moderately
useful
pIGFBP-1 test and cervical length 20-30mm
1 (Danti Prospective Serious1 No serious Very Very NR 41 75% (15- 98.8% (95- 61.5 (3.5 to 0.25 (0.02 Very
2011) cohort inconsistency serious2,3 serious7 100)a 100) a 1083) a to 2.79) a low
Very useful Moderately
useful
pIGFBP-1 test and cervical length ≤30mm
1 (Danti Prospective Serious1 No serious Serious2,3 Serious4 NR 60 50% (7-93) 70% (56- 1.65 (0.57 0.72 (0.27 Low
2011) cohort inconsistency 81) to 4.74) to 1.94)
Not useful Not useful
NR Not reported, pIGFBP-1 phosphorylated insulin-like growth factor binding protein-1
1. Clinicians were not blinded to transvaginal ultrasound results. Women were admitted to hospital if cervical length ≤ 30mm (n=60) and offered a pIGFBP-1 test. Clinicians
were blinded to the results of the pIGFBP-1 test (index test) therefore subsequent clinical management, such as use of tocolytics, which was decided by the attending clinician,
was not influenced by index test results
3. Tocolytics and corticosteroids were administered at the discretion of the attending clinician. 22/60 (37%) of women with cervical length ≤30mm and 5/42 (12%) women with
cervical length >30mm received tocolysis. 28/60 (47%) of women with cervical length ≤30mm and 4/42 (10%) women with cervical length >30mm received corticosteroids.
Corresponding information for subgroups of women with cervical length <20mm or 20-30mm are not reported
not reported)
Table 38: GRADE profile for evaluation of fetal fibronectin greater than 50 nanograms/ml plus cervical length to diagnose preterm birth
within 48 hours or 7 days
Other
considerations
Number (percentage Positive Negative
of Risk of with previous Sample likelihood likelihood
studies Design bias Inconsistency Indirectness Imprecision preterm birth) size Sensitivity Specificity ratio ratio Quality
Fetal fibronectin > 50 nanograms/ml and cervical length to diagnose birth within 48 hoursa
Fetal fibronectin and cervical length <15mm
1 Prospective Serious1 No serious inconsistency Very Serious4 13% 215 41.2% 95.5% 9.06 (3.32 0.62 (0.40 Very
(Gomez cohort serious2,3 (20.9 to (93.7 to to 22.07)b to 0.84)b low
2005) 61.6)b 97.2)b Moderately Not useful
useful
2005) 79.8)b 87.7)b Not useful Moderately
useful
Fetal fibronectin > 50 nanograms/ml plus cervical length to diagnose birth within 7 daysa
1 Prospective Serious1 No serious inconsistency Very Serious4 13% 215 42.9% 97.7% 20.04 0.58 (0.48 Very
(Gomez cohort serious2,3 (28.4 to (95.7 to (6.60 to to 0.75)b low
2005) 52.2)b 99.3)b 69.99)b Not useful
Very
useful
Fetal fibronectin and cervical length≤ 15mm
1 (Van Prospective Serious1 No serious inconsistency Very Serious4 12% 714 88.7% 26.7% 1.21 (1.01 0.40 (0.18 Very
Baaren cohort serious2,3 (77.0 to to (16.1 to to 1.45)b to 1.01.)b low
2014) 95.7)b 39.7)b Not useful Moderately
useful
Fetal fibronectin and cervical length< 30mm
2005) 76.2)b 91.1)b Moderately Moderately
useful useful
Fetal fibronectin and cervical length 15–20mm

Other
considerations
Number (percentage Positive Negative
of Risk of with previous Sample likelihood likelihood
studies Design bias Inconsistency Indirectness Imprecision preterm birth) size Sensitivity Specificity ratio ratio Quality
1 (Van Prospective Serious1 No serious inconsistency Very No serious 12% 714 100% 47.7% 1.91 (1.56 0.00 Very
Baaren cohort serious2,3 imprecision (42.9 to (36.8 to to 2.34)b Very low
2014) 76.2)b 58.7)b Not useful useful
1 (Van Prospective Seriou No serious inconsistency Very Serious7 12% 714 72.73% 54.1% 1.59 (1.05 0.50 (0.19 Verylow
Baaren cohort s1 serious2, (39.1 to (44.3 to to 2.40)b to 1.34)
2014) 3
93.6)b 63.7)b Not useful Moderately
useful
1 (Van Prospective Seriou No serious inconsistency Very Very 12% 714 80% (28.8 58.4% 1.93 (1.15 0.34 (0.06 Verylow
Baaren cohort s1 serious2, serio to 96.7)b (46.6 to to 3.21)b to 2.0)
2014) 3
us5 69.6)b Not useful Moderately
useful
Fetal fibronectin and cervical length ≥30mm
1 (Van Prospective Seriou No serious inconsistency Very No 12% 714 100% 76.3% 4.22 (3.38 0.00 Verylow
Baaren cohort s1 serious2, serious (70.5 to to 5.26)b Very
2014) 3
imprecisi 81.4)b Not useful useful
on
a. The test for fetal fibronectin was performed prior to transvaginal sonography and a digital examination carried out to ascertain cervical dilation and effac ement.
b. Calculated by the NCC- WCH technical team
1. Clinicians were not blinded to the results of the index test therefore subsequent clinical management, such as use of tocolytics where decided by the attending obstetrician,
may have been influenced by index test results and have affected when birth (the reference standard) occurred.
2. Women included in the study were a mixed population where some received tocolysis and some did not.
5. Evidence was downgraded by 2 due to 95% confidence interval for the negative likelihood ratio ranging from very useful (0-0.1) to not useful (>0.5)
Tocolysis

Likelihood ratios are reported as the primary measure of diagnostic accuracy. The positive
likelihood ratio reports the number of times more likely women in preterm labour are to have
that symptom than women not in preterm labour. The higher the value, the more likely it is
that a woman with a positive test is actually in preterm labour. By convention, a value
between 5 and 10 is regarded as moderately useful and a value of over 10 is very useful.
Tests where the likelihood ratios lie close to 1 have little practical significance.
The negative likelihood ratio indicates whether the absence of a sign, score or a combination
of tests is a good way of determining that a woman is not in preterm labour. The lower the
value, the more likely it is that a woman with a negative test is not in preterm labour. In this
case, the lower the value reported in the GRADE table the better the test may be for this
diagnosis. By convention, a value of less than 0.1 is regarded as very useful and a value of
0.1 to 0.2 is moderately useful. Again, a negative likelihood ratio close to 1 demonstrates that
a negative test is equally likely for women in preterm labour and those who are not. Hence
tests where the likelihood ratios lie close to 1 have little practical significance.
9.5.1 Diagnosis using clinical examination
9.5.1.1 Bishop score

Very low quality evidence from 1 prospective cohort study (n=70) found a not useful positive
likelihood ratio and a moderately useful negative likelihood ratio for a Bishop score of 4 to 6
in diagnosing birth within 48 hours or 7 days of admission to hospital.
Evidence from a prospective cohort study (n=25) found a not useful positive likelihood ratio
and a moderately useful negative likelihood ratio for a Bishop score of more than 2 in
diagnosing birth within 7 days of admission to hospital. The quality of the evidence was very
low.
Evidence from a large prospective cohort study (n=395) found a not useful positive likelihood
ratio for a Bishop score of 4 or above in diagnosing birth within 48 hours or 7 days but a
moderately useful and very useful negative likelihood ratio for birth within 48 hours or 7 days,
respectively. The quality of the evidence was low.
The same study found a very useful positive likelihood ratio and a not very useful negative
likelihood ratio for a Bishop score of 8 or above to diagnose birth within 48 hours or 7 days.
The quality of the evidence was very low to low.
9.5.2 Diagnosis using biochemical tests
9.5.2.1 Phosphorylated insulin-like growth factor binding protein-1 (pIGFBP-1)

Four prospective cohort studies (n=448) found that pIGFBP-1 had a not useful positive
likelihood ratio and negative likelihood ratios that ranged from not useful to very useful in
diagnosing birth within 48 hours of testing. The quality of the evidence was very low to low.
@ 2015 National Collaborating Centre for Women’s and Children’s Health

161
Tocolysis
Low to very low quality evidence from 7 prospective cohort studies (n=567) investigating
pIGFBP-1 testing reported positive likelihood ratios that ranged from not useful to very useful
and negative likelihood ratios that were moderately or very useful in diagnosing birth within 7
days of testing.
1.1.1.3. Fetal fibronectin
Two prospective cohort studies (n=333) examined fetal fibronectin test to diagnose birth
within 48 hours. One study found that the test was not useful while the other reported that the
test was moderately useful when considering likelihood ratios. The quality of the evidence
was very low.
Twelve prospective cohort studies (n=3688) found that a fetal fibronectin test had a not
useful positive likelihood ratio for diagnosing preterm labour within 7 days of admission while
7 studies found it was moderately useful. Fourteen studies found a moderately useful, 3
studies a very useful and 1 study a not useful negative likelihood ratio for diagnosing preterm
labour within 7 days of admission. The quality of the evidence was very low to low.
9.5.2.2 Fetal fibronectin before or after digital cervical examination

One prospective cohort study (n=50) found a not useful positive likelihood ratio for a fetal
fibronectin test both before and after digital cervical examination. The negative likelihood
ratio was very useful and moderately useful for a fetal fibronectin test before and after digital
cervical examination, respectively. The quality of the evidence was low for fetal fibronectin
before cervical examination and very low for fetal fibronectin after cervical examination.
9.5.3 Diagnosis using ultrasound features
9.5.3.1 Cervical length measured using transvaginal ultrasound

Two prospective cohort studies (n=725) used a cervical length of less than 15 mm to
diagnose birth within 48 hours. Both studies found moderately useful positive likelihood
ratios; negative likelihood ratios were moderately useful and very useful. The quality of the
evidence was low and very low respectively.
Two prospective cohort studies (n=610) used a cervical length of less than 30 mm to
diagnose birth within 48 hours. Positive likelihood ratios were not useful and negative
likelihood ratios moderately useful. The quality of the evidence was very low.
Low to very low quality evidence from 3 other cervical lengths of less than 5 mm, less than
10 mm and less than 25 mm were also used to diagnose birth within 48 hours. Positive
likelihood ratios were very useful for less than 5 mm and less than 10 mm and moderately
useful for less than 25 mm. Negative likelihood ratios were moderately useful for less than 10
mm and less than 25 mm and not useful for less than 5 mm.
Eight prospective cohort studies (n=1614) used a cervical length of less than 15 mm to
diagnose birth within 7 days and found both positive and negative likelihood ratios to range
from not useful to very useful. Sensitivity ranged from low to high and specificity ranged from
moderate to high. The quality of evidence was very low.
Five prospective studies (n=714) used a cervical length of less than 25 mm to diagnose birth
within 7 days. Positive likelihood ratios were not useful in all but 1 study which identified a
very useful positive likelihood ratio for this test. Negative likelihood ratios were all moderately

162
Tocolysis
useful. The quality of the evidence was low in 2 studies for this test and very low in the
remaining 3.
Three prospective studies (n=712) used a cervical length of less than 30 mm to diagnose
birth within 7 days. Positive likelihood ratios were found to be not useful and negative
likelihood ratios moderately useful. The quality of evidence was very low for this test in all
studies.
Two further cervical lengths of less than 5 mm and less than 10 mm were used to diagnose
birth within 7 days. Positive likelihood ratios were found to be very useful and negative
likelihood ratios were not useful and moderately useful, respectively. Sensitivity was low and
specificity high for both tests. The quality of the evidence was low and very low, respectively.
One prospective cohort study (n=122) identified evidence for change in cervical length
greater than 20% at different cervical lengths to diagnose birth within 7 days. Tests included
change score alone and in combination with cervical lengths less than 15 mm, greater than
15 mm and less than 25 mm. Positive likelihood ratios ranged from not useful to very useful.
Negative likelihood ratios were not useful for all tests. Sensitivity was low and specificity high
for all tests. The quality of the evidence was very low for all tests.
Subgroup analyses
One prospective study (n=116) provided evidence for subgroup analysis on cervical length
(cut-offs of less than 15 mm and less than 25 mm) in women admitted before 32 weeks’
gestation or 32 weeks’ gestation and above to diagnose birth within 7 days. Positive and
negative likelihood ratios were moderately useful and not useful. The quality of the evidence
was very low to low.
9.5.4 Diagnosis using combination of tests

1.1.1.4. Clinical examination plus ultrasound features
One prospective cohort study (n=213) found for that the combination of Bishop score of 4 to
7 and cervical lengths (measured using transvaginal ultrasound) of less than 20 mm, less
than 25 mm and less than 30 mm provided not useful positive likelihood ratios to diagnose
birth within 48 hours or 7 days. Negative likelihood ratios were not useful for this combination
of tests for cervical lengths of less than 20 mm but moderately useful for cervical length less
than 25 mm and less than 30 mm. The quality of the evidence was very low for all tests.
One prospective cohort study (n=213) found that a selective test that combined Bishop score
of 4 to 7 and cervical length (measured using transvaginal ultrasound) of less than 30 mm
provided not useful positive likelihood ratios and moderately useful negative likelihood ratios
for diagnosis of preterm labour within 48 hours or 7 days.
1.1.1.5. Biochemical tests plus clinical examination
One prospective cohort study (n=50) reported that the combination of a fetal fibronectin test
greater than 50 nanogram/ml and a Bishop score of greater than 2 provided very useful
positive and negative likelihood ratios to diagnose birth within 7 days. The quality of the
evidence was very low.
9.5.4.1 Biochemical tests plus ultrasound features

pIGFBP1

163
Tocolysis
One prospective cohort study (n=19) found that the combination of a pIGFBP-1 test and
cervical length of less than 20 mm or 30 mm or less provided positive and negative likelihood
ratios that were not useful for diagnosis of preterm labour within 7 days. Sensitivity and
specificity were both low. The quality of this evidence was very low to low.
However, 2 prospective cohort studies (n=92) found that with cervical length measurements
of less than 25 mm or 20–30 mm the same combination of tests provided positive likelihood
ratios that were very useful and negative ratios were moderately useful for diagnosis of
preterm labour within 7 days. The quality of this evidence was very low.
Fetal fibronectin
One study (n=215) found that the combination of fetal fibronectin and cervical length less
than 15 mm provided moderately useful and very useful positive likelihood ratios, and not
useful negative likelihood ratios, to diagnose birth within 48 hours and 7 days, respectively.
Sensitivity was low and specificity high for at both time points.
However, the same study found that with cervical length measurements of less than 30 mm
fibronectin and cervical length measurements provided not useful and moderately useful
positive likelihood ratios, and moderately useful negative likelihood ratios, to diagnose birth
within 48 hours and 7 days, respectively. One other study (n=714) found that the combination
of fetal fibronectin and cervical length of 15 mm or less provided not useful positive likelihood
ratios, and moderately useful negative likelihood ratios with moderate sensitivity and low
specificity, to diagnose birth within 7 days, The same study found that with cervical length
measurement of 15–20 mm, 20–25 mm, 25–30 mm and 30 mm or more this combination test
provided not useful positive likelihood ratio and very useful and moderately useful negative
likelihood ratios. The quality of the evidence was very low for all tests.

This question was prioritised for health economic analysis.
A search was undertaken for health economic evidence on the diagnostic accuracy of the
various tests (clinical assessment, biochemical tests and ultrasound) used alone or in
combination to identify preterm labour leading to preterm birth in women presenting with
intact membranes. A total of 229 articles were identified by the search. After reviewing titles
and abstracts, 15 full copies of papers were obtained but these were all excluded. Therefore
no relevant economic evidence was identified for this question.
In order to assess the cost effectiveness of alternative diagnostic strategies it is necessary to
consider also the resources and interventions through which diagnosis can lead to improved
health outcomes. Therefore the analysis undertaken for this question utilised the output of
the health economic model produced for the tocolytic review, as that is a treatment that could
be offered as the result of a diagnostic assessment for women with suspected preterm labour
and intact membranes.
The Guideline Committee considered 3 types of treatment to be relevant to improve neonatal
and maternal outcomes following a diagnosis of preterm labour in women with intact
membranes. These are tocolysis to delay preterm birth, magnesium sulfate for
neuroprotection and corticosteroids for fetal lung maturation. The clinical evidence reviewed
for this guideline did not include studies which assessed these treatments provided in
combination and therefore, for pragmatic reasons, the analysis undertaken for this question
utilised the output of the health economic model produced for the tocolytic review, as

164
Tocolysis
tocolysis is a treatment that could be offered as the result of a diagnostic assessment for
women with suspected preterm labour and intact membranes. It should be noted that this
approach may underestimate the benefits of diagnosis as the actual treatment benefit is not
restricted to that available from tocolysis.
The new health economic evaluation undertaken for this guideline took the form of a cost
utility analysis and aimed to compare alternative diagnostic strategies in women to identify
preterm labour in women with suspected preterm labour and intact membranes between the
gestational ages of 24 and 34 weeks. Due to the limitations of the diagnostic accuracy review
studies included in the clinical review, the evaluation took a ‘what-if’ approach to diagnostic
accuracy. This involved taking all 10,201 combinations of sensitivity and specificity between
0% and 100% with 1 percentage point increments and comparing their cost-utility in a
‘diagnostic test plus treatment if indicated’ strategy relative to strategies of ‘treat all without
diagnosis’ or ‘no diagnosis and no treat’ and determining what combinations of sensitivity
and specificity were cost effective for a given prevalence, diagnosis and treatment cost. Such
an analysis could allow the committee to see what, if any, diagnostic strategies could be
considered as cost-effective options based on their reported diagnostic accuracy and the
diagnostic accuracy threshold required for cost effectiveness.
In this ‘what-if’ model, the treatment benefit is only obtained by ‘true positives’ (which in this
case means those in actual preterm labour who are treated). The absolute risk of adverse
outcomes for ‘true positives’ was modelled using the relative treatment effect of calcium
channel blockers, which were assessed in Section 16.4 as being the most cost-effective
tocolytic, applied to the baseline risk of these outcomes in the absence of treatment. ‘False
negatives’ (those in preterm labour not treated, either as a direct result of the strategy or a
negative test result) are assumed to have the baseline risk of adverse outcomes. ‘False
positives’ (those not in preterm labour but treated, either as a direct result of the strategy or a
positive test result) do not receive any benefit from treatment but do incur the relevant
treatment costs.
The baseline data used in this guideline suggested that the baseline risk of adverse
outcomes varied with gestational age with, as expected, declining risk with increasing
gestational age (see Table 110, Table 111 and Table 112).
It was assumed in the tocolytic model that the relative treatment effect would be constant
across the different gestational ages. However, the difference in baseline risk means that the
absolute treatment benefit, a key component of cost effectiveness, declines with increasing
gestational age.
Although the model supports recommendations which use a ‘treat all’ strategy at lower
gestational age and the use of a diagnostic test to determine treatment at higher gestational
ages, the limitations of the diagnostic accuracy evidence means that the model does not give
a definitive gestational age at which the strategy should change. Both studies of transvaginal
ultrasound using a cervical length of 15 mm or less have diagnostic accuracy figures that are
sufficient to make treatment based on a diagnostic test cost effective relative to ‘treat all’ at
30 weeks. Using transvaginal ultrasound and cervical length of 10 mm or less also has
diagnostic accuracy figures that would support a recommendation when compared with ‘treat
all’, but this is only based on a single study and an element of clinical judgement and
pragmatism was used to inform the recommendations to ‘treat all’ below a gestational age of
30 weeks and to use transvaginal ultrasound and cervical length of 15 mm or less as a
diagnostic test to determine treatment where gestational age is 30 weeks and above. The
model does not show that ‘treat all’ is a cheap strategy but rather that the additional costs are
worth the reduction in adverse outcomes at lower gestational ages.

165
Tocolysis
Sensitivity analysis suggested that the cost of the diagnostic test (within plausible ranges)
was not an important driver of cost-effective thresholds for ‘treat all’, ‘treat based on
diagnostic test’ and ‘no diagnosis and no treatment’. The inclusion of a cost for false
negatives was also found to have little impact on model results. However, the analysis did
suggest that model conclusions about cost-effective combinations of sensitivity and
specificity could be sensitive to relatively small changes in prevalence.
There is also a concern that the implications of a ‘treat all’ strategy might require some units
to transfer women out of their hospital and therefore a sensitivity analysis was undertaken
where the treatment cost was increased by £300 per woman to allow for the costs of such
transfers. As expected this change lowers the threshold for diagnostic accuracy to be
considered cost effective relative to ‘treat all’ and increases the threshold for diagnostic
accuracy to be considered cost effective relative to ‘no diagnosis and no treat’. At the lowest
gestational ages the higher treatment cost has a relatively small impact on the diagnostic
threshold (see Figure 43) but this increases with increasing gestational age.
The overall impact of this sensitivity analysis would be to tend to push down the gestational
age at which the cost-effective strategy would change from ‘treat all’ to treatment based on a
diagnostic test. However, given the uncertainty with respect to the diagnostic accuracy of the
tests reviewed, the committee, on balance, did not consider that this sensitivity analysis had
a sufficiently large impact on the diagnostic accuracy threshold to justify using a diagnostic
test at gestational age lower than 30 weeks.
The model is described in greater detail in Chapter 16.

The committee considered the following measures of diagnostic accuracy for decision-
making for this topic: sensitivity, specificity, and positive and negative likelihood ratios. The
committee considered the relative importance of having a high false positive and high false
negative result in the diagnosis of preterm labour and the consequences for further
management of women and babies.
Likelihood ratios were considered the most critical measures of diagnostic accuracy of
different tests for preterm labour and for the committee’s decision-making. The committee
agreed that if a woman had her baby within 7 days of a positive test then she was ‘truly’ in
labour and the test is a useful predictor.
Preterm labour (reference standard) was defined in relation to both 48 hours and 7 days. The
outcome of preterm delivery at 48 hours was considered important because it is related to
the decision-making regarding the timing of steroid and magnesium sulfate administration.
The diagnosis of preterm labour in 7 days was considered equally important because if
negative then the clinicians can be fairly confident that this woman is unlikely to deliver
preterm in either 7 days or 48 hours and that would change the management strategy
allowing discharge of women from hospital.
The committee discussed in depth the importance of extending pregnancy even by days in
early gestations to improve survival.

166
Tocolysis
Avoiding false negative diagnoses was considered more important by the committee than
false positives, because the risks associated with preterm birth outside hospital and the
harms of not giving steroids and magnesium sulfate where indicated are likely to outweigh
the harms of over-treatment of women incorrectly believed to be in preterm labour.
Additionally, false negative diagnoses disproportionally impact women who live far from a
tertiary centre and those at very early gestations.
The committee also discussed the need for vaginal examination and concerns associated
with the actual procedure as an invasive technique, as well as the role of vaginal examination
in decision-making. The women’s own views and circumstances are important in this
decision-making (although not captured by diagnostic accuracy measures) as they are
undergoing uncomfortable procedures (Bishop score and speculum exam) which might
reduce the uptake rate.

A number of studies were included in the review which considered different measurements.
The Bishop score was not found to be a helpful test for diagnosis of preterm labour unless
the score was over 8. The committee discussed the interpretation of these results and
concluded that there was not much to gain from this test in terms of diagnosing women at
preterm labour. In relation to potential harms, besides missing a diagnosis of preterm labour,
the committee discussed the risk of infection and discomfort associated with the invasive
nature of this test. If used prior to sampling for fibronectin, then it could compromise the
accuracy of this biochemical test. In summary, Bishop scoring was not considered an
appropriate option to be used alone for diagnosis of preterm labour. However, the committee
did not feel a ‘do not do’ recommendation is warranted because of the complexity of the
decision, but wanted to consider the Bishop score as a ‘last resort’ test if other
measurements for the diagnosis of preterm labour are not available.
The evidence showed that a short cervical length (less than 15 mm) appears to have a
moderately or very useful positive and negative likelihood ratio to diagnose women with
preterm delivery at 48 hours whereas a cervical length less than 5 mm had a very useful
negative likelihood ratio. Additional usefulness of this test was found in relation to 20 mm and
below for accurately diagnosing preterm labour at 7 days (useful positive likelihood ratio) and
above 5 mm for ruling out women without preterm labour (useful negative likelihood ratio)
although confidence intervals are wide and results should be interpreted with caution.
There were mixed results in relation to the use of fetal fibronectin to diagnose preterm labour.
This test was found to be more useful to rule out preterm delivery and the committee
discussed that fetal fibronectin should be used for the preterm labour diagnosis only if
transvaginal ultrasound measurement of cervical length is indicated but is not available or not
acceptable. Given that the majority of studies included in the evidence review used the
threshold of 50 nanograms/ml, the committee considered this quantitative cut-off point of
assessing positive results in fibronectin (more than 50 nanograms/ml) or negative results (50
nanograms/ml or less).
The combination of either Bishop score or fibronectin testing and cervical length
measurement were also found to be not useful to diagnose preterm labour and the
committee decided upon a strong recommendation of not using these tests in combination
for the diagnosis of preterm labout. Although the committee chose not to make a research
recommendation, they commented that further research may be necessary because looking

167
Tocolysis
at individual tests would only be part of the full assessment for diagnosis of preterm labour
and so combining them might be helpful.
pIGFBP1 plus cervical length was also not found to be a clinical relevant tool for diagnosis of
preterm labour and of particular note was that this combination was of no more use than the
use of cervical length measurements alone.
The committee discussed the invasive nature of these techniques. They noted that in line
with the NICE guideline on intrapartum care, an initial clinical assessment should include a
clinical history and observations of the women and her baby (Recommendation 1.4.2).The
committee concluded that if biochemical or ultrasound testing is being performed it should be
performed at the same time as the initial vaginal examination, to minimise any additional
discomfort associated with these diagnostic procedures. In addition, the committee
strengthened the importance of clinical assessment in the decision-making for the treatment
plan after diagnosis of preterm labour.

Diagnosis requires resources and therefore has opportunity costs, as those resources cannot
be deployed elsewhere in the healthcare system with a concomitant loss in health related
quality of life in those who could have benefited from such an alternative deployment.
Therefore, for diagnosis to be cost effective there is usually a minimum requirement that
diagnosis can lead to improved outcomes. Therefore, the cost effectiveness of diagnostic
strategies are normally linked to the effectiveness of any treatment arising from a positive
diagnosis and a consideration of health benefits and resource uses should usually be
considered as part of a diagnostic and treatment package. In the analysis undertaken for this
guideline diagnosis was linked to tocolysis and the costs of that treatment are included here
in the consideration of health benefits and resource uses.
An economic evaluation compared alternative diagnostic strategies to identify preterm
labour in women with suspected preterm labour and intact membranes. In addition to the
various diagnostic tests and thresholds, strategies of ‘treat all without diagnosis’ and ‘no
diagnosis and no treatment’ were also included as comparators. There was considerable
uncertainty with respect to diagnostic accuracy and therefore a “what-if” approach was used
to identify cost-effective combinations of sensitivity and specificity. This approach allowed the
committee to make an assessment of the cost-effectiveness of the various diagnostic
strategies based on the reported diagnostic accuracy and the diagnostic accuracy thresholds
required for cost-effectivenessThe ‘treat all’ strategy can be considered as analogous to a
diagnostic test that has a sensitivity of 100% and a specificity of 0%. This means that no
woman with suspected preterm labour would miss the benefits of treatment as a result of an
incorrect false negative result. However, it would also mean that all women who were not in
actual preterm labour would be treated for no actual benefit. Conversely, a strategy of ‘no
diagnosis and no treat’ can be considered analogous to a diagnostic test with a sensitivity of
0% and a specificity of 100%. No women in suspected preterm labour would derive the
benefit of treatment but resources would not be wasted on women who would not derive
benefit. Actual diagnostic tests fall between these two extremes, missing a number of actual
cases and leading to the treatment of a proportion of cases who would not benefit, with the
actual numbers falling into these categories dependent on the diagnostic accuracy of the test
in question
‘True positives’ (those women in actual preterm labour who are treated) obtain a benefit from
treatment. This benefit was modelled by calculating the absolute risk of adverse outcomes

168
Tocolysis
using the relative treatment effect of calcium channel blockers (which were assessed as the
most cost-effective tocolytic, see Sections 12.9 and 16.4) applied to the baseline risk of these
outcomes in the absence of treatment. ‘False negatives’ (those in preterm labour not treated,
either as a direct result of the strategy or a negative test result) are assumed to have the
baseline risk of adverse outcomes. ‘False positives’ (those not in preterm labour but treated,
either as a direct result of the strategy or a positive test result) do not receive any benefit
from treatment but do incur the relevant treatment costs. The model took into account that at
low gestational ages the absolute risks of adverse outcomes are much larger and therefore
the false negative rate can be particularly important in determining the most cost- effective
strategy. This is because, at low gestational ages, false negatives can result in large losses
of health related quality of life and, in this context, expensive lifetime NHS costs for adverse
neurodevelopmental outcomes. However, absolute risks fall with increasing gestational age
and therefore the relative benefits of treatment fall. As a result the false positive rate can be
increasingly important as the ‘wasted’ resources from those who derive no treatment benefit
from it are spread over fewer gains in health related quality of life. At the extremes a ‘treat all’
strategy minimises the false negatives and therefore is more likely to be cost effective at
lower gestational ages. Conversely, a ‘do not diagnose, do not treat’ strategy minimises the
number of false positives and therefore is more likely to be cost effective at higher
gestational ages. The modelling undertaken for this guideline tended to bear out this logic
and provided strong evidence that the cost-effective approach could vary with gestational
age.
The ‘what if’ model showed that at earlier gestational ages when the absolute risks are high
treating all women with suspected preterm labour and intact membranes can be cost
effective even when allowing for the fact that 90% of those treated might not derive any
treatment benefit. This is because the diagnostic accuracy of the tests is unlikely to produce
a good enough trade-off in terms of reduced false positives to offset the high opportunity
costs of missing false negatives at low gestational ages. At higher gestational ages,
treatment can remain cost effective at higher gestational ages when absolute risks are lower
providing a diagnostic test can be applied with sufficiently good diagnostic accuracy, as
additional benefit can be achieved without an unacceptable increase in cost arising from
false positives.
It was assumed in the tocolytic model that the relative treatment effect of different
interventions would be constant across the different gestational ages. However, the
difference in baseline risk means that the absolute treatment benefit, a key component of
cost effectiveness, declines with increasing gestational age.
Conceptually it follows that the cost effectiveness of alternative diagnostic strategies could
vary with gestational age. Other things being equal in the model, sensitivity is of greater
relative importance at lower gestational ages where the absolute effects of missing cases is
greatest. A ‘treat all’ strategy may be optimal despite the high costs of treating ‘false
positives’ as the absolute treatment effect is maximised. In addition, there are not any costs
associated with testing and a reduction in the cost of ‘downstream’ adverse events may to
some extent offset higher treatment costs. However, a diagnostic test may still be preferred
over ‘treat all’ if the test sensitivity is sufficiently high and relatively large savings in treatment
cost are generated by reducing the amount of unnecessary treatment. The importance of
sensitivity means that relatively large increases in test specificity are needed on the cost
effectiveness threshold to compensate for any reduction in test sensitivity.
At higher gestational ages a ‘no diagnosis and no treat’ may be most cost effective as it
avoids entirely costs associated with diagnosis and treatment with only a relative small loss

169
Tocolysis
in absolute treatment benefits from missed treatment. A diagnostic test may still be preferred
especially if it has a relatively low false positive rate while capturing sufficient absolute
benefit from identifying cases. The increased relative importance of specificity at higher
gestational ages means that a relatively smaller increase in test specificity is needed to
compensate for declining sensitivity on the cost effectiveness threshold
Although a change in diagnostic strategy according to gestational age was indicated by the
analysis undertaken for this guideline, the gestational age at which this change should take
place is difficult to precisely identify, given the uncertainty with respect to the diagnostic
accuracy of the various tests. Nevertheless, the ‘what-if’ data and results from the clinical
review suggested that 30 weeks and above may be reasonable gestational age at which to
require treatment to be guided by a positive diagnostic test, and thereby reduce
inconvenience to women and costs to the health service when absolute risks are, relatively,
lower. At 30 weeks there was some suggestion from the diagnostic studies reviewed that
transvaginal ultrasound using a cervical length of 15 mm or less could have sufficient
diagnostic accuracy to be considered cost effective relative to ‘treat all’, ‘do not diagnose, do
not treat’ or other diagnostic tests or combinations of tests which do not have a cost-effective
sensitivity/specificity combination.
In addition, further sensitivity analysis suggested that the cost of the diagnostic test (within
plausible ranges) was not an important driver of cost-effective thresholds for the decisions to
‘treat all’, ‘treat based on diagnostic test’ and ‘no diagnosis and no treatment’. The inclusion
of a cost for false negatives was also found to have little impact on the model results.
However, the analysis did suggest that model conclusions about cost-effective combinations
of sensitivity and specificity could be sensitive to relatively small changes in prevalence. The
committee was concerned that the implications of a ‘treat all’ strategy might require some
units to transfer women out of their hospital and therefore a sensitivity analysis was
undertaken where the treatment cost was increased per woman to allow for the costs of such
transfers. As expected, this change lowers the threshold for diagnostic accuracy to be
considered cost effective relative to ‘treat all’ and increases the threshold for diagnostic
accuracy to be considered cost effective relative to ‘no diagnosis and no treat’. At the lowest
gestational ages the higher treatment cost has a relatively small impact on the diagnostic
threshold (see Section 16.2.3) but this increases with increasing gestational age.

The majority of evidence contributed to this section was moderate to low as assessed by the
Quality Assessment of Diagnostic Accuracy Studies (QUADAS) checklist. Overall quality
assessment was made based on a modified GRADE approach. The thresholds of
measurements were not selected based on clinical considerations but the results reported as
in the studies. The studies varied considerably in terms of their populations’ characteristics
and baseline risk, and therefore pooling the result was not appropriate, but this is not unusual
for diagnostic studies. Although the majority of studies for fibronectin had used only 1
threshold of more than 50 nanograms/ml, the baseline characteristics of women across
studies were too different to allow pooling of their results and therefore their interpretation is
difficult. In addition, no complete information was given for all diagnostic measures (for
example 95% confidence interval of sensitivity and specificity) to allow a diagnostic meta-
analysis. In addition, the results on fibronectin should be interpreted with caution as sample
sizes were small and quality of evidence was low.

170
Tocolysis

The National Screening Committee has determined that there is not enough evidence to
justify routine screening for cervical length in low-risk women. The Guideline Committee
noted that transvaginal ultrasound scanning is not available across the NHS because of
limitations of equipment or expertise, and that investment in technology and training may be
required for its universal implementation in the NHS. They also noted the importance of staff
training to ensure that ultrasound measurements of cervical length were performed using
consistent and standard criteria.

The committee concluded that measuring cervical length using transvaginal ultrasound is the
most accurate way to diagnose preterm labour when used alone for women over 30 weeks in
pregnancy. Fibronectin was also useful if cervical length measurement not available or not
acceptable but not as good a diagnostic tool as cervical length. The committee noted the
importance of false positives and negatives and the associated harm with either missing
women at risk of preterm birth who are deprived of the benefits of treatment, or identifying
wrongly that women are at risk of preterm birth resulting in unnecessary management.
They acknowledged the need for women to understand different diagnostic testing options,
including their associated benefits and harms, and the interpretation of the results to guide
possible subsequent management strategies.
The committee members made recommendations based on their interpretation of the
evidence and on their expert clinical opinions.
Please see the also the health economics profile of the tocolysis section in Section 10.10.
9.8 Recommendations
updated guideline.
10 Tocolysis
10.1 Introduction
This review considers the clinical and cost effectiveness of medicines given to women who
are in suspected or diagnosed preterm labour with the aim of delaying birth to improve
outcomes. By definition, a drug which stops or delays the progress of labour once it is
believed to have started is a ‘tocolytic’. However, there is some debate about which drugs
should be classified as tocolytic. Drugs which are used to ‘prevent’ preterm labour are not
considered tocolytics. Progesterone is a good example of a drug which is normally used as a
prophylactic agent for women at risk of preterm labour during the antenatal period (see
Section 4.2). However, progesterone is also used increasingly in the intrapartum context for
its potential ‘tocolytic effect’. In contrast, ethanol, one of the first agents used as a tocolytic,
would no longer be considered a therapeutic option even if found to be effective for neonatal
outcomes because of known maternal side effects. This chapter will review the relative

171
Tocolysis
effectiveness of all medicines which have been used to delay or stop preterm labour and will
refer to them as tocolytics.
There is currently variation in clinical practice with respect to the use of tocolytic medicines,
both in terms of the choice of medicine and the selection of women who receive treatment
(all pregnant women in preterm labour or a selected sub-group).
Where multiple treatment options exist, it is very difficult to determine which intervention is
most effective in improving outcomes based on the results of conventional pair-wise meta-
analyses of direct evidence. The challenge of interpretation of direct evidence for assessing
the most effective intervention for improving outcomes arises for 2 reasons:
• Some pairs of alternative interventions may not have not been directly
compared in arandomised controlled trial (RCT) (for example, in the
case of tocolytics there are no studies that have directly compared
oxytocin receptor blocker to magnesium sulfate).
• A head-to-head analysis usually only provides information about the
relative effect of amaximum of 2 treatments; it does not provide an estimate
of the relative effects acrossmultiple treatment options.
To overcome these issues, mixed treatment comparison (MTC) meta-analytic techniques,
also termed network meta-analysis (NMA), were performed. One advantage of performing
this type of analysis is that it allows the synthesis of data from direct and indirect
comparisons without breaking randomisation, to produce measures of treatment effect and
ranking of different interventions. If treatment A has never been compared with treatment B
head to head, but these 2 interventions have been compared with a common comparator
(treatment C), then an indirect treatment comparison can use the relative effects of the 2
treatments versus the common comparator. This is also the case whenever there is a path
linking 2 treatments through a set of common comparators. All the randomised evidence is
considered within the same model. NMA is a generalisation of standard pairwise meta-
analysis for A versus B trials to data structures that include, for example, A versus B, B
versus C, and A versus C trials. A basic assumption of NMA methods is that direct and
indirect evidence estimate the same parameter; that is, the relative effect between A and B
measured directly from an A versus B trial is the same as the relative effect between A and B
estimated indirectly from A versus C and B versus C trials. This is often termed the
consistency assumption and should be assessed and taken into account when interpreting
the results of an NMA. NMA techniques strengthen inference concerning the relative effect of
2 treatments by including both direct and indirect comparisons between treatments and, at
the same time, allow simultaneous inference on all treatments while respecting
randomisation.
A second advantage of MTC (or NMA) is that for every intervention in a connected network,
a relative effect estimate (with its 95% credible intervals) can be estimated versus any other
intervention. These estimates provide a useful clinical summary of the results and facilitate
the formation of recommendations based on all of the best available evidence, while
appropriately accounting for uncertainty. Furthermore, these estimates will be used to
parameterise treatment effectiveness in the new cost effectiveness modelling.
For details of the methods, results and interpretation of the NMA, see ‘Methods’ below and
Appendix J.
Given the interventional nature of this review question, the only RCTs considered were of
women at high risk of or suspected to be in preterm labour that compared different

172
Tocolysis
interventions as tocolytics between each other or between a tocolytic and a placebo or usual
care for delaying preterm delivery. Trials that compared combination interventions as
tocolytics were excluded for the scope of this review question. The Guideline Committee
decided upon the following outcomes (maternal and neonatal) with a hierarchy that reflects
their importance for decision-making (the smaller the number, the higher this outcome’s
importance):
1. maternal mortality
2. neonatal mortality
3. perinatal mortality
4. maternal infection
5. delay of birth by at least 48 hours
6. neonatal sepsis
7. chronic lung disease (CLD)/bronchopulmonary dysplasia
8. intraventricular haemorrhage (IVH)
9. mothers with adverse events requiring cessation of treatment
10. neurodevelopmental disability (combined outcome including: developmental
delay; intellectual, gross motor, visual or hearing impairment; cerebral palsy;
learning difficulties)
11. periventricular leucomalacia (PVL)/white matter injury
12. gestational age at birth
13. respiratory distress syndrome (RDS)
14. quality of life.
However, given the paucity of data for some outcomes and the time constraints in the
guideline development, the committee prioritised the following outcomes for the NMA:
• RDS
• IVH
• mothers with adverse events requiring cessation of treatment
• neonatal sepsis
A class effect model was adopted for the new NMA because it was hypothesised that
treatments within class would borrow similar clinical characteristics and mechanisms of
effect. In other words, results for one member of the class in relation to efficacy and side
effects were considered to be generalisable to other members of that same class. For that
reason, trials with any interventions not licensed in the UK were included in the NMA to allow
the maximum use of available evidence and borrow strength from indirect evidence in the
network only if there was another trial that included UK licensed (for preterm labour or for
other conditions) interventions for the same class. The committee confirmed that they would
only consider for decision-making UK licensed medicines from each class depending on their
clinical and cost effectiveness analysis. Other considerations also affected the design of the
NMA:

173
Tocolysis
• The committee discussed that although dosage, mode of administration and

timing of treatment may influence the effectiveness of different tocolytics
interventions, it was considered unlikely for this factor to change the
direction of relative effect for the differentinterventions tested in the analysis.
Therefore the committee decided not to consider anyconfounding effect of
these factors in the NMA.
• Some of the included studies examined medicines that are not licensed as
tocolytics for use in pregnancy (including nylidrin and barusiban). These
medicines were included in theNMA to increase the size of the network, and
because it is not uncommon for medicines that are not licenced for pregnancy
indications to be prescribed for use in this context.
• It is anticipated that the economic analysis will be based on the lowest cost
treatment in aclass that the committee would be willing to recommend.
• The committee decided to have separate classes for alcohol/ethanol and
combination treatments (classed as ‘other’) in the new NMA. The only study
that considered the hormone human chorionic gonadotropin (HCG) was
excluded because it is not used in current practice and did not form a loop to
any other trials in the network. Both the new separate alcohol/ethanol and
‘others’ classes were removed from final ranking and healtheconomic
analysis as the committee did not aim to consider making recommendations
about these treatment options.
• Placebo was used as the reference treatment in the NMA as there is no
universally recognised ‘standard’ tocolytic. The effect in the placebo group is
used as an indicator forthe effect when a tocolytic has not been
administered.
Further details on the protocol for this review question are given in Appendix D.
10.2 Review question

What is the clinical and cost effectiveness of tocolytics given to women with suspected or
diagnosed preterm labour to improve outcomes:
• progesterone/progestogens
• beta-sympathomimetics
• oxytocin receptor antagonists
• calcium channel blockers
• cyclo-oxygenase enzyme inhibitors
• non-steroidal anti-inflammatory drugs
• magnesium sulfate?
10.3 Description of included evidence

Nine studies were included for this review question overall (Haas 2012, Houtzager 2006, Jaju
2011, Kashanian 2014, Klauser 2012, Klauser 2014, Nankali 2014, Nikbakht 2014, Salim
2012). One was a systematic review (SR) and NMA (Haas 2012) of 95 RCTs from a variety
of settings. The other 8 were individual primary RCTs; 1 from the Netherlands (Houtzager

174
Tocolysis
2006), 1 from India (Jaju 2011), 1 from Israel (Salim 2012), 2 from the USA (Klauser 2012,
Klauser 2014) and 3 from Iran (Kashanian 2014, Nankali 2014, Nikbakht 2014).
10.3.1 Summary of included studies

Details on the nature of included studies and their own characteristics are given in Table 43.
The mean gestational age of women across all included studies was 26 weeks and the
population was predominately women with no preterm premature rupture of membranes (P-
PROM). For full details of included studies, see the evidence tables in Appendix H.
Table 39:Summary of included studies

Type of study
design (sample Interventions (number
Included studies size) ofstudies) Outcomes
Haas 2012 SR and NMA of 95 • betamimetics (ritodrine, • delayed deliveryby
RCTs (n=10,860) terbutaline, nylidrin, 48 hours (primary
salbutamol, fenoterol, outcome)
hexoprenaline, • neonatal mortality
isoxsuprine)(60 RCTs)
• neonatal
• calcium channel blockers respiratory
(nifedipine, nicardipine) distress
(29RCTs) syndrome
• magnesium sulfate • maternal adverse
(29RCTs) events (all cause)
• nitrates (nitroglycerin,
nitricoxide) (4 RCTs)
• oxytocin receptor
blockers(atosiban,
barusiban) (13 RCTs)
• others (alcohol, human
chorionic gonadotropin)
(5RCTs)
• prostaglandin inhibitors
(indomethacin,
indomethacinplus alcohol*,
celecoxib, sulindac,
ketorolac, rofecoxib) (18
RCTs)
• placebo (25 RCTs)
Klauser 2012 RCT (n=301) • prostaglandin • neonatal adverse
inhibitors effects
Klauser 2014 (indomethacin)
(furtheranalysis • magnesium sulfate
of Klauser 2012) • calcium channel
blockers(nifedipine)
Houtzager 2006 Follow up study of • calcium channel • long-term
Paptsonis 1997 blockers(nifedipine) psychosocial and
and 2000 included • betamimetics (ritodrine) motor effects on
in Haas 2012 children
(n=102)

175
Tocolysis
Type of study
design (sample Interventions (number
Included studies size) ofstudies) Outcomes
Jaju 2011 RCT (n=210) • calcium channel • adverse events
• betamimetics (ritodrine)
Salim 2012 RCT (n=145) • calcium channel • adverse events
• oxytocin receptor
blockers(atosiban)
Kashanian 2014 RCT (n=120) • calcium channel
• nitrates (nitro-glycerine
[NG])
Nankali 2014 RCT (n=84) • nitrates (glyceryl
trinitrate [GTN])
• placebo
Nikbakht 2014 RCT (n=100) • magnesium sulfate • efficacy
nifedipine
• placebo
NMA network meta-analysis, RCT randomised controlled trial, SR systematic review
* Indomethacin plus alcohol was not detailed in the published paper but was included in Haas’s analysis and is
recorded here for completeness.
10.4 Introduction to the new network meta-analysis

A standard network meta-analysis (NMA) model in WinBUGS was carried out by the National
Institute for Health and Clinical Excellence (NICE) Technical Support Unit (TSU) at Bristol
University with the support of the technical team at National Collaborating Centre for
Women’s and Children’s Health (NCC-WCH).
The NMA for this review analysis was structured around the database of the published NMA
(Haas 2012) with the following changes:
• Addition of new data from the additional studies (Klauser 2012, Klauser 2014,
Jaju 2011, Salim 2012, Kashanian 2014, Nankali 2014, Nikbakht 2014) that were
not originally included in Haas 2012.
• Further data relating to additional outcomes prioritised by the committee were
extracted and added to the data set from 8 studies that had already been included
by Haas (2012).
• Alcohol and human chorionic gonadotropin (HCG) were separated as different
class interventions whereas these were placed in 1 class of medicines (‘other’
class) in Haas (2012).
• Three studies were removed from the original dataset for the following reasons:
o Grignaffini (2007) as it was an observational study
o Lorzadeh (2007) because the included intervention (human chorionic
gonadotropin [HCG]) is not used in current practice and it was only the only
study for that loop in the network
o Roy (1992) because it only reported results for 1 outcome of interest and for
that outcome no events were reported in 1 arm of the trial (see below).

176
Tocolysis
Following these changes, the final dataset from Haas (2012) NMA was based on 91 RCTs
and included data for 35 different medicines across 9 out of 10 classes of interest as follows:
• placebo (placebo or usual or standard care without a tocolytic medicine)

• beta mimetics (ritodrine, terbutaline, nylidrin, salbutamol, fenoterol, hexoprenaline,
isoxsuprine)
• calcium channel blockers (nifedipine, nicardipine)
• magnesium sulfate
• nitrates (nitroglycerin, nitric oxide)
• oxytocin receptor blockers (atosiban, barusiban)
• others (treatments defined as ‘tocolysis’ and ‘other tocolytics’ by study authors)
• prostaglandin inhibitors (indomethacin, celecoxib, sulindac, ketorolac, rofecoxib)
• alcohol/ethanol.
The available data allowed for an NMA to be undertaken for 8 out of the 13 outcomes
prioritised by the committee as follows (separate NMAs were carried out for each outcome):
• RDS
• IVH
• mothers with adverse events sufficient to require cessation of treatment
• neonatal sepsis
Limited data were available for 4 out of the 5 remaining outcomes and were analysed
using conventional pair-wise meta-analysis (see below).
10.4.1 Methods
An NMA class model was used to estimate the relative effects of each treatment class
compared with placebo/control. Since there was no evidence of within-class variability (see
Appendix J) for any of the outcomes considered, all the results presented assume that all
treatments in a class have the same relative effect.
A binomial/logit model was used to model all outcomes other than gestational age at birth,
and a normal model with identity link was used to model estimated gestational age (EGA).
The final dataset consisted of data from 93 trials comparing 35 treatments, although not all
trials report all the outcomes of interest. Studies reporting zero events on all arms were
removed from the NMA as they do not contribute information on the relative treatment
effects. A Bayesian framework is used to estimate all parameters, using Markov chain Monte
Carlo simulation methods implemented in WinBUGS 1.4.3. (WinBUGS is the software used
for performing the computation). Under this framework, and unlike in standard meta-analysis
packages, it is not necessary to add a continuity correction (add 0.5 to arms of studies that
report zero events in 1 arm). For detailed description of methods (baseline variability, relative
effects model, NMA model for binary and continuous data) and sample WinBUGS code see
Appendix J.

177
Tocolysis
10.4.2 Limitations in the data

Most comparisons were only made in 1 or 2 trials. Furthermore, not all trials report all
outcomes. Some networks were very sparse in terms of patient numbers contributing to each
loop.
Because some studies included multiple births, allowing more than 1 infant per mother, it
was not always clear which was the most appropriate number of individuals to consider for
outcomes on the infant. Where available we used the number of infants as the denominator.
Although this does not account for the expected correlation in outcomes of infants from the
same mother, it prevents double counting of infants from the same mother who may both
have had an event.
10.5 Introduction to pair-wise meta-analysis

There was insufficient data to undertake NMA for the other outcomes set up in the review
protocol. Limited data allowed for conventional meta-analysis to be undertaken for the
following outcomes:
• neurodevelopmental disability (developmental delay, intellectual, gross motor,

visual or hearing impairment, learning difficulties) – results for this outcome were
derived from Houtzager (2006)
• periventricular leucomalacia (PVL) – results for this outcome were derived from
studies included in Haas (2012)
• chronic lung disease (CLD) – results for this outcome were derived from studies
included in Haas (2012) and have been presented separately
• maternal infection – results for this outcome were derived from studies included in
Haas (2012) and have been presented separately
• three studies reported maternal mortality; however, all 3 studies reported zero
events in each arm and so no further analysis was possible.

Firstly, the results of the updated NMA are presented by outcome (Sections 10.6.1 to 10.6.8).
For each outcome included in the NMA, the following information is presented:
• a text description of the studies included in the outcome-specific network

• a network diagram providing a graphic representation of the treatments compared
in the outcome-specific network and the ‘weight’ of the data in terms of the
number of included studies and number of participants
• a table summarising the relative effects derived from the direct and the overall
NMA for each outcome
• a table summarising the probability rankings of effectiveness for the medicine
classes included in the NMA, excluding the classes of ‘alcohol/ethanol’ and
‘others’
• a set of rankograms that provide a graphic representation of the probability
rankings of effectiveness for the medicine classes included in the NMA

178
Tocolysis
• a modified GRADE profile summarising the quality assessment of the studies

included in the outcomes-specific network.
Evidence for the pairwise comparisons for the outcomes of neurodevelopmental
disability, PVL maternal infection and CLD is presented by comparison in GRADE
profiles.
A full description of the characteristics and results of the included studies can be
found in the evidence tables in Appendix H.
10.6.1 Neonatal mortality

Out of the 98 studies included in the review, 59 studies reported neonatal mortality as an
outcome:
• 9 studies observed no events and were removed
• 1 study only reported events in 1 arm and was also removed.
The remaining 49 studies (Adam 1966, Al-Omari 2006, Besinger 1991, Bisits 2004, Cararach
2006, Caritis 1984, Cotton 1984, Cox 1990, CPLIG 1992, Essed 1978, European 2001, Fan
2003, French/Australian 2001, Glock 1993, Goodwin 1996, Holleboom 1996, Kashanian
2011, Klauser 2012, Koks 1998, Kupferminc 1993, Kurki 1991, Laohapojanart 2007, Larson
1980, Lauersen 1977, Leveno 1986, Lyell 2007, Ma 1992, Maitra 2007, McWhorter 2004,
Merkatz 1980, Mittendorf MAGnet 2002, Morales 1989, Morales 1993, Moutquin 2000,
Nassar 2009, Niebyl 1980, Panter 1999, Papatsonis 1997/ 2000, Parilla 1997, Rayamajhi
2003, Romero 2000, Shim 2006, Spearing 1979, Spellacy 1979, Surichamorn 2001,
VandeWater 2008, Weiner 1988, Zhu 1996, Zuckerman 1984) examined 19 medicine
treatments allowing for 9 out of the 9 treatment classes to be assessed against each other.
Figure 3 shows the results of this assessment. The lines represent trials comparing 2 classes
of medicine. The thickness of the lines is proportional to the number of studies contributing to
the comparison. The size of the dots is proportional to the number of participants randomised
to the treatment.
Figure 3: Graphic representation of tocolytics trials for the NMA for neonatal mortality

179
Tocolysis
In Table 44 the values shown in the upper diagonal (cells above those shaded dark grey) are the log odds-ratios (ORs) for the columns header
versus the row header and are derived from the NMA. Given that this table relates to an adverse outcome, values lower than 1 favour the
column defining treatment and values higher than 1 favour the row defining the treatment. The data in the upper diagonal is also presented in
the forest plots in Appendix I.
The values shown in the lower diagonal (cells below those shaded dark grey) are the log ORs for the row header versus the column header and
are derived from the direct comparison analysis. Given that this table relates to an adverse outcome, values higher than 1 favour the column
defining treatment and values lower than 1 favour the row defining the treatment.
Table 40: Posterior median of the log odds ratios (OR) and 95% credible intervals (CrI) for neonatal mortality
Calcium Oxytocin
Placebo/ Prostaglandin Magnesium channel receptor Alcohol/ Other
control inhibitors sulfate Betamimetics blockers Nitrates blockers ethanol treatments
Placebo/control 1.13 (0.39 to 1.49 (0.56 to 1.02 (0.49 to 0.62 (0.21 to 0.98 (0.02 to 0.73 (0.23 to 2.33 (0.41 to 0.56 (0.11 to
3.40) 4.09) 2.15) 1.80) 62.47) 2.19) 13.99) 2.60)
Prostaglandin 1.08 (0.15 to 1.32 (0.45 to 0.90 (0.32 to 0.55 (0.16 to 0.86 (0.01 to 0.64 (0.16 to 2.06 (0.29 to 0.49 (0.08 to
inhibitors 7.80) 3.81) 2.41) 1.70) 55.83) 2.37) 13.93) 2.74)
Magnesium sulfate 1.18 (0.23 to 1.42 (0.35 to 0.68 (0.26 to 0.42 (0.13 to 0.65 (0.01 to 0.49 (0.13 to 1.56 (0.24 to 0.37 (0.06 to
5.58) 7.13) 1.75) 1.23 42.22) 1.72) 10.17) 1.83)
Betamimetics 0.79 (0.31 to 1.15 (0.21 to 0.91 (0.14 to 0.61 (0.25 to 0.96 (0.02 to 0.71 (0.26 to 2.28 (0.44 to 0.55 (0.10 to
1.97) 6.02) 6.53) 1.43) 56.79) 1.83) 12.21) 2.56)
Calcium channel - 0.25 (0.03 to 2.33 (1.27 to 0.59 (1.18 to 1.57 (0.02 to 1.17 (0.36 to 3.74 (0.60 to 0.89 (0.15 to
blockers 1.62) 4.87) 1.74) 106.20) 3.84) 25.10) 5.07)
Nitrates 0 cell - - 0.95 (0.02 to - 0.74 (0.01 to 2.39 (0.03 to 0.57 (0.01 to
58.44) 49.77) 192.10) 43.68)
Oxytocin receptor 4.98 (0.45 to - - 0.43 (0.13 to 1.15 (0.11 to - 3.21 (0.49 to 0.76 (0.12 to
blockers 74.44) 1.29) 12.54) 22.37) 4.70)
Alcohol/ethanol - - - 3.74 (0.63 to - - - 0.24 (0.03 to
26.36) 1.65)
Other treatments 0.69 (0.06 to - 0 cell 2.97 (0.18 to - - - -
8.18) 58.38)
Tocolysis
Table 45 shows the probability rankings for medicines for neonatal mortality, ranking them in
order of best medicine classes for improving the outcome. Rows are arranged in the
decreasing order of estimate effect with the best treatment at the top and the worst at the
bottom of the table.
Table 41: Probability rankings for medicines for neonatal mortality

Probability of being
the best treatment Rank 95%
option to improve credible
Class the outcome Mean rank Median rank interval
Nitrates 36% 4.0 4 (1 to 7)
Calcium channel 32% 2.3 2 (1 to 6)
blockers
Oxytocin receptor 21% 2.9 3 (1 to 7)
blockers
Placebo/control 4% 4.1 4 (1 to 7)
Prostaglandin 4% 4.6 5 (1 to 7)
inhibitors
Betamimetics 1% 4.3 4 (2 to 7)
Magnesium sulfate 1% 5.7 6 (2 to 7)

181
Tocolysis
Figure 4 shows a graphic representation of each medicine’s effectiveness probability ranking

for reducing neonatal death. In these rankograms the numbers on the y axis indicate
probability and the numbers on the x axis show the potential effectiveness ranks from most
effective (1) to least effective (7). The line indicates the probability that the medicine will
achieve each rank; for example placebo/control has a 0.01 probability of being ranked most
effective and 0.2 probability of being ranked fifth most effective.
Figure 4: Graphic representation of each medicine’s effectiveness probability

ranking for reducing neonatal death

182
Tocolysis
Table 42: Quality assessment of the evidence contributing to the NMA for neonatal
death
Quality assessment
Number of Design Risk of bias Inconsisten Indirectnes Imprecision Other Quality
studies cy s considerati
ons
Respiratory distress syndrome
1 NMA of 49 49 RCTsa Serious1 Serious2 Serious3 Serious4 None Very low
studies
(Original
data from
Haas 2012)
CrI credible interval, NMA network meta-analysis, OR odds ratio, p probability, RCT randomised controlled trial,
RDS respiratory distress syndrome
a. Cotton 1984, Klauser 2012, Goodwin,1996, Niebyl 1980, Panter 1999, Zuckerman 1984, Cox 1990, Spellacy
1979, Merkatz 1980, Leveno 1986, CPLIG 1992, Romero 2000, Weiner 1988, Morales 1993, Parilla 1997,
Morales 1989, Kurki 1991, McWhorter 2004, Lyell 2007, Essed 1978, Holleboom,1996, Caritis 1984, Maitra 2007,
Cararach 2006, VandeWater 2008, Papatsonis 1997/ 2000, Shim 2006, Moutquin 2000, Lauersen 1977,
French/Australian 2001, Laohapojanart 2007, European 2001, Nassar 2009, Al-Omari 2006, Koks 1998,
Kupferminc 1993, Fan 2003, Rayamajhi 2003, Bisits 2004, Mittendorf MAGnet 2002, Glock 1993, Surichamorn
2001, Zhu 1996, Kashanian 2011, Besinger 1991, Larson 1980, Adam 1966, Ma 1992, Spearing 1979.
1. Analysis was based on the class therefore different doses and co-treatment were combined together
2. There were some evidence of inconsistency (conflict between direct and indirect evidence) in comparisons of
placebo/control v oxytocin receptor blockers:
Bayesian p-value = 0.022
Direct O =4.95 (95% CrI 0.83 to 40.45)
Indirect OR=0.44 (95% CrI 0.17 to 1.14)
3. Women with multiple pregnancy were included in 23/59 studies
4. Wide and very wide CrI across all comparisons

183
Tocolysis
10.6.2 Perinatal mortality

Out of the 98 studies included in the review, 47 studies reported perinatal mortality as an
outcome. Three studies observed no events and were removed. The remaining 44 studies
examined 20 medicine treatments allowing for 9 out of the 9 treatment classes to be
assessed against each other.
Figure 5 shows a graphic representation of tocolytics trials for the NMA for perinatal
mortality. Lines represent trials comparing 2 classes of medicine. The thickness of the lines
is proportional to the number of studies contributing to the comparison. The size of the dots
is proportional to the number of participants randomised to the treatment.
Figure 5: Graphic representation of tocolytics trials for the NMA for perinatal mortality.

184
Tocolysis
Table 47 shows the posterior median of the log ORs and 95% CrIs for perinatal death.
Values shown in the upper diagonal are the log ORs for the column header versus the row header and are derived from the NMA. Given that
this table relates to an adverse outcome, values lower than 1 favour the column defining treatment and values higher than 1 favour the row
defining the treatment. Upper diagonal data is also presented in the forest plots in Appendix I.
Values in the lower diagonal are the log ORs for the row header versus the column header and are derived from the direct comparison analysis.
Given that this table relates to an adverse outcome, values higher than 1 favour the column defining treatment and values lower than 1 favour the
row defining the treatment.
Table 43: Posterior median of the log odds ratios (OR) and 95% credible intervals (CrI) for perinatal death
Calcium Oxytocin
Placebo/control 0.72 (0.22 to 1.19 (0.35 to 1.01 (0.48 to 0.76 (0.25 to 0.10 (0.00 to 0.86 (0.25 to 2.59 (0.50 to 2.00 (0.41 to
2.28) 3.73) 1.99) 2.24) 1.07) 2.59) 13.84) 9.74)
Prostaglandin 0.78 (0.13 to 1.65 (0.44 to 1.40 (0.43 to 1.07 (0.25 to 0.14 (0.00 to 1.19 (0.25 to 3.59 (0.54 to 2.80 (0.41 to
inhibitors 4.96) 6.34) 4.49) 4.34) 1.86) 5.39) 25.48) 18.54)
Magnesium sulfate 2.05 (0.36 to 2.97 (0.46 to 0.85 (0.28 to 0.64 (0.18 to 0.08 (0.00 to 0.72 (0.16 to 2.20 (0.34 to 1.67 (0.26 to
10.82) 28.36) 2.736) 2.48 1.14) 3.29) 15.72) 11.72)
Betamimetics 0.89 (0.35 to 0.93 (0.14 to 2.62 (0.13 to 0.75 (0.31 to 0.10 (0.00 to 0.85 (0.28 to 2.56 (0.57 to 1.98 (0.42 to
1.87) 5.63) 120.7) 1.83) 1.05) 2.42) 12.94) 9.75)
Calcium channel - - 3.99 (0.32 to 0.61 (0.23 to 0.13 (0.00 to 1.13 (0.27 to 3.41 (0.60 to 2.63 (0.43 to
blockers 143.2) 1.51) 1.65) 4.35) 21.24) 16.10)
Nitrates 0 cell - - 0.24 (0.01 to - 8.42 (0.65 to 26.41 (1.57 to 20.1 (1.23 to
3.80) 308.20) 1163.00) 874.00)
Oxytocin receptor 2.44 (0.36 to - - 0.51 (0.14 to - - 3.03 (0.50 to 2.32 (0.37 to
blockers 16.93) 1.78) 21.71) 16.03)
Alcohol/ethanol - - - 3.70 (0.76 to - - - 0.77 (0.13 to
21.61) 4.53)
Other treatments 0.70 (0.07 to - - 5.12 (0.53 to - - - -
6.37) 53.36)
Tocolysis
Tocolysis
Table 48 shows probability rankings for medicines for perinatal mortality, ranking probability
of best medicine classes for improving the outcome. Rows are arranged in the decreasing
order of estimate effect, with the best treatment at the top and the worst at the bottom.
Table 44: Probability rankings for medicines for perinatal mortality

Probability of being the
best treatment option to Rank 95% credible
Class improve the outcome Mean rank Median rank interval
Nitrates 89% 1.3 1 (1 to 5)
inhibitors
blockers
blockers
Magnesium sulfate 1% 5 6 (2 to 7)
Betamimetics 0% 5 5 (2 to 7)
Figure 6 shows a graphic representation of each medicine’s effectiveness probability ranking

for reducing perinatal mortality. The numbers on the y axis indicate probability while the
numbers on the x axis show the potential effectiveness ranks from most effective (1) to least
effective (7). The line indicates the probability that medicine will achieve each rank.

187
Tocolysis
Figure 6: Graphic representation of the each medicine’s effectiveness probability

ranking for reducing perinatal mortality

188
Tocolysis
Table 45: Quality assessment of the evidence contributing to the network analysis for
perinatal death
Quality assessment
ons
RDS
1 NMA of 49 44 RCTsa Sserious1 No serious Serious2 Serious3 None Very low
studies inconsistenc
(Original y
data from
Haas 2012)
CrI credible interval, NMA network meta-analysis, RCT randomised controlled trial
a. Cotton 1984, Niebyl 1980, Panter 1999, Zuckerman 1984, Cox 1990, Spellacy 1979, Leveno 1986, CPLIG
1992, Romero 2000, Weiner 1988, Morales 1993, Parilla 1997, Morales 1989, Kurki 1991, McWhorter 2004,
Lyell2007, Essed 1978, Holleboom 1996, Caritis 1984, Cararach 2006, VandeWater 2008, Papatsonis
1997/2000, Shim 2006, Moutquin 2000, Lauersen 1977, French/Australian 2001, European 2001, Koks 1998,
Fan 2003, Rayamajhi 2003, Bisits 2004, Glock 1993, Besinger 1991, Larson 1980, Larson 1986, Smith 2007 ,
Floyd 1995, Gummerus 1983, Sirohiwal 2001, Trabelsi 2008, Adam 1966, Spearing 1979, Jaju 2011
3. Wide and very wide CrI across all comparisons

189
Tocolysis
10.6.3 Delay of birth by more than 48 hours

Out of the 98 studies included in the review, 69 studies reported delay of birth by more than
48 hours as an outcome. Two studies observed same events rate in both arms and were
removed. The reminding 67 studies examined 26 medicine treatments allowing for 9 out of
the 9 treatment classes to be assessed against each other.
Figure 7 shows a graphic representation of tocolytics trials for the NMA for delay by 48
hours. Lines represent trials comparing 2 classes of medicine. The thickness of the lines is
proportional to the number of studies contributing to the comparison. The size of the dots is
proportional to the number of participants randomised to the treatment.
Figure 7: Graphic representation of tocolytics trials for the NMA for delay by 48
hours

190
Tocolysis
Table 50 shows the posterior median of the ORs and 95% CrIs for delay of birth by more than 48 hours.
In the upper diagonal values shown are the ORs for the column headers versus the row headers and are derived from the NMA. Given that this
table relates to a positive outcome, values higher than 1 favour the column defining the treatment and values lower than 1 favour the row
defining the treatment. Upper diagonal data is also presented in the forest plots in Appendix I.
In the lower diagonal values shown are the ORs for the row headers versus the column headers and are derived from the direct comparison
analysis. Given that this table relates to a positive outcome, values lower than 1 favour the column defining treatment and values higher than 1
favour the row defining the treatment.
Table 46: Posterior median of the odds ratios (OR) and 95% credible intervals (CrI) for delay birth by more than 48 hours
Calcium Oxytocin
Placebo/control 3.14 2.10 2.04 2.02 0.89 1.93 0.83 1.10
(1.45,7.05) (1.10,4.07) (1.17,3.59) (1.11,3.76) (0.40,2.02) (1.02,3.65) (0.12,5.64) (0.38,3.24)
Prostaglandin 14.51 0.67 0.65 0.64 0.28 0.61 0.26 0.35
inhibitors (2.87,86.23) (0.33,1.33) (0.32,1.29) (0.31,1.30) (0.10,0.78) (0.26,1.40) (0.04,1.89) (0.10,1.19)
Magnesium sulfate 2.65 0.88 0.97 0.96 0.43 0.92 0.40 0.53
(0.91,7.85) (0.36,2.10) (0.57,1.66) (0.56,1.65) (0.17,1.07) (0.45,1.86) (0.06,2.73) (0.17,1.65)
Betamimetics 2.68 0.32 0.90 0.99 0.44 0.95 0.41 0.54
(1.22,6.15) (0.09,1.12) (0.41,1.98) (0.65,1.50) (0.19,1.01) (0.54,1.63) (0.06,2.63) (0.18,1.57)
Calcium channel 1.68 2.06 1.24 0.91 0.44 0.96 0.41 0.54
blockers (0.30,9.26) (0.63,6.82) (0.56,2.73) (0.56,1.49) (0.19,1.03) (0.52,1.74) (0.06,2.73) (0.18,1.63)
Nitrates 0.35 - - 0.74 1.85 2.16 0.93 1.23
(0.13,0.93) (0.15,3.62) (0.46,7.27) (0.85,5.47) (0.12,7.02) (0.34,4.52)
Oxytocin receptor 1.51 - - 1.02 1.06 - 0.43 0.57
blockers (0.70,3.15) (0.50,2.06) (0.36,3.13) (0.06,2.96) (0.19,1.78)
Alcohol/ethanol - - - 0.38 - - - 1.33
(0.05,2.74) (0.21,8.59)
Other treatments 1.12 - - 0.48 - - - -
(0.33,3.90) (0.06,3.56)
Tocolysis
Tocolysis
Table 51 shows probability rankings for medicines by class to increase delay of birth by more
than 48 hours, ranking probability of best medicine classes for improving the outcome. Rows
arranged in the decreasing order of estimate effect with best treatment at the top and the
worst at the bottom of the table.
Table 47: Probability rankings for medicines by class to increase delay of birth by
greater than 48 hours
the best treatment
option to improve Rank 95% credible
inhibitors
blockers
blockers
Nitrates 0% 6.46 7 (4 to 7)
Figure 8 shows a representation of each medicine’s effectiveness probability ranking for

improving outcome. The numbers on the y axis indicate probability. The numbers on the x
axis show the potential effectiveness ranks from most effective (1) to least effective (7).

193
Tocolysis

ranking for improving outcome

194
Tocolysis
Table 48: Quality assessment of the evidence contributing to the outcome of delay of
birth by more than 48 hours
Quality assessment
ons
RDS
studies
(Original
data from
Haas 2012)
CrI credible interval, NMA network meta-analysis, OR odds ratio, p probability, RCT randomised controlled trial
a. Thornton 2009, Cotton 1984, Goodwin,1996, Panter 1999, Zuckerman 1984, Cox 1990, Larsen 1986, CPLIG
1992, Goodwin 1994, Romero 2000, Weiner 1988, Morales 1993, Morales 1989, Kurki 1991, Lyell 2007, Maitra
2007, Cararach 2006, VandeWater 2008, Shim 2006, Lin 2009, Moutquin 2000, Trabelsi 2008, French/Australian
2001, Laohapojanart 2007, European 2001, Nassar 2009, Al-Omari 2006, Ma 1992, Spearing 1979, Husslein
2007, Larson 1980, Ingemarsson 1976, Koks 1998, Smith 2007, Borna 2007, Kashanian 2011, Besinger 1991,
McWhorter 2004, Larmon 1999, Papatsonis 1997, Papatsonis 2000, Al-Qattan 2000, Kupferminc 1993, Fan 2003:
Rayamajhi 2003, Bisits 2004, Holleboom,1996, Kashanian 2011, Amorim 2009, Weerakul 2002, Mawaldi 2008,
Motazedian 2010, Beall 1985, Haghighi 1999, Taherian 2007, Chau 1992, Surichamorn 2011, Aramayo
1990, Wilkins 1988, Tchilinguirian 1984, Garcia-Velasco 1998, Jaju 2011, Nankali 2014, Kashanian 2014, Salim
2012, Nikbakhat 2014.
2 There were some evidence of inconsistency (conflict between direct and indirect evidence) in comparisons of
placebo/control vspProstaglandin inhibitors
Direct OR=14.51 (95% CrI 2.87 to 86.23) 2 small studies compare these classes directly
The direct and indirect are not contradictory to each other (both on the same side of one)
Strong evidence of inconsistency (conflict between direct and indirect evidence) in comparisons of placebo/control
v nitrates
Direct OR=0.35 (95% CrI 0.13 to 0.93) 2 medium studies compare these classes directly
Some evidence of inconsistency (conflict between direct and indirect evidence) in comparisons of prostaglandin
inhibitors v calcium channel blockers
The direct and indirect are contradictory to each other (both on opposite sides of one). Results from this network
should therefore be treated with caution.
Some evidence of inconsistency (conflict between direct and indirect evidence) in comparisons of calciumchannel
blockers v nitrates
The direct and indirect are contradictory to each other (both on opposite sides of one). Results from this
network should therefore be treated with caution.
4. Wide and very wide CrIs across all comparisons except one direct comparison (placebo/control vs
prostaglandin inhibitors)

195
Tocolysis
10.6.4 Neonatal sepsis

Out of the 98 studies included in the review, 22 studies reported neonatal sepsis as an
outcome. Three studies observed no events and were removed. The remaining 19 studies
examined 12 medicines allowing for 7 out of the 9 treatment classes to be assessed against
each other.
Figure 9 shows a graphic representation of tocolytics trials for the NMA for neonatal sepsis.
Lines represent trials comparing 2 classes of medicine. The thickness of the lines is
proportional to the number of studies contributing to the comparison. The size of the dots is
proportional to the number of participants randomised to the treatment
Figure 9: Graphic representation of tocolytics trials for the NMA for neonatal sepsis

196
Tocolysis
Table 53 shows the posterior median of the ORs and 95% CrIs for neonatal sepsis.
table relates to an adverse outcome, values lower than 1 favour the column defining the treatment and values higher than 1 favour the row defining
the treatment. Upper diagonal data is also presented in the forest plots in Appendix I.
analysis. Given that this table relates to an adverse outcome, values higher than 1 favour the column defining treatment and values lower than 1
Table 49: Posterior median of the odds ratios (OR) and 95% credible intervals (CrI) for neonatal sepsis
Prostaglandin Calcium channel Oxytocin receptor
Placebo/control inhibitors Magnesium sulfate Betamimetics blockers blockers Other treatments
Placebo/control 1.59 (0.33,9.33) 1.93 (0.43,10.77) 1.15 (0.25,6.56) 0.83 (0.18,4.75) 1.16 (0.22,7.15) 1.31 (0.21,8.05)
Prostaglandin 0.00 (0.00,0.52) 1.21 (0.63,2.37) 0.72 (0.29,1.77) 0.52 (0.23,1.14) 0.73 (0.25,2.10) 0.81 (0.07,8.96)
inhibitors
Magnesium sulfate 7.91 (1.42,69.55) 0.96 (0.47,1.94) 0.59 (0.26,1.35) 0.43 (0.21,0.86) 0.60 (0.22,1.62) 0.67 (0.06,7.04)
Betamimetics 0.00 (0.00,0.18) 1.00 (0.02,39.17) 0.72 (0.42,1.23) 1.01 (0.55,1.87) 1.13 (0.09,12.27)
Calcium channel 0.68 (0.28,1.57) 0.54 (0.10,2.36) 0.54 (0.29,0.98) 1.40 (0.65,3.03) 1.56 (0.13,16.94)
blockers
Oxytocin receptor 1.06 (0.56,2.05) 0.96 (0.16,5.80) 1.12 (0.09,12.97)
blockers
Other treatments 1.30 (0.21,8.09)
Tocolysis
Table 54 shows probability rankings for medicines by class to reduce neonatal sepsis,
ranking probability of best medicine classes for improving the outcome. Rows are arranged
in the decreasing order of estimate effect with best treatment at the top and the worst at the
bottom of the table.
Table 50: Probability rankings for medicines by class to reduce neonatal sepsis
the best treatment
blockers
blockers
inhibitors
Figure 10 shows a representation of each medicine’s effectiveness probability ranking for

improving outcomes. The numbers on the y axis indicate probability. The numbers on the x
axis show the potential effectiveness ranks from most effective (1) to least effective (6). The
line indicates the probability that medicine will achieve each rank.

198
Tocolysis

Table 51: Quality assessment of the evidence contributing to the outcome of neonatal
sepsis
Quality assessment
ons
Neonatal sepsis
1 NMA of 19 19 RCTsa Serious1 No serious Serious2 Serious3 None Very low
(Original y
data from
Haas 2012)
CrI credible interval, NMA network meta-analysis, OR odds ratio, RCT randomised controlled trial
a. Cotton 1984, Klauser 2012, Goodwin,1996, Niebyl 1980, Weiner 1988, Stika 2002, Al-Omari 2006, Kurki 1991,
McWhorter 2004, Lyell 2007, Holleboom 1996, Maitra 2007, VandeWater 2008, Papatsonis 1997/2000, Moutquin
2000, French/Australian 2001, European 2001, Nassar 2009, Salim 2012.

199
Tocolysis
2. Women with multiple pregnancy were included in 7/19 studies 4 wide and very wide CrI across all comparisons
except on placebo/control vs magnesium sulfate
3. There were strong evidence of inconsistency (conflict between direct and indirect evidence) in comparisons of
magnesium sulfate vs calcium channel blockers
Some evidence of inconsistency (conflict between direct and indirect evidence) in comparisons of betamimetics
vs calcium channel blockers
Bayesian p-value=0.025
The direct and indirect are not contradictory to each other (both on the same side of 1)

200
Tocolysis
10.6.5 Intraventricular haemorrhage

Out of the 98 studies included in the review, 34 studies reported IVH as an outcome. Four
studies observed no events and were removed. The remaining 30 studies examined 14
medicines allowing for 8 out of the 9 treatment classes to be assessed against each other.
Figure 12 shows a graphic representation of tocolytics trials for the NMA for IVH. Lines
represent trials comparing 2 classes of medicine. The thickness of the lines is proportional to
the number of studies contributing to the comparison. The size of the dots is proportional to
the number of participants randomised to the treatment.
Figure 11
Figure 12: Graphic representation of tocolytics trials for the NMA for IVH

201
Tocolysis
Table 56 shows the posterior median of the ORs and 95% CrIs for intraventricular haemorrhage. In the upper diagonal values shown are the ORs
for the column headers versus the row headers and are derived from the NMA. Given that this table relates to an adverse outcome, values lower
than 1 favour the column defining treatment and values higher than 1 favour the row defining the treatment. Upper diagonal data is also presented
in the forest plots in Appendix I.
Table 52 :Posterior median of the odds ratios (OR) and 95% credible intervals (CrI) for intraventricular haemorrhage
Oxytocin
Prostaglandin Magnesium Calcium channel receptor
Placebo/control inhibitors sulfate Betamimetics blockers Nitrates blockers Other treatments
Placebo/control 0.76 (0.35 to 0.69 (0.33 to 0.79 (0.51 to 0.40 (0.21 to 0.34 (0.08 to 0.82 (0.48 to 0.14 (0.02 to
1.59) 1.43) 1.22) 0.74) 1.13) 1.37) 0.77)
Prostaglandin - 0.91 (0.54 to 1.05 (0.33 to 0.53 (0.27 to 0.45 (0.10 to 1.08 (0.48 to 0.19 (0.02 to
inhibitors 1.54) 2.06) 1.01) 1.72) 2.44) 0.94)
Magnesium sulfate 0.76 (0.22 to 0.91 (0.52 to 1.15 (0.58 to 0.58 (0.30 to 0.49 (0.11 to 1.19 (0.53 to 0.21 (0.03 to
2.40) 1.60) 2.29) 1.11) 1.89) 2.66) 0.95)
Betamimetics 0.66 (0.39 to 1.08 (0.32 to - 0.50 (0.30 to 0.43 (0.11 to 1.03 (0.63 to 0.18 (0.02 to
1.11) 3.66) 0.83) 1.39) 1.71) 0.96)
Calcium channel - 0.60 (0.25 to 0.59 (0.07 to 0.44 (0.24 to 0.85 (0.20 to 2.06 (1.04 to 0.36 (0.04 to
blockers 1.34) 3.95) 0.79) 3.05) 4.08) 1.88)
Nitrates 2.55 (0.19 to - - 0.20 (0.03 to - 2.43 (0.68 to 0.42 (0.04 to
81.45) 0.86) 10.28) 3.66)
Oxytocin receptor 0.84 (0.41 to - - 1.06 (0.55 to - - 0.17 (0.02 to
blockers 1.70) 2.08) 0.98)
Other treatments - - 0.21 (0.03 to - - - -
0.95)
Tocolysis
Tocolysis
Table 57 shows probability rankings for medicines by class to reduce IVH, ranking probability
of best medicine classes for improving the outcome. Rows are arranged in the decreasing
order of estimate effect with best treatment at the top and the worst at the bottom of the
table.
Table 53: Probability rankings for medicines by class to reduce IVH

the best treatment
Nitrates 59% 1.9 1 (1 to 7)
blockers
Prostaglandin inhibitor 1% 4.6 4 (2 to 7)
blockers
Figure 13 shows a graphic representation of the each medicine’s effectiveness probability

ranking for improving outcome. The numbers on the y axis indicate probability. The numbers
on the x axis show the potential effectiveness ranks from most effective (1) to least effective
(7). The line indicates the probability that medicine will achieve each rank.

204
Tocolysis

Table 54: Quality assessment of the evidence contributing to the outcome of IVH
Quality assessment
ons
IVH
(Original y
data from
Haas 2012)
a. Cotton 1984, Klauser 2012, Goodwin 1996, Panter 1999, Cox 1990, Leveno 1986, CPLIG 1992, Smith 2007,
Romero 2000, Morales 1993, Parilla 1997, Morales 1989, Besinger 1991, Kurki 1991, Schorr 1998, McWhorter
2004, Lyell 2007, Mittendorf MAGnet 2002, Bisits 2004, Holleboom 1996, Maitra 2007, VandeWater 2008,
Papatsonis 1997, Papatsonis 2000, Shim 2006, Moutquin 2000, French/Australian 2001, Laohapojanart 2007,
European 2001, Nassar 2009, Salim 2012
3.Wide and very wide CrI across all comparisons

205
Tocolysis

206
Tocolysis
10.6.6 Discontinuation of treatment due to maternal adverse events

Out of the 98 studies included in the review, 62 studies reported neonatal sepsis as an
outcome. Twenty studies observed no events and were removed. Six studies only reported
events on 1 arm and were also removed. The remaining 36 studies examined 6 out of the 9
treatment classes to be assessed against each other.
Figure 14 shows a graphic representation of tocolytics trials for the NMA for treatment
discontinued due to maternal side effect. Lines represent trials comparing 2 classes of
medicine. The thickness of the lines is proportional to the number of studies contributing to
the comparison. The size of the dots is proportional to the number of participants randomised
to the treatment.
Figure 14: Graphic representation of tocolytics trials for the NMA for
treatment discontinued due to maternal side effect.

207
Tocolysis
Table 59 shows the posterior median of the odds ratios (OR) and 95% credible intervals (CrI) for discontinuation of treatment due to maternal
adverse events. In the upper diagonal values shown are the ORs for the column headers versus the row headers and are derived from the NMA.
Given that this table relates to an adverse outcome, values lower than 1 favour the column defining treatment and values higher than 1 favour the
row defining the treatment. Upper diagonal data is also presented in the forest plots in Appendix I.
In the lower diagonal values shown are the OR for the row headers versus the column headers and are derived from the direct comparison
Table 55: Posterior median of the odds ratios (OR) and 95% credible intervals (CrI) for discontinuation of treatment due to maternal
adverse events
Calcium channel Oxytocin receptor
Placebo/control Magnesium sulfate Betamimetics blockers Nitrates blockers
Placebo/control 16.21 (1.89 to 175.70) 132.20 (18.52 to 5.22 (0.35 to 56.55) 5.58 (0.26 to 165.00) 3.15 (0.31 to 23.18)
1284.00)
Magnesium sulfate Zero cell 8.05 (2.23 to 34.05) 0.32 (0.04 to 1.39) 0.34 (0.01 to 9.19) 0.19 (0.02 to 1.15)
Betamimetics 109.84 (2.67 to 8.82 (1.01 to 90.65) 0.04 (0.01 to 0.14) 0.02 (0.00 to 1.20) 0.02 (0.00 to 0.09)
23623.56)
Calcium channel - 0.37 (0.01 to 12.76) 0.02 (0.00 to 0.15) 1.10 (0.04 to 48.83) 0.59 (0.07 to 5.25)
blockers
Nitrates 2.69 (0.01 to 817.29) 0.86 (0.00 to 323.11) - - 0.54 (0.01 to 14.60)
Oxytocin receptor 4.78 (0.04 to 601.24) - 0.01 (0.00 to 0.09) - -
blockers
Tocolysis
Table 60 shows probability rankings for medicines by class for discontinuation of medicine
because of maternal side effect. Rows are arranged in the decreasing order of estimate
effect with the best treatment at the top and the worst at the bottom of the table.
Table 56: Probability rankings for medicines by class for discontinuation of medicine
because of maternal side effect.
the best treatment
Nitrates 11% 3.3 3 (1 to 6)
blockers
blockers

(6). The line indicates the probability that medicine will achieve each rank

209
Tocolysis
Figure 15: Graphic representation of the each medicine’s

effectiveness probability ranking for improving
outcome
Table 57: Quality assessment of the evidence contributing to the outcome of

discontinuation of treatment due to side effect
Quality assessment
ons
Discontinuation due to side effect
(Original y
data from
Haas 2012)
a. Cotton 1984, Goodwin1996, Holleboom 1996, Papatsonis 1997/2000, Moutquin 2000, French/Australian 2001,
Nassar 2009, Cox 1990, Laveno 1986, Larsen 1986, Smith 2007, Glock 1993, Essed 1978, Sirohiwal 2001,
Rayamajhi, 2003, Al-Qattan 2000, Cararach 2006, Shim 2006, Trablsi 2008, Weerakul 2002, Beall 1985, Romero

210
Tocolysis
2000, Hollander 1987, Wilkins 1988, Miller 1982, Surichamorn 2001, Chau 1992, Larmon 1999, Floyd 1995, El-
Sayed 1999, Caritis 1984, Garcia-Velasco 1998, Van de Water 2008, Maitra 2007, Motazedian 2010, European
2001.
3. Wide and very wide CrI across all comparisons except on placebo/control vs magnesium sulfate)

211
Tocolysis
10.6.7 Estimated gestational age at birth

Out of the 98 studies included in the review, 53 studies reported mean gestational age as an
outcome. Two studies reported mean gestational age only on 1 arm of the study and were
removed. The remaining 51 studies examined 19 medicines allowing for 7 out of the 9
treatment classes to be assessed against each other.
Figure 16 shows a graphic representation of tocolytics trials for the NMA for estimated
gestational age. Lines represent trials comparing 2 classes of medicine. The thickness of the
lines is proportional to the number of studies contributing to the comparison. The size of the
dots is proportional to the number of participants randomised to the treatment.
Figure 16: Graphic representation of tocolytics trials for the NMA for
estimated gestational age

212
Tocolysis
Table 62 shows the posterior median of the mean difference and 95% CrIs for gestational age at birth in weeks.
In the upper diagonal values shown are the mean difference for the column headers versus the row headers and are derived from the NMA. Given
that this table relates to a positive and continuous outcome, values higher than 0 favour the column defining treatment and values lower than 0
favour the row defining the treatment. Upper diagonal data is also presented in the forest plots in Appendix I.
In the lower diagonal values shown are the mean difference for the row headers versus the column headers and are derived from the direct
comparison analysis. Given that this table relates to an adverse outcome, values lower than 0 favour the column defining treatment and values
higher than 0 favour the row defining the treatment.
Table 58: Posterior median of the mean difference and 95% credible Intervals (CrI) for gestational age at birth in weeks
Oxytocin
Prostaglandin Magnesium Calcium channel receptor
Placebo/control inhibitors sulfate Betamimetics blockers Nitrates blockers
Placebo/control 2.32 (1.27,3.35) 1.29 (0.29,2.27) 1.25 (0.40,2.07) 1.69 (0.69,2.66) 1.65 (0.52,2.78) 0.68 (−1.32,2.67)
Prostaglandin 3.27 (1.68,4.78) −1.04 (−2.01, −1.08 (−2.08, −0.64 (−1.68,0.42) −0.67 (−1.97,0.67) −1.65 (−3.76,0.52)
inhibitors −0.04) −0.05)
Magnesium −0.14 (−1.60,1.28) 0.92 (−1.73,3.57) -0.04 (−0.99,0.91) 0.40 (−0.51,1.31) 0.36 (−0.88,1.63) −0.61 (−2.69,1.50)
sulfate
Betamimetics 1.91 (0.90,2.90) −0.24 (−1.46,0.97) - 0.44 (−0.32,1.20) 0.40 (−0.54,1.37) −0.57 (−2.58,1.47)
Calcium channel - −1.56 (−3.42,0.28) - - −0.03 (−1.16,1.10) −1.01 (−2.98,0.99)
blockers
Nitrates 1.09 (−1.79,4.00) −0.53 (−2.32,1.25) −0.19 (−2.78,2.45) - 0.80 (−0.08,1.67) −0.98 (−3.15,1.21)
Oxytocin −0.51 (−3.00,2.01) 0.92 (−1.73,3.57) −0.02 (−1.25,1.22) - - 0.58 (−0.47,1.67)
receptor
blockers
Tocolysis
Table 63 ranks probability of best medicine classes for improving the estimated gestational
age. Rows are arranged in the decreasing order of estimate effect, with best treatment at the
top and the worst at the bottom of the table.
Table 59: Ranking probability of best medicine classes for improving the estimated
gestational age
the best treatment
inhibitors
Nitrates 13% 3.04 3 (1 to 6)
blockers
blockers

(7). The line indicates the probability that medicine will achieve each rank

214
Tocolysis
Figure 17: Graphic representation of the each medicine’s effectiveness

probability ranking for improving outcome

215
Tocolysis
Table 60:Quality assessment of the evidence contributing to the outcome of estimated

Quality assessment
ons
Mean estimated gestational age
studies
(Original
data from
Haas 2012)
CrI credible interval, NMA network meta-analysis, OR odds ratio, p probability, RCT randomised controlled trial
a. Cotton 1984, Klauser 2012, Goodwin 1996, Niebyl 1980, Weiner 1988, Stika 2002, Al-Omari 2006, Kurki 1991,
McWhorter 2004, Lyell 2007, Holleboom 1996, Papatsonis 1997/2000, Moutquin 2000, French/Australian 2001,
Nassar 2009, Sawdy 2003, Zuckerman 1984, Panter 1999, Cox 1990, How 2006, Casapo 1977, CPLIG 1992,
merkatz 1980, Laveno 1986, Larsen 1986, Smith 2007, Borna 2007, Rasanen 1995, Parilla 1997, Besinger 1991,
Kashanian 2011, Schorr 1998, Surichamorn 2001, Larmon 1999, Taherian 2007, Glock 1993, Essed 1978,
Sirohiwal 2001, Rayamajhi 2003, Al-Qattan 2000, Cararach 2006, Fan 2003, Koks 1998, Lin 2009, Shim 2006,
Neri 2009, Jannet 1997, Trablsi 2008, Weerakul 2002, Kashanian 2014, Salim 2012.
3. Some evidence of inconsistency (conflict between direct and indirect evidence) in comparisons of
placebo/control v magnesium sulfate
Bayesian p-value=0.015
The direct and indirect are contradictory to each other (both on opposite sides of one). Results from this network
should be considered with caution
4. Wide and very wide CrI across all comparisons except on (placebo/control vs magnesium sulfate)

216
Tocolysis
10.6.8 Respiratory distress syndrome (RDS)

Out of the 98 studies included in the review, 57 studies reported RDS as an outcome. One
study observed no events and was removed. Seven studies only reported events on 1 arm
and were also removed. The remaining 49 studies examined 22 medicines allowing for 8 out
of the 9 treatment classes to be assessed against each other.
Figure 18 shows a graphic representation of tocolytics trials for the NMA for RDS. Lines
represent trials comparing 2 classes of medicine. The thickness of the lines is proportional to
the number of studies contributing to the comparison. The size of the dots is proportional to
the number of participants randomised to the treatment.
Figure 18: Graphic representation of tocolytics trials for

the NMA for RDS

217
Tocolysis
Table 65 shows the posterior median of the ORs and 95% CrIs for respiratory distress syndrome.
table relates to an adverse outcome, values lower than 1 favour the column defining treatment and values higher than 1 favour the row defining the
treatment. Upper diagonal data is also presented in the forest plots in Appendix I.
Table 61: Posterior median of the odds ratios (OR) and 95% credible Intervals (CrI) for respiratory distress syndrome
Oxytocin
Prostaglandin Magnesium Calcium channel receptor Other
Placebo/control inhibitors sulfate Betamimetics blockers blockers Alcohol/ethanol treatments
Placebo/control 1.13 (0.68 to 1.20 (0.76 to 0.88 (0.65 to 0.81 (0.50 to 0.96 (0.66 to 2.55 (0.78 to 0.75 (0.26 to
1.86) 1.90) 1.23) 1.34) 1.43) 9.05) 2.21)
Prostaglandin 1.01 (0.33, to 1.06 (0.69 to 0.78 (0.49 to 0.71 (0.41 to 0.85 (0.52 to 2.25 (0.65 to 0.75 (0.26 to
inhibitors 3.10) 1.66) 1.28) 1.29) 1.42) 9.05) 2.21)
Magnesium sulfate 1.26 (0.58 to 1.02 (0.64 to 0.73 (0.47 to 0.67 (0.41 to 0.80 (0.51 to 2.12 (0.62 to 0.63 (0.19 to
2.72) 1.65) 1.15) 1.12) 1.29) 7.86) 2.02)
Betamimetics 0.72 (0.50 to 1.25 (0.48 to 0.53 (0.05 to 0.92 (0.61 to 1.08 (0.77 to 2.88 (0.92 to 0.85 (0.28 to
1.04) 3.25) 4.31) 1.39) 1.54) 9.75) 2.59)
Calcium channel - - 0.85 (0.41 to 0.74 (0.45 to 1.19 (0.73 to 3.14 (0.93 to 0.93 (0.28 to
blockers 1.72) 1.22) 1.90) 11.33) 3.01)
Oxytocin receptor 1.50 (0.90 to 0.60 (0.29 to - 1.00 (0.65 to 0.77(0.27 to 2.14) 2.65 (0.80 to 0.79 (0.25 to
blockers 2.95) 1.22) 1.62) 9.46) 2.41)
Alcohol/ethanol - - - 2.88 (0.96 to - - 0.29 (0.06 to
9.45) 1.46)
Other treatments 0.75 (0.26 to - - - - - -
2.16)
Tocolysis
Tocolysis
Table 66 shows probability rankings for medicines by class to reduce respiratory

distress syndrome (RDS). Rows are arranged in the decreasing order of estimate effect
with the best treatment at the top and the worst at the bottom of the table.
Table 62: Probability rankings for medicines by class to reduce respiratory distress
syndrome (RDS
the best treatment
Calcium channel 55% 2. 1 (1 to 5)
blockers
blockers
inhibitors

220
Tocolysis
Figure 19 shows a graphic representation of each medicine’s effectiveness probability

(6). The line indicates the probability that medicine will achieve each rank.
Figure 19: Graphic representation of each medicine’s effectiveness probability


221
Tocolysis
Table 63: Quality assessment of the evidence contributing to the outcome of RDS
Quality assessment
ons
RDS
studies
(Original
data from
Haas 2012)
CrI credible interval, NMA network meta-analysis, OR odds ratio, p probability, RCT randomised controlled trial,
RDS respiratory distress syndrome
a. Thornton 2009, Cotton 1984, Klauser 2012, Goodwin 1996, Niebyl 1980, Panter 1999, Zuckerman 1984, Cox
1990, Spellacy 1979, Larsen 1986, Merkatz 1980, Leveno 1986, CPLIG 1992, Goodwin 1994, Romero 2000,
Weiner 1988, Stika 2002, Rasanen 1995, Morales 1993, Parilla 1997, Morales 1989, Kurki 1991, Schorr 1998,
McWhorter 2004, Miller 1982, Floyd 1995, Lyell 2007, Essed 1978, Gummerus 1983, Holleboom 1996, Caritis
1984, Maitra 2007, Cararach 2006, VandeWater 2008, Al-Qattan 2000, Papatsonis 1997, Papatsonis 2000, Shim,
2006, Lin 2009, Moutquin 2000, Lauersen 1977, Trabelsi 2008, French/Australian 2001, Laohapojanart 2007,
European 2001, Nassar 2009, Al-Omari 2006, Jaju 2011, Salim 2012.
1.Analysis was based on the class therefore different doses and co-treatment were combined together
2.There were some evidence of inconsistency (conflict between direct and indirect evidence) in comparisons of
placebo/control vs. betamimetics and placebo/control vs. oxytocin receptor blockers:
Placebo/control v betamimetics Bayesian p-value=0.034
Direct OR=0.72 95% CrI (0.51 to 1.02)
Indirect OR=1.48 95% CrI (0.84 to 2.56) Placebo/control v oxytocin receptor blockers Bayesian p-value=
0.015
Direct OR=1.49 95% CrI (0.91 to 2.72)
Indirect OR=0.63 95% CrI (0.40 to 1.02)
4. Wide and very wide CrIs across all comparisons except 2 (magnesium sulfate vs betamimetics and magnesium
sulfate vs calcium channel blockers)

222
Tocolysis
10.7 Evidence profiles for the pairwise comparisons
Table 64: GRADE profile for the comparison of placebo versus indomethacin
Summary of findings
Frequencya (%)/ mean
Quality assessment (SD) Effect
Relative Absolute effect
No. of Risk of (95% CI) (95% CI)
studies Design bias Inconsistency Indirectness Imprecision Other consideration Placebo Indomethacin Quality
Maternal infection
2 RCT Serious No serious Serious2 Very None 2/33 4/31 RR 0.48 67 fewer per 1000 Very low
studies 1
inconsistency serious3 (0.09 to (from 117 fewer to
(Niebyl 2.46) 188 more)
1980 &
Panter
1999)
Chronic lung disease (CLD)
2 RCT Serious No serious Serious2 Very None 4/35 5/35 RR 0.80 29 fewer per 1000 Very low
studies 1
inconsistency serious3 (0.23 to (from 110 fewer to
(Niebyl 2.73) 247 more)
1980 &
Panter
1999)
CI confidence interval, MID minimally important difference, RCT randomised controlled trial, RR risk ratio, SD standard deviation
1. No clear inclusion/exclusion criteria hence high risk of selection bias
3. Evidence was downgraded by 2 due to serious imprecision as 95% confidence interval crossed 2 default MID
a. Denominator for the outcome of maternal infection was the number of women whereas for the outcome of chronic lung disease the number of babies.
Tocolysis
Summary of findings
Quality assessment No. of babies Effect
Table
Absolute effect
65:
No. of Risk of Other Relative/ size (95% CI)
studies Design bias Inconsistency Indirectness Imprecision considerations Placebo Nitrates (95% CI) Quality
1 study RCT No No serious Serious1 Very None 7/79 RR 7.00 76 more per 1000 Very low
(Smith serious inconsistency serious2 1/74 (0.13 to (from 11 fewer to 31
2007) risk of 3.44) more)
bias
Periventricular leucomalacia (PVL)
1 study RCT No No serious Serious1 Very None Placebo 2/79 Nitrates RR 4.81 160 more per 1000 Very low
(Smith serious inconsistency serious2 0/74 (0.23 to (from 32 fewer to
2007) risk of 101.79) 1000 more)
bias
GRADE profile for the comparison of placebo versus nitrates
CI confidence interval, MID minimally important difference, RCT randomised controlled trial, RR risk ratio
Table 66: GRADE profile for the comparison of betamimetics versus nitrates
Summary of findings
Absolute
No. of Risk of Other Beta- Relative/ (95% CI)
studies Design bias Inconsistency Indirectness Imprecision considerations mimetics Nitrates (95% CI) Quality
1 study RCT No No serious serious1 very serious2 None 9/116 9/120 RR 1.03 2 more per Very low
(Bisits serious inconsistency (0.43 to 1000 (43
2004) risk of 2.51) fewer to 113
bias more)
Tocolysis
Tocolysis
Table 67: GRADE profile for comparison of nifedipine versus ritodrine

Summary of findings
Quality assessment Frequency (%)/ mean

(SD) Effect
Relative
RR/mea
n
differen
No. of Risk of ce (95% Absolute
studies Design bias Inconsistency Indirectness Imprecision Other consideration Nifedipine Ritodrine CI) effect Quality
1 study RCT Serious1 No serious No serious Very serious2 None 1/95 5/90 RR 0.18 46 fewer Very low
(Paptsonis inconsistency indirectness (0.02 to per 1000
2000) 1.58) (from 54
fewer to 24
more)
Behaviour emotional functioning (follow-up at age 9–12 year) – measured using child behaviour checklist– higher score represents more psychosocial problema
1 study RCT Serious No serious No serious Serious 4 None 50 (11.9) 52 (11.6) MD -2 (- - Low
(Houtzager 1,2,3 inconsistency indirectness 6.57 to
2006) 2.57)
Behaviour emotional functioning (follow-up at age 9–12 year) – measured using teacher report form– higher score represents more psychosocial problema-
1 study RCT Serious No serious No serious Serious 4 None 49 (10) 50 (9.9) MD -1 (- - Low
2006) 2.87)
Children’s physical quality of life (QoL) (follow-up at age 9–12 year) – measured using teacher report form– higher score represents more favourable QoLa
2006) 0.92)
Children’s motor quality of life (QoL) (follow-up at age 9–12 year) – measured using teacher report form– higher score represents more favourable QoLa
1 study RCT Serious No serious No serious No serious None 30 (3.1) 30 (2.5) MD 0 (- - Modera
(Houtzager 1,2,3 inconsistency indirectness imprecision 1.10 to te
2006) 1.10)
Tocolysis
Summary of findings
Quality assessment Frequency (%)/ mean

(SD) Effect
Relative
RR/mea
n
differen
No. of Risk of ce (95% Absolute
studies Design bias Inconsistency Indirectness Imprecision Other consideration Nifedipine Ritodrine CI) effect Quality
Children’s autonomy quality of life (QoL) (follow-up at age 9–12 year)- measured using teacher report form– higher score represents more favourable QoLa
1 study RCT Serious No serious No serious No serious None 31 (1.2) 31 (1.6) MD 0 (- - Modera
(Houtzager 1,2,3 inconsistency indirectness imprecision 0.55 to te
2006) 0.55)
Children’s cognitive quality of life (QoL) (follow-up at age 9–12 year)- measured using teacher report form– higher score represents more favourable QoLa
1 study RCT Serious No serious No serious No serious None 28 (4) 28 (3.8) MD 0 (- Modera te
(Houtzager 1,2,3 inconsistency indirectness imprecision 1.52 to -
2006) 1.52)
Children’s positive emotion quality of life (QoL) (follow-up at age 9–12 year) – measured using teacher report form – higher score represents more favourable QoLa
2006) 0.00)
Children’s negative emotion quality of life (QoL) (follow-up at age 9–12 year) – measured using teacher report form– higher score represents more favourable QoLa
2006) 0.02)
Motor quality (follow-up at age 9–12 year) – movement ABC – higher score represents more motor problema
1 study RCT Very No serious No serious Serious 4 None mean 5 mean 9.3 MD - - Very low
(Houtzager serious1,2,3 inconsistency indirectness (6.9) (17.2) 4.30 (-
2006) 9.29 to
0.69)
CI confidence interval, MD mean difference, MID minimally important difference, QoL quality of life, RCT randomised controlled trial, RR risk ratio
a. The child long term outcomes were assessed using the Dutch version of the Child Behaviour Checklist (CBCL) completed by parents. The child' teacher completed the Teacher
Report Form (TRF). High score on the CBCL and TRF represent more problematic behaviour. Total score is for internalising problem such as anxiety, depression, or social
behaviour, non-compliance, or hyper activity.
The child quality of life (QoL) was assessed using the Dutch TNO AZL Children's Quality of Life Questionnaire (TACQOL). High score represent a more favourable QoL.
1. No clear inclusion/exclusion criteria hence high risk of selection bias
Tocolysis
2. Low response rate; 65% in nifedipine and 55% in ritodrine group

3. Unclear if evaluation tools were validated
Tocolysis
Table 68: GRADE profile for the comparison of placebo versus indomethacin
Summary of findings
Relative/ Absolute
No. of Risk of Other RR (95% (95% CI)
studies Design bias Inconsistency Indirectness Imprecision considerations Placebo Indomethacin CI) Quality
1 study RCT No No serious Serious1 Very serious2 None 0/20 1/19 RR 0.30 37 fewer Very low
(Panter serious inconsistency (0.01 to per 1000
1999) risk of 7.85) (from 52
bias fewer to 36
more)
1.Multiple pregnancy included
2.Evidence was downgraded by 2 due to serious imprecision as 95% confidence interval crossed 2 default MID
Table 69: GRADE profile for comparison of indomethacin versus magnesium sulfate
Summary of findings
Relative/ Absolute
No. of Risk of Other Indomethacin Magnesium RR (95% (95% CI)
studies Design bias Inconsistency Indirectness Imprecision considerations sulfate CI) Quality
1 study RCT No No serious Serious1 No serious None 2/103 0/95 NC NC Moderate
(Klauser serious inconsistency imprecision
2012) risk of
bias
CI confidence interval, NC not calculable, RCT randomised controlled trial, RR risk ratio
Tocolysis
Table 70: GRADE profile for comparison of indomethacin versus magnesium sulfate
Summary of findings
Relative/ Absolute
No. of Risk of Other Indomethacin RR (95% (95% CI)
studies Design bias Inconsistency Indirectness Imprecision considerations Nifedipine CI) Quality
1 study RCT No No serious Serious1 No serious None 2/103 0/119 NC NC Moderate
(Klauser serious inconsistency imprecision
2012) risk of
bias
CI confidence interval, NC not calculable, RR risk ratio
1.Multiple pregnancy included
Maternal corticosteroids
10.8 Evidence statements on NMA results

Neonatal mortality
Randomised very low quality evidence from the NMA on 36 tocolytics treatments from 9
classes (placebo, prostaglandin inhibitors, magnesium sulphate, betamimetics, calcium
channel blockers, nitrates, oxytocin receptor blockers, alcohol/ethanol, other treatments) with
a total sample size of almost 7000 women in diagnosed or suspected preterm labour showed
that no medicine class was significantly better than placebo for reducing neonatal mortality.
However, results should be interpreted with caution due to high uncertainty in the effect
estimates and the inconsistency between direct and indirect analysis for a medicine
comparison.
Perinatal mortality
Randomised very low quality evidence from the NMA on 35 tocolytics treatments from 9
classes (placebo, prostaglandin inhibitors, magnesium sulphate, betamimetics, calcium
channel blockers, nitrates, oxytocin receptor blockers, alcohol/ethanol, other treatments) with
a total sample size of over 6000 women in diagnosed or suspected preterm labour showed
that:
• Nitrates were more effective than all other medicine classes (including placebo)
for reducing perinatal mortality, though there was considerable uncertainty in
estimates of their efficacy. They had the highest probability (89%) of being the
most effective medicine class for this outcome.
• Prostaglandin inhibitors had the next highest probability of being best (5%).
The results of this model showed some inconsistencies between direct and indirect analyses.
Delay birth by more than 48 hours
Randomised very low quality evidence on 36 treatments from 9 classes (placebo,
prostaglandin inhibitors, magnesium sulphate, betamimetics, calcium channel blockers,
nitrates, oxytocin receptor blockers, alcohol/ethanol, other treatments) in the NMA with a total
sample size of almost 8000 women in diagnosed or suspected preterm labour showed that:
• Prostaglandin inhibitors were both more effective than all other medicine classes
(including placebo) for delaying birth by 48 hours and had the highest probability
(76%) of being the best treatment class compared with other medicine classes.
• Oxytocin receptors blockers and magnesium sulfate had the next highest
probability of being ranked best (8 and 7%).
• Calcium blockers were not found to significantly improve this outcome compared
with placebo.
The results from direct comparisons between medicine classes were not always consistent
with those from NMA.
Neonatal sepsis
Very low quality evidence from 21 RCTs investigating 7 classes (placebo, prostaglandin
inhibitors, magnesium sulphate, betamimetics, calcium channel blockers, oxytocin receptor
blockers, other treatments) in the NMA with almost 2500 women in diagnosed or suspected
preterm labour found that:
• Calcium channel blockers were more effective than all other tocolytic medicines
compared with placebo for reducing neonatal sepsis and also had the highest

231
probability (45%) of being the most effective medicine class compared with other
medicine classess
• Placebo/control and oxytocin receptor blockers had the next highest probability of
being the most effective with 38% and 11% probability respectively.
The results from direct comparisons was not always consistent with those from NMA.
Intraventricular haemorrhage
Very low quality evidence from 28 RCTs investigating 8 classes (placebo, prostaglandin
inhibitors, magnesium sulphate, betamimetics, calcium channel blockers, nitrates, oxytocin
receptor blockers, other treatments) in the NMA with over 5000 women in diagnosed or
suspected preterm labour found that:
• Nitrates were more effective than all other medicine classes compared with
placebo for reducing intraventricular haemorrhage and also had the highest
probability (59%) of being the most effective medicine class compared with other
medicine classes.
• Calcium channel blockers had the next highest probability of being the most
effective medicine class compared with other medicine classes (38%).
The result from direct was not always consistent with those from meta-analysis.
Discontinuation of treatment due to adverse events effect
Very low quality evidence from 15 RCTs investigating 6 classes (placebo, magnesium
sulphate, betamimetics, calcium channel blockers, nitrates, oxytocin receptor blockers) in the
NMA with 4000 women in diagnosed preterm labour and 410 women in suspected preterm
labour found that:
• Placebo/control were more effective than all other medicine classes for not
causing maternal adverse events and discontinuation of treatment and that
placebo/control treatment also had the highest probability (73%) of being the most
effective medicine class compared with other medicine classes.
• Nitrates and oxytocin receptor blocker had the next highest probability of being
the most effective with 11% and 10% probability respectively.
The results from direct comparisons were consistent with those from meta-analysis.
Estimated gestational age at birth
Randomised very low quality evidence from 28 treatments investigating 7 classes (placebo,
nitrates, oxytocin receptor blockers) in the NMA with a total sample size of over 5500 women
in diagnosed or suspected preterm labour concluded that:
• Prostaglandin inhibitors were more effective than all other medicine classes
compared with placebo for increasing estimated gestational age and had the
highest probability (64%) of being the most effective medicine class compared
with other medicine classes.
• Nitrates and calcium channel blockers had the next highest probability of being
the most effective with 21% and 9% probability respectively.
The results from direct comparisons were not always consistent with those from NMA.
Very low quality randomised evidence from 28 treatments investigating 8 classes (placebo,

232
oxytocin receptor blockers, alcohol/ethanol, other treatments) contributed to the NMA with a
total sample size over 5500 women in diagnosed or suspected preterm labour showed that:
• Calcium channel blockers were more effective than all medicine classes
compared with placebo for reducing respiratory distress syndrome and had the
highest probability (55%) of being the best medicine class compared with other
medicine classes.
• Beta-mimetics and oxytocin receptor blockers had the next highest probability of
being the most effective with 20% and 11% probability respectively.
The results from direct comparisons were not always consistent with those from NMA.
10.9 Evidence statements on pair-wise comparisons

Evidence from meta-analysis of 2 RCTs with over 60 participants found no significant
difference for the outcomes of maternal infection and chronic lung disease (CLD) in women
treated with indomethacin compared with those who received placebo. The quality of the
evidence was of very low quality.
Very low quality evidence from 10 RCTs with 112 participants found no significant difference
in the rate of CLD between babies whose mothers were treated with nitrates compared with
those on placebo.
Evidence from 10 other RCTs with 236 participants found no significant difference in the rate
of CLD in babies whose mothers were treated with beta-mimetics compared with those
whose mothers were treated with nitrates.The quality of the evidence was of moderate
quality.
Very low quality evidence from 2 studies with 192 participants found no significant difference
in the rate of periventricular leucomalacia (PVL) in babies whose mothers were treated with
placebo compared with those whose mothers were treated with indomethacin or nitrates.
Moderate evidence from 2 other studies with 407 participants found no difference in the rate
of PVL in babies whose mothers were treated with nifedipine compared with those whose
mothers were treated with ritodrine, and those whose mothers were treated with
indomethacin compared with magnesium sulfate.
Long-term psychosocial functioning (follow-up at age 9–12 years)
Evidence from a follow-up of an RCT with 102 participants found no significant difference
between children whose mothers were treated with nifedipine compared with ritodrine for the
behaviour of emotional functioning and quality of life (physical, motor, autonomy, cognitive,
positive emotion). Children whose mothers were treated with nifedipine had significantly
lower negative emotion quality of life score compared with children whose mothers were
treated with ritodrine. The quality of the evidence was moderate to low.

A new health economic model was developed using the evidence from the NMA undertaken
for this guideline on neonatal mortality, intraventricular haemorrhage and respiratory distress
syndrome. As far as we are aware this is the first economic evaluation of tocolysis where
relative treatment effects are based on the results of an NMA.
Using a cost-utility analysis approach a range of tocolytics (prostaglandin inhibitors,
betamimetics, calcium channel blockers, magnesium sulfate, nitrates, oxytocin receptor

233
blockers) were compared as well as standard care (no tocolytic) in women at between 24+0
and 34+0 weeks of pregnancy in suspected or diagnosed preterm labour.
The base-case analysis found that calcium channel blockers were the most cost-effective
treatment across all gestational ages considered in the model as reflected by its net mean
benefit which was the highest across 10,000 Monte Carlo simulations. Oxytocin receptor
blockers had the second highest net mean benefit. Nitrates actually had a slightly higher
probability of being cost effective than calcium channel blockers (for example 36% versus
34% at a gestational age of 24 weeks) but this reflects the wider confidence intervals for
nitrates.
A sensitivity analysis suggested that changing the assumptions with respect to the loss in
quality adjusted life years (QALYs) from RDS and IVH had a negligible impact on the model
results. It was also clear from the net mean benefit achieved with calcium channel blockers
relative to the alternatives, that treatment costs were not an important driver of the cost-
effectiveness results. Although oxytocin receptor blockers had the highest cost of all
treatment options they were found to have the second highest net mean benefit across all
gestational ages included in the model.
The model is described in greater detail in Chapter 16.

The Guideline Development Committee considered both neonatal and maternal outcomes for
this review question. The following were considered in relation to neonatal outcomes:
• neonatal and perinatal mortality

• neonatal sepsis
• chronic lung disease
• gestational age at birth
• periventricular leucomalacia
• neurodevelopmental disability
Neonatal and perinatal mortality and respiratory distress syndrome were considered the most
critical outcomes for decision-making.
Among the maternal outcomes, the committee included the following as the most important
outcomes:
• maternal infection
• maternal mortality
• discontinuation of treatment due to maternal adverse events.
Discontinuation of treatment due to maternal adverse events was the only maternal outcome
prioritised in the NMA.

In relation to the most critical neonatal outcomes for decision-making (neonatal and perinatal
mortality and respiratory distress syndrome), the evidence reviewed in the NMA showed that
calcium channel blockers had the highest probability of being the best medicine for reducing

234
respiratory distress syndrome and were more effective for this outcome than the other
tocolytics used in the NMA (placebo, prostaglandin inhibitors, magnesium sulfate,
betamimetics, oxytocin receptor blockers, alcohol and other treatments) when given to
women in suspected or diagnosed preterm labour. No single class of treatment reviewed was
found to be more effective than placebo for reducing neonatal mortality whereas nitrates
were found to be more effective than all other drugs and had the highest probability of being
the best medicine to reduce perinatal mortality.
With regard to other outcomes, NMA results showed that calcium blockers were also
beneficial in terms of protecting preterm babies from neonatal sepsis and were the second
best treatment for improving intraventricular haemorrhage (IVH) and the third best treatment
for increasing gestational age at birth. Prostaglandin inhibitors were found to be the most
beneficial treatment in terms of delaying birth by more than 48 hours and for increasing
estimated gestational age, and the second most effective treatment for reducing perinatal
mortality. However, prostaglandin inhibitors were not found to be significantly better than
calcium channel blockers at delaying birth by more than 48 hours. In addition, prostaglandin
inhibitors were not found to improve the ‘harder’ outcomes such as neonatal mortality,
respiratory distress syndrome and neonatal sepsis in all of which they scored very low in the
ranking of best treatments. The committee was also aware of other harms thought to be
associated with prostaglandin inhibitors, such as premature closure of the ductus arteriosus.
Therefore the committee did not consider them as a tocolytic option for women in suspected
or diagnosed preterm labour.
The NMA results also found that nitrates were the most effective treatment for IVH and the
second best, along with oxytocin receptors, at reducing the risk of discontinuation of
treatment due to adverse events and at increasing estimated gestational age, but the
committee discussed that these benefits from nitrates need to be balanced against the
potential harm to the fetus. In addition, the number of trials including this treatment was small
(only 6) and therefore results should be interpreted with caution. Nitrates also did not connect
to the NMA network for the outcome of respiratory distress syndrome (10 of the selected
critical outcomes) and therefore the committee was uncertain on the effectiveness of this
intervention at improving this outcome. Nitrates were also not found to be significantly better
than placebo at reducing chronic lung disease in pair-wise comparisons.The use of oxytocin
receptor blockers for reduction of maternal side effects and for increasing gestational age
has to be balanced against its poor efficacy in reducing IVH and RDS and its modest effect
on perinatal mortality. Therefore the committee decided that this should not be the first option
of tocolytic treatment.
In relation to maternal outcomes, evidence from the NMA showed all reviewed treatments
had an unfavourable effect on discontinuation of treatment due to adverse effects. The
evidence from the pair-wise comparison showed that indomethacin was not significantly
more harmful than placebo on the outcome of maternal infection.
Betamimetics did not score highly in terms of clinical effectiveness for any of the outcomes
reviewed in the NMA or pair wise meta-analysis and the committee confirmed that their use
should not be considered for tocolytic treatment for suspected or diagnosed women at
preterm labour.
The committee members discussed that in their own clinical experience, the most frequent
clinical case scenario of women in diagnosed or suspected preterm labour would be to
administer magnesium sulfate to improve the baby’s neuroprotection. The question the
committee aimed to address is whether there is any additional clinical benefit, with minimal
harms for both the baby and the mother, from adding another tocolytic (in addition to
magnesium sulfate). The included evidence did not provide further information on whether
magnesium sulfate had been already prescribed for neuroprotection prior to a decision being
made about the use of other tocolytics. However, the committee highlighted that magnesium

235
sulfate has only been in routine use for this reason for approximately the last 5 years, so in
older studies this would not have been relevant. Based on their clinical experience, no
adverse interaction (for example increasing the frequency of adverse events) was anticipated
by using a combination of magnesium sulfate and another tocolytic medicine.
The average gestational age profile of women included in the evidence for this section was
26 weeks but the range was wider and covered women between 24 and 36 weeks of
gestation. The committee discussed the role of tocolytics by gestational age and recognised
the lack of data for the effectiveness and/or harm of tocolytics on the fetus at a gestational
age below 26 weeks.
The committee also recognised that the clinical decision to start tocolytic treatment needs to
take into consideration a range of maternal factors such as the woman’s status in the care
pathway (whether in suspected or diagnosed preterm labour) and the coexistence of other
features such as bleeding and infection, in which circumstances delaying preterm labour
would be contraindicated. In relation to neonatal considerations, the decision to offer tocolytic
treatment should assess the likely benefit of maternal corticosteroids, the gestational age
and the impact of prolonging birth, as well as the availability of neonatal care in the care
setting or the need for transfer to another hospital unit. Women’s preference on starting
tocolytics should also be taken into consideration in the planning of care.
The committee discussed the different types and dosages of calcium blockers used in the
studies included in the NMA. It was noted that the majority of evidence on calcium blockers
was derived from trials which included nifedipine. Given that nifedipine is the most widely
used calcium blocker in clinical practice and nicardipine (the other calcium blocker included
in the trials reviewed) is associated with significant side effects, the Committee
recommended the use of nifedipine for tocolysis among the calcium blockers. The dosage of
nifedipine in the largest RCT included in the NMA (Klauser 2012) was given in a loading
dose of 30 mg orally followed by 20–30 mg every 4–6 hours until contractions abated.
However, the recommended dosage of nifedipine by the Royal College of Obstetricians and
Gynaecologists (RCOG) (in the Green-top guideline on Preterm Labour, Tocolytic Drugs) is
an initial oral dose of 20 mg followed by 10–20 mg 3 to 4 times daily, adjusted according to
uterine activity for up to 48 hours. The same guideline also highlights that a total dose of
nifedipine above 60 mg appears to be associated with a 3- to 4-fold increase in adverse
events such as headache and hypotension. The committee agreed that the dose of
nifedipine recommended by the RCOG would be the one most commonly used in clinical
practice and therefore endorsed it.
10.11.3 Consideration of economic benefits and harms

The committee discussed that there is clear potential for effective tocolysis to be cost
effective as the adverse outcomes of preterm birth are associated with significant losses in
health related quality of life and large subsequent healthcare expenditure.
The health economic model produced for this review question focused on 3 of the outcomes
prioritised in the NMA (neonatal, perinatal mortality and respiratory distress syndrome) which
were also considered to have the greatest impact on health related quality of life. The
decision to focus on these 3 outcomes was to avoid double counting, as would be the case
for some other outcomes (for example the impact of medicines on delaying birth by 48 hours
and increasing gestational age). The relative treatment effect derived from the results of
NMA on neonatal mortality was used to estimate the absolute mortality risk across the entire
perinatal and neonatal period with the various treatment alternatives. This model found that
calcium channel blockers were the most cost-effective treatment for women of all gestational
ages.

236
One advantage of health economic evaluation is that it allows for benefits from different
outcomes to be synthesised into a single measure, in this case the QALY. This allows explicit
trade-offs to be made across outcomes in a way that is conceptually difficult when comparing
different outcomes in isolation. The 3 NMAs underpin the cost-effectiveness findings but it is
worth highlighting that although the result takes uncertainty into account through the tool of
probabilistic sensitivity analysis, it is not governed by the classical rules of statistical
inference. So in the network analyses included in the health economic model, calcium
channel blockers were either the ‘best’ treatment or one of the ‘best’. Being the best does not
necessarily mean that it is possible to reject a null hypothesis of no difference against an
alternative using the usual, but arbitrary, 5% statistical significance level for a particular
outcome but it does mean that when averaged across the simulations this will have a very
important bearing on the cost-effectiveness result. Furthermore, the implication of not
recommending the most cost-effective treatment will be that a treatment (or no treatment)
alternative which is less likely to be cost effective is used instead.
Although the model was not particularly sensitive to treatment costs, calcium channel
blockers are one of the cheapest tocolytics and may be cost saving at the lower gestational
ages as ‘downstream’ savings from averted adverse outcomes more than offset treatment
cost.
The committee felt that based on the available economic evidence, it would be reasonable to
recommend calcium channel blockers and specifically nifedipine as a first line tocolytic
treatment. The committee discussed that this would not deviate from current practice in many
settings providing care for women at risk of preterm labour. In addition, the committee
thought that oxytocin receptor blockers should be offered to those for whom nifedipine was
contraindicated as the model provided evidence that they were the most cost-effective
treatment option after calcium channel blockers.

The quality of evidence included in the NMA was very low mainly due to indirectness, as
almost half of the included studies involved multiple pregnancies and imprecision around the
effect size. There were considerable inconsistencies observed between direct and indirect
evidence for most of the outcomes selected in the NMA.
There were limitations in the data included in the NMA. Most comparisons were only made in
1 or 2 trials. Furthermore, not all trials report all outcomes so some networks were very
sparse in terms of patient numbers contributing to each loop. In addition, some trials had
zero events in all arms and could not contribute to the estimation of treatment effects and
were removed. The variation in results between different NMAs also make overall
assessment of treatment options difficult for clinical interpretation.
Because some studies in the NMA included multiple births, allowing more than 1 infant per
mother, it was not always clear which was the most appropriate number of individuals to
consider for neonatal/ infant outcomes. Where available we used the number of infants as
the denominator. Although this does not account for the expected correlation in outcomes of
infants from the same mother, it prevents double counting of infants (twins) from the same
mother where both have had an event.
The evidence included in the pair-wise comparisons (outcomes in the protocol that were not
prioritised in the NMA) was also of moderate to very low quality given the high risk of
selection bias, imprecision and indirectness (a proportion of women in studies had multiple
pregnancies).

237

The committee stressed the importance of information provision to women in suspected or
diagnosed preterm labour about the effect of different tocolytics on child and mother
outcomes. It was discussed how different women may place different importance on the
different maternal and neonatal outcomes and this may be affected by any previous similar
preterm delivery experience. Therefore, the committee emphasised the role of provision of
information on the role of tocolytics on delaying birth and improving neonatal outcomes.

The committee concluded that:
• Calcium blockers were found to be the most clinical and cost-effective tocolytic
medicine for women in suspected or diagnosed preterm labour with intact
membranes.
• Oxytocin receptor blockers were also found effective for some other outcomes but
were not the most effective option overall.
• Prostaglandin inhibitors may produce a protective effect for delaying birth by more
than 48 hours.
• There is limited data on the long-term consequences of tocolytics for both babies
and their mothers.
10.12 Recommendations
updated guideline.
11 Maternal corticosteroids
11.1 Introduction
It has been recognised for many years that antenatal administration of corticosteroids to a
mother prior to preterm birth reduces the severity of lung disease of prematurity and of other
associated complications for her baby (Roberts 2006). This includes the severity of lung
disease in babies of women with preterm birth associated with diabetes in pregnancy,
hypertension in pregnancy and multiple pregnancy (see the NICE guideline on multiple
pregnancy). However, there remains uncertainty regarding the effectiveness of antenatal
corticosteroids at the extremes of gestations of preterm birth.
In addition, women at risk of preterm birth who are presenting with symptoms of preterm
labour or in suspected preterm labour do not always go on to deliver within the next 4–7
days, but may remain at high risk of preterm delivery. For these women, there is uncertainty
about whether repeat courses of corticosteroids give additional benefit for fetal lung
maturation, and if so, whether the risks of additional doses of corticosteroid (to both the fetus
and the mother) may outweigh any benefit.
This section covers 2 aspects of maternal corticosteroids with regard to their clinical
effectiveness for fetal lung maturation: their impact on neonatal outcomes given at different
gestations; and whether a repeated or single course is the most effective treatment option.

238
11.2 Single course of maternal corticosteroids at

different gestations
What is the clinical effectiveness of a single course of maternal corticosteroids for fetal lung
maturation given at different gestations in improving preterm neonatal outcomes?

Two sources of evidence are included in the review for this question; a Cochrane review
(SR) and meta-analysis (Roberts 2013) and 1 RCT (Porto 2011) from Brazil. Roberts (2013)
comprised 21 component randomised controlled trials (RCTs) (3885 women, 4269 babies)
from a variety of locations: 10 trials from the USA, 2 from Finland and 1 each from New
Zealand, the UK, The Netherlands, South Africa, Canada, Brazil, Jordan, Tunisia and Spain.
Additional data were obtained for 4 of the trials included in the SR, including individual
participant data (IPD) from the largest trial of corticosteroids for fetal lung maturation.
In the Cochrane review (Roberts 2013), the following subgroup analyses were conducted:
• gestational age at delivery (less than 28 weeks, less than 30 weeks, less than 32
weeks, less than 34 weeks, less than 36 weeks, at least 34 weeks, at least 36
weeks)
• entry to delivery interval (less than 24 hours, less than 48 hours, 1–7 days,
greater than 7 days)
• prelabour rupture of membranes (at trial entry, more than 24 hours before
delivery, more than 48 hours before delivery
• pregnancy-induced hypertension syndromes
• type of glucocorticoid (betamethasone, dexamethasone, hydrocortisone).
Post hoc subgroup analysis was performed for gestational age at entry to trial (less than 26
weeks, between 26 and 29+6 weeks, between 30 and 32+6 weeks, between 33 and 34+6
weeks, between 35 and 36+6 weeks, greater than 37 weeks).
Fourteen of the included trials in the SR compared corticosteroids with placebo, whereas the
remaining trials compared corticosteroids with expectant management. The choice of
corticosteroid for the majority of trials in the SR (15 RCTs) was betamethasone and only 6
trials used dexamethasone (1 trial did not report the corticosteroid used). The route and
dosage of corticosteroid also varied between the trials, with the most common protocol of
administration being 12 mg betamethasone intramuscularly divided into 2 doses, 24 hours
apart (6 trials). Eight trials in the SR allowed repeated courses of corticosteroids in their
study protocols, although no clear information was given on the proportion of women who
actually received repeated corticosteroids and, if so, how many repeated courses per
woman. Therefore, subgroup analysis for those women with single or repeated course of
corticosteroids was not feasible.
The literature reports data based both on gestational age at delivery and gestational age at
trial entry (that is, gestational age at time of first corticosteroid administration). The Guideline
Development Committee felt that both types of data would be helpful to inform clinical
decision-making.
In order to estimate whether the effect of maternal corticosteroids is biased by the inclusion
of studies with repeated courses of corticosteroids, sensitivity analysis by excluding these
studies was performed and the results were compared with the results from the overall meta-
analysis.

239
The committee preselected subgroup analysis at the protocol stage based on the following
factors:
• gestational age at delivery (less than 24+0 weeks, less than 26+0 weeks, less
than 28+0 weeks, less than 30+0 weeks, less than 32+0 weeks, less than 34+0
weeks, less than 36+0 weeks, at least 34+0 weeks, at least 36+0 weeks )
• gestational age at trial entry
• with and without intact membranes
• entry to delivery interval (less than 24 hours, less than 48 hours, 1 to 7 days,
more than 7 days)
• planned/spontaneous preterm birth.
The range of women’s gestational age varied considerably in the included RCTs from 24
weeks to 37 weeks. The mean gestational age at the trial entry ranged from 25.1 weeks
(standard deviation [SD] 1.4 weeks) for 11 RCTs to 32.0 weeks (SD 3.2 weeks) for 5 RCTs
and ranging from 26.6 weeks (SD 1.3 weeks) to 33.6 weeks (SD 4.6 weeks) for the rest of
the trials included in the Cochrane review. The mean time between corticosteroid
administration and birth was often not clearly reported, with only a minority of studies (5)
reporting the proportion of births within 7 days (from 50% to 76%).
The study population for this review question included women who were in spontaneous
preterm labour, for whom a preterm birth was planned or who had prelabour premature
rupture of membranes (P-PROM). Eight trials included all women with P-PROM whereas 4
studies included mixed populations with 23–63% of their total population having P-PROM.
Details of tocolysis administration were reported in 11 trials: use of tocolytic drugs varied
between trials and the percentage of women receiving tocolysis ranged from 23% to 100%.
Lastly, 10 out of 21 trials included only women with a singleton pregnancy. The remaining
trials included mixed populations of single and twin pregnancies with the proportion of twin
pregnancies ranging from 2% to 20%. There was no data available for sensitivity analysis on
singleton pregnancies versus mixed pregnancies.
11.2.2.1 Additional source of evidence on antenatal corticosteroids before 26

weeks’ gestation
One SR (Onland 2011) aimed to evaluate the effectiveness of antenatal corticosteroids
before 26 weeks’ gestation (extreme prematurity) and was included in this section. The
review included 9 trials whose eligibility criteria for entry in the study allowed for a lower
gestational age cut-off of less than 26 weeks. The majority of these studies (8 out of 9) were
also included in the SR by Roberts (2013) while the ninth study studied combined
corticosteroids and vitamin K therapy. Five of the 9 included trials permitted repeat courses
of corticosteroids and 5 of the 9 trials included women with a multiple pregnancy.

Study quality was assessed using the GRADE methodology. As RCTs were considered an
appropriate study design for addressing this question, they were initially assigned a quality
rating of ‘high’ and subsequently downgraded based on potential sources of bias.
The evidence is presented in the following GRADE table:
• Table 75:GRADE profile for comparison of corticosteroids versus placebo or expectant
management.

240
In this table the following subgroup factors were examined:

o women with premature rupture of membranes at first dose of corticosteroids
o different gestations at birth
o different gestations at first dose of corticosteroisin different gestations at birth (less than
28+0, 30+0, 32+0, 34+0 and 36+0 weeks)

241
Table 71: GRADE profile for comparison of corticosteroids versus placebo or expectant management
Number of women or
Quality assessment babies Effect
Placebo
or
Other expectan
Inconsiste Indirectne Impreci considerati Corticost t manage Relative
Number of studies Design Risk of bias ncy ss sion ons eroids ment (95% CI) Absolute (95% CI) Quality
Fetal and neonatal deathsa
1 meta-analysis of 14 studies Randomised No serious Serious1 No Serious2 Repeat 262/1957 344/1945 RR 0.77 41 fewer per 1000 Low
(Porto 2011; Roberts 2013) trials risk of bias serious course: (13.4%) (17.7%) (0.66 to (from 21 fewer to 60
indirectn 4 studies 60/465 99/453 0.88) fewer)
ess Multiples: 8
studies
Fetal and neonatal deaths – P-PROM at first dose (subgroup analysis)
(Roberts 2013) trials risk of bias serious courses: (14.9%) (24.1%) (0.46 to (from 43 fewer to
indirectn 2 studies 21/95 42/89 0.82) 130 fewer)
ess Multiples: 2
studies
Fetal and neonatal deaths – gestational age at birth <28 weeks (subgroup analysis)
1 meta-analysis of 2 studies Randomised No serious No serious No Serious2 Repeat 39/60 53/69 RR 0.81 146 fewer per 1000 Modera
(Roberts 2013) trials risk of bias3 inconsisten serious courses: (65%) (76.8%) (0.65 to (from 269 fewer to 8 te
cy indirectn 0 studies 1.01) more)
ess Multiples: 2
studies
1 study Randomised No serious No serious No Serious Repeat 59/99 71/102 RR 0.86 97 fewer per 1000 Modera
(Roberts 2013) trial risk of bias inconsisten serious courses: (59.6%) (69.6%) (0.7 to (from 209 fewer to te
cy indirectn 0 studies 35 more)
1.05)
ess Multiples: 1
study
(Roberts 2013) trial risk of bias serious courses: (35.7%) (49.3%) (0.57 to (from 59 fewer to
indirectn 0 studies 212 fewer)
0.88)
ess
Number of women or
Placebo
or
Other expectan
Multiples: 2
studies
1 study Randomised No serious No serious No Serious2 Repeat 90/312 113/286 RR 0.73 107 fewer per 1000 Modera
cy indirectn 0 studies 0.91) 166 fewer)
ess Multiples: 1
study
1 meta-analysis of 2 studies Randomised No serious Very No Serious2 Repeat 107/498 135/471 RR 0.75 72 fewer per 1000 Very
(Roberts 2013) trials risk of bias serious4 serious courses: (21.5%) (28.7%) (0.61 to (from 17 fewer to low
indirectn 0 studies 0.94) 112 fewer)
ess Multiples: 2
studies
Fetal and neonatal deaths – mean gestational age <28 weeks at trial entry (subgroup analysis)
1 meta-analysis of 2 studies Randomised No serious NR No Very Repeat NR NR RR 0.98 NC Low
(Onland 2011) trials risk of bias serious serious6 courses: 2 (0.57 to
indirectn studies 1.67)
ess Multiples:
2 studies
Fetal and neonatal deaths– first dose given before 26 weeks gestation (subgroup analysis)
1 study Randomised No serious No serious No Very None 15/23 17/26 RR 1.00 0 fewer per 1000 Low
(Roberts 2013) trials risk of bias inconsisten serious serious6 (65.2%) (65.4%) (0.66 to (from 222 fewer to
cy indirectn 1.50) 327 more)
ess
Fetal and neonatal deaths – first dose given between 26 and <30 weeks gestation (subgroup analysis)
1 study Randomised No serious No serious No Serious2 None 50/140 54/121 RR 0.80 89 fewer per 1000 Modera
(Roberts 2013) trial risk of bias inconsisten serious (35.7%) (44.6%) (0.59 to (from 183 fewer to te
ess
Fetal and neonatal death – first dose given between 30 and <33 weeks gestation (subgroup analysis)
Number of women or
Placebo
or
Other expectan
(Roberts 2013) trial risk of bias inconsisten serious (11.5%) (19.5%) (0.35 to (from 127 fewer to 2 te
cy indirectn 1.01) more)
ess
Fetal and neonatal death – first dose given between 33 and <35 weeks gestation (subgroup analysis)
1 study Randomised No serious No serious No Very None 18/168 18/185 RR 1.10 10 more per 1000 Low
(Roberts 2013) trial risk of bias inconsisten serious serious6 (10.7%) (9.7%) (0.59 to (from 40 fewer to
ess
Fetal and neonatal death - first dose given between 35 and <37 weeks gestation (subgroup analysis)
(Roberts 2013) trial risk of bias inconsisten serious serious6 (3.4%) (2.8%) (0.25 to (from 21 fewer to
ess
Fetal and neonatal death - first dose given after 36 weeks gestation (subgroup analysis)
1 study Randomised No serious No serious No Very None 3/18 0/24 RR 9.21 NC Low
(Roberts 2013) trial risk of bias inconsisten serious serious6 (16.7%) (0%) (0.51 to
cy indirectn 167.82)
ess
Intraventricular haemorrhage – all grades
1 meta-analysis of 13 studies Randomised No serious Serious1 No No Repeat 88/1445 155/1427 RR 0.54 50 fewer per 1000 Modera
(Roberts 2013) trials risk of bias serious serious courses: 6 (6.1%) (10.9%) (0.43 to (from 34 fewer to 62 te
indirectn imprecisi studies 0.69) fewer)
ess on Multiples: 7
studies
Intraventricular haemorrhage - premature rupture of membranes at first dose (subgroup analysis)
(Roberts 2013) trials risk of bias inconsisten serious courses: (4.2%) (8.6%) (0.28 to (from 18 fewer to 62 te
cy indirectn 3 studies 15/195 31/182 0.79) fewer)
ess Multiples: 2
studies
Intraventricular haemorrhage - gestational age at birth <28 weeks (subgroup analysis)
Number of women or
Placebo
or
Other expectan
cy indirectn 0 studies 0.86) 369 fewer)
ess Multiples: 1
study
cy indirectn 0 studies 1.1) 26 more)
ess Multiples: 1
study
(Roberts 2013) trial risk of bias inconsisten serious courses: (9%) (17.3%) (0.28 to (from 2 fewer to 125 te
cy indirectn 0 studies 0.99) fewer)
ess Multiples: 1
study
(Roberts 2013) trial risk of bias inconsisten serious courses: (5.9%) (11.2%) (0.29 to (from 6 fewer to 79 te
cy indirectn 0 studies 0.95) fewer)
ess Multiples: 1
study
(Roberts 2013) trial risk of bias inconsisten serious courses: (4.1%) (7.2%) (0.31 to (from 50 fewer to 1 te
cy indirectn 0 studies 1.02) more)
ess Multiples: 1
study
Intraventricular haemorrhage (all grades) – mean gestational age <28 weeks at trial entry in non-intervention arm (subgroup analysis)
Number of women or
Placebo
or
Other expectan
1 meta-analysis of 2 studies Randomised No serious NR No Very Repeat NR NR RR 0.90 NC Low
(Onland 2011) trials risk of bias serious serious6 courses: 2 (0.45 to
ess Multiples: 2
studies
Intraventricular haemorrhage - first dose given before 26 weeks gestation (subgroup analysis)
(Roberts 2013) trials risk of bias inconsisten serious serious6 (20%) (16.7%) (0.24 to (from 127 fewer to
ess
Intraventricular haemorrhage – first dose given between 26 to 29+6weeks gestation (subgroup analysis)
(Roberts 2013) trial risk of bias inconsisten serious (7.4%) (16.7%) (0.21 to (from 8 fewer to 132 te
cy indirectn 0.95) fewer)
ess
Intraventricular haemorrhage – first dose given between 30 and <33 weeks gestation (subgroup analysis)
1 study Randomised No serious No serious No Very None 1/155 4/140 RR 0.23 22 fewer per 1000 Low
(Roberts 2013) trial risk of bias inconsisten serious serious6 (0.65%) (2.9%) (0.03 to (from 28 fewer to 29
ess
(Roberts 2013) trial risk of bias inconsisten serious serious6 (1.9%) (1.7%) (0.23 to (from 13 fewer to 74
ess
1 study Randomised No serious No serious No Very none 0/85 0/106 NC NC Low
(Roberts 2013) trial risk of bias inconsisten serious serious10 (0%) (0%)
cy indirectn
ess
Intraventricular haemorrhage – first dose given after 36 weeks gestation (subgroup analysis)
Number of women or
Placebo
or
Other expectan
1 study Randomised No serious No serious No Very None 0/18 0/24 NC NC Low
(Roberts 2013) trial risk of bias inconsisten serious serious10 (0%) (0%)
cy indirectn
ess
Intraventricular haemorrhage – grades 3 or 4b (subgroup analysis)
1 meta-analysis of 4 studies Randomised No serious No serious No No Repeat 6/186 30/187 RR 0.22 125 fewer per 1000 High
(Garite 1992; Lewis 1996; trials risk of bias inconsisten serious serious courses: 4 (3.2%) (16%.0%) (0.10 to (from 82 fewer to
Morales 1989; Silver 1996) cy indirectn imprecisi studies 3/125 15/117 0.49) 144 fewer)
ess on Multiples:
2 studies
Intraventricular haemorrhage grades 3 or 4 – mean gestational age <28 weeks at trial entry in non-intervention arm (subgroup analysis)
1 meta-analysis of 2 studies Randomised No serious NR No No Repeat NR NR RR 0.20 NC Modera
(Onland 2011) trials risk of bias serious serious courses: 2 (0.06 to te
indirectn imprecisi studies 0.64)
ess on Multiples: 2
studies
Chronic lung disease
1 meta-analysis of 6 studies Randomised No serious Very No Serious2 Repeat 48/413 50/405 RR 0.86 17 fewer per 1000 Very
(Roberts 2013) trials risk of bias serious4 serious courses: 4 (11.6%) (12.3%) (0.61 to (from 48 fewer to 27 low
indirectn studies 1.22) more)
ess Multiples:
3 studies
Chronic lung disease – premature rupture of membranes at first dose (subgroup analysis)
1 study Randomised Serious5 No serious No Serious2 Repeat 23/87 41/78 RR 0.5 263 fewer per 1000 Low
(Roberts 2013) trial inconsisten serious courses: (26.4%) (52.6%) (0.33 to (from 126 fewer to
cy indirectn 1 study 0.76) 352 fewer)
ess Multiples: 0
studies
Bronchopulmonary dysplasia at 28 days postnatal age
Number of women or
Placebo
or
Other expectan
1 meta-analysis of 2 studies Randomised No serious NR No Serious2 Repeat NR NR RR 1.18 NC Modera
(Onland 2011) trials risk of bias serious courses: 2 (0.78 to te
ess Multiples: 2
studies
Need for mechanical ventilation/CPAP
(Porto 2011; Roberts 2013) trials risk of bias inconsisten serious courses: 2 (14.9%) (22.5%) (0.54 to (from 20 fewer to te
cy indirectn studies 35/301 58/284 0.91) 103 fewer)
ess Multiples: 2
studies
Need for mechanical ventilation/CPAP – premature rupture of membranes at first dose (subgroup analysis)
1 study Randomised No serious No serious No Very Repeat 15/105 16/101 RR 0.9 16 fewer per 1000 Low
(Roberts 2013) trial risk of bias inconsisten serious serious6 courses: (14.3%) (15.8%) (0.47 to (from 84 fewer to
ess Multiples: 1
study
Neonatal sepsis
(Porto 2011; Roberts 2013) trials risk of bias inconsisten serious courses: 2 (4.7%) (8.3%) (0.39 to (from 14 fewer to 51 te
cy indirectn studies 19/267 37/253 0.83) fewer)
ess Multiples: 3
studies
Neonatal sepsis – premature rupture of membranes at first dose (subgroup analysis)
1 meta-analysis of 2 studies Randomised No serious No serious No Very Repeat 11/128 11/123 RR 0.96 4 fewer per 1000 Low
(Roberts 2013) trials risk of bias inconsisten serious serious6 courses: (8.6%) (8.9%) (0.44 to (from 50 fewer to
cy7 indirectn 1 study 2.12) 100 more)
ess Multiples: 1
study
Neonatal sepsis – mean gestational age <28 weeks at trial entry in non-intervention arm (subgroup analysis)
Number of women or
Placebo
or
Other expectan
1 study (Onland 2011) Randomised No serious No serious No Very Repeat NR NR RR 0.40 NC Low
trial risk of bias inconsisten serious serious6 courses: (0.04 to
cy indirectn unclear 3.70)
ess Multiples:
Unclear
Cerebral palsy in childhood (at 2-year follow up)
1 meta-analysis of 5 studies Randomised No serious No serious No Serious2 Multiples: 4 20/490 28/414 RR 0.6 27 fewer per 1000 Modera
(Roberts 2013) trials risk of bias inconsisten serious studies (4.1%) (6.8%) (0.34 to (from 45 fewer to 2 te
ess
Cerebral palsy in childhood (at 2-year follow up) (subgroup analysis with studies without repeated courses of corticosteroids
Roberts 2013 Randomised No serious No serious No Very Repeat 1/60 2/34 RR 0.28 42 fewer per 1000 Low
trials risk of bias inconsisten serious serious6 courses: 1 (0.03 to (from 57 fewer to
cy indirectn study 3.01) 123 more)
ess
Visual impairment in childhood (at 2-year follow up)
(Roberts 2013) trials risk of bias inconsisten serious courses: (9%) (16.7%) (0.24 to (from 127 fewer to te
ess Multiples: 2
studies
Hearing impairment in childhood (at 2-year follow up)
(Roberts 2013) trials risk of bias inconsisten serious serious6 courses: (1%) (1.5%) (0.04 to (from 15 fewer to
cy7 indirectn 0 studies 9.87) 134 more)
ess Multiples: 2
studies
Neurodevelopmental delay in childhood (at 24-month follow up ; defined as tetraplegic cerebral palsy and/or a score <70 on Bayley Scales for 2-year children)c
Number of women or
Placebo
or
Other expectan
(Roberts 2013) trial risk of bias inconsisten serious serious6 courses: (6%) (9.4%) (0.14 to (from 81 fewer to
ess Multiples: 1
study
Developmental delay in childhood (at 18 to 24-month follow up ; defined as Psychomotor Developmental Index of the Bayley Scales at 18 months of age (50 ≤ Index ≤ 67)d
(Roberts 2013) trials risk of bias inconsisten serious courses: (4.1%) (7.5%) (0.24 to 1) (from 57 fewer to 0 te
1 study remains when cy indirectn 1 study 4/60 7/34 RR 0.32 more)
multiples removed – no ess Multiples: 1 (0.10 to
effect remains, confidence study 1.03)
interval widens
Intellectual impairment in childhood (at 18 to 24-month follow up ; defined in Collaborative 1981 as Mental Developmental Index of the Bayley Scales at 18 months of age (50 ≤ Index
≤ 67); Liggins 1972 ≤ 70 on Stanford-Binet Intelligence Scale and in Schutte 1980 as <70 on Weschler Intelligence Scale for Children-Revised full-scale IQ)
1 meta-analysis of 3 studies Randomised No serious No serious Serious8 Serious2 Repeat 16/409 17/369 RR 0.86 6 fewer per 1000 Low
(Roberts 2013) trials risk of bias inconsisten courses: (3.9%) (4.6%) (0.44 to (from 26 fewer to 32
cy 0 studies 1.69) more)
Multiples: 3
studies
Behavioural/learning difficulties in childhood (at 24-month follow up ; defined as children who had to repeat a class or required special educationf
(Roberts 2013) trial risk of bias inconsisten serious serious6 courses: (16.7%) (19.4%) (0.35 to (from 126 fewer to
ess Multiples: 1
study
Maternal death (where all women had severe preeclampsia) g
(Roberts 2013) trials risk of bias inconsisten serious serious6 courses: (0.53%) (0.56%) (0.06 to (from 5 fewer to 82
cy9 indirectn 1 study 15.5) more)
ess Multiples: 2
studies
Side-effects of therapy in women
Number of women or
Placebo
or
Other expectan
1 study Randomised No serious No serious No Very Repeat 0/50 0/51 NC NC Low
(Roberts 2013) trial risk of bias inconsisten serious serious8 courses: (0%) (0%)
cy indirectn 0 studies
ess Multiples: 1
study
Puerperal sepsis
1 meta-analysis of 8 studies Randomised No serious Serious1 No serious2 Repeat 57/496 44/507 RR 1.35 30 more per 1000 Low
(Roberts 2013) trials risk of bias serious courses: (11.5%) (8.7%) (0.93 to (from 6 fewer to 82
indirectn 4 studies 1.95) more)
ess Multiples: 2
studies
Puerperal sepsis – premature rupture of membranes at first dose (subcommittee analysis)
1 meta-analysis of 4 studies Randomised No serious Serious1 No Very Repeat 16/242 14/235 RR 1.11 7 more per 1000 Very
(Roberts 2013) trials risk of bias serious serious6 courses: 2 (6.6%) (6%) (0.55 to (from 27 fewer to 74 low
indirectn studies 2.25) more)
ess Multiples: 2
studies
CI confidence interval, CPAP continuous positive airway pressure, MID minimally important difference, NC not calculable, NR not reported, RR relative risk
1. Evidence was downgraded by 1 due to serious heterogeneity (chi-squared p<0.1, I-squared inconsistency statistic of 50%-74.99%)
3. In one trial (contributes 15% to meta-analysis) significantly lower gestational age in control than experimental group at entry to study and birth, significantly more women in
control group received tocolysis
4. Evidence was downgraded by 2 due to very serious heterogeneity (chi-squared p<0.1, I-squared inconsistency statistic of >75%)
5. Unclear method of randomisation and allocation concealment
6. Evidence was downgraded by 2 due to very serious imprecision as 95% confidence interval crossed 2 default MIDs 8
7. Each of the 3 trials used a different scale to measure intellectual impairment (3/3 trials included women with a multiple pregnancy)
8. Confidence interval could not be calculated: zero events in both arms of the trial so effect estimate cannot be calculated; trial underpowered for outcome
Magnesium sulfate for neuroprotection
11.2.4.1 Neonatal outcomes

Fetal and neonatal deaths
In the overall analysis, findings from a meta-analysis of 14 RCTs (n=3902) suggested there
were significantly fewer fetal and neonatal deaths in the group of women who had received
corticosteroids compared with women who had placebo or expectant management.
Subgroup analysis in women with P-PROM had also showed significantly fewer fetal and
neonatal deaths in babies born to women who had received corticosteroids compared with
women who had placebo or expectant management.
Subgroup analysis by gestational age at birth suggested that there were significantly fewer
fetal and neonatal deaths in babies born at less than 32 weeks, less than 34 weeks and less
than 36 weeks to women who had corticosteroids compared with placebo or expectant
management, but there was no significant difference in fetal and neonatal deaths in babies
born at less than 28 weeks or less than 30 weeks. Subgroup analysis by gestational age at
first dose of corticosteroids suggested that there was no significant difference in fetal and
neonatal death in the steroid versus the placebo group at any gestational age cut-off.
The evidence was of moderate to low quality across all outcomes.
Intraventricular haemorrhage
Evidence from a meta-analysis of 13 RCTs (n=2872) suggested there were significantly
fewer any grade of intraventricular haemorrhage and a significantly lower proportion of grade
III or IV intraventricular haemorrhage in babies born to women who had received
Subgroup analysis in women with P-PROM had also showed fewer intraventricular
haemorrhage in babies born to women who had received corticosteroids compared with
women who had placebo or expectant management. Subgroup analysis by gestational age
at birth from 1 study suggested that there were fewer babies with intraventricular
haemorrhage born at less than 28 weeks, less than 32 weeks and less than 34 weeks to
women who had corticosteroids compared with placebo or expectant management, but
subgroup analysis from 1 other study showed no significant difference in intraventricular
haemorrhage in babies born at less than 30 weeks and less than 36 weeks.
Fewer babies who were between 26 weeks’ and 30 weeks’ gestation at first corticosteroid
dose had intraventricular haemorrhage compared with placebo or expectant management.
Further analysis of trials where the mean gestational age of babies was less than 28 weeks
found that a significantly lower proportion of babies born to women who had received
corticosteroids had an intraventricular haemorrhage grade III and IV compared with babies
born to women who had received placebo or expectant management. No significant
difference was found in the number of babies born with any grade of intraventricular
haemorrhage at 28 days.
The quality of the evidence for all outcomes ranged from moderate to low (with the majority
rated moderate).
Bronchopulmonary dysplasia/chronic lung disease

Evidence from a meta-analysis of 2 RCTs (number of participants not reported) suggested
there was no significant difference in bronchopulmonary dysplasia at 28 days following birth

252
in babies born to women who had received corticosteroids compared with women who had
placebo or expectant management. Evidence from meta-analysis of 5 RCTs (n=818)
suggested there was no significant difference in chronic lung disease in babies born to
women who had received corticosteroids compared with women who had placebo or
expectant management. However, subgroup analysis in women with P-PROM showed
significantly lower rate of chronic lung disease in babies born to women who had received
corticosteroids compared with women who had placebo or expectant management. The
evidence was of low to very low quality.
Need for mechanical ventilation
Moderate and low quality evidence from meta-analysis of 6 RCTs (n=844) suggested there
was significantly less need for mechanical ventilation in babies born to women who had
received corticosteroids compared with women who had placebo or expectant management.
However, subgroup analysis in women with P-PROM showed no significant difference in
need for mechanical ventilation in babies born to women who had received corticosteroids
compared with women who had placebo or expectant management.
Neonatal sepsis
Moderate and low quality evidence from meta-analysis of 6 RCTs (n=1594) suggested there
were significantly lower rates of neonatal sepsis in babies born to women who had received
However, subgroup analysis in women with P-PROM and by gestational age at less than 28
weeks showed no significant difference in neonatal sepsis in babies born to women who had
received corticosteroids compared with women who had placebo or expectant management.
Neurodevelopmental disability
There was no difference in the proportion of babies who developed cerebral palsy at 2 year
follow-up (5 studies, n=904) or the proportion of children with visual or hearing impairment (2
studies, n=166), neurodevelopmental delay (1 study, n=82), developmental delay (2 studies,
n=518), intellectual impairment (3 studies, n=778) or behavioural/learning difficulties (1 study,
n=90) born to women who had corticosteroids compared with women who had placebo or
expectant management. The quality of the evidence ranged from moderate to very low.
Maternal outcomes
Low quality evidence from meta-analysis of 3 RCTs (n=365) showed there was no significant
difference in maternal death (where all women had severe pre-eclampsia) and in puerperal
sepsis (8 studies, n=1003) between women who had corticosteroids compared with women
who had placebo or expectant management. One study (n=101) reported that there were no
side effects in both women who had corticosteroids and women who had placebo or
expectant management. The evidence was of low quality.
11.2.4.2 Further sensitivity analyses

Single and repeat course trials
Sensitivity analysis was conducted to test any differences in the direction of results from only
single-course corticosteroid trials. When we excluded the studies with repeat-course
corticosteroids studies from the analysis, we found the following differences in the results of
the main analysis:
• The beneficial effect of corticosteroids on fetal and neonatal deaths and intraventricular
haemorrhage remained significant.

253
• The beneficial effect of corticosteroids on the need for mechanical ventilation and
neonatal sepsis was no longer significant.
• The lack of difference observed in effect on chronic lung disease, cerebral palsy,
developmental delay and puerperal sepsis remained.
Singleton and multiple pregnancy trials
Sensitivity analysis was done to compare the results of trials with only singleton pregnancies
and trials with mixed populations (both single and multiple pregnancies). When trials that
only included women with single pregnancies were considered, the beneficial effect of
corticosteroids on fetal and neonatal deaths, intraventricular haemorrhage (all grades and
grades 3 and 4), need for mechanical ventilation and neonatal sepsis remained significant.
When trials that included women with multiple pregnancies were excluded from the analysis,
corticosteroids were shown to have a beneficial effect in reducing chronic lung disease.

A single search was undertaken for health economic evidence on a course or repeat courses
of maternal corticosteroids for fetal lung maturation given at different gestations in improving
preterm neonatal outcomes. A total of 136 articles were identified by the search. After
reviewing titles and abstracts, 9 full papers were obtained and 1 was included for review.
An old UK study (Mugford 1991) used a decision analytic approach to assess the cost
effectiveness of antenatal corticosteroids relative to no treatment in order to prevent death of
a preterm infant and to prevent respiratory distress syndrome in surviving preterm infants.
The model considered the cost effectiveness of treatment across 2 different populations of
women with threatened preterm labour: less than 31 weeks’ gestation and less than 35
weeks’ gestation. The authors reported that antenatal corticosteroids was dominant (cheaper
and more effective) across the 2 different populations.
This question was initially prioritised for health economic analysis although new analysis was
ultimately not undertaken as there were more important topics to be addressed in terms of
health economic analysis. Furthermore, corticosteroids are relatively cheap and a single
course is current practice. The review did not find clinical evidence to suggest that repeat
courses were beneficial and therefore there cannot be economic evidence that would justify
the routine use of repeat courses.

Given that the primary intention of administering maternal corticosteroids is to accelerate
fetal lung maturation in babies who are likely to be born preterm, the Guideline Development
Committee prioritised measures that would be likely to indicate whether the drug had been
successful in achieving this end. This included any pulmonary-specific adverse event or need
for mechanical ventilation and bronchopulmonary dysplasia/chronic lung disease, as well as
proxy measures of neonatal lung disease (mortality, all death up to 1 year),
neurodevelopmental disability and intraventricular haemorrhage, or white matter injury
(periventricular leucomalacia [PVL]). Although the committee agreed that all manifestations
of neurodevelopmental disability could be reported as a single outcome, the data were
available for particular developmental disabilities and results were presented individually for
each type of disability.
Neonatal sepsis was prioritised as an important outcome because corticosteroids are used in
other contexts to suppress the body’s normal immune response to infection. Therefore the

254
committee considered that a rise in infection rates might be a potential, unintended adverse
outcome of steroid use.
In terms of maternal outcomes, the committee prioritised mortality because they felt any
change in the incidence of this outcome would affect clinical decision-making. They also
agreed that the composite of all maternal adverse events was important. They were aware of
the possible effect of maternal corticosteroids on blood sugar control in women with diabetes
and the potential for adverse events related to the immunosuppressive effect of
corticosteroids.

There was consistent evidence from randomised trials that maternal corticosteroids are
beneficial from 26 weeks’ gestation in terms of reducing neonatal morbidity including
intraventricular haemorrhage (IVH, all grades and grade III or IV), need for mechanical
ventilation and neonatal sepsis. Given that some of these benefits (specifically reduced IVH
and lung disease of prematurity) were found for gestations down to 26 weeks, the Guideline
Development Committee considered that they would also probably apply to babies born
before 26 weeks’ gestation.
There was also some evidence that maternal corticosteroids may reduce fetal and neonatal
mortality from 26 weeks’s gestation but the committee interpreted these results with caution
due to low quality of the evidence and the concerns around its generalisation. Also these
studies are likely to be underpowered for low event rate outcomes such as these.
The committee was reassured that any concerns regarding the potential risk of neonatal
sepsis were not supported by the included evidence. Infection rates were reduced in the
intervention group (corticosteroids) compared with the ‘no treatment’ arm. It was noted that
the reduction in the incidence of neonatal sepsis was less marked in the sub-group of women
who had preterm prelabour rupture of membranes (P-PROM). However, the committee
concluded that the potential risk of infection in this sub-group does not negate the benefits of
antenatal corticosteroids.
There was limited evidence available for any of the outcomes when maternal corticosteroids
were given before 26 weeks’ gestation. There were only data for 2 of the prioritised
outcomes: fetal and neonatal deaths, and intraventricular haemorrhage. Nevertheless, the
committee considered it reasonable to extrapolate from later gestational ages down to
23 weeks because the physiology of lung disease and IVH are the same at 23–25 weeks’
gestation as at 26 weeks’ gestation. Therefore the committee considered that it would be
plausible to expect the same benefits to be seen at earlier gestations as were reported for
babies of 26 weeks’ gestation and above. However, it was acknowledged that any benefits
from the use of maternal corticosteroids in this group may be relatively small in the context of
the significant morbidities and mortality rate of babies who are born at such extremely
preterm gestations. The committee noted that in current clinical practice the decision to give
maternal corticosteroids at extreme preterm gestations is often influenced by the decision on
whether or not to offer neonatal life support at birth, following discussion with parents
regarding the risks and likely outcomes. The committee considered that this should be
discussed carefully on an individual basis with the woman and her partner and with reference
to other relevant guidance such as that published by the British Association of Perinatal
Medicine (2008) and the Royal College of Obstetrics and Gynaecology (2014).
There was no reliable evidence of benefit of antenatal corticosteroids in terms of fetal or
neonatal death, intraventricular haemorrhage or chronic lung disease after 36 weeks’
gestation. The effectiveness of antenatal corticosteroids in reducing requirement for
ventilation or pressure support has not been reported. The committee considered that any

255
potential short-term benefits to the baby of antenatal corticosteroid administration should be

balanced against potential risk of delaying delivery at these gestations.

The committee noted that corticosteroids are relatively inexpensive when viewed in isolation
but it was necessary to take into account the cost of the other associated management
strategies (such as the use of drugs to delay birth to allow corticosteroids to be administered)
when assessing costceffectiveness. The cost of the combined strategy was considered in the
section on tocolysis.
However, the committee considered that the use of maternal corticosteroids in women whose
babies were most likely to benefit had the potential to reduce the number of ventilated days
and that this would result in a cost saving that would outweigh the initial costs incurred.

The evidence included in the review was mostly of moderate or low quality. The Guideline
Development Committee had some reservations about the appropriateness of included
studies. The committee considered that although the participants in the included trials
reflected the study population of the guideline, within these trials were many sub-populations
who might potentially vary in terms of their response to corticosteroids (for example women
in spontaneous preterm labour and those having a planned preterm birth). The committee
acknowledged that multiple sub-group analyses would not give any precision in the estimate
of effects. The sensitivity analysis excluding women with multiple pregnancy indicated that
findings were broadly the same as those in overall analysis and hence separate
recommendations were not warranted.
The committee had concerns regarding the relevance of data derived from trials that were
conducted in the 1970s. The committee noted that clinical practice had changed significantly
in the interim and therefore these findings may not be relevant to current practice. For
example, the availability of neonatal interventions in contemporary practice may lessen the
impact of maternal corticosteroid administration to prevent death in preterm neonates.
However, the committee considered that, because more preterm babies are surviving, the
effects of corticosteroids on neonatal complications may have become more important. In
light of these considerations, the committee considered that it was reasonable to consider
this evidence in the context of the meta-analyses where data from more recent studies
contributed an equal or dominant amount of data to the overall result.
The committee also noted that the SR by Onland (2011) added little value in terms of specific
information regarding the effectiveness of steroids a lower gestational ages (before 26
weeks’ gestation) because the data were only reported in terms of 2 gestational age groups
– those with a mean gestational age of 28 weeks or less and those with a mean gestational
age of over 28 weeks. However, although the data were included to give an estimate of
subgroup analysis for lower gestational ages, results should be interpreted with caution given
the effect of data double counting (as some of the data were also included in the SR by
Roberts (2013). It was highlighted that the length of follow-up in the included studies was too
short to report accurately outcomes such as cerebral palsy and learning difficulties, but the
consensus view of the committee was that IVH grades 3 and 4 are likely to lead to these
outcomes and could therefore be accepted as useful short-term proxies.

The Guideline Development Committee noted the lack of available evidence on which to
judge the optimal timing of administration of corticosteroids in relation to the time of birth and
particularly the ‘latest point’ at which the drug could most effectively be given. The committee

256
acknowledged that this had not been prioritised as an aim of the review but, taking account
of the drug’s pharmacological mechanism of action, the committee suspected that any
benefits would be likely to be transferred even if there was only a limited amount of time
(such as less than 24–48 hours) between administration and time of birth. However, the
committee could not make any recommendations to this effect.
The committee was aware of the existence of 2 other studies (EPICURE and EPIPHASE)
that did not meet the inclusion criteria but might provide further information about use of
corticosteroids at low gestations.

There was sufficient evidence of benefit without concomitant harm to justify a strong
recommendation for the use of corticosteroids in women who are thought to be in
spontaneous preterm labour, having planned a preterm birth or have preterm prelabour
rupture of membranes between 26 and 34 weeks’ gestation. The Guideline Development
Committee concluded that some of these benefits would be seen in babies born at lower and
higher gestational ages, but that the evidence was less robust at these gestations. The
extrapolation of findings to groups outside the gestational age range of 26–34 weeks was
more complex in terms of clinical effectiveness, which warranted less strong
recommendations at gestations below 26 weeks and above 34 weeks.
11.2.13 Recommendations
The recommendations on corticosteroids are in Section 11.4.
11.3 Repeat course corticosteroids for fetal lung

maturation
updated guideline.
updated guideline.
12 Magnesium sulfate for

neuroprotection
12.1 Introduction
With advances in neonatal care in recent years, more babies born preterm are surviving and
in particular those born at extremely preterm gestations now commonly survive the neonatal
period (over 90% of those born before 28 weeks in high income countries, Blencow 2013).
These children frequently suffer long-term complications of prematurity. Neurological effects
are common and may cause severe disability. They include cerebral palsy (which is
associated with intraventricular haemorrhage and periventricular leukomalacia),
developmental delay, cognitive problems, and behavioural and learning difficulties. Agents to
protect the developing fetal brain would therefore have great potential to reduce disability.

257

What is the clinical and cost effectiveness of magnesium sulfate given to women at high risk
of giving birth preterm (defined as those suspected to be in preterm labour or diagnosed as
being in preterm labour and those having planned preterm birth) for preventing cerebral palsy
and other neurological disorders in babies born at different preterm gestations?

Six studies were included in this review (Crowther 2003, Marret 2007, Marret 2008,
Mittendorf 2002, Rouse 2008, Doyle 2014). Four of them were randomised controlled trials
(RCTs) (Crowther 2003, Marret 2007, Mittendorf 2002, Rouse 2008) and 2 studies (Marret
2008, Doyle 2014) were follow-up studies of two of the included RCTs (Marret 2007,
Crowther 2003).
The settings of studies were France (Marret 2007, Marret 2008), the USA (Mittendorf 2003,
Rouse 2008), and Australia and New Zealand (Crowther 2003, Doyle 2014).
All of the included studies evaluated the use of magnesium sulfate (MgSO4) given to women
at high risk of preterm labour, either because they were already thought to be in labour (with
or without ruptured membranes) or they were having a planned preterm birth within 24 hours
(only 1 study did not define the timing of planned preterm birth). Only trials where
magnesium sulfate was given for the purposes of neuroprotection of the baby were included.
One trial included women at between over 24 and less than 34 completed weeks’ gestation
(Mittendorf 2003), 1 trial included all women at less than 33 weeks’ gestation (Marret 2007),
1 trial included women at 24 to 31 weeks’ gestation (Rouse 2008) and 1 trial only included
women at less than 30 weeks’ gestation (Crowther 2003). We reported separate results from
sub-group analyses by the number of weeks of gestation at randomisation as a proxy to
estimate any difference in the effect of magnesium sulfate by gestation of babies at birth. The
administration of the intervention (magnesium sulfate) started from 16 to 36 weeks of
gestation.
The initial (loading) dose of magnesium sulfate was 4 gm in 3 trials (Crowther 2003, Marret
2007, Mittendorf 2002) and 6 gm in 1 trial (Rouse 2008). In 2 trials, women then received a
maintenance infusion of 1 gm/hour (Crowther 2003) or 2 gm/hour (Rouse 2008). For full
details see the evidence tables in Appendix H.

The following GRADE profiles present the comparison of magnesium sulfate with no
magnesium sulfate:
• Table 78:GRADE profile for comparison of magnesium sulfate (MgSO4) with no
magnesium sulfate
magnesium sulfate (long term child outcomes)
magnesium sulfate (maternal outcomes)
In 4 of the studies (Crowther 2003, Marret 2008, Rouse 2008, Doyle 2014) long-term
outcomes (cerebral palsy, gross motor dysfunction, developmental delay, cognitive
dysfunction, vision and hearing) were reported after excluding babies who died from the

258
denominator. Therefore, the risks reported by the authors represent the risk of the outcome
among children alive and available for follow-up, rather than being a reflection of the actual
risk of the specific outcome occurring following the decision to administer magnesium sulfate
or not (intention to treat analysis). Where possible, the denominator was changed to include
the babies who died (stillbirths + neonatal deaths before discharge + neonatal/paediatric
deaths after discharge) in the denominators, in order to provide an accurate reflection of
long-term risk to inform decision-making.
A full description of the characteristics and results of the included studies can be found in the

259
Table 72: GRADE profile for comparison of magnesium sulfate (MgSO4) with no magnesium sulfate (neonatal outcomes)
Number of women or
Other
Risk of Inconsistenc Indirect conside No Relative
Number of studies Design bias y ness Imprecision rations MgSO4 MgSO4 (95% CI) Absolute (95% CI) Quality
Stillbirth (randomised before 34 weeks)
1 meta-analysis of 3 studies Randomised No serious No serious Serious1 No serious None 16/2160 22/2214 RR 0.74 3 fewer per 1000 Modera
(Crowther 2003, Marret trials risk of bias inconsistency imprecision (0.74%) (0.99%) (0.39 to (from 6 fewer to 4 te
2007, Rouse 2008) 1.4) more)
Stillbirth (randomised before 30 weeks) (subgroup analysis)
Crowther 2003 Randomised No serious No serious Serious1 Very None 9/629 11/626 RR 0.81 3 fewer per 1000 Very
trials risk of bias inconsistency serious3 (1.4%) (1.8%) (0.34 to (from 12 fewer to 17 low
1.95) more)
Neonatal mortality: before discharge
2007, Rouse 2008) 1.18) more)
Neonatal mortality: before discharge (randomised before 30 weeks) (subgroup analysis)
Crowther 2003 Randomised No serious No serious Serious1 Serious2 None 76/629 92/626 RR 0.82 26 fewer per 1000 Very
trials risk of bias inconsistency (12.1%) (14.7%) (0.62 to (from 56 fewer to 13 low
1.09) more)
Neonatal/paediatric mortality: between discharge and follow-upa
1 meta-analysis of 2 studies Randomised No serious No serious Serious1 No serious None 20/1808 21/1878 RR 1 0 fewer per 1000 Modera
(Crowther 2003, Rouse trials risk of bias inconsistency imprecision (1.1%) (1.1%) (0.55 to (from 5 fewer to 9 te
2008) 1.84) more)
Neonatal/paediatric mortality: between discharge and follow-upa (randomised before 30 weeks) (subgroup analysis)
2.71) more)
Total perinatal, neonatal and paediatric mortalityb
2008, Mittendorf 2002, 1.13) more)
Rouse 2008)
Total perinatal, neonatal and paediatric mortalityb (randomised before 30 weeks) (subgroup analysis)
Number of women or
Other
Crowther 2003 Randomised No serious No serious Serious1 Serious2 None 87/629 107/626 RR 0.81 32 fewer per 1000 Low
trials risk of bias inconsistency (13.8%) (17.1%) (0.62 to (from 65 fewer to 9
1.05) more)
Total perinatal, neonatal and paediatric mortalityb (randomised at or after 28 weeks) a (subgroup analysis)
1 study Randomised No serious No serious Serious1 Very None 21/599 15/599 RR 1.4 10 more per 1000 Very
(Rouse 2008) trials risk of bias inconsistency serious3 (3.5%) (2.5%) (0.73 to (from 7 fewer to 42 low
2.69) more)
Total perinatal, neonatal and paediatric mortalityb (randomised before 28 weeks)a (subgroup analysis)
1 study Randomised No serious No serious Serious1 Serious2 None 78/442 78/496 RR 1.12 19 more per 1000 Low
(Rouse 2008) trials risk of bias inconsistency (17.6%) (15.7%) (0.84 to (from 25 fewer to 77
1.49) more)
Grades III or IV intracranial haemorrhagec (findings on cranial ultrasound)
1 meta-analysis of 3 studies Randomised No serious No serious Serious1 Serious2 None 72/1738 90/1799 RR 0.81 10 fewer per 1000 Low
(Crowther 2003; Mittendorf trials risk of bias inconsistency (4.1%) (5%) (0.6 to (from 20 fewer to 5
2002; Rouse 2008) 1.09) more)
Grades III or IV intracranial haemorrhagec (findings on cranial ultrasound) (randomised before 28 weeks) (subgroup analysis)
1.4) more)
Periventricular leukomalaciac (findings on cranial ultrasound)
1 meta-analysis of 3 studies Randomised No serious No serious Serious1 Very None 44/1738 48/1799 RR 0.94 2 fewer per 1000 Very
(Crowther 2003; Mittendorf trials risk of bias inconsistency serious3 (2.5%) (2.7%) (0.63 to (from 10 fewer to 11 low
2002; Rouse 2008) 1.4) more)
Periventricular leukomalaciac (findings on cranial ultrasound) (randomised before 30 week a) (subgroup analysis)
Crowther 2003 Randomised No serious No serious Serious1 Very None 22/596 21/586 RR 1.03 1 more per 1000 Very
1.85) more)
Cerebral palsy: anyd
1 meta-analysis of 4 studies Randomised No serious No serious Serious1 Serious2 None 102/2130 146/2184 RR 0.71 19 fewer per 1000 Low
(Crowther 2003; Marret trials risk of bias inconsistency (4.8%) (6.7%) (0.56 to (from 6 fewer to 29
2007; Mittendorf 2002; 0.91) fewer)
Rouse 2008)
Cerebral palsy: anyd (randomised before 30 week a) (subgroup analysis)
Number of women or
Other
1.32) more)
MDI Mental Development Index. RR risk ratios, SD standard deviation
a. These are the deaths reported as occurring between the point of initial discharge and later follow-up (1 year in the case of Rouse 2008 and 2 years in the case of Crowther
2003. (Note: Because of the way the data are split between the trial and follow-up papers, these data are not reported for the Marret trial)
b. Deaths are reported up to the age of 1 year in Rouse 2008 and 2 years in Crowther 2003 and Marret 2008. There was no long-term follow-up in Mittendorf 2002 and the point
at which deaths occurred is not reported.
c. Reported as a proportion of babies who received a cranial ultrasound
d. Mittendorf 2002 did not have long term follow-up. Follow-up was at 2 years in Crowther 2003, Marret 2008 and Rouse 2008.
e. Rouse 2008 reported this outcome for pregnancies rather than babies, and insufficient data are reported to convert it. Crowther 2003 reported the data for moderate and
severe cerebral palsy separately, and these were pooled by the NCC-WCH technical team.
f. Crowther 2003 reported data for minimal and substantial gross motor dysfunction separately, and these were pooled by the NCC-WCH technical team. 18/616 (2.9%) of
babies in the magnesium sulfate arm and 34/620 (5.5%) of babies in control arm had substantial gross motor dysfunction.
g. Developmental delay was defined according to Mental Development Index (MDI) scores. It was classified as: mild (MDI - 2 SDs to less than - 1 SD), moderate (MDI - 3 SDs
to - 2 SDs) or severe (MDI <3 SDs).
h. Children were classified as blind if their vision in both eyes was worse than 6/60
i. Children were classified as deaf if they required hearing aids
1. All trials in the meta-analysis included a proportion of women with multiple pregnancy
Table 73: GRADE profile for comparison of magnesium sulfate (MgSO4) with no magnesium sulfate (long term child outcomes)
Number of women
Quality assessment or babies Effect
studies Design Risk of bias Inconsistency Indirectness Imprecision considerations MgSO4 MgSO4 (95% CI) (95% CI) Quality
Cerebral palsy: moderate or severe (at 2 years)e
1 meta-analysis Randomised trials No serious No serious Serious1 Serious2 None 35/1661 59/1715 RR 0.61 13 fewer per Moderate
of 2 studies risk of bias inconsistency (2.1%) (3.4%) (0.4 to 1000
(Crowther 2003; 0.92) (from 3 fewer
Rouse 2008) to 21 fewer)
Number of women
Cerebral palsy: moderate or severe (at 2 years)e (randomised before 30 week) (subgroup analysis)
Crowther 2003 Randomised trials No serious No serious Serious1 Very serious3 None 15/620 21/620 RR 0.71 10 fewer per Very low
risk of bias inconsistency (2.4%) (3.4%) (0.37 to 1000
1.37) (from 21
fewer to 13
more)
Cerebral palsy: moderate or severe (at 2 years)e (Randomised at or after 28 week) (subgroup analysis)
1 study Randomised trials No serious No serious Serious1 Very serious3 None 8/599 8/599 RR 1 0 fewer per Very low
(Rouse 2008) risk of bias inconsistency (1.3%) (1.3%) (0.38 to 1000
2.65) (from 8 fewer
to 22 more)
Cerebral palsy: moderate or severe (at 2 years)e (randomised before 28 weeka) (subgroup analysis)
1 study Randomised trials No serious No serious Serious1 Serious2 None 12/442 30/496 RR 0.45 33 fewer per Low
(Rouse 2008) risk of bias inconsistency (2.7%) (6%) (0.23 to 1000
0.87) (from 8 fewer
to 47 fewer)
Cerebral palsy: school-age (6–11 years of age)
1 study (Doyle Randomised trials No serious No serious Serious1 Very serious3 None 23/295 21/314 RR 1.17 11 more per
2014) risk of bias inconsistency (7.8%) (6.7%) (0.66 to 1000 (from Very Low
2.06) 23 more to
71 more)
Gross motor dysfunction (at 2 years)f
1 meta-analysis Randomised trials No serious No serious Serious1 Sserious2 None 157/963 171/951 RR 0.91 16 fewer per Low
of 2 studies risk of bias inconsistency (16.3%) (18%) (0.74 to 1000
(Crowther 2003; 1.1) (from 47
Marret 2008) fewer to 18
more)
Gross motor dysfunction (at 2 years)f (randomised before 30 week) (subgroup analysis)
Crowther 2003 Randomised trials No serious No serious Serious1 Very serious3 None 102/616 107/620 RR 0.96 7 fewer per Very low
risk of bias inconsistency (16.6%) (17.3%) (0.75 to 1000
1.23) (from 43
fewer to 40
more)
Motor function: school-age (6–11 years of age)
Number of women
1 study (Doyle Randomised trials No serious No serious Serious1 Serious2 None 80/297 80/300 RR 1.01 267 fewer
2014) risk of bias inconsistency (26.9%) (26.7%) (0.77 to per 1000 Low
1.32) (from 267
more to 267
more)
Developmental delay: any (at 2 years)g
1 study Randomised trial No serious No serious Serious3 No serious None 176/581 170/585 RR 1.04 12 more per Moderate
(Crowther 2003) risk of bias inconsistency imprecision (30.3%) (29.1%) (0.87 to 1000
1.24) (from 38
fewer to 70
more)
Cognitive dysfunction (at 2 years)
1 study Randomised trial Serious4,5,6 No serious Serious7 Serious2 None 57/347 62/331 RR 0.88 22 fewer per Very low
(Marret 2008) inconsistency (16.4%) (18.7%) (0.63 to 1000
1.22) (from 69
fewer to 41
more)
Vision: blindness (at 2 years)h
1 study Randomised trial No serious No serious Serious3 No serious None 1/620 1/621 RR 1 0 fewer per Moderate
15.98) (from 2 fewer
to 24 more)
Hearing: deafness (at 2 years)i
1 study Randomised trial No serious No serious Serious3 No serious None 8/620 7/621 RR 1.14 2 more per Moderate
3.14) (from 7 fewer
to 24 more)
CI confidence interval, MDI Mental Development Index, MgSO4 magnesium sulfate, MID minimally important difference, NC not calculable, RR relative risk, SD standard
deviation
e. Rouse 2008 reported this outcome for pregnancies rather than babies, and insufficient data are reported to convert it. Crowther 2003 reported the data for moderate and
severe cerebral palsy separately, and these were pooled by the NCC-WCH technical team.
f. Crowther 2003 reported data for minimal and substantial gross motor dysfunction separately, and these were pooled by the NCC-WCH technical team. 18/616 (2.9%) of
babies in the magnesium sulfate arm and 34/620 (5.5%) of babies in control arm had substantial gross motor dysfunction.
g. Developmental delay was defined according to Mental Development Index (MDI) scores. It was classified as: mild (MDI - 2 SDs to less than - 1 SD), moderate (MDI - 3 SDs
to - 2 SDs) or severe (MDI < 3 SDs).
Table 74: GRADE profile for comparison of magnesium sulfate (MgSO4) with no magnesium sulfate (maternal outcomes)
Number of women
Maternal death
1 meta-analysis Randomised No serious risk No serious Serious1 Very serious None 0/1917 1/1950 RR 0.32 0 fewer per Very low
of 3 studies trials of bias inconsistency 4 (0%) (0.05%) (0.01 to 1000
(Crowther 2003; 7.92) (from 1 fewer
Marret 2007; to 4 more)
Rouse 2008)
Maternal adverse effects: any
1 meta-analysis Randomised Serious9 Serious3 Serious1 No serious None 1309/16 339/165 RR 3.82 579 more per Low
of 2 studies trials imprecision 13 2 (1.38 to 1000
(Crowther 2003; (81.2%) (20.5%) 10.59) (from 78 more
Rouse 2008) to 1000 more)
Maternal adverse effects: leading to stopping of infusion
1 meta-analysis Randomised No serious risk No serious Serious1 No serious None 123/161 44/1652 RR 2.81 48 more per Moderat
of 2 studies trials of bias inconsistency imprecision 3 (2.7%) (2.01 to 1000 e
(Crowther 2003; (7.6%) 3.93) (from 27 more
Rouse 2008) to 78 more)
Maternal adverse effects: cardiac or respiratory arrest
1 meta-analysis Randomised No serious risk No serious Serious1 No serious None 0/821 0/805 NC NC Moderat
of 2 studies trials of bias inconsistency imprecision (0%) (0%) e
(Crowther 2003;
Marret 2007)
Maternal adverse effects: drop in diastolic blood pressure of more than 15 mmHg
Number of women
1 study Randomised No serious risk No serious Serious5 Serious2 None 77/535 52/527 RR 1.46 45 more per Low
(Crowther 2003) trial of bias inconsistency (14.4%) (9.9%) (1.05 to 1000
2.03) (from 5 more
to 102 more)
Maternal adverse effects: hypotension
1 study Randomised Serious12 No serious Serious7 Very None 3/286 0/278 RR 6.8 NC Very low
(Marret 2007) trial inconsistency serious4 (1%) (0%) (0.35 to
131.14)
CI confidence interval, MDI minimally important difference, MgSO4 magnesium sulfate, NC not calculable, RR relative risk, SD standard deviation
e. Rouse 2008 reported this outcome for pregnancies rather than babies, and insufficient data are reported to convert it. Crowther 2003 reported the data for moderate andsevere
cerebral palsy separately, and these were pooled by the NCC-WCH technical team.
f. Crowther 2003 reported data for minimal and substantial gross motor dysfunction separately, and these were pooled by the NCC-WCH technical team. 18/616 (2.9%) ofbabies
in the magnesium sulfate arm and 34/620 (5.5%) of babies in control arm had substantial gross motor dysfunction.
g. Developmental delay was defined according to Mental Development Index (MDI) scores. It was classified as: mild (MDI - 2 SDs to less than - 1 SD), moderate (MDI - 3 SDsto -
2 SDs) or severe (MDI < 3 SDs).
3. Evidence was downgraded by 1 due to serious heterogeneity (chi-squared p<0.1, I-squared inconsistency statistic of 50%-74.99%)
5. 16.7% of women had a multiple pregnancy

Moderate quality evidence from 3 RCTs with a total sample size of over 3000 babies showed
that those whose mothers were treated with magnesium sulfate during pregnancy were
significantly less likely to be diagnosed with cerebral palsy (CP) compared with those who
were not treated. However, subgroup analysis demonstrated that this protective effect of
magnesium sulfate on the risk of cerebral palsy was not significant for those who were
randomised before 30 weeks’ gestation (very low quality evidence). No other neonatal
outcome (still birth, neonatal, perinatal and paediatric mortality, intracranial haemorrhage and
periventricular leukomalaciac) was found significantly different between the groups.
The risk of cerebral palsy at 2 years was also found significantly lower for those children
whose mothers were treated with magnesium sulfate during pregnancy compared with
control group (moderate quality) and this finding remained significant in the subgroup
analysis by gestational age only for the subgroup of children whose mothers were
randomised at the intervention/control arms before 28 weeks of gestation. No other long-term
developmental outcome (gross motor dysfunction, developmental delay, hearing and vision
difficulties) was found significantly different between children whose mothers received
treatment with magnesium sulfate and control groups (moderate to very low quality
evidence).
Low to moderate quality evidence from 3 RCTs with a total sample size of over 3000 women
showed that significantly more women who were treated with magnesium sulfate were likely
to experience adverse effects (including any adverse effect, leading to stopping of infusion or
dropping of diastolic pressure by more than 15 mmHg) compared with those who were not
treated. No significant difference was found for the maternal outcome of hypotension (very
low quality) between the 2 groups.
No evidence was found on studies comparing the different effectiveness of different doses of
magnesium sulfate.

A systematic search found 2 studies (Cahill 2011, Bickford 2013) which considered the cost
effectiveness of magnesium for neuroprotection in women with imminent or threatened
preterm labour. These studies both reported that magnesium sulfate for neuroprotection
dominated no magnesium sulfate for neuroprotection, meaning that savings in the cost of
adverse outcomes more than offset treatment costs and that the intervention resulted in
gains in health related quality of life. These studies are reported in more detail in Appendix H.
In addition an original health economic model was developed using the evidence from the
clinical review undertaken for this guideline on neonatal mortality, cerebral palsy,
periventricular leukomalacia and intraventricular haemorrhage. This took the form of a cost–
utility analysis and compared magnesium sulfate for neuroprotection against no magnesium
sulfate for neuroprotection in women with between 24+0 and 34+0 weeks of pregnancy and
at high risk of preterm birth. The base-case analysis reached the same conclusion as the 2
identified published studies, which was that magnesium sulfate for neuroprotection
dominated no magnesium sulfate being cheaper when considering lifetime costs and also
offering increases in health related quality of life. Probabilistic sensitivity analysis suggested
that magnesium sulfate for neuroprotection had an 86% probability of being cost effective
when compared with no magnesium sulfate. The committee decided to recommend

267
magnesium sulfate for babies born before 34 weeks in the absence of evidence of harm and
in the belief that the reduction in long-term adverse effects would be cost effective.
Given this finding, sensitivity analyses were designed to subject this conclusion to challenge,
for example by finding the threshold for input parameters when magnesium sulfate for
neuroprotection would cease to be cost effective even if the input value fell outside a
plausible range. These sensitivity analyses found that the base-case input values were
markedly below these thresholds for cost effectiveness, suggesting that the model results
were robust with respect to uncertainty not directly related to treatment effect size, the
uncertainty of which was assessed with probabilistic sensitivity analysis using Monte Carlo
simulation.
An additional sensitivity analysis was undertaken to evaluate the cost effectiveness of
magnesium sulfate for neuroprotection at a gestational age of 34 weeks, as the baseline risk
of adverse effects is considerably less at this gestational age when compared with earlier
gestational ages. The sensitivity analysis suggested that magnesium sulfate for
neuroprotection is cost effective up to a gestational age of 34 weeks providing that the
relative treatment effect size on cerebral palsy is maintained.
The model is described in detail in Chapter 16.

The Guideline Development Committee selected both any form of neonatal mortality and
stillbirth as the priority outcomes for this review. The rationale for this was that the
mechanisms leading to stillbirth might be different from those that cause death after birth and
therefore the committee felt that it was feasible that the drug might be more beneficial or
harmful in relation to one outcome than for the other.
Although the committee considered many different developmental outcomes for this review
question, it ultimately concluded that a reduction in the incidence of cerebral palsy would be
the most useful measure. It was hypothesised that if there was a significant reduction in the
risk of cerebral palsy in an RCT, it would be reasonable to attribute this difference to the
action of magnesium sulfate. They acknowledged that cerebral palsy covers a wide spectrum
of disability and therefore a reduction in the more severe forms would show a particular
benefit for the use of magnesium sulfate. Another concern that prevented the committee from
prioritising other developmental outcomes over cerebral palsy was that assessment
techniques used in research to measure the severity of such developmental problems are
often unreliable and vary from study to study.
In terms of maternal outcomes the committee felt that it was important to consider maternal
adverse effects as well as maternal mortality because in their clinical experience there is an
appreciable incidence of adverse maternal effects from the use of magnesium sulfate. They
did not, however, think it would be particularly helpful to know what proportion of women
suffering adverse effects had chosen to stop the infusion because they thought that any
results might be affected by the trial setting. In addition, this outcome would vary from what
might be expected in usual clinical circumstances. This is due to the fact that women who are
taking part in a trial would be likely to have access to a higher level of information and
support than might be the case outside a trial setting and thus be more likely to tolerate
adverse effects. Also, in their experience, women would tolerate a high level of adverse
effects if they felt that this was likely to improve outcomes for the baby.

268

The Guideline Development Committee noted that there was consistent evidence of benefit
to the baby in terms of significantly lowering the risk of cerebral palsy resulting from the use
of magnesium sulfate. Although there was overall evidence of benefit when women with
babies up to 34 weeks’ gestation were included in these trials, subgroup analysis by
gestational age showed significant benefit only for babies born before 28 weeks. However,
the committee decided to recommend magnesium sulfate for babies born before 34 weeks in
the absence of evidence of harm and in the belief that this would widen any potential long-
term benefit to a larger group of babies.
The committee also noted that the use of magnesium sulfate showed no difference in effect
between the experimental and control groups in terms of perinatal, neonatal and paediatric
mortality and concluded from this that the use of magnesium sulfate is not associated with a
higher risk of harm to the baby.
With regards to maternal outcomes, limited available data showed no firm conclusions about
the risk of cardiac or respiratory arrest or maternal mortality. However, the committee noted
that there was some evidence of harm to the woman treated with magnesium sulfate in terms
of adverse effects and these results were in keeping with their clinical experience. While they
acknowledged that the results showed a statistically significant difference in blood pressure,
with a drop of more than 15 mmHg for women receiving magnesium sulfate compared with
controls, the committee was uncertain as to whether this difference is significant in clinical
terms. These findings support current practice that women being treated with magnesium
sulfate should continue to routinely have their vital signs monitored.
The committee observed that, in their clinical experience, different women had different
responses, including adverse events to magnesium sulfate, and therefore it was difficult to
capture the real balance of risks and benefits for the individual pregnant woman but that
there would always need to be some clinical judgement involved about the decision to initiate
treatment with magnesium sulfate.
In light of all of these considerations the committee felt that the benefits to the baby
outweighed the potential harms to the woman and that the evidence supported a strong
recommendation to offer magnesium sulfate for neonatal neuroprotection to women in labour
before 30 weeks’ gestation, and women who werbefore 30 weeks’ gestation having a
planned preterm birth within the next 24 hours.
The committee acknowledged that the review had not been designed to look at the
comparative effectiveness of different doses of magnesium sulfate. Given the lack of
evidence for a specific dose of this treatment, it would be preferable and pragmatic in terms
of improving safety and reducing the likelihood of errors being made if the recommendations
could be consistent with other established drug protocols. In light of this the committee
decided to recommend the dosage of magnesium sulfate used in the treatment of pre-
eclampsia (4 grams bolus, 1 gm/hour intravenously [IV]) because this seemed clinically
applicable and was the one used in many of the studies.
The committee also concluded that the potential risk of harm to the woman warranted further
recommendations being made to ensure that appropriate monitoring of this treatment is
carried out. For the reasons given above, it was agreed that maternal and fetal monitoring
should be consistent with the monitoring protocols for magnesium toxicity used in women
who are receiving magnesium sulfate for the treatment of pre-eclampsia. In addition, the
frequency of monitoring for toxicity in the presence of symptoms of oliguria or other signs of
renal failure and adjustment on magnesium’s toxicity was noted by the committee as a
separate recommendation.

269
The committee also discussed whether repeated doses of magnesium sulfate should be
offered and, if so, whether both the bolus and/or the intravenous infusion should be repeated
as these were areas of clinical uncertainty in current practice. The committee noted that the
review had not been designed to look at the effectiveness of repeat courses and so did not
make any recommendations to this effect, but was aware that repeated administration
sometimes happens in practice.

Magnesium sulfate is a relatively inexpensive treatment and the evidence identified in the
clinical review demonstrated a significant clinical benefit to the baby in terms of a significantly
reduced incidence of cerebral palsy.
The Guideline Development Committee was aware that the need to monitor women who are
being treated with magnesium sulfate makes the intervention more costly than the cost of the
drug alone. Nevertheless, given that the management of cerebral palsy is extremely costly,
the committee considered that overall the health benefits not only justified the resource use,
but that the initial costs incurred would be likely to be offset by large cost savings
downstream and the health economic model produced for this guideline provided support for
this viewpoint. A sensitivity analysis also provided cost-effectiveness evidence to support a
recommendation for magnesium sulfate for neuroprotection up to a gestational age of 34+0
weeks providing that the relative treatment effect size for cerebral palsy incidence is
maintained at this gestational age.

The majority of randomised evidence included for this review was of moderate to low quality
and not overly subject to bias.
The Guideline Development Committee noted that blinding of the participants and the
assessors would be difficult because the adverse effects of magnesium sulfate can be very
obvious in some women. This concern was mitigated somewhat by the fact that some
adverse effects were also reported in the placebo group. Overall, the committee felt that
neither the concerns about blinding nor the observed placebo effects would change their
interpretation of the results.

It was also noted that the length of follow-up in the trials (2 years) meant that the results were
unlikely to capture the impact of the drug on cognitive function as this aspect of development
is unlikely to become clear until the child reaches school age. Therefore, the interpretation of
results in the short term should be interpreted with caution.
updated guideline.
4. What is the clinical effectiveness of a bolus plus infusion of

magnesium sulfate compared with a bolus alone for preventing
Research question neurodevelopmental injury in babies born preterm?
Why this is needed

270
4. What is the clinical effectiveness of a bolus plus infusion of

magnesium sulfate compared with a bolus alone for preventing
Research question neurodevelopmental injury in babies born preterm?
Importance to ‘patients’ or There is evidence from randomised studies that magnesium sulfate
the population has neuroprotective properties for the baby when given to women
who will deliver preterm up to 34+0 weeks of pregnancy. However,
there is uncertainty about the best method of administering
magnesium sulfate for this purpose, with different studies using
different strategies. There are significant advantages for the woman
and for reducing healthcare costs if a bolus is as effective as a bolus
plus infusion, because magnesium sulfate has side effects for the
woman, and more monitoring is needed for infusion, with additional
associated healthcare costs. A randomised controlled trial would best
address this question by assessing the effects of each method on
neonatal and maternal outcomes.
Relevance to NICE The current guideline recommends the use of magnesium sulfate for
guidance neuroprotection, and the research is unlikely to change the
recommendation, but will address the best protocol for administration.
Relevance to the NHS If a bolus is as effective as a bolus plus infusion this will reduce
healthcare costs, because additional monitoring is needed for
infusion
National priorities N/A
Current evidence base There is existing evidence of the effectiveness of magnesium sulfate
in neuroprotection but uncertainty about the best method of
administration, with different studies using different protocols.
Equality The population is defined as pregnant woman at imminent risk of
preterm birth before 34+0 weeks gestational age.
Feasibility The Committee decided that a RCT would best address this research
recommendation. This could realistically be carried out within a
reasonable timescale and cost.
There are no particular ethical issues beyond other perinatal trials,
and no technical issues since participating centres will have
availability of monitoring.
Other comments N/A

271
Fetal monitoring
13 Fetal monitoring
13.1 Introduction
Babies in the uterus derive oxygen from the mother via the placenta and umbilical cord.
During contractions of the uterus in labour this oxygen exchange can be intermittently
interrupted. Well babies at term during normal labour are not adversely affected by this.
However, this is not always the case, and fetal hypoxia and then acidosis can occur. In
theory, the preterm fetus is more vulnerable than fullterm babies to this risk of intrapartum
hypoxia-acidosis.
Surveillance for fetal hypoxia in labour is undertaken by fetal heart rate monitoring. The fetal
heart rate can be monitored using either intermittent auscultation (listening in to the baby’s
heart using a handheld device) or by a continuous electronic recording. Continuous
electronic recording can be undertaken using either an external ultrasound transducer,
positioned on the mother’s abdomen to pick up the fetal heart rate, or a fetal scalp electrode
(a small clip introduced through the mother’s vagina and attached to the baby’s head).
The outputs of both electronic methods are displayed as a cardiotocograph (CTG) trace. The
tocograph is a simultaneous recording of the uterine contractions, so the CTG trace provides
a visual continuous record of fetal heart rate and uterine contractions. There are features that
can indicate the baby is well, for example accelerations, and features that are not, in
isolation, normally of concern, for example transient slowing of the fetal heart during a
contraction in labour. There are other features that may indicate a serious emergency (for
example development of a persistent bradycardia following cord prolapse or placental
abruption).
13.2 Interpretation of the fetal heart rate (FHR) pattern

13.2.1 Introduction
The 4 features of the fetal heart rate that are scrutinised in the CTG trace at term are: the
baseline heart rate, the baseline variability, the presence or absence of decelerations and the
presence of accelerations. All of these are used to indicate whether fetal hypoxia-acidosis is
developing. The physiological changes of hypoxia-acidosis result in alterations in these
parameters. Whether the recognised features of the fetal heart rate at term can be
extrapolated to the preterm fetus is uncertain.

What are the criteria for best interpreting the preterm fetal heart rate trace at different
gestational ages for unborn babies whose mothers are in suspected or diagnosed preterm
labour?

Twelve studies were included in this review (Althaus 2005, Aina-Mumuney 2007, Kariniemi
1984, Bowes 1980, Matsuda 2003, Holmes 2001, Martin 1974, Rayburn 1987, Burrus 1994,
Braithwaite 1986, Douvas 1984, Nisenblat 2006).

272
Fetal monitoring
Seven included studies are from the USA (Burrus 1994, Bowes 1980, Althaus 2005, Martin
1974, Aina-Mumuney 2007, Douvas 1984, Rayburn 1987), 2 from Canada (Braithwaite 1986,
Holmes 2001) and 1 each from from Finland (Kariniemi 1984), Israel (Nisenblat 2006) and
Japan (Matsuda 2003).
This review question was designed to test the predictive value of fetal heart rate features for
neonatal adverse outcomes including neonatal acidemia, intraventricular haemorrhage and
neonatal death. A more inclusive approach was adopted and different types of observational
studies (either retrospective or prospective cohort studies, case-control studies or
consecutive or non-consecutive case series) were included. Two main types of analyses
were conducted:
• Observational studies looking at the predictive value of features of fetal heart rate for
neonatal adverse outcomes; tachycardia and bradycardia were assessed in 2 studies
(Althaus 2005, Aina-Mumuney 2007), accelerations and decelerations were assessed in
5 studies (Kariniemi 1984, Matsuda 2003, Bowes 1980, Martin 1974, Holmes 2001) and
defined CTG classification systems were assessed in 6 studies (Rayburn 1987, Kariniem
1991, Douvas 1984, Nisenblat 2006, Burrus 1994, Braithwaite 1986).
• Observational studies testing the association between fetal heart rate baseline variability
and neonatal respiratory distress syndrome, neonatal death and/or metabolic acidosis
were assessed in 4 studies (Kariniemi 1984, Althaus 2005, Bowes 1980, Aina-Mumuney
2007).
The mean gestational age of babies in 7 included studies ranged from 26 to 30 weeks
(Althaus 2005, Kariniemi 1984, Nisenblat 2006, Rayburn 1987, Holmes 2001, Bowes 1980,
Burrus 1994). Two studies included women giving birth at less than 36 weeks’ gestation
(Aina-Mumuney 2007, Matsuda 2003). Two studies included women with gestational age of
less than 35 and 30 weeks (Martin 1974, Braithwaite 1986) and a further 1 study included
babies with birth weight less than 1800 g and did not report the specific gestational age
(Douvas 1984).
The use of tocolytics was not reported in 8 studies. In 4 studies women received tocolysis but
the proportion of women receiving these medicines was not reported.

Data is reported in modified GRADE profiles below for the following CTG parameters:
• fetal heart rate:
o Table 81:GRADE profile for predictive value of bradycardia and tachycardia for adverse
neonatal outcomes
o Table 82:GRADE profile for association between tachycardia and systemic fetal
inflammation
• baseline variability:
o Table 83: GRADE profile for predictive value of fetal heart rate baseline variability for
neonatal adverse outcomes
o Table 84 GRADE profile for association between fetal heart rate baseline variability and
neonatal adverse outcomes or umbilical artery blood gas values
• accelerations:
o Table 85:GRADE profile for predictive value of absence of fetal heart rate accelerations
(non-reactive CTG) for adverse neonatal outcomes
• decelerations:

273
Fetal monitoring
o Table 86:GRADE profile for predictive value of fetal heart rate late, ‘prolonged’ and
‘severe variable’ decelerations for adverse neonatal outcomes
o Table 87:GRADE profile for association between variable fetal heart rate decelerations
and adverse neonatal outcome
• categorization/classification of CTGs:
o Table 88: GRADE profile for predictive value of published categorisation of CTGs for
adverse neonatal outcomes
o Table 89: GRADE profile for association between categorisation of CTGs and adverse
neonatal outcomes
The grading of evidence from prospective comparative observational studies or prospective
consecutive case series started at high quality and was then downgraded if there were any
issues identified that would undermine the trustworthiness of the findings. Evidence from
retrospective comparative observational studies or retrospective consecutive case series
started at moderate quality and was then downgraded if there were any issues. Evidence
from non-consecutive case series started at low quality and was then downgraded if there
were any issues.
The classifications of CTGs used and reported in 6 of the studies (Rayburn 1987, Kariniem
1991, Douvas 1984, Nisenblat 2006, Burrus 1994, Braithwaite 1986) are detailed in the
The stage of labour was considered an important additional piece of information for the
interpretation of results and was included in the GRADE table along with the study’s sample
size. Although the most appropriate measures of assessing the predictive ability of the
criteria for interpreting preterm fetal hearth rate are the positive and negative likelihood
ratios, supplementary information was reported for sensitivity and specificity. For the studies
that tested associations of features of fetal heart trace with neonatal outcomes, odd ratios
(ORs) (with 95% confidence intervals [CIs]) were considered as best measures of these
associations.

274
Fetal monitoring
Table 75: GRADE profile for predictive value of bradycardia and tachycardia for adverse neonatal outcomes
Quality assessment Measure of diagnostic accuracy (95% CI)a
Total
Number Stage number of Positive Negative
of Risk of Definition of women & likelihood likelihood
studies Design bias Inconsistency Indirectness Imprecision of outcome labour baby pairs Sensitivity Specificity ratio ratio Quality
Tachycardia (>160 bpm) (mean duration 35.2 min SD 22.8)
1 study Case No No serious Serious1 Serious2 Cerebral 1 hour 246 17.24% 78.5% 0.80 1.05 Low
(Althaus control serious inconsistency white matter before (11.48 to (70.1 to (0.49 to (0.94 to
2005) risk of injury a birth 24.39) 85.4) 1.31) Not 1.19) Not
bias useful useful
Bradycardia (<110 bpm) (NICHD classification) (duration > 2 min)
1 study Case Serious3 No serious Serious1 Serious2 Cerebral 1 hour 246 4.8% 95.8% 1.16 0.99 Very
(Althaus control inconsistency white before (1.79 to (90.6 to (0.36 to (0.94 to low
2005) matter birth 10.16) 98.6) 3.71) Not 1.05) Not
injury a useful useful
Bradycardia (<110 bpm) episodesbc
1 study Case Serious1 No serious Serious1 Serious2 Cerebral 1 hour 246 4.80% 92.56% 0.65 1.03 Very
(Althaus control inconsistency white before (1.79 to (86.3 to (0.24 to (0.96 to low
2005) matter birth 10.16) 96.5) 1.76) Not 1.10) Not
injury a useful useful
bpm beats per minute, CI confidence interval, LR likelihood ratio, NICHD National Institute of Child Health and Human Development, SD standard deviation
a. Diagnosed by neonatal head ultrasound; first at 24–72 hours after birth, second at 10–14 days of life and third at 6 weeks of life.
b. The number of bradycardia episodes lasting >2 minutes reported. There were 6 bradycardia episodes in cases and 9 in controls.
c. Bradycardia mean nadir (bpm): cases 87.3 (SD 4.1), control: 83.3 (SD 23.4). Bradycardia mean duration (minutes): cases: 5.88 (SD 4.1), controls: 5.02 (SD 2.20)
1. n=27 (29.8%) of cases and n=10 (15.8%) of controls had multiple gestations
2. Wide CI (LRs)
Table 76: GRADE profile for association between tachycardia and systemic fetal inflammation
Quality assessment
Number of Definition of Stage of Number of babies with OR (95%
studies Design Risk of bias Inconsistency Indirectness Imprecision outcome labour tachycardia >160 bpm CI) Quality
Tachycardia (>160 bpm) (NICHD classification)
1 study Case No serious No serious Serious1 No serious Systemic fetal 2 hours 150 OR 1.38 Moderate
control risk of bias inconsistency imprecision inflammationa before birth (0.30 to
Fetal monitoring
Quality assessment
Number of Definition of Stage of Number of babies with OR (95%
studies Design Risk of bias Inconsistency Indirectness Imprecision outcome labour tachycardia >160 bpm CI) Quality
(Aina- 6.42)
Mumuney
2007)
bpm beats per minute, CI confidence interval, NICHD National Institute of Child Health and Human Development, OR odds ratio
a. Systemic fetal inflammation was diagnosed by histologically confirmed chorioamnionitis and funisitis
Table 77: GRADE profile for predictive value of fetal heart rate baseline variability for neonatal adverse outcomes
Quality assessment Measure of diagnostic accuracy (95% CI)*
Total
number of Positive Negative
Number of Risk of Inconsist Indirectn Imprecisi Definition of Stage of women & Sensitivi Specifici likelihoo likelihoo
studies Design bias ency ess on outcome labour baby pairs ty ty d ratio d ratio Quality
Reduced variability (“silent” pattern: FHR variability <5 bpm >5 min)
1 study Case Serious1,2, No Serious4 Serious5 Neonatal death NR 74 .42.3% 29.2% 0.60 0.77 Very low
(Kariniemi 1991) control 3 serious (23.4 to (16.9 to (0.37 to (0.14 to
inconsiste 63.0) 44) 0.97) Not 3.43) Not
ncy useful useful
1 study Case Serious1,2, No Serious4 Serious5 Respiratory NR 74 50.0% 18.9% 0.73 2.15 Very low
(Kariniemi 1991) control 3 serious distress (31.9 to (8.0 to (0.53 to (0.98 to
inconsiste syndromea 68.1) 35.1) 1.01) Not 4.72) Not
ncy useful useful
Baseline variability <5 bpm (NICHD classification)
1 study Case No No Serious6 Serious5 Cerebral white 1 hour 246 19.2% 75.2% 0.77 1.07 Low
(Althaus 2005) control serious serious matter injuryb before (12.7 to (66.5 to (0.48 to (0.94 to
risk of inconsiste birth 27.2) 82.6) 1.25) Not 1.23) Not
bias ncy useful useful
Baseline variability <5 bpm (duration 20 minutes)
1 study Case Serious1,2 No No Serious5 Neonatal 1 hour 61 10.0% 82.3% 6.57 1.09 Very
(Bowes 1980) control serious serious death before (1.66 to (69.1 to (0.08 to (0.86 to low
inconsiste indirectne birth 44.5) 91.5) 3.99) 1.39) Not
ncy ss Moderate useful
ly useful
Fetal monitoring
Quality assessment Measure of diagnostic accuracy (95% CI)*

Total
1 study Case Serious1,2 No No Serious5 Respiratory 1 hour 61 12.0% 85.3% 0.82 1.03 Very
(Bowes 1980) control serious serious distress before (2.69 to (86.9 to (0.21 to (0.84 to low
inconsiste indirectne syndromec birth 31.25) 94.9) 3.10) Not 1.26) Not
ncy ss useful useful
1 study Case Serious1,2 No No No Central 1 hour 61 0.00% 81.8% 0.00 1.22 Low
(Bowes 1980) control serious serious serious nervous before (69.1 to Not (1.08 to
inconsiste indirectne imprecisi system birth 90.9) useful 1.38) Not
ncy ss on haemorrhaged useful
1 study Case Serious1,2 No No Serious5 Umbilical cord 1 hour 61 50.0% 92.3% 6.50 0.54 Very
(Bowes 1980) control serious serious pH < 7.20 before (18.9 to (74.8 to (1.50 to (0.29 to low
inconsiste indirectne birth 81.1) 98.3) 28.23) 1.02)
ncy ss Moderate Not
ly useful useful
CI confidence interval, FHR fetal heart rate, bpm beats per minute, LR likelihood ratio, MID minimally important difference, NICHD National Institute of Child Health and Human
Development
a. Respiratory distress syndrome was defined as in the presence of tachypnoea, retraction and granting, hypoxaemia in room air and air bronchogram and
reticulogranularpattern in X-ray when symptoms appears 6 hours after birth and lasted 24 hours.
b. The diagnosis of white matter injury was made by neonatal head ultrasonograms at 24–72 hours after birth, 10–14 days of life and at 6 weeks.
c. Respiratory distress syndrome (RDS) was diagnosed if the all following were present: arterial Po2 was <50 mmHg in room air, increased ambient oxygen, continuous positive
airway pressure or ventilation required >24 hours to support respiration, chest X-ray evidence, no evidence of other disease caused RDS
d. Central nervous system CNS haemorrhage was diagnosed in babies who exhibited: seizures, fullness of anterior fontanelle, a decreased in the haematocrit, and blood in the
cerebral spinal fluid
1. The traces were evaluated by only 1 of the study’s authors
2. No clear inclusion/exclusion criteria, hence high risk of selection bias
3. No clear definition of FHR pattern. Unclear in what stage of labour the traces were obtained and evaluated
4. Most babies delivered by caesarean section before labour started
5. Confidence interval crossed 1 default MIDs (LRs)
Fetal monitoring
Table 78: GRADE profile for association between fetal heart rate baseline variability and neonatal adverse outcomes or umbilical artery
blood gas values
Quality assessment
Number of
babies with
Number of Risk of Inconsist Indirectn Imprecisi Definition of Stage of defined CTG OR (95% CI) or mean
studies Design bias ency ess on outcome labour pattern (SD) Quality
Decreased short term variability (<5 bpm NICHD classification)
1 study Cohort No No Serious1 No Systemic fetal 1st stage 150 OR 0.71 (0.34 to Moderate
(Aina-Mumuney serious serious serious inflammation 1.50)
2007) risk of inconsiste imprecisio
bias ncy n
Reduced reactivity (NICHD classification)
1 study Cohort No No Serious1 No serious Systemic 1st stage 150 OR 0.96 (0.49 to Moderate
(Aina-Mumuney serious serious imprecision fetal 1.87)
2007) risk of inconsiste inflammation
bias ncy
Increase reactivity (NICHD classification)
1 study Case No No Serious2 No Umbilical cord 1 hour 246 Non-reactive: Moderate
(Althaus 2005) control serious serious serious pH and before pH: 7.29±0.10
risk of inconsiste imprecision baseexcess birth Base excess: −2.7
bias ncy (3.8)
Reactive:
pH: 7.31±0.08
Base excess: −2.9
(3.4)
Both p=NS
BD base deficit, bpm beats per minute, CI confidence interval, CTG cardiotocograph, FHR fetal heart rate, NICHD National Institute of Child Health and Human
Development,NS no significant difference, OR odds ratio
2. 17% of cases and 6% of controls had multiple gestations
Table 79: GRADE profile for predictive value of absence of fetal heart rate accelerations (non-reactive CTG) for adverse neonatal
outcomes
Total
Non-reactive CTG (<2 accelerations >15 bpm in 30 min)
Fetal monitoring

Total
1 study Case Serious1,2, No Serious4 No Neonatal death NR 74 50.0% 14.6% 0.59 0.77 Very low
(Kariniem 1984) control 3 serious serious (29.9 to (6.10 to (0.39 to (1.56 to
inconsiste imprecisi
70.0) 27.7) 0.87) Not 7.52) Not
ncy on
useful useful
1 study Case Serious1,2, No Serious4 No Respiratory NR 74 68.7% 24.3% 0.91 1.28 Very low
(Kariniem 1984) control 3 serious serious distress (49.9 to (11.8 to (0.68 to (0.60 to
inconsiste imprecisi syndrome 83.8) 41.2) 1.22) Not 2.76) Not
ncy on useful useful
bpm beats per minute, CI confidence interval, CTG cardiotocograph, FHR fetal heart rate, LR likelihood ratio, MID minimally important difference, NR not reported
a. Respiratory distress syndrome was defined as the presence of tachypnoea, retraction and grunting, hypoxaemia in room air and air bronchogram and reticulogranular
patternin X-ray when symptoms appeared 6 hours after birth and lasted 24 hours.
3. No clear definition of FHR pattern. Unclear in what stage of labour the traces were obtained and evaluated
4. Most babies delivered by caesarean section before labour started
Table 80: GRADE profile for predictive value of fetal heart rate late, ‘prolonged’ and ‘severe variable’ decelerations for adverse neonatal
outcomes
Number
of
Number Stage women Positive Negative
of Definition of of & baby likelihood likelihood
studies Design Risk of bias Inconsistency Indirectness Imprecision outcome labour pairs Sensitivity Specificity ratio ratio Quality
Late decelerations (not defined)
1 study Case Serious1,2,3 No serious Serious4 Serious5 Neonatal NR 74 53.8% 16.67% 0.65 (0.44 0.77 (0.30 Very
(Kariniem control inconsistency death (33.4 to (7.50 to to 0.94) to 5.60) low
1984) 73.4) 30.2) Not useful Not useful
1 study Case Serious1,2,3 No serious Serious4 Serious5 Respiratory NR 74 59.3% 18.9% 0.73 (0.53 2.15 (0.98 Very
(Kariniem control inconsistency distress (40.6 to (8.0 to to 1.01) to 4.72) low
1984) syndromea 76.2) 35.1) Not useful Not useful
Fetal monitoring

Number
of
“Combined distress” pattern (decelerations and “silent” pattern in 30 minutes)
1 study Case Serious1,2,3 No serious Serious4 No serious Neonatal NR 74 19.2% 35.4% 0.30 (0.13 2.28 (1.49 Very
(Kariniem control inconsistency imprecision death (6.63 to (22.2 to to 0.67) to 3.49) low
1984) 39.4) 50.4) Not useful Not useful
1 study Case Serious1,2,3 No serious Serious4 Serious5 Respirat NR 74 37.5% 40.5% 0.63 (0.37 2.54 (0.96 Very
(Kariniem control inconsistency ory (21.1 to (24.7 to to 1.06) to 2.48) low
1984) distress 56.3) 57.9) Not useful Not useful
syndrom
eb
Late decelerations with loss of variability for <30 minutes
1 study Cohort Serious2,6,7,8 No serious No serious No serious Umbilical 2 772 28.6%b 86.4% b 2.10 b Not 0.82 b Not Low
(Matsuda inconsistency indirectness imprecision artery pH hours useful useful
2003) <7.1 before
birth
1 study Cohort Serious2,6,7,8 No serious No serious No serious Umbilical 2 772 85.7% b 68.2% b 2.69 b Not 0.20 b Low
(Matsuda inconsistency indirectness imprecision artery pH hours useful Moderatel
2003) <7.1 before y useful
birth
1 study Cohort Serious2,6,7,8 No serious No serious No serious Umbilical 2 772 100% b 45.5% b 1.83 b Not 0.0 b very Low
2003) <7.1 before
birth
”Prolonged” decelerations with loss of variability <30 minutes (prolonged decelerations not defined)
birth
“Prolonged” decelerations with loss of variability <60 minutes (prolonged decelerations not defined)
birth
“Prolonged” decelerations with loss of variability <90 minutes (prolonged decelerations not defined)
Fetal monitoring

Number
of
1 study Cohort Serious2,6,7,8 No serious No serious No serious Umbilical 2 772 100% b 16.2% b 1.19 b Not 0.0 b Very Low
2003) <7.1 before
birth
“Severe variable late” decelerations (ominous periodic changes not defined)
1 study Case Serious1,2 No serious No serious No serious Neonatal 1 hour 61 12.0 % 86.11% 0.86 (0.23 1.02 Low
(Bowes control inconsistency indirectness imprecision death before (2.69 to (70.04 to to 3.29) (0.84to
1980) birth 31.2) 95.2) Not useful 1.24) Not
useful
1 study Case Serious1,2 No serious No serious Serious5 Central 1 hour 61 16.7% 12.73% 0.19 (0.03 6.55 Very low
(Bowes control inconsistency indirectness nervous before (2.76 to (5.30 to to 1.15) (3.00 to
1980) system birth 63.90) 24.5) Not useful 14.27) Not
haemorr useful
hage
c
1 study Case Serious1,2 No serious No serious Serious5 Respirat 1 hour 61 0.00 84.3% 0.00 Not 1.19 Very low
(Bowes control inconsistency indirectness ory before (71.4 to useful (1.05 to
1980) distress birth 93) 1.34) Not
syndrom useful
ed
1 study Case Serious1,2 No serious No serious Serious5 Umbilical 1 hour 61 60.0% 100% 0.00 Not 0.40 Very low
(Bowes control inconsistency indirectness cord pH before (26.3 to (86.6 to useful (0.19 to
1980) <7.20 birth 87.7) 98.3) 0.85)
Moderatel y
useful
CI confidence interval, CTG cardiotocograph, LR likelyhood ratio, MID minimally important difference, NR not reported,
a. Respiratory distress syndrome was defined as in the presence of tachypnoea, retraction and granting, hypoxaemia in room air and air bronchogram and
reticulogranularpattern in X-ray when symptoms appears 6 hours after birth and lasted 24 hours.
b. Insufficient data reported to calculate CI
c. Central nervous system (CNS) haemorrhage was diagnosed in babies who exhibited: seizures, fullness of anterior fontanelle, a decreased in the haematocrit, and blood in
the cerebral spinal fluid
d. Respiratory distress syndrome (RDS) was diagnosed if the all following were present: arterial Po2 was < 50 mm Hg in room air, increased ambient oxygen,
continuouspositive airway pressure or ventilation required >24 hours to support respiration, chest x-ray evidence, no evidence of other disease caused RDS
3. No clear definition of CTG pattern. Unclear in what stage of labour the traces were obtained and evaluated
4. Most babies delivered by caesarean section before labour start
Fetal monitoring

6. Unclear how and by whom data were analysed
7. Unclear if the assessors were blinded to outcomes
8. Women’s characteristics not reported
Table 81: GRADE profile for association between variable fetal heart rate decelerations and adverse neonatal outcome
Quality assessment
Number of babies Number of babies with
with defined CTG defined outcome or
Number Definition of Stage of patterns mean outcome value or
of studies Design Risk of bias Inconsistency Indirectness Imprecision outcome labour (gestation) p value Quality
Variable decelerations a
1 study Case No serious No serious No serious No serious Neonatal death 1 hour 82 Cases:2/41 Controls: 0/41 Moderate
(Holmes control risk of bias inconsistency indirectness imprecision before p=NS
2001) birth
1 study Case No serious No serious No serious No serious Umbilical cord artery 1 hour 79 Cases:0/38 Controls: 2/41 Moderate
(Holmes control risk of bias inconsistency indirectness imprecision pH <7.1 before p=NS
2001) birth
1 study Case No serious No serious No serious No serious Resuscitation 1 hour 82 Cases:1/41 Controls: 2/41 Moderate
(Holmes control risk of bias inconsistency indirectness imprecision (cardiac massage or before p=NS
2001) drug therapy) birth
1 study Case No serious No serious No serious No serious Intraventricular 1 hour 82 Cases:4/41 Controls: 0/41 Moderate
(Holmes control risk of bias inconsistency indirectness imprecision haemorrhage grade before p=0.04
2001) III or IV birth
1 study Case No serious No serious No serious No serious Periventricular 1 hour 82 Cases:1/41 Controls: 0/41 Moderate
(Holmes control risk of bias inconsistency indirectness imprecision leukomalacia before p=NS
2001) birth
“Severe variable” decelerations with late component
1 study Case Serious1,2,3 No serious No serious No serious Neonatal death 1st 73 Severe variable Moderate
(Martin series inconsistency indirectness imprecision stage deceleration with late
1974) component: 10/11
Mild/moderate variable
decelerations without late
component: 1/11
p=0.05
BD base deficit, bpm beats per minute, CI confidence interval, FHR fetal heart rate, NICHD National Institute of Child Health and Human Development, OR odds ratio, NS
nosignificant difference
Fetal monitoring
a. Cases consisted of traces with ≥ 3 variable decelerations in the hour prior to birth and controls consisted of traces with ≤2 variable decelerations. Variable
decelerationdefined as an abrupt decrease in FHR of at least 15 bpm lasting for between 15 seconds and 2 minutes according to the National Institutes of Child Health
and Human Development (NICHD)
1. Unclear if the assessors were blinded to the outcomes
2. No clear exclusion criteria, hence high risk of selection bias
Table 82: GRADE profile for predictive value of published categorisation of CTGs for adverse neonatal outcomes
Quality assessment Total Measure of diagnostic accuracy (95% CI)a
numb
er of
wome
Definition n& Positive Negative
Number Desig of Stage of baby Sensitivi Specifici likelihoo likelihoo
of studies n Risk of bias Inconsistency Indirectness Imprecision outcome labour pairs ty ty d ratio d ratio Quality
“Reassuring” CTG (normal pattern with/without occasional mild/moderate variable decelerations)
1 study Case Serious1,2 No serious No serious Serious3 Intraventric Minimum of 72 55.2% 47.3% 1.05 Not 0.94 Not Very
(Rayburn contro inconsistency indirectness ular 20 min of (38.3 to (31 to useful useful low
1987) l haemorrha tracing during 71.3) 61.1) (0.69 to (0.58 to
ge the first stage 1.59) Not 1.54) Not
of labour useful useful
Fischer classification (abnormal vs normal)a
1 study Case No serious No serious No serious Serious3 Neonatal Last 30 min 383 86.5% 48.39% 1.71 0.24 Modera
(Braithwai contro risk of bias inconsistency indirectness death of tracing (69.8 to (30.17 to (0.19 to (0.08 to te
te 1986) l during the 77.7) 66.9) 2.48) Not 0.73) Not
first stage of useful useful
labour
“Abnormal” CTG (not defined)
1 study Case Serious4,5,6 No serious Serious7 Serious3 Neonatal NR 74 80.7% 8.33% 0.88 2.31 Very
(Kariniemi contro inconsistency death (60.6 to (2.37 to (0.72 to (0.68 to low
1991) l 93.3) 20.2) 1.08) Not 7.86) Not
useful useful
1 study Case Serious4,5,6 No serious Serious7 Serious3 Respiratory NR 74 81.2% 8.11% 0.88 2.31 Very
(Kariniemi contro inconsistency distress (63.5 to (1.80 to (0.73 to (0.63 to low
1991) l syndrome 92.7) 21.9) 1.07) Not 8.51) Not
useful useful
“Abnormal” CTGb
Fetal monitoring
Quality assessment Total Measure of diagnostic accuracy (95% CI)a

numb
er of
wome
Definition n& Positive Negative
Number Desig of Stage of baby Sensitivi Specifici likelihoo likelihoo
of studies n Risk of bias Inconsistency Indirectness Imprecision outcome labour pairs ty ty d ratio d ratio Quality
1 study Case Serious8,9 No serious No serious Serious3 Birth NR 89 72.7% 94.6% 13.5 0.29 Very
(Douvas series inconsistency indirectness asphyxiad (54.4 to (85.1 to (4.43 to (0.16 to low
1984) 86.7) 98.9) 41.6) 0.50)
Very Moderate
useful ly useful
1 study Case Serious8,9 No serious No serious Serious3 Respiratory NR 89 66.7% 88.1% 5.62 0.38 Very
(Douvas series inconsistency indirectness distress (47.2 to (77 to 95) (2.68 to (0.23 to low
1984) syndromec 82.7) 11.78) 0.63)
Moderate Moderate
ly useful ly useful
“Pathological” CTGd
1 study Cohor No serious No serious No serious No serious Neurodevel 1 hour before 111 27% 74% 1.03 Not 0.98 Not Low
(Nisenblat t risk of bias inconsistency indirectness imprecision opmental birth useful useful
2006) abnormality
“Suspicious” CTGe
1 study Case Serious1,2 No serious No serious Serious3 Intraventric Minimum of 72 29.1% 69.2% 0.95 1.02 Very
(Rayburn contro inconsistency indirectness ular 20 min of (12.6 to (48.2 to (0.41 to (0.71 to low
1987) l haemorrha tracing during 51) 85.6) 2.22) Not 1.47) Not
ge the first stage useful useful
of labour
“Ominous” CTGf
1 study Case Serious1,2 No serious No serious Serious3 Intraventric Minimum of 72 45.1% 66.67% 1.13 0.91 Very
(Rayburn contro inconsistency indirectness ular 20 min of (27.3 to (40.6 to (0.63 to (0.59 to low
1987) l haemorrha tracing during 63.9) 77.3) 2.03) Not 1.41) Not
ge the first stage useful useful
of labour
bpm beats per minute, CI confidence interval, CTG cardiotocography, FHR fetal heart rate,, LR likelihood ratio, MID minimally important difference, NR not reported
a. Normal CTG in 30 min period was defined as: baseline 120–160 bpm, variability >6 bpm, accelerations present, and no decelerations (extracted from Fischer
1976,translated from German)
b. CTG considered as abnormal if any of the following were seen late decelerations defined as persistent decelerations following 50% of the contractions over a 30 minute
perio, severe variable decelerations defined as decelerations <70 bpm asting for >60 second, absent or minimal beat to beat variability, defined as <5 bpm over a 30
minuteperio, prolonged bradycardia defined as FHR <100 bpm persistently over a period of >3 minutes
c. The measure of asphyxia was based on the 1 of the following: Apgar score <3 at 1 minute or <6 at 5 minute, immediate resuscitation requiring positive pressure oxygen for
>1 minute, pH<7.25 on arrival in the neonatal intensive care unit
d. “Pathological” CTG was defined as: baseline fetal heart rate >160 bpm or <110 bpm, absence of FHR variability (amplitude range undetectable), either recurrent late
Fetal monitoring
decelerations (deceleration is associated with the uterine contraction, with nadir of the deceleration occurring after peak of the contraction) or recurrent severe variable
decelerations (decrease in FHR below 70 bpm lasting longer than 60 seconds or other decelerations with slow return to baseline, associated with the uterine contractions, the
onset, depth, and duration vary with successive uterine contractions) classified as mild, moderate, or severe on the basis of umbilical artery base deficit (cut off >12
mmol/litre)and neonatal encephalopathy and other organ system complications
e. “Suspicious” CTG: intermittent late decelerations, decreased variability or tachycardia present
f. “Ominous” CTG: consistent with repetitive severe variable or late decelerations or repetitive prolonged decelerations (>2
min)“Suspicious” or “ominous” patterns that were continuous and repetitive for >30 min were considered indicative of “fetal
distress”
1. No clear exclusion criteria, hence high risk of selection bias
2. No clear definition of CTG features
4. The CTGs were evaluated by only 1 of the study’s authors
6. No clear definition of CTG pattern. Unclear in what stage of labour the traces were obtained and evaluated
7. Most babies delivered by caesarean section before labour start
8. No clear inclusion and exclusion criteria, hence high risk of selection bias
Table 83: GRADE profile for association between categorisation of CTGs and adverse neonatal outcomes
Quality assessment
Number of Number of babies with
Number of Risk of Definition of Stage of woman and defined outcome and p
studies Design bias Inconsistency Indirectness Imprecision outcome labour baby pairs value Quality
“Normal” versus “abnormal” CTGb
1 study Case Serious1,2 No serious No serious No serious Neonatal death 1 hour 41 n=13/41 Low
(Burrus series inconsistency indirectness imprecision before normal pattern n=3/22
1994) birth abnormal pattern
n=10/19
p = 0.007
1 study Case Serious 1,2
No serious No serious No serious Intraventricular 1 hour 41 n=8/41 Low
(Burrus series inconsistency indirectness imprecision haemorrhage before normal pattern n=2/22
1994) birth abnormal pattern n=6/19
p=NS
1 study Case Serious1,2 No serious No serious No serious >42 days on 1 hour 41 n=4/41 Low
(Burrus series inconsistency indirectness imprecision assisted ventilation before normal pattern n=2/22
p=NS
Fetal monitoring
Quality assessment
Number of Number of babies with
Number of Risk of Definition of Stage of woman and defined outcome and p
studies Design bias Inconsistency Indirectness Imprecision outcome labour baby pairs value Quality
1 study Case Serious1,2 No serious No serious No serious >90 days of 1 hour 41 n=6/41 Low
(Burrus series inconsistency indirectness imprecision hospitalisation before normal pattern n=5/22
p=NS
1 study Case Serious 1,2
No serious No serious No serious Cerebral palsy at 1 1 hour 41 n=3/41 Low
(Burrus series inconsistency indirectness imprecision year before normal pattern n=1/22
p=NS
bpm beats per minute, CTG cardiotocography, FHR fetal heart rate, NS no significant difference, NR not reported
a. “Ominous” CTG defined as repetitive pattern of late deceleration and pronounced variable decelerations (>40 seconds duration and/or >60 beats loss)
b. “Normal” and “abnormal” CTG defined by Kubli 1969 as: normal baseline (FHR 120–160 bpm), bradycardia (FHR 100–120 bpm) and severe bradycardia (FHR<100
bpm)Variability was defined as normal variability (amplitude range > 5 bpm), moderately reduced variability (2–5 bpm), severely reduced variability (<2 bpm), a salutatory
or hypervariable pattern was diagnosed if amplitude range exceeded 25 bpm
Decelerations defined as mild variable deceleration (lasted <30 sec irrespective of level, if the nadir was >80 bpm irrespective of duration, or if their nadir was 70–80 bpm
iflasting <60 sec), moderate variable deceleration (lasted 30 to 60 sec with the nadir <60 bpm, or lasted >60 sec but with a nadir between 70–80 bpm), severe variable
deceleration (lasted >60 sec with a nadir. <70 bpm, occasional (2 or fewer in a 10 minute window) or frequent (3 or more)
1. Small study with low statistical power
2. Poor level of agreement between 2 CTG evaluators
Fetal monitoring

Fetal heart rate: bradycardia, tachycardia
Evidence from 2 observational studies with approximately 400 participants found that fetal
baseline tachycardia or bradycardia had not useful positive and negative likelihood ratios for
predicting neonatal cerebral white matter injury at birth. There was no evidence of a
difference in the risk of systemic fetal inflammation between babies with intrapartum
tachycardia compared with those without intrapartum tachycardia. The evidence was of
moderate to very low quality.
Baseline variability
Evidence from 3 observational studies with over 400 participants indicated that reduced
baseline variability had low to moderate positive likelihood ratios for predicting poor neonatal
outcomes. There was no evidence of a difference in the risk of systemic fetal inflammation
between babies with reduced baseline variability compared with those without reduced
baseline variability. The evidence was of moderate quality.
Absence/presence of accelerations (reactivity)
Very low quality evidence from 1 observational study with 74 participants showed no
difference in the risk of systemic fetal inflammation or poor umbilical cord blood gases for
babies with non-reactive traces compared with those with reactive traces.
The diagnostic value of absence of accelerations was low for respiratory distress syndrome
and neonatal death (across all diagnostic parameters).
Late, ‘prolonged’ and ‘severe variable late’ decelerations
Evidence from 6 observational studies (number of participants ranged from 61 to 772) for all
diagnostic values of late decelerations for poor neonatal outcomes showed that alone they
were not useful for predicting poor neonatal outcome (very low quality evidence). Late
decelerations with loss of variability lasting less than 30 minutes, less than 60 minutes and
less than 90 minutes had not useful positive and negative likelihood ratios (low quality
evidence). ‘Prolonged’ decelerations with loss of variability lasting less than 30 minutes, less
than 60 minutes and less than 90 minutes had generally not useful positive and negative
likelihood ratios. ‘Severe variable late’ decelerations also had not useful positive and
negative likelihood ratios for predicting poor neonatal outcomes, although the study was very
old with no clinical relevance. The evidence was of low and very low quality.
Variable decelerations
Evidence from 2 observational studies with over 150 participants showed that variable
decelerations (defined as ‘not severe’) had no significant association with neonatal death,
umbilical cord artery pH less than 7.1, or the need for neonatal resuscitation.
There was some evidence of an association between variable decelerations and grade III or
IV intraventricular haemorrhage. There was some evidence that variable decelerations with a
depth of 30–60 bpm had a higher degree of association with fetal acidosis when compared
with variable decelerations with a depth of less than 30 bpm. However, this association was
only seen for babies under 34 weeks’ gestation and the numbers of cases in the studies was
very small. There was some evidence that variable decelerations classified as severe with a
late component had a high degree of association with neonatal death, while mild/moderate
variable decelerations without a late component had a low degree of association with
neonatal death, but again, the numbers of cases in the study was very small. The evidence
was of low quality.

287
Fetal monitoring
Categorisation/classification of fetal heart rate traces

Evidence from 5 observational studies (number of participants ranged from 72 to 383) for the
diagnostic values of a range of different categorisations of CTGs were conflicting, ranging
from high to low across all diagnostic parameters. Studies describing CTGs as “reassuring”,
“pathological”, “suspicious” and “ominous” all found low and not useful diagnostic accuracy
across all diagnostic parameters for adverse neonatal outcomes. Two studies describing the
CTG as “abnormal” found moderate to low sensitivity and high to low specificity, high to low
positive likelihood ratios and low negative likelihood ratios for adverse neonatal outcomes.
The evidence was predominately of low and very low quality.
There was some evidence of a high association between CTGs categorised as abnormal
(however defined) and neonatal death but there was no association with other poor neonatal
outcomes. The evidence was of low quality.

No search for health economic evidence was undertaken for this question as it was thought
to be concerned with how to interpret a trace rather than making decisions between
alternative courses of actions.

The Guideline Development Committee prioritised neonatal death and fetal acidosis as the
most important outcomes for this question. The committee felt it was important to assess
how effective the cardiotocograph is at identifying babies with fetal hypoxia that may lead to
acidosis and other adverse outcomes both in terms of identifying true positives and ruling out
false negatives.

The Guideline Development Committee discussed that the physiological control of fetal heart
rate differs in preterm and term fetuses. The additional impact of immaturity of control of fetal
heart rate in preterm fetuses makes the CTG interpretation different from term. They noted
that some characteristics and patterns of the FHR are dependent on gestational age as they
reflect the development and maturity of the central nervous system as well as the
cardiovascular system. In a term fetus certain FHR features may be pathological but in a
preterm fetus they could be physiological. The committee noted that the mean gestational
age of women in most of the included studies ranged from 26 to 30 weeks. It was
disappointing that no subgroup analysis was performed in the included studies for the
significance of fetal heart rate patterns at different gestations.
The committee found the definition of a normal CTG for preterm fetuses challenging as the
evidence was limited. Although the category of abnormal CTG seems to be the best predictor
of birth asphyxia, this result relates to the least useful outcome in the results because its
incidence is very high in preterm babies.
The committee noted that in very premature labour (24 to 26 weeks) there is a high risk of
neonatal morbidity and mortality, and survival is dependant more on fetal weight and maturity
than on intrapartum hypoxia and mode of delivery.

288
Fetal monitoring
The committee also recognised that it is well established that the baseline fetal heart rate in
preterm fetuses is at the higher end of the normal range for a term fetus for physiological
reasons, but that this reverts to the range more consistent with term fetuses as gestation
advances. However, they felt that any rate more than 160 bpm should be defined as
tachycardia across all preterm gestational ages. The committee was concerned that women
in preterm labour were at increased risk of infection (such as chorioamnionitis) and that this
may present as a persistent fetal tachycardia, giving rise to a risk of misinterpretation of the
CTG.
The committee discussed that the baseline variability may be reduced at preterm gestations
for physiological reasons. However, at term, fetal heart rate variability is an important clinical
indicator of fetal acid base balance and oxygenation of the autonomic nerve centres within
the brain. In the term setting, sustained absent variability is predictive of cerebral asphyxia. In
this review, there was some evidence of this in 1 study with moderate specificity and
moderate useful likelihood ratio for adverse neonatal outcomes. However, given a possible
physiological explanation for a higher baseline and reduced variability, these features should
not be considered alone as indications for operative interventions in the preterm setting.
The committee also considered whether accelerations in the fetal heart rate of very preterm
babies may not be present or their height and frequency may be significantly reduced. The
committee agreed that fetal heart rate decelerations are common and normal at very early
preterm gestations (26 weeks and less) reflecting immature development of cardioregulatory
mechanisms. They discussed that the presence of shallow or short-lived decelerations in
very preterm babies should not be considered necessarily as an indicative of hypoxia when
all other CTG features are reassuring. From 27 weeks onwards the frequency and height of
accelerations increases and decelerations are normally not physiological. Importantly,
anticipated survival following preterm delivery also improves.
The committee believed that although electronic fetal monitoring guidelines for term fetuses
(see the NICE guideline intrapartum care) cannot be always applied during labour to preterm
fetuses, they can be considered as relevant after 32 weeks, as physiological maturity of the
cardiovascular and neurological systems from this gestational age is comparable with that of
term fetuses. Thus, from 32 weeks, baseline fetal heart rate and variability should be similar
to that in term fetuses and accelerations with an amplitude of more than 15 beats from the
baseline should be present as an indicator of fetal wellbeing. Decelerations can be
interpreted as for the term fetus. The committee discussed that theoretically, compared with
term fetuses, preterm fetuses tend to have lower reserves and may deteriorate more quickly
than term fetuses. Thus earlier and/or more prompt intervention may be required than for
term fetuses.
The committee believed that a normal CTG is reassuring that the baby is in good condition.
An abnormal CTG does not always indicate that the outcome for the baby will be poor. There
is considerable variation between individuals in what is considered normal and abnormal
CTG. The committee was aware that there are risks to the mother and fetus if an abnormal
CTG is used as the sole indication for intervention.
The committee commented that clinical staff should not focus only on the CTG when caring
for the woman in preterm labour, but should take the full clinical picture into account.
The committee noted that the evidence from this review showed that the use of CTG is only
moderately useful at best in predicting poor fetal/neonatal outcomes, with the majority of
studies showing it to be not useful (not useful positive likelihood ratios).
The data showed that only a few CTG features have some limited evidence supporting their
usefulness in predicting fetal outcome:

289
Fetal monitoring
• Fetal heart rate (bradycardia, tachycardia) did not seem to be of value (see Table 81 and
Table 82).
• Abnormal baseline variability) of less than 5 bpm for 20 minutes in 1 hour before birth
was moderately useful positive likelihood ratio in 1 study for identifying cord pH less than
7.20 and neonatal death (see Table 83 and Table 84).
• Accelerations did not seem to be of value (see Table 85).
• Decelerations:
o Late, ‘prolonged’ and ‘severe variable late’ decelerations seemed to predict low cord
pH values and absence of this feature was associated with normal pH (see Table 86).
o Variable decelerations had a positive association with intraventricular haemorrhage
but the evidence was from a single very old study (see Table 87).
Different categorisations/classifications of CTGs were associated with conflicting evidence of
their value in predicting adverse outcomes or reassurance (see Table 88 and Table 89).The
committee felt that individual parameters in CTG could not be viewed and interpreted alone.
The available evidence does not support the assumption that the CTG tracing can be viewed
precisely. The evidence presented in the studies takes no account of the gestational age and
the degree of prematurity and the potential for physiological changes of the fetal heart rate to
be considered as pathological. It is for these reasons that the committee felt that the
classification should be less complex and less rigid. They felt that there is a need to consider
the CTG as part of a bigger picture and CTG alone should not be the basis for intervention
decisions. The committee emphasised that the potential for harm arising from a false positive
result in preterm labour is higher than in labour at term.
The committee agreed that women should be fully consulted before performing continuous
fetal heart rate monitoring.
The committee commented that clinical staff should not focus only on the CTG when caring
for the woman in preterm labour, but should take the full clinical picture into account.

As this question looked at the diagnostic accuracy of different features of fetal heart rate
traces, there were no resource use issues to consider.

The quality of the evidence reviewed varied from moderate to very low. The Guideline
Development Committee noted many limitations of the research findings.
In most studies the sample was small: much larger numbers of cases would be needed to
show a significant effect, particularly in terms of long-term neurodevelopment.
Clinicians were not, and could not be blinded in these observational studies, so would have
offered treatment based on the cardiotocograph. This ‘treatment effect’ is a serious problem:
it would have undermined the validity of the estimates of diagnostic accuracy because an
abnormal CTG is likely to prompt steps to intervene to delivery to improve neonatal outcome.
Therefore when a case of adverse outcome was avoided by intervention before any harm
occurred, it may well have erroneously been counted as a ‘false positive’.
The committee also noted that fetal hypoxia was not likely to be the only or the principal risk
for many preterm babies in labour. The interaction of these factors and their effect on the
CTG could be complex.
Much of the evidence reviewed came from old studies. Clinical practice has changed to such
an extent that the findings might not be relevant to current clinical practice. In particular, in 1

290
Fetal monitoring
study reported in 1978 the assessment of ‘central nervous system haemorrhage’ was not
robust as no cranial ultrasound was performed.

The Guideline Development Committee considered that the interpretation of CTG might be
different between healthcare professionals and therefore care should be taken when
interpreting CTGs so that appropriate action is taken where there are concerning signs.
The committee discussed that it would be more appropriate to establish principles rather than
precise parameters by which to assess the CTG.
The recommendations on interpreting fetal heart rate are in Section 13.6.
13.3 Monitoring options: cardiotocography and

intermittent auscultation
13.3.1 Introduction
Intermittent auscultation involves the healthcare professional listening to the fetal heart rate
at regular intervals. At term it is recommended that intermittent auscultation is undertaken
every 15 minutes in the first stage of labour and every 5 minutes in the second stage of
labour and for 1 minute after a contraction.
Continuous electronic recording can be undertaken using either an external ultrasound
transducer, positioned on the mother’s abdomen to pick up the fetal heart rate, or a fetal
scalp electrode (a small clip introduced through the mother’s vagina and attached to the
baby’s head). The outputs of both electronic methods are displayed as a cardiotocograph
(CTG) trace. The tocograph is a simultaneous recording of the uterine contractions, so the
CTG trace provides a visual continuous record of fetal heart rate and uterine contractions.
This section reviews the evidence comparing the effectiveness of these 2 monitoring
techniques to improve outcomes for babies born preterm.

What is the effectiveness of electronic fetal monitoring compared with intermittent
auscultation at different gestational ages for unborn babies whose mothers are in suspected
or diagnosed preterm labour?

As this question was set out to assess the comparative effectiveness of 2 interventions,
randomised controlled trials (RCTs) were selected as the best study design to answer this
review question. Comparative observational studies were considered when there were
limited RCT data. Two studies were included in this review (Luthy 1987, Shy 1988). Both
included studies were from the USA. One study was a multicentre RCT with 246 participants
(Luthy 1987) and the other was a retrospective cohort study with 304 participants carried out
in 14 hospitals that provided obstetric care (Shy 1988).
Both included studies examined the association between use of electronic fetal heart rate
monitoring (EFM) and intermittent fetal heart rate auscultation with neonatal outcomes in
women with suspected or diagnosed preterm labour. The term cardiotocography (CTG) is

291
Fetal monitoring
used throughout this review since this more accurately describes the monitoring carried out
during labour which simultaneously records the fetal heart rate and uterine contractions.
The mean pregnancy gestation of babies in 1 included study (Luthy 1987) was 26 weeks in
both study arms. No major differences were observed in this study between the CTG group
and intermittent auscultation group for maternal age, marital status, race, postnatal care and
birthweight. The other study (Shy 1988) included babies with birthweight from 700 to 1500 g
and did not analyse data based on the gestational age. Although women’s characteristics
were not reported in this study, the analysis was adjusted for birthweight, community hospital
birth, rupture of membranes and non-cephalic presentation.
The use of tocolytics was reported in both studies. In 1 study (Shy 1988) over 50% of women
in both intermittent auscultation and electronic fetal monitoring groups were exposed to
tocolytics, while in the other study tocolytics were only given to women with intact
membranes, although the number of women exposed to tocolytics is not reported (Luthy
1987). Fetal distress was the most common indication for caesarean section in 1 study
(Luthy 1987) in which it was reported that caesarean section for fetal distress was performed
for 8.2% of women with electronic fetal heart monitoring compared with 5.6% of women with
intermittent auscultation.

Data is reported in 2 GRADE profiles.
• Table 90: GRADE profile for comparison of cardiotocography versus intermittent
auscultation – RCTs
• Table 91: GRADE profile for comparison of cardiotocography versus intermittent
auscultation – observational studies.
The grading of evidence from the RCT started at high quality and was then downgraded if
there were any issues identified that would undermine the trustworthiness of the findings.
Evidence from the retrospective comparative observational study started at low quality and
was then downgraded if there were any issues identified.
Although the protocol was set up to investigate the effect of different gestational ages on
outcomes as subgroup analyses, one of the included studies looked at the role of babies’
birthweight and results are presented for this factor in a subgroup analysis as a surrogate of
gestational age.

292
Fetal monitoring
Table 84: GRADE profile for comparison of cardiotocography versus intermittent auscultation – RCTs
Number Other
of Risk of Inconsiste consider Cardiotocog Intermittent
studies Design bias ncy Indirectness Imprecision ations raphy auscultation Relative (95% CI) Absolute (95% CI)
Perinatal mortality
Perinatal mortality - birthweight 500–1500 grams
1 study Rando Very No serious Serious2 Serious3 None 14/122 14/124 RR 1.02 (0.51 to 2.04) 2 more per 1000 Very
(Luthy mised serious1 inconsisten (11.5%) (11.3%) (from 40 fewer to 84 low
1987) trial cy more)
Intracranial haemorrhage grade III/IVd
Total intracranial haemorrhage - birthweight 501–1750 grams
1 study Rando Serious1 No serious Serious2 Very serious4 None 20/122 16/124 RR 1.25 (0.68 to 2.30) 32 more per 1000 (from Very
(Luthy mised inconsisten (16.1%) (12.9%) 41 fewer to 168 more) low
1987) trial cy
Intracranial haemorrhage - birthweight 1101–1750 grams
1 study Rando Serious1 No serious Serious2 Very None 4/122 6/124 RR 0.67 (0.19 to 2.3) 16 fewer per 1000 (from Very
(Luthy mised inconsisten serious4 (3.2%) (4.8%) 39 fewer to 63 more) low
1987) trial cy
1 study Rando Serious1 No serious Serious2 Serious3 None 16/122 10/124 RR 1.6 (0.76 to 3.39) 48 more per 1000 (from Very
(Luthy mised inconsisten (12.9%) (8.1%) 19 fewer to 193 more) Low
1987) trial cy
Intracranial haemorrhage grade I/IIa
Total Intracranial haemorrhage - birthweight 501–1750 grams
1 study Rando Serious1 No serious Serious2 Serious3 None 19/122 27/124 RR 0.7 (0.41 to 1.2) 66 fewer per 1000 (from Very
(Luthy mised inconsisten (15.6%) (22.1%) 131 fewer to 44 more) Low
1987) trial cy
1 study Rando Serious1 No serious Serious2 Very serious4 None 12/122 16/124 RR 0.75 (0.37 to 1.52) 33 fewer per 1000 (from Very
1987) trial cy
(Luthy mised inconsisten (5.7%) (9%) 67 fewer to 53 more) low
1987) trial cy
Fetal monitoring
Number Other
studies Design bias ncy Indirectness Imprecision ations raphy auscultation Relative (95% CI) Absolute (95% CI)
Severe respiratory distress syndrome
Severe respiratory distress syndrome - birthweight 501–1750 grams
(Luthy mised inconsisten (27%) (28.2%) 102 fewer to 124 more) low
1987) trial cy
1987) trial cy
1987) trial cy
Umbilical cord arterial pH <7.20
1987) trial cy
Umbilical cord arterial pH ≥7.20
1 study Rando Serious1 No serious Serious2 Serious3 None 74/122 72/124 RR 1.04 (0.85 to 1.28) 23 more per 1000 (from Low
(Luthy mised inconsisten (60.7%) (58.1%) 87 fewer to 163 more)
1987) trial cy
Maternal outcomes
Mode of birth – Caesarean birth
1987) trial cy
Mode of birth – Spontaneous birth
1 study Rando Serious1 No serious Serious2 No serious None 88/122 97/124 RR 0.92 (0.80 to 1.07) 63 fewer per 1000 (from Low
(Luthy mised inconsisten imprecision (72%) (78%) 156 fewer to 55 more)
1987) trial cy
a. Outcome not defined in the paper

1. Unclear how outcomes were ascertained, diagnosed or verified
2. Data was not analysed based on gestational age
3. Confidence interval crossed 1 default MIDs
Fetal monitoring
Table 85: GRADE profile for comparison of cardiotocography versus intermittent auscultation – observational studies
Number Other
studies Design bias ncy Indirectness Imprecision ations raphy auscultation Relative (95% CI) Absolute (95% CI) Quality
Perinatal mortality
1 study Cohort Very No serious No serious Very serious4 None 82/213(39%) 49/91(54%) Adjustedb RR 0.91 48 fewer per 1000 Very
(Shy serious1,2, inconsisten indirectness a
(0.65 to 1.3) (from 188 fewer to 162 low
1988) 3 cy more)
1 study Cohort Very No serious No serious No serious None 30/136(22%) 10/37(27%) Adjustedb RR 0.82 49 fewer per 1000 Very
(Shy serious5,6, inconsisten indirectness imprecisio (0.39 to 1.7) (from 165 fewer to 189 low
7
1988 cy n more)
1 study Cohort Very No serious No serious No serious None 52/77 (68%) 39/54(72%) Adjustedb RR 0.94 43 fewer per 1000 Very
(Shy serious5,6, inconsisten indirectness imprecisio (0.63 to 1.4) (from 267 fewer to 289 low
7
1988 cy n more)
CI confidence interval, EFM electronic fetal monitoring, IA intermittent auscultation, MID minimally impo rtant difference, RR risk ratio
a. Reported wrongly as 31% in the published paper.
b. Adjusted for birth-weight, community hospital birth, rupture of membranes, and non-cephalic presentation
1. Unclear on what basis women allocated to have IA or EFM
3. No standard protocol for intermittent auscultation used in 14 participating hospitals
5. Unclear how outcomes were ascertained, diagnosed or verified
6. Data was not analysed based on gestational age
Fetal monitoring

Maternal and neonatal outcomes
Evidence from 1 RCT (n=246) showed no significant difference in the risk of perinatal
mortality, intracranial haemorrhage, severe respiratory distress syndrome, seizure and
umbilical artery pH at birth in preterm babies born to women monitored with CTG compared
with those receiving intermittent auscultation. The evidence across all outcomes was of very
low quality.
Evidence from the same RCT (n=246) showed no significant difference in rates of caesarean
section or spontaneous vaginal birth between women who received intrapartum CTG and
women who received intermittent auscultation. The evidence was of very low and low quality.
Very low quality evidence from 1 retrospective observational study (n=304) also showed no
significant difference in the risk of perinatal mortality in preterm babies born to women
monitored with CTG compared with those receiving intermittent auscultation.
No evidence was found for the outcomes of: trauma/injury to the baby; periventricular
leucomalacia (PVL) or white matter injury; neonatal sepsis; need for mechanical ventilation;
and length of stay in neonatal intensive care unit or neonatal unit.

A search was undertaken for health economic evidence on electronic fetal monitoring
compared with intermittent auscultation at different gestational ages for unborn babies whose
mothers are in suspected or diagnosed preterm labour. A total of 54 articles were identified
by the search. After reviewing titles and abstracts, no papers were obtained. Therefore, no
relevant economic evidence was identified for this question.
This question was not identified as a priority for health economic analysis as it was thought
by the Guideline Development Committee that any recommendations would have a relatively
low cost, especially as the equipment is readily available on labour wards.

In this review, the Guideline Development Committee aimed to investigate to find whether
the use of continuous CTG in labour was any more effective than intermittent auscultation in
identifying fetuses who are at greater risk of poor outcomes arising as a complication of
hypoxia-acidosis during preterm birth. The key outcomes of interest were fetal and neonatal
death and the rates of more serious morbidities such as intracranial haemorrhage and
respiratory distress syndrome.

There was limited evidence in this area with only 2 studies (1 randomised, 1 comparative
retrospective) of 550 preterm labours contributing to this review. The evidence showed there
were no significant differences in any of the identified clinical outcomes in the CTG group
compared with the group monitored by intermittent auscultation (IA). Further subgroup
analysis by birthweight showed no significant outcomes for intracranial haemorrhage,
respiratory distress syndrome and perinatal mortality when analysed between the 2 groups.
However, the Guideline Development Committee felt that there was likely insufficient power
to detect such outcomes, as most of the informative cases (event rates) would be among

296
Fetal monitoring
babies with the lowest gestational ages. Furthermore, the committee acknowledged that
although the caesarean section rates were not different between the two groups, they could
not be sure how much emphasis should be placed on this finding, given the limitations of the
studies. They also noted that the caesarean section rates were relatively low in both groups
compared with current rates but speculated that this probably was a reflection of changes in
practice rather than anything to do with the actual monitoring strategy.
The committee therefore concluded that the evidence from the 2 studies failed to
demonstrate any benefit or harm from the use of CTG compared with IA in preterm labour.
They argued that, in contrast to the term setting, the outcome in preterm babies was more
likely to be determined by factors such as gestational age, birthweight and whether steroids
were administered rather than intrapartum hypxoxia-acidosis (which is what intrapartum fetal
monitoring is intended to detect). The committee did not feel they could make a strong
recommendation about the use of one method over the other. The committee was aware that
many women in preterm labour would have additional risks that would prompt the use of
CTG and were also aware that even in the absence of risk factors, use of CTG is in common
practice for most women in preterm labour.
The committee felt that women’s views should be taken into account when the decision of
fetal monitored is made and that it was important to provide ongoing information and support
for mothers when using either CTG or IA. Some women will have a preference for IA
because the intervention means that there will be greater interaction with a midwife (who
would be present continuously though her labour) undertaking IA correctly every 15 minutes
during labour. However, some women may get more reassurance by not only the presence
of the midwife but also having the more detailed information from a CTG.
The committee also recognised that although in the majority of preterm labours monitoring
the fetal heart rate would be considered standard practice, in certain circumstances an active
decision may be taken not to monitor (for example with extreme prematurity if the results of
monitoring would not inform the obstetric or neonatal management). The committee noted
that in very premature labour (less than 26 weeks) there is a high risk of neonatal morbidity
and mortality, and survival is dependant more on fetal weight and maturity than on
intrapartum hypoxia and mode of delivery. The committee did not discuss this in detail as
they felt that this should be approached on a case-by-case basis taking account of the
woman’s views (and her partner’s or family’s preferences as appropriate) and the individual
circumstances surrounding the pregnancy. Overall the committee felt that, especially at low
gestations, the decision-making process regarding the decision to monitor and the method by
which to monitor was a difficult one and therefore a senior clinician needs to be involved in
the discussion.
With regards to continuous CTG monitoring, the committee recognised the importance of a
good quality recording and that in circumstances where an external transducer may not
provide this (for example where the woman has a high BMI), there would be the potential
need for use of a fetal scalp electrode (FSE). The committee acknowledged that the formal
review of FSE had not been able to identify any evidence that met the protocol. The
committee had a consensus view that FSE had the potential to cause harm in preterm babies
and concluded that in the absence of evidence for or against its use, the potential risks and
benefits of FSE in the preterm fetus should be considered and discussed with women on a
case-by-case basis (see Section 13.4 on use of FSE).
With regard to monitoring by IA, the committee agreed that in the absence of any evidence to
the contrary, this should be carried out in accordance with the guidelines on monitoring term
fetuses. They believed that although electronic fetal monitoring guidelines for term fetuses
(see the NICE guideline on intrapartum care) cannot be always applied during labour to
preterm babies, they can be considered as relevant after 32 weeks, as physiological maturity
of the cardiovascular and neurological systems from this gestational age is comparable to

297
Fetal monitoring
that of term babies. Thus, from 32 weeks, baseline fetal heart rate and variability should be
similar to that in term fetuses and accelerations with an amplitude of more than 15 beats
from the baseline should be present as an indicator of fetal well-being. Decelerations can be
interpreted as for the term fetus. The Committee discussed that theoretically, compared to
than term fetuses. Thus earlier and/or more prompt intervention may be required compared
to term fetuses. With regards to monitoring by IA, the Committee agreed that in the absence
of any evidence to the contrary, this should be carried out in accordance with the guidelines
on monitoring the term fetus. The Committee felt that women’s views should be taken into
account when the decision of fetal monitoring is made and that it was important to provide
ongoing information and support for mothers when using either CTG or IA. They believed
that although electronic fetal monitoring guidelines for term fetuses (see the NICE guideline
on intrapartum care) cannot be always applied during labour to preterm babies, they can be
considered as relevant after 32 weeks’ gestation, as physiological maturity of the
cardiovascular and neurological systems from this gestational age is comparable with that of
term babies. Thus, from 32 weeks, baseline fetal heart rate and variability should be similar
to that in term fetuses and accelerations with an amplitude of more than 15 beats from the
baseline should be present as an indicator of fetal wellbeing. Decelerations can also be
interpreted as for the term fetus. The committee discussed that theoretically, compared with
than term fetuses. Thus earlier and/or more prompt intervention may be required compared
with term fetuses.
With regards to monitoring by IA, the Committee agreed that in the absence of any evidence
to the contrary, this should be carried out in accordance with the NICE guideline on
Intrapartum Care. The committee felt that women’s views should be taken into account when
the decision of fetal monitoring is made and that it was important to provide ongoing
information and support for mothers when using either CTG or IA. The committee
acknowledged that performing IA is intensive and emphasised that its effectiveness as a
method of monitoring would be highly dependent on availability and competence of staff to
auscultate the fetal heart for 1 minute every 15 minutes in the first stage and every 5 minutes
in the second stage.
The committee noted the importance of auscultation for 1 minute immediately after a
contraction in order to confirm the absence of late decelerations as this is reassuring (in IA
and CTG) and is highlighted in the recommendations.
The committee also recognised that the issue of woman’s mobility during fetal monitoring if
continuous cardiotocography is needed has been covered by the NICE guideline on
Intrapartum Care and this set of recommendations in the guideline cross refer to the relevant
sections of that guideline.

The clinical evidence showed no difference in the outcomes when CTG was compared with
IA when used in preterm labour. Both technologies are readily available in most settings
already and thus it is unlikely that this intervention would have major cost implications.

The quality of the evidence was predominately very low for the different outcomes. The
Guideline Development Committee noted that the proportion of high-risk women in each arm
of the cohort study might not have been comparable and both studies were old and
underpowered to detect important outcomes like perinatal mortality. However, in principle,
the adverse outcomes that fetal monitoring is designed to prevent are more common in
preterm babies than in normally grown singleton babies at term.

298
Fetal monitoring
Women in the IA group with abnormal fetal heart monitoring were assigned to have
caesarean section, and this does not reflect current clinical practice in that it is unlikely that
women would go straight from IA to caesarean section without EFM inbetween.
Overall, the committee did not feel sufficiently confident in the findings to make strong
recommendations in favour of either CTG or IA.

Despite the paucity of research evidence relating to the method of fetal heart monitoring in
preterm labour, the Guideline Development Committee felt it was nevertheless important to
monitor fetal heart rate by some means during preterm labour. The committee discussed the
fact that physiological reserves available to combat hypoxia are less than those available to
term babies. Hence, a preterm baby may become hypoxic sooner than its term counterpart.
The committee felt that it was important to inform women of the lack of evidence of benefit of
CTG versus IA prior to offering monitoring.
The recommendations on monitoring options are in Section 13.6.
13.4 Fetal scalp electrode

13.4.1 Introduction
Fetal scalp electrode (FSE) placement is used in electronic fetal heart monitoring (EFM) to
assess the fetal heart rate (FHR) pattern when external monitoring cannot be used or when
the signal quality is poor. There is uncertainty about risks and benefits of using an FSE to
perform EFM in preterm labour. There is concern that application of FSE is associated with
an increase in the risk of trauma and infection in preterm babies. Evidence is needed to
determine if the benefits of using an FSE to perform EFM outweigh any risks and if there are
gestational ages at which the risks of FSE outweigh any potential benefits.

At what gestational age can a fetal scalp electrode be used for unborn babies whose
mothers are in diagnosed preterm labour?

No studies were identified which met the inclusion criteria for this review question.

No evidence profile was generated.

No evidence was identified that addressed this question.
13.4.6 Health Economics profile

No health economic studies were identified and no health economic modelling was planned
for this question.

299
Fetal monitoring

The aim of fetal heart rate monitoring, whether it is by intermittent auscultation or continuous
electronic fetal monitoring, performed either externally using an ultrasound transducer or by a
fetal scalp electrode, is to detect ‘fetal distress’ and inform the use of interventions (primarily
birth by caesarean section) before hypoxia-acidosis occurs and cause harm to the unborn
baby.
In order to establish whether the use of an FSE was of benefit in preterm labour, the
Guideline Development Committee identified poor neonatal outcomes pertaining to hypoxia
and acidosis (including mortality and intraventricular haemorrhage/periventricular
leucomalacia) as priority outcomes, as well as cord blood gas values at birth as another
measure of acidosis.
Balanced against this was the committee’s wish to understand whether FSE is associated
with iatrogenic harm to the fetus and/or the woman. The committee felt that the attachment of
an electrode to the preterm fetus’s soft scalp could theoretically cause trauma as well as
local or more widespread infection in the fetus. These fetal outcomes were therefore
prioritised. In terms of the woman, the committee was interested to know whether the use of
FSE was associated with an increase or a reduction in delivery by caesarean section, and
hence mode of birth was prioritised. Maternal mortality was also measured as it was
acknowledged that this would always be of concern as a potential harm.
Length of stay in a neonatal intensive care unit or neonatal unit and need for mechanical
intervention were also prioritised as these were considered to be potentially relevant to both
poor outcomes arising from the presence of acidosis and iatrogenic harm caused by the use
of FSE.

There was no evidence that met the protocol which could demonstrate whether FSE usage
was of benefit or harm to preterm babies, but there was consensus in the Guideline
Development Committee (based on their knowledge of preterm fetal anatomy) that attaching
a scalp electrode had the potential to cause complications.
The committee also noted that it is not current practice to routinely use FSE in preterm
fetuses less than 34 weeks.
There was agreement that it is not always possible to obtain a good quality fetal heart rate
trace from continuous electronic fetal monitoring performed externally using an ultrasound
transducer (for example it can be difficult in women with high BMI).
The committee was conscious of the fact that when a fetal heart rate cannot be detected
adequately, the likely result is that a caesarean section will be offered, and if this was the
case it might be preferable to apply an FSE rather than deliver the fetus when potentially
there is nothing wrong. The committee was also aware, however, of the limited evidence
base for both the effectiveness of any form of continuous fetal monitoring (see Section 13.3)
and the predictive value of the CTG trace (see Section 13.2) so felt that it was only
appropriate to make a weak recommendation that FSE should be considered in such
circumstances.
They also agreed that if FSE was considered then it was important to explain to the woman
that the risks and benefits of FSE were unknown and that there were possible alternatives to
EFM with FSE, including the option not to monitor at all.

300
Fetal monitoring

The Guideline Development Committee noted that the use of FSE is potentially costly given
the level of skill required to deliver the intervention. It is not current practice below 34 weeks’
gestation and if harm was caused through its use, this harm could potentially be serious and
associated with high treatment costs. However, the committee was also aware that any
decision to use FSE would be taken with a view to preventing other serious harms to the
fetus or the woman which would themselves be associated with high losses in quality
adjusted life years (QALYs), so overall they felt that balance between the potential benefits
and resource use supported a weak recommendation.

No evidence was identified for inclusion in the review.

There were no other considerations.
13.4.7.6 Key conclusions

FSE should only be considered when continuous electronic fetal monitoring using an external
ultrasound transducer cannot be performed. The decision to use an FSE should be taken in
discussion with the woman and a senior obstetrician. The Guideline Development Committee
felt that in particular, it should be explained to the woman that the risks of FSE are unknown
but that in the absence of being able to monitor externally, these must be balanced against
the alternatives of no monitoring or expedited delivery.
The recommendations on use of FSE are in Section 13.6.
13.5 Fetal blood sampling

13.5.1 Introduction
Fetal blood sampling is an invasive procedure used to obtain fetal blood for measurement of
either pH or lactate. In preterm labour, as in term labour, it can be used as an adjunct to
electronic fetal monitoring to establish whether an abnormal heart rate pattern is due to
hypoxaemia/acidosis. It is unclear in the preterm setting whether fetal blood sampling may
confer additional benefit by reducing the risk of a false positive result from electronic fetal
monitoring. This must be balanced against the potential adverse effects on the mother and
fetus. A systematic review is needed to address the question of efficacy of fetal blood
sampling in preterm labour and to consider how this may vary for different gestational ages.

What is the utility of fetal blood sampling as an adjunct to intrapartum fetal heart rate
monitoring at different gestational ages?

No studies were identified which met the inclusion criteria for this review question.

301
Fetal monitoring

No evidence profile was generated.

No evidence was identified that addressed this question.
13.5.6 Health Economics profile

No health economic studies were identified and no health economic modelling was planned
for this question.

The aim of fetal heart rate monitoring is to detect ‘fetal distress’ and inform the use of
interventions (primarily birth by caesarean section) before hypoxia or acidosis occur and
cause harm to the unborn baby. Fetal blood sampling (FBS) is intended as an adjunct to
continuous electronic fetal monitoring with the aim of confirming or refuting the presence of
hypoxia or acidosis in those fetuses with an abnormal fetal heart rate pattern.
In order to establish whether FBS was effective, the Guideline Development Committee
therefore identified poor neonatal outcomes pertaining to hypoxia and acidosis (including
mortality and intraventricular haemorrhage/periventricular leucomalacia) as priority
outcomes, as well as cord blood gas values at birth.
Balanced against this was the committee’s wish to understand whether FBS is associated
with iatrogenic harm to the baby and/or the woman. The committee felt the necessity to
pierce the scalp of a baby with an immature skull with large fontanelles to take a blood
sample meant that the use of FBS theoretically could cause trauma and neonatal sepsis;
these fetal outcomes were therefore prioritised. In terms of the woman, their main concern
was whether the use of FBS was associated with unnecessary caesarean section so mode
of birth was prioritised. Maternal mortality was also measured as it was acknowledged that
this would always be of concern as a potential harm.
Length of stay in neonatal intensive care unit or neonatal unit and need for mechanical
intervention were also prioritised as these were considered to be potentially relevant to both
poor outcomes arising from the presence of acidosis or iatrogenic harm caused by the use of
FBS.

There was no evidence that met the protocol that showed FBS was of benefit to preterm
fetuses but there was consensus in the committee (based on their knowledge of preterm fetal
anatomy) that piercing the fetus’s scalp in order to take a blood sample was potentially
associated with risks such as bleeding, infection and dural puncture.
The committee also noted that it is not current practice to routinely use FBS in preterm
fetuses less than 34 weeks’ gestation.
The committee was aware of some observational studies (that did not conform to the
protocol and thus were not included in the review) in which women were recruited from as
early as 26 weeks’ gestation and FBS was used. However, these studies did not provide any
information about adverse outcomes and were unhelpful in addressing the issues of risk.

302
Fetal monitoring
However, they did confirm that FBS is used in some circumstances at these low gestational
ages.
The committee was aware that the NICE guideline on intrapartum care recommends the use
of FBS as an adjunct to electronic fetal monitoring in term babies but did not feel that they
could extrapolate this to the entire preterm population because of the theoretical
consequences of anatomical and physiological differences between term and preterm
babies, especially at lower gestational ages.
The committee also noted that FBS would only be used as an adjunct to electronic fetal
monitoring and so if electronic fetal monitoring is not being used (for example if intermittent
auscultation is being carried out instead), then there is no justification for using FBS.
In light of all these considerations the committee felt that FBS should not be used in fetuses
below 34 weeks’ gestation on the basis of theoretical concerns about complications below
this gestation. Their main concerns related to the risk of excessive bleeding and of cerebro-
spinal fluid leakage resulting from accidental dural puncture via the anterior fontanelle. These
risks are likely to be mitigated by advancing gestation and therefore the committee felt, after
consultation with a senior obstetrician and the woman, that FBS could be considered after 34
weeks’ gestation if the likely benefits outweigh the risks. The committee felt that it was
reasonable to recommend that if an FBS is performed, it should be done in accordance with
the advice given in the NICE guideline on intrapartum care.

The Guideline Development Committee noted that the use of FBS is potentially costly given
the level of skill required to deliver the intervention, the fact that it is not current practice
below 34 weeks’ gestation and the fact that if harm was caused through its use, this harm
could potentially be serious and associated with high treatment costs. Overall they felt that
the balance between the potential benefits and resource use only supported a weak
recommendation for the use of FBS in the specific subgroup of fetuses between 34+0 and
36+6 gestational weeks.

No evidence was identified for inclusion in the review.
13.5.7.5 Key conclusions

FBS should not be performed before 34 gestational weeks. The use of FBS in fetuses
between 34+0 and 36+6 gestational weeks should be discussed with a consultant
obstetrician and with the woman. If FBS is performed it should be done in accordance with
the advice given in the NICE guideline on intrapartum care.
updated guideline.

5. Is intermittent auscultation or electronic fetal monitoring effective
Research question in the preterm fetus?
Why this is needed

303
Fetal monitoring
5. Is intermittent auscultation or electronic fetal monitoring effective

Research question in the preterm fetus?
Importance to Intermittent auscultation of the fetal heart creates a very different patient
‘patients’ or the experience to continuous electronic recording, but the relative impacts of
population these modes of fetal monitoring in terms of clinical decision-making and
infant outcome has not been tested in a population of women in preterm
labour. Fetal monitoring is ingrained as part of term deliveries and there is
a lack of evidence to allow extrapolation from term birth to preterm birth,
both in terms of which mode to use and how to interpret the results.
Identification of the fetus at risk of hypoxia-acidosis would reduce
unnecessary intervention whilst ensuring delivery of the compromised fetus
before neurological damage or death ensues.
Relevance to NICE High priority, as there is currently very poor quality evidence on which to
guidance base Committee recommendations.
Relevance to the NHS The experience of women in labour at any gestation is an important metric
of the quality of maternity services. Improved detection of the fetus at risk
of hypoxia-acidosis would reduce mortality and morbidity from preterm
birth.
National priorities NHS outcomes framework 2015-6, #4, Ensuring that people have a
positive experience of care.
NHS Outcomes Framework #1: Preventing people from dying prematurely
Current evidence Existing trials reviewed by the Committee have small numbers and serious
base methodological flaws.
Equality There are no obvious equality issues; the population is defined by
gestational age.
Feasibility There is no reason in principle why more adequately designed and
powered trials should not be carried out. The ethical issues are not in
principle different from those affecting other perinatal trials.
Other comments None

304
Mode of birth
14 Mode of birth
14.1 Introduction
The potential health risks facing babies born preterm may be compounded by complications
occurring at the time of delivery. In clinical practice, caesarean section (CS) delivery is
performed if there are fetal indications that this would be safer than vaginal birth (for example
evidence of hypoxia-acidosis in labour when vaginal delivery is not imminent and rapid
delivery may prevent permanent neurological damage). CS would also be preferable to a
traumatic vaginal delivery (for example, with a preterm footling breech presentation, most
clinicians would offer CS rather than vaginal delivery). However, the value of CS in
comparison with vaginal birth in the absence of clinical indications is uncertain for women in
suspected or diagnosed preterm labour.

For women who present in suspected or diagnosed preterm labour (who have not planned
antenatally to give birth by caesarean section (CS) and for whom there are no other known
indications for CS birth), what is the clinical effectiveness of deciding to carry out a CS
compared with deciding to allow vaginal birth?
This review question aims to assess whether there is any difference in maternal and
neonatal outcomes when CS (which has not been planned before the onset of labour for
other indications) is compared with vaginal birth for women in suspected or diagnosed
preterm labour. As this question was set out to assess the comparative effectiveness of 2
interventions, randomised controlled trials (RCTs) were selected as the best study design.
The Guideline Development Committee also preselected subgroup analysis at the protocol
stage based on the following factors: breech presentation; cephalic presentation;
instrumental birth; and gestational age.
The committee had extensive discussion when the protocol was developed about the
description of the type of CS to be captured. The committee agreed that this question is not
about planned mode of birth decided antenatally but seeks to answer what is the optimum
mode of birth for women who present in preterm labour. It does not examine the emergency
use of CS for acute fetal or maternal compromise in women attempting vaginal delivery,
because they felt that there was overwhelming belief that emergency CS in this scenario was
beneficial. The differences between the focus of this review and much of the data in the
literature (which refer to emergency CS) was taken into consideration in the interpretation of
evidence and its influence in the decision-making.

One Cochrane systematic review (SR) and meta-analysis (Alfirevic 2013) with 4 component
RCTs (Penn 1996, Viegas 1985, Wallace 1984, Zlatnik 1993) is included in this review.
Studies which investigated a pre-planned mode of birth for women in suspected or
diagnosed preterm labour were excluded for the purposes of this review.
Two studies included in the Cochrane SR were conducted in in the USA (Wallace 1984,
Zlatnik 1993), 1 in the UK (Penn 1996) and 1 in Singapore (Viegas 1985).
All included trials examined the impact of the mode of birth (immediate CS or vaginal birth)
on neonatal outcomes in preterm and very low birth weight babies (gestational age across all
studies ranged from 26 weeks to 33 weeks) with cephalic or breech presentation. Three

305
Mode of birth
studies included preterm babies with only breech presentation (Penn 1996, Zlatnik 1993,
Viegas 1985) and 1 study included only babies with cephalic presentation (Wallace
1984).The quality assessment of the included trials was downgraded due to study design
(outcome assessors were unblinded), incomplete outcome data and small sample size. In
addition, recruitment in all 4 included trials was stopped early.
The main issue with the interpretation of results in this systematic review is that a large
number of women randomised to one type of mode of birth or another actually gave birth by
the other method (in other words a crossover effect with women ending in a different group
from the one to which they were randomised). More specifically, 3 trials (Penn 1996, Zlatnik
1993, Wallace 1984) in the SR included 20% (9/46) of women who were allocated to the
caesarean section group but subsequently gave birth vaginally because birth was too rapid
to allow a caesarean to be performed, and 21% (9/43) of women allocated to vaginal birth
group who actually gave birth by caesarean section for fetal or maternal indications.

Data is reported in 2 GRADE profiles separately for neonatal and maternal outcomes:
• Table 92: GRADE profile for comparison of CS (which has not been planned before the
onset of labour for other indications) versus vaginal birth - neonatal outcomes
• Table 93: GRADE profile for comparison of CS (which has not been planned before the
onset of labour for other indications) versus vaginal birth – maternal outcomes
The grading of evidence from the systematic review was assessed at high quality and then
downgraded if there were any issues identified that would undermine the confidence in the
findings.

306
Mode of birth
Table 86: GRADE profile for comparison of CS (which has not been planned before the onset of labour for other indications) versus
vaginal birth - neonatal outcomes

Risk of Other Vaginal Relative
No of studies Design bias Inconsistency Indirectness Imprecision considerations CS Birth (95% CI) Absolute Quality
Perinatal death
1 meta-analysis Randomised Serious1 No serious Serious2 Serious3 None 2/46 (4.3%) 8/43 RR 0.29 132 fewer per Very
of 3 studies trials inconsistency (18.6%) (0.07 to 1000 (from 173 low
(Alfirevic 2013) 1.14) fewer to 26
more)
Perinatal death – Breech (subgroup analysis)
1 meta-analysis Randomised Serious1 No serious Serious4 Serious5 None 1/23 6/28 RR 0.28 154 fewer per Very
of 2 studies trials inconsistency (4.3%) (21.4%) (0.05 to 1000 (from 204 low
more)
Perinatal death – Cephalic (subgroup analysis)
1 study Randomised Serious1 No serious No serious Serious6 None 1/23 2/15 RR 0.33 89 fewer per Low
(Alfirevic 2013) trials inconsistency indirectness (4.3%) (13.3%) (0.03 to 1000 (from 129
3.29) fewer to 305
more)
Brain injury- Hypoxic ischemic encephalopathy
1 study Randomised Very No serious Very serious4 Very None 1/5 0/7 RR 4 (0.2 NC Very
(Alfirevic 2013) trials serious1,7 inconsistency serious6 (20%) (0%) to 82.01) low
Intracranial haemorrhage
1 meta-analysis Randomised Very No serious Serious8 Very None 4/56 4/54 RR 0.92 6 fewer per 1000 Very
of 4 studies trials serious1,5,7 inconsistency serious6 (7.1%) (7.4%) (0.27 to (from 54 fewer to low
(Alfirevic 2013) 3.14) 159 more)
Intracranial haemorrhage – Breech (subgroup analysis)
1 meta-analysis Randomised Very Serious9 Serious4 Very None 1/33 3/39 RR 0.58 32 fewer per Very
of 3 studies trials serious1,5,8 serious6 (3%) (7.7%) (0.12 to 1000 (from 68 low
more)
Intracranial haemorrhage – Cephalic (subgroup analysis)
Mode of birth

Risk of Other Vaginal Relative
No of studies Design bias Inconsistency Indirectness Imprecision considerations CS Birth (95% CI) Absolute Quality
1 study Randomised serious1,5,7 No serious No serious Very None 3/23 1/15 RR 1.96 64 more per Very
(Alfirevic 2013) trials inconsistency indirectness serious6 (13%) (6.7%) (0.22 to 1000 (from 52 low
17.1) fewer to 1000
more)
Abnormal follow-up in childhood (outcome not defined) – Cephalic
1 study Randomised Very No serious No serious Very None 4/23 4/15 RR 0.65 93 fewer per Very
(Alfirevic 2013) trials serious1,5,7 inconsistency indirectness serious6 (17.4%) (26.7%) (0.19 to 1000 (from 216 low
2.22) fewer to 325
more)
1 meta-analysis Randomised Serious1 No serious Serious2 Serious3 None 9/53 16/50 RR 0.55 144 fewer per Very
of 3 studies trials inconsistency (17%) (32%) (0.27 to 1000 (from 234 low
more)
Respiratory distress syndrome – Breech (subgroup analysis)
1 meta-analysis Randomised Serious1,5 No serious Serious4 Very None 6/30 12/35 RR 0.57 147 fewer per Very
of 2 studies trials inconsistency serious6 (20%) (34.3%) (0.25 to 1000 (from 257 low
more)
Respiratory distress syndrome – Cephalic (subgroup analysis)
1 study Randomised Very No serious No serious Very None 3/23 4/15 RR 0.49 136 fewer per Very
(Alfirevic 2013) trials serious1,5,7 inconsistency indirectness serious6 (13%) (26.7%) (0.13 to 1000 (from 232 low
1.88) fewer to 235
more)
Need for mechanical ventilation
1 study Randomised Very No serious Serious4 Very None 4/5 3/7 RR 1.87 373 more per Very
(Alfirevic 2013) trials serious1,5,7 inconsistency serious6 (80%) (42.9%) (0.71 to 1000 (from 124 low
4.88) fewer to 1000
more)
CI confidence interval, CS caesarian section, MID minimally important difference, RR risk ratio
1. Outcomes assessors were not blinded to the group allocation
2. Participants in 2 studies had breech presentation
3. Evidence was downgraded by 2 due to very serious imprecision as 95% CI crossed 2 default MIDs
4. All participants had breech presentation
5. Detail of allocation concealment not reported
6. Evidence was downgraded by 2 due to very serious imprecision as 95% CI crossed 2 default MIDs
7. Unclear detail of randomisation
Mode of birth

9. Evidence was downgraded by 1 due to serious heterogeneity (chi-squared p<0.1, I-squared inconsistency statistic of 50%–74.99%) and no plausible explanation was found
Table 87: GRADE profile for comparison of CS (which has not been planned before the onset of labour for other indications) versus
vaginal birth – maternal outcomes

No of Risk of Other Vaginal Relative
studies Design bias Inconsistency Indirectness Imprecision considerations CS birth (95% CI) Absolute Quality
Postpartum haemorrhage
1 meta- Rando Serious3 No serious Serious1 Very serious2 None 1/54 (1.9%) 0/51 (0%) RR 3.69 NC Very low
analysis of 3 mised inconsistency (0.16 to
studies trials 83.27)
(Alfirevic
2013)
Postpartum haemorrhage – Breech (subgroup analysis)
1 meta- Rando Serious3 No serious Serious4 Serious2 None 1/30 0/35 RR 3.69 NC Very low
analysis of 2 mised inconsistency (3.3%) (0%) (0.16 to
studies trials 83.27)
(Alfirevic
2013)
Postpartum haemorrhage – Cephalic (subgroup analysis)
1 study Rando Serious3 No serious No serious Very serious2 None 0/23 0/15 NC NC Very low
(Alfirevic mised inconsistency indirectness (0%) (0%)
2013) trials
Maternal wound infection
1 meta- Rando Serious3 No serious Serious1 Very serious2 None 1/53 1/50 RR 1.16 3 more per Very low
analysis of 3 mised inconsistency (1.9%) (2%) (0.18 to 1000 (from 16
studies trials 7.7) fewer to 134
(Alfirevic more)
2013)
Maternal wound infection – Breech (subgroup analysis)
Mode of birth

No of Risk of Other Vaginal Relative
studies Design bias Inconsistency Indirectness Imprecision considerations CS birth (95% CI) Absolute Quality
analysis of 2 mised inconsistency (3.3%) (2.9%) (0.18 to 1000 (from 23
studies trials 7.7) fewer to 191
(Alfirevic more)
2013)
Maternal wound infection – Cephalic (subgroup analysis)
1 study Rando Serious3 No serious No serious No serious None 0/23 0/15 RR 1.0 0 more per Moderate
(Alfirevic mised inconsistency indirectness imprecision (0%) (0%) 1000 (from 0
2013) trial more to
0more)
Other maternal infection (outcome not defined)
analysis of 3 mised inconsistency (18.9%) (8%) (1.02 to 1000 (from 2
studies trials 6.78) more to 462
(Alfirevic more)
2013)
Other maternal infection (outcome not defined) – Breech (subgroup analysis)
1 meta- Rando Serious3 No serious Serious1 Serious5 None 10/30 4/35 RR 2.63 186 more per Very low
analysis of 2 mised inconsistency (33.3%) (11.4%) (1.02 to 1000 (from 2
studies trials 6.78) more to 661
(Alfirevic more)
2013)
Other maternal infection (outcome not defined) – Cephalic (subgroup analysis)
1 study Rando Serious3 No serious No serious No serious None 0/23 0/15 RR 1.0 0 more per Moderate
(Alfirevic mised inconsistency indirectness imprecision (0%) (0%) 1000 (from 0
2013) trials more to
0more)
CI confidence interval, CS caesarian section, MID minimally important difference, NC not calculable, RR risk ratio
2. Evidence was downgraded by 2 due to very serious imprecision as 95%CI crossed 2 default MIDs
3. Outcomes assessors were not blinded to the group allocation
4. All participants had breech presentation
5. Evidence was downgraded by 1 due to serious imprecision as 95% CI crossed 1 default MID
Timing of cord clamping for preterm-babies

14.5.1 Neonatal outcomes
Perinatal mortality
Evidence from 1 SR of 3 RCTs (n=89) showed no significant difference in perinatal death in
preterm babies born by caesarean section (not planned before the onset of labour)
compared with those born by vaginal birth. A subgroup analysis by breech for cephalic
presentation confirmed the direction of results from the overall analysis. The evidence was of
low or very low quality.
Brain injury
Evidence from 1 small RCT (n=13) showed no significant difference in the proportion of
preterm babies with brain injury (hypoxic ischemic encephalopathy) when born by caesarean
section (not planned before the onset of labour) compared with those born by vaginal birth. A
subgroup analysis by the babies’ presentation found that the risk of brain injury was higher
with CS than with vaginal birth but not significantly.
Evidence from 1 SR of 4 studies (n=110) also showed no significant difference in the
incidence of intracranial pathology in preterm babies born by caesarean section (not planned
before the onset of labour) compared with those born by vaginal birth.
The evidence across all studies for these outcomes was of very low quality.
Neurodevelopment follow-up in childhood
There was very low quality evidence from 1 study (n=38) that showed no difference in the
incidence of abnormal follow-up in childhood in preterm babies born by caesarean section
compared with those born by vaginal birth. The evidence was from a small study and
abnormal follow-up was not defined by the study so results should be interpreted with
caution.
The evidence from 1 SR of 3 RCTs (n=103) showed no significant difference in the outcome
of respiratory distress syndrome in preterm babies born by caesarean section compared with
those born by vaginal birth. Evidence from 1 study (n=15) also showed no significant
difference in the need for mechanical ventilation in preterm babies born by caesarean section
compared with those born by vaginal birth. The evidence across all studies was of low and
very low quality.
14.5.2 Maternal outcomes

The evidence from 1 SR of 3 RCTs (n=105) showed no significant difference in the incidence
of postpartum haemorrhage in women whose babies were born by caesarean section
compared with those whose babies were born by vaginal birth. The evidence across all
studies was of very low quality.
Maternal infection

311
The evidence from 1 SR of 3 RCTs (n=103) showed no significant difference in the incidence
of maternal wound infection in women whose babies were born by caesarean section
compared with those whose babies were born by vaginal birth. However, evidence from the
same RCTs (n=103) showed significantly higher incidence of other maternal infection in
women whose babies were born by caesarean section compared with those whose babies
were born by vaginal birth. The evidence across all studies was of very low quality.

A search was undertaken for health economic evidence on women who present in suspected
or diagnosed preterm labour (who have not planned antenatally to give birth by caesarean
section [CS] and for whom there are no other known indications for CS birth) regarding the
clinical effectiveness of deciding to carry out a CS compared with deciding to allow vaginal
birth. A total of 82 articles were identified by the search. After reviewing titles and abstracts, a
single paper was obtained and then excluded as it was based on a fetal indication for
caesarean section. Therefore, no relevant economic evidence was identified for this
question.
This question was identified as a medium priority for health economic analysis as there are
potentially large resource implications. However, it was thought that the cost effectiveness
would simply reinforce the clinical evidence if that suggested the superiority of one mode of
birth. Conversely, if the clinical evidence is ambiguous then the committee thought that
economic analysis might provide little added value to aid decision-making. Ultimately, no new
health economic analysis was undertaken and the clinical evidence review did not
demonstrate the superiority of a particular mode of birth.

The Guideline Development Committee considered neonatal outcomes of critical importance
in answering the question whether CS which was not planned before the onset of labour
would be more effective than vaginal birth for women in suspected or diagnosed preterm
labour.
In relation to neonatal outcomes, neonatal mortality, long-term neurodevelopmental delay
and respiratory distress were considered the most critical outcomes. Caesarean section is
known to be associated with an increased risk of respiratory morbidity in term neonates,
probably because the process of vaginal delivery is associated with a more effective
transition to postnatal life.
With regard to maternal outcomes, infections acquired during delivery, although not
considered critical for this review question, were still considered to be important. The risk of
postnatal infection (pyrexia, endometritis, puerperal sepsis), thrombosis and pulmonary
embolism, and excessive blood loss are higher after caesarean section than vaginal delivery.
However, some of these complications could also be due to the underlying causes that lead
to the need for caesarean section, and not necessarily be CS-induced complications.
Although the committee considered that CS is a relatively safe procedure, maternal mortality
was still an important outcome to be considered.
The committee noted that CS for extremely preterm babies may pose technical difficulties
and require incision in the upper uterine segment. The committee was aware that a history of
CS birth, especially those that involve an incision of the upper uterine segment, will have
implications for the selection of mode of birth for future pregnancies.

312

The included evidence on neonatal outcomes in women in whom CS had not already been
planned before the onset of labour concluded that CS was neither beneficial nor harmful
compared with vaginal birth. Although there was an indication that the risk of perinatal
mortality for babies may be lower for those who were delivered by CS compared with vaginal
birth, this evidence was inconclusive.
In addition, there was no clear direction of effect for neonatal outcomes from investigating
which is the optimal mode of delivery of preterm babies in subgroup analysis by cephalic or
breech presentation of the baby.
However, there was some evidence showing that there may be an adverse effect of CS in
increasing ‘other’ maternal infection.
The committee was aware that in current clinical practice the selection of the mode of birth
for preterm babies is often extrapolated from full-term babies. For example, if the baby has a
breech presentation then CS would be the most favoured mode of birth. There was no
evidence in this review to suggest that following current practice for term babies (such as
delivery by CS for breech presentation) would be harmful to the baby. Indeed, the committee
noted that the point estimate for reduction in perinatal death in babies with breech
presentation was a risk ratio (RR) of 0.28 (confidence interval [CI] 0.05 to 1.49), which is a
considerable reduction in risk but with low precision given the wide confidence intervals.
Given that all women included in the studies were at 26–36 weeks’ gestation, the committee
decided to make no recommendations about the optimal mode of birth for pregnant women
at below 26 weeks’s gestation.

A planned caesarean section is usually a more expensive procedure than a planned vaginal
birth but an unplanned emergency caesarean section following a planned vaginal birth is
more expensive still. However, the committee did not think that the costs of a particular mode
of birth would be an important driver of cost effectiveness if a particular mode of birth
produced better maternal and neonatal outcomes, as the cost of adverse outcomes and
complication would more than offset any differential in the cost of birth itself. However, the
evidence did not demonstrate the superiority of any one mode of birth and therefore the
implications for resource uses and health benefits, if any, is uncertain.

The evidence across all studies was of very low quality mainly due to serious limitations on
risk of bias and imprecision which gives less confidence to the direction of effects.
The main methodological challenge for this review was the very limited data from
randomised trials (due to low numbers of women recruited) and the high proportion of babies
(20%) who were not delivered by the planned (randomised) mode of birth. This can restrict
the generalisation of results as the crossover effect of moving from one randomised arm to
another can introduce bias. Crossover may reflect rapid progress of preterm labour which
ends in vaginal birth despite caesarean section being planned and, conversely, problems
developing during labour may require caesarean section despite aiming for vaginal
birth.Three out of 4 studies in the meta-analysis included preterm babies with only breech
presentation who are also at higher risk of developing complications than preterm babies
with cephalic presentation.

313

The Guideline Development Committee was aware of the evidence regarding CS at term, as
reviewed in the NICE guideline on caesarean section. They felt that the significant maternal
effects (such as perineal and abdominal pain during birth, and 3 days post-partum, injury to
vagina, early postpartum haemorrhage and obstetric shock) of CS would be similar at term
and preterm, although preterm CS would be more likely to require a vertical uterine incision,
after which most clinicians would advise caesarean delivery in the next pregnancy. They
noted the adverse effects of increased blood loss and risk of wound infection and venous
thromboembolism following surgery.
The Committee had less confidence about extrapolating the neonatal effects of CS from term
to preterm, other than that upper segment incision has implications for future delivery
compared with standard lower segment CS. Nevertheless, the committee noted that babies
born following CS at term had an increased chance of admission for respiratory distress
compared with babies born vaginally.

There was inconclusive evidence about the difference in neonatal and maternal outcomes for
either CS or vaginal birth for women in suspected or diagnosed preterm labour.
updated guideline.
15 Timing of cord clamping for preterm

babies
15.1 Introduction
There has been a recent change of practice in management of the third stage of labour in
term deliveries away from immediate clamping of the umbilical cord to deferred cord
clamping. In healthy term babies the evidence supports deferred clamping (RCOG Scientific
Impact Paper No 1 2015). Immediate clamping of the cord reduces blood flow from the
placenta to the baby and this could impact upon the transition from fetal to neonatal
circulation. The loss of blood volume also lessens the baby’s iron stores and thus increases
the risk of anaemia after birth. Both these effects could be particularly important in preterm
babies, but need to be set against the possible risks of delayed clamping, such as increased
risk of jaundice or delay in resuscitation at a critical time for both mother and baby. This
chapter examines the evidence relating to preterm babies and their mothers.

In preterm birth, does later or delayed cord clamping in active management of third stage
improve maternal and neonatal outcomes compared to earlier or immediate cord clamping?

314

Four studies were included in this review (Rabe 2012, March 2011, Elimian 2014, Ranjit
2014).
The included studies consisted of 1 systematic review (SR) with 15 component trials from a
variety of locations in developed countries (Rabe 2012), 2 randomised controlled trials
(RCTs) from the USA (March 2011, Elimian 2014) and 1 RCT from India (Ranjit 2014).
All included trials evaluated the effect of the timing of umbilical cord clamping of preterm
infants on neonatal outcomes. Only 1 included study reported results for maternal outcomes
(Ranjit 2014). The timing of cord clamping varied between the studies. “Earlier” cord
clamping was defined as immediate clamping of the cord ranging from 5 seconds to 30
seconds after the birth of the baby. “Later” cord clamping was defined as clamping the cord
from 30 seconds to 3 minutes after the birth of the baby. One study in the SR (Rabe 2012)
and 1 of the individual RCTs (March 2011) compared earlier cord clamping plus cord milking
with earlier cord clamping without cord milking.
Four RCTs in the SR included only women giving birth at less than 30 weeks’ gestation, 7
trials included women giving birth at less than 33 weeks’ gestation, 4 trials included women
giving birth at less than 35 weeks’ gestation, 2 trials only included low birthweight babies
under 2000 g and 1 trial included babies less than 36 weeks’ gestation. Although some of the
trials reported sub-group analyses by gestation at randomisation, they did not report
outcomes analysed by gestation of babies at birth and therefore this information could not be
further explored.
Not all the included studies in the SR covered the active management of third stage in the
cord clamping regarding the administration of a uterotonic. Five out of 15 studies in the SR
specified that a uterotonic was used intravenously (IV) at birth. The type of uterotonic, the
dose used, and the timing of administration varied between the studies. The use of a
uterotonic was not reported in the other 10 studies included in the SR, nor was it reported in
the 2 out of 3 separately reviewed clinical trials. For details of study interventions and
comparisons please see Table 95 and Table 96 and the evidence tables in Appendix H.
Seven of the studies in the SR provided information on the choice of mode of birth: in 1 trial
all babies were born by caesarean section; in 3 all babies were born vaginally; and in 3 RCTs
there was a mixed population in which an approximately equal number of babies were born
by vaginal or caesarean delivery. Mode of birth was not specified in the remaining 3 RCTs of
the SR nor in the 2 out of 3 separate trials.
Babies born by vaginal birth were held 10 to 15 inches below the level of the introitus in the
later cord clamping groups in most studies, except in 1 study where babies were held at the
level of the uterus. Babies born by caesarean section were also held below the level of
incision, apart from in 3 studies where the babies were held above the uterus either beside
the woman’s legs or on her thighs.
A summary of the main characteristics of each study in the SR and the individual trials is
given in Table 94 as these characteristics varied considerably.

315
Table 88: Summary of baseline characteristics of included studies

Definition of Position of the
delayed cord baby at time of
Definition of clamping (DCC) or clamping in
Study author early cord details of cord Intervention Uterotonic Gestational age Additional
and date clamping (ECC) milking group (DCC) (type/route/time) Country or birth weight Mode of birth comments
Aladagandy Immediately after 30–90 seconds Infant held as low Syntocinon 5 IU IV at UK (Glasgow) 24+0 – 32+6 Vaginal/caesarean Intention to
2006 (included birth. after birth as the cord’s birth of baby’s head Caesarean section: treat analysis.
in Rabe 2012) length permitted ECC: n=12/23 n=3/23
DCC: n=14/23 allocated to
DCC had early
clamping (1
due to short
cord, 2 asked
for by
neonatologist).
Baezinger 2007 Immediately after 60–90 seconds Infant held as low Syntocinon IV Switzerland 24+0–32+6 weeks Vaginal/caesarean N =3 infants in
(included birth (<20 after birth as possible for immediately after birth Caesarean section: ECC group
in Rabe 2012) seconds) vaginal births, and in DCC group ECC: n=16/24 died within 72
15 cm below the DCC: n=11/15 hours after
placenta at birth (n =1
caesarean from sepsis, n
section. =2 from
hyaline
membrane
disease)
Elimian 2014 Within 5 seconds 30–35 seconds Not reported Not reported USA 24–34 weeks Not reported
of birth after birth
Hofmeyr 1988 Immediately after Intervention 1: 60 Not reported Ergometrine at birth in South Africa < 35 weeks Not reported
(included in birth seconds after birth second intervention
Rabe 2012) Intervention 2: 60 group (route not
seconds after birth reported)
Hofmeyr 1993 Shortly after birth, 60–120 seconds At level of uterus Not reported < 2000 g Not reported n=8 in DCC
(included in according to after birth at vaginal births or (gestational age had cord
Rabe 2012) usual practice above not reported) clamped early
level of uterus on because of
the woman’s cord around
the neck or
thighs at
need for
caesarean section
resuscitation

Hosono 2008 Immediately after Umbilical cord Below or at the Not reported Japan 24–28 weeks Vaginal/caesarean n=2 infants in
(included in birth milked vigorously level of placenta Caesarean section: the milking
Rabe 2012) towards umbilicus and about 20cmof ECC: n=14/20 (70%) group died; 1
2–3 times the placenta DCC: n=14/20 (70%) at the 26 day
(estimated speed due to
20 cm/second) intestinal
perforation and
1 at 42 days
owing to
sepsis
n=3 deaths in
control group;
the reason not
reported
Kinmond 1993 Mean time to cord 30 seconds after 20cm below the Not reported UK (Cardiff) 27–33 weeks All vaginal birth: n=36 In ECC group
(included in clamping 10 birth introitus clamping
Rabe 2012) seconds performed
within 20
seconds for
n=18/19 and at
25 seconds for
n=1.
Kugelman 2007 < 10 seconds 30–45 seconds 20–30 cm below No uterotonic used Israel <35 weeks Vaginal/caesarean Intention to
(included after birth after birth level of introitus at Caesarean section: treat analysis.
in Rabe 2012) vaginal ECC: n=23/35 High risk
births or below DCC: n=20/30 pregnancy:
level of the ECC: n=13/35
incision at DCC: n=9/30
caesarean section Multiple
pregnancy:
ECC: n=8/35
DCC: n=9/30
March 2011 Immediately after Approximately Not reported USA 24+0–28+6 weeks Not reported
birth 20cm of umbilical
cord was milked
toward the baby
immediately
following birth.

McDonnell Immediately after 30 seconds after Between woman’s Syntocinon IV at birth Australia 26–33 weeks. Not reported Intention to
1997 (included birth birth legs treat analysis
in Rabe 2012) (on 3
occasions the
protocol was
broken to allow
resuscitation)
4 twin
pregnancies
were included,
each twin
randomised
separately.
Mercer 2003 5–10 seconds 30–45 seconds 10–15 inches No uterotonic used USA 24+0–31+6 weeks Vaginal/caesarean Intention to
(included in after birth after birth below level of the before cord clamping Caesarean section: treat analysis
Rabe 2012) placenta at ECC: n=6/16 (n=2 ECC had
vaginal births or DCC: n=9/16 DCC)
below the level of
the incision at
caesarean section
Mercer 2006 5–10 seconds 30–45 seconds 10–15 inches No uterotonic used USA 24+0–31+6 weeks Vaginal/caesarean Intent to treat
(included in after birth after birth below level of before cord clamping analysis
Rabe 2012) placenta at No infant
vaginal births or needed
below the incision resuscitation
at caesarean
section.
Nelle 1998 Immediately after 30 seconds after 30cm below Not reported Germany <1500 g All caesarean section
(included in birth birth placenta ( gestational age (n=19)
Rabe 2012) not reported)
Oh 2002 <5 seconds after 30–45 seconds Not reported Not reported USA 24 –28 weeks Pilot study
(included in birth after birth Powered to
Rabe 2012) detect change
in human
chorionic
gonadotrophin
levels

Rabe 2000 20 seconds after 45 seconds after Below the level of Oxytocin IV at birth of Germany <33 weeks Vaginal/caesarean n=1 baby in
(included in birth birth placenta if the first shoulder Mean gestational Caesarean section: ECC died on
Rabe 2012) possible age: ECC: n=19/20 3rd day of life
DCC: n=15/20 because of
ECC:
severe
n=29.48±1.96
necrotising
DCC: enterocolitis
n=30.01±1.57
Ranjit 2014 Immediately after After 2 minutes Mothers abdomen Not specified India 30+0–36+6 weeks Caesarean section: No intention to
birth following birth in vaginal birth ECC: n=25/50 DCC: treat analysis
and on mother’s n=20/45
thigh in caesarean
birth
Strauss 2008 “Within 2–5 60 seconds after 10–12 inches Not specified USA 30–36 weeks Not reported Intention to
(included in seconds of birth birth below introitus at treat analysis
Rabe 2012) (not exceeding 15 vaginal birth or
seconds)” beside the
woman’s thigh at
caesarean section
Ultee 2008 Within 30 180 seconds after On woman’s Not reported Netherland 34+0–36+6 weeks All vaginal birth All Caucasian
(included in seconds of birth birth abdomen in both mothers
Rabe 2012) (mean groups n=2 with
13.4 seconds - protocol
(SD 5.6 seconds) violation were
excluded, 1 in
each arm
DCC delayed cord clamping, ECC early cord clamping, IV Intravenous

The search strategies for this chapter can be found in Appendix E, the excluded studies in Appendix G, the evidence tables in Appendix H and the
forest plots in Appendix I.
The findings for the effect of “later” versus “earlier” cord clamping on neonatal outcomes are reported in 2 GRADE profiles. The first includes also
the subgroup analysis for use of uterotonic and the second compares different strategies for increasing placental transfusion (later cord clamping
vs cord milking.
Table 89: GRADE profile for comparison of later cord clamping versus earlier cord clamping - neonatal and maternal outcomes: overall
and with sub-group analysis
Quality assessment Number of babies Effect
Later Early Relative

Risk of Other cord cord (95%
Number of studies Design bias Inconsistency Indirectness Imprecision considerations clamping clamping CI) Absolute Quality
Infant death (up to discharge/variable)
1 meta-analysis of 13 Randomise Very No serious Serious2 Serious3 None 10/363 22/399 RR 0.51 27 fewer per Very low
studies d trials serious inconsistency (2.8%) (5.5%) (0.26 to 1000
1
(Rabe 2012) and 1 1.01) (from 41
study (Ranjit, 2014) fewer to 1
more)
Infant death (up to discharge/variable) – Uterotonic used at birth (subgroup analysis)
1 meta-analysis of 4 Randomise Very No serious No serious Very serious5 None 5/77 5/78 RR 0.92 5 fewer per Very low
studies d trials serious inconsistency indirectness (6.5%) (6.4%) (0.29 to 1000 (from
(Rabe 2012) 4
2.95) 46 fewer to
125 more)
Severe intraventricular haemorrhage
1 meta-analysis of 6 Randomise Very No serious Serious2 Very serious5 None 5/154 7/151 RR 0.68 15 fewer per Very low
(Rabe 2012) 6
1.96) (from 36
fewer to 45
more)
Severe intraventricular haemorrhage – Uterotonic used at birth (subgroup analysis)

1 meta-analysis of 2 Randomise Very No serious No serious Very serious5 None 2/42 0/33 (0%) RR 2.92 NC Very low
studies d trials serious inconsistency indirectness (4.8%) (0.15 to
(Rabe 2012)
7
56.51)
Intraventricular haemorrhage (all grades)
1 meta-analysis of 10 Randomise Very No serious Serious2 Serious3 None 35/304 57/329 RR 0.59 47 fewer per Very low
(Rabe 2012) and 1
8
0.84) (from 18
study (Ranjit 2014) fewer to 104
fewer)
Intraventricular haemorrhage (all grades) – Uterotonic used at birth (subgroupe analysis)
1 meta-analysis of 3 Randomise Very No serious No serious Serious3 None 12/54 14/46 RR 0.61 119 fewer per Very low
studies d trials serious inconsistency indirectness (22.2%) (30.4%) (0.34 to 1000 (from
(Rabe 2012)
9
1.08) 201 fewer to
24 more)
Ventilated for respiratory distress syndrome
1 meta-analysis of 5 Randomised Very No serious No serious Serious3 None 54/163 57/196 RR 0.93 123 fewer per Very low
studies trials serio inconsistency indirectness (27.6%) (29.1%) (0.695 1000 (from
(Rabe 2012) and 1 us 10
to 1.25) 230 fewer to
230more)
study (Ranjit 2014)
Hypoxic ischemic encephalopathy
1 study (Ranjit 2014) Randomised serious 2
No serious No serious Very serious5 None 0/44 (0%) 1/50 (2%) RR 0.57 17 fewer per Very low
trials inconsistency indirectness (0.05 to 1000 (from 38
6.05) fewer to 201
more)
Ventilated for respiratory distress syndrome – Uterotonic used at birth (subgroup analysis)
1 meta-analysis of 3 Randomised Very No serious No serious Very serious5 None 18/83 28/107 RR 0.82 47 fewer per Very low
studies trials serio inconsistency indirectness (21.7%) (26.2%) (0.5 to 1000 (from
(Rabe 2012) us 11
1.33) 131 fewer to
86 more)
Hyperbilirubinemia (treated)
1 meta-analysis of 3 Randomised Very No serious No serious Serious3 None 51/82 51/98 RR 109 more per Very low
studies trials serio inconsistency indirectness (62.2%) (52%) 1.21(0.9 1000 (from 31
(Rabe 2012) us 12
4 to fewer to 286
1.55) more)
Hyperbilirubinemia (treated) – Uterotonic used at birth (subgroup analysis)

1 study Randomised Very No serious No serious Very serious5 None 12/19 12/20 RR 1.05 30 more per Very low
(Rabe 2012) trials serio inconsistency indirectness (63.2%) (60%) (0.64 to 1000 (from
us 1.73) 216 fewer to
13
438 more)
Transfused for anaemia
1 meta-analysis of 8 Randomised Very No serious Serious2 Serious3 None 61/207 91/223 RR 0.66 139 fewer per Very low
studies trials serio inconsistency (29.5%) (40.8%) (0.52 to 1000
(Rabe 2012) and 1 us 0.82) (from 75
14
study (March 2011) fewer to 196
fewer)
Transfused for anaemia – Uterotonic used at birth (subgroup analysis)
1 meta-analysis of 2 Randomised Very No serious No serious Serious3 None 13/42 22/43 RR 0.61 200 fewer per Very low
studies trials serio inconsistency indirectness (31%) (51.2%) (0.37 to 1000 (from
(Rabe 2012) us 1) 322 fewer to
15
0 more)
Anaemia of prematurity
1 study Randomise Very No serious No serious Serious3 None 36/99 48/101 RR 0.77 109 fewer per Very low
(Elimian 2014) d trials serious inconsistency indirectness (36.4%) (47.5%) (0.55 to 1000
16
1.07) (from 214
fewer to 33
more)
Apgar score at 5 minutes <8
studies d trials serious inconsistency indirectness (18.1%) (20.2%) (0.45 to 1000
(Rabe 2012) 17
1.62) (from 111
fewer to 125
more)
Apgar score at 5 minutes <8 – Uterotonic used at birth (subgroup analysis)
studies d trials serious inconsistency indirectness (15%) (27%) (0.29 to 1000 (from
(Rabe 2012) 18
1.96) 178 fewer to
240 more)
Haematocrit at 4 hours of life (%)

1 meta-analysis of 5 Randomise Very No serious No serious No serious None 82 91 NC MD 5.40 Low
studies d trials serious inconsistency indirectness imprecision higher
(Rabe 2012) 19
(3.62 higher
to 7.17
higher)
Haematocrit at 4 hours of life (%) – Uterotonic used at birth (subgroup analysis)
studies d trials serious inconsistency indirectness imprecision higher (0.35
(Rabe 2012) 10
higher to 6.77
higher)
Haematocrit at 24 hours of life (%)
studies d trials serious inconsistency indirectness imprecision higher
(Rabe 2012) and 1 20
(3.91 higher
study (Ranjit 2014) to 6.76
higher)
Haematocrit at 24 hours of life (%) – Uterotonic used at birth (subgroup analysis)
1 study Randomise Very No serious No serious No serious None 15 23 NC MD 6.19 Low
(Rabe 2012) d trials serious inconsistency indirectness imprecision higher (1.2
21
higher to
11.18 higher)
Maternal outcomes
1 study Randomise Serious No serious No serious Very serious5 None 1/44 (2%) 1/50 (2%) RR 1.14 3 more per Very low
(Ranjit 2012) d trials 22 inconsistency indirectness (0.0 to 1000 (from
17.6) 19 fewer to
333 more)
CI confidence interval, MD mean difference, MID minimally important difference, NC not calculable, RR relative risk
1. High risk of bias for blinding in n=4 studies, high risk of bias for incomplete outcome data in n=3 studies and high risk of bias for allocation concealment and selective reporting in
n=2 studies. Unclear risk of bias for randomisation and allocation concealment in n=10 studies, unclear risk of bias for blinding in n=8 studies, unclear risk of bias for incomplete
outcome data in n=3 studies and unclear risk of bias for selective reporting data n=9 studies
2. Immediate cord clamping was compared with milking the umbilical cord in n=1 study
4. High risk of bias for incomplete outcome data reporting and selective reporting in n=1 study, for blinding in n=1 study
6. High risk of bias for blinding in n=1 study. Unclear risk of bias for randomisation in n=4 studies. Immediate cord clamping was compared with milking the umbilical cord in n=1
study
7. Unclear risk of bias for randomisation, allocation concealment and blinding in n=2 studies
8. High risk of bias for blinding in n=4 studies. High risk of bias for selective reporting in n=2 studies. Unclear risk of bias for randomisation in n=7 studies. Immediate cord clamping
was compared with milking the umbilical cord in n=1 study
9. High risk of bias in blinding in n=1 study and unclear risk of bias for randomisation, allocation concealment in n=2 studies
10. High risk of bias for blinding in n=2 studies. Unclear risk of bias for allocation concealment and randomisation in all 5 studies, uneven number of participants present in the earlier
and later cord clamping groups in n=2 studies. 1 study was part of a multicentre trial, and the outcome reported was collected just for this subject.
11. High risk of bias for incomplete outcome data reporting and selective reporting in n=1 study, for blinding in n=1 study
12. High risk of bias for allocation concealment and blinding in n=1 study. High risk of bias for incomplete data in n=2 studies. Unclear risk of bias for randomisation in n=2 studies.
13. Unclear risk of bias for randomisation, allocation concealment and blinding
14. High risk of bias for blinding in n=3 studies, high risk of bias for incomplete outcome data and selective reporting in n=1 study. Unclear risk of bias for randomisation and allocation
concealment in n=6 studies. In 1 study n=3/36 early deaths reported in the immediate cord clamping group and this group was then excluded from the analysis as they were no
longer eligible to experience outcomes.
15. Unclear risk of bias for randomisation, allocation concealment and blinding in n=1 study and high risk of bias for blinding in n=1 study
16. A published conference abstract with very limited data reported
17. Unclear risk of bias for blinding in n=2 studies, unclear risk of bias for randomisation, allocation concealment and selective reporting in n=2 studies. In 1 study 57% (n=8) of babies
allocated to delayed cord clamping had the cord clamped early, either due to cord round the neck, or need for resuscitation.
18. Unclear risk of bias for randomisation, allocation concealment and blinding in n=2 studies
19. High risk of bias for incomplete outcome data and selective reporting in n=1 study, for blinding n=2 studies
20. High risk of bias for incomplete outcome data reporting and selective reporting in n=2 studies
21. High risk of bias for incomplete outcome data reporting and selective reporting
22. No intention to treat analysis performed
Table 90: GRADE finding for comparison of later cord clamping versus earlier cord clamping on neonatal outcomes (cord milking)
Later cord
clamping/more Earlier
Number of Risk of Other placental cord Relative
studies Design bias Inconsistency Indirectness Imprecision considerations transfusion clamping (95% CI) Absolute Quality
Infant death (up to discharge/variable)
1 meta- Randomised Very No serious No serious Very None 8/299 (2.7%) 14/329 RR 0.62 16 fewer Very
analysis of 12 trials serious1 inconsistency indirectness serious2 (4.3%) (0.28 to per 1000 low
studies 1.36) (from 31
(Rabe 2012) fewer to 15
more)
Infant death (up to discharge/variable) – Cord milkinga (subgroup analysis) g
1 study Randomised Serious3 No serious serious4 Very None 2/20 (10%)b 3/20 (15%)c RR 0.67 49 fewer Very
(Rabe 2012) trial inconsistency serious2 (0.12 to per 1000 low
Later cord
clamping/more Earlier
Number of Risk of Other placental cord Relative
studies Design bias Inconsistency Indirectness Imprecision considerations transfusion clamping (95% CI) Absolute Quality
3.57) (from 132
fewer to 386
more)
Severe intraventricular haemorrhage
1 meta- Randomised Very No serious No serious Very None 3/134 (2.2%) 3/131 RR 0.85 18 fewer Very
analysis of 5 trials serious5 inconsistency indirectness serious2 (2.2%) (0.20 to per 1000 low
studies 3.66) (from 18
(Rabe 2012) fewer to 61
more)
Severe intraventricular haemorrhage – Cord milkinga (subgroup analysis)
1 study Randomised Serious3 No serious Serious4 Very None 2/20 (10%) 4/20 (20%) RR 0.50 100 fewer Very
(Rabe 2012) trial inconsistency serious2 (0.10 to per 1000 low
2.43) (from 180
fewer to 286
more)
Transfused for anaemia
1 meta- Randomised Serious6 No serious No serious serious7 None 37/166 (22.2 %) 61/186 RR 0.63 199 fewer Very
analysis of 6 trials inconsistency indirectness (32.7%) (0.46 to per 1000 low
studies (Rabe 0.87) (from 70
2012) fewer to 291
fewer)
Transfused for anaemia – Cord milkinga (subgroup analysis)
1 meta- Randomised Serious6 No serious Serious4 Serious7 None 24/47 (51%) 30/37 RR 0.70 162 fewer Very
analysis of 2 trials inconsistency (81%) (0.53 to per 1000 low
studies (March 0.94) (from 32
2011 and Rabe fewer to 253
2012 ) fewer)
b. Umbilical cord milked vigorously towards umbilicus 2-3 times
c. One baby died at 26 days after birth due to intestinal perforation and 1 baby died at 42 days owing to sepsis
d. The reason for death is not reported
1. High risk of bias for blinding in n=5 studies, high risk of bias for incomplete outcome data in n=2 studies and high risk of bias for allocation concealment and selective reporting
n=2 studies. Unclear risk of bias for randomisation and selective reporting in n=9 studies, unclear risk of bias for blinding in n=6 studies, unclear risk of bias for incomplete outcome
data in n=3 studies

3. High risk of bias in blinding
4. Immediate cord clamping was compared with milking the umbilical cord
5. Unclear risk of bias for randomisation and selection bias in n=4 studies.
6. High risk of bias for blinding in n=3 studies, high risk of bias for incomplete outcome data and selective reporting in n=1 study. Unclear risk of bias for randomisation and allocation
concealment in n=6 studies. In 1 study n=3/36 early deaths reported in the immediate cord clamping group and this group was then excluded from the analysis as they were no
longer eligible to experience outcomes.
8. March 2011 is a published conference abstract with very limited data reported

Overall analysis
Very low quality evidence from 1 SR of 15 RCTs and 1 trial (n=976) found that there was no
significant difference in the risk of infant death, mechanical ventilation for respiratory distress
syndrome, hyperbilirubinemia, severe intraventricular haemorrhage and Apgar score (less
than 8) at 5 minutes in preterm neonates allocated to receive later cord clamping compared
with those allocated to receive earlier cord clamping. However, the rate of intraventricular
haemorrhage (all grades) and the numbers transfused for anaemia were significantly lower in
neonates allocated to receive later cord clamping compared with those allocated to receive
earlier cord clamping. Low quality evidence found that the levels of haematocrit at 4 hours
and 24 first hours were significantly higher in neonates allocated to receive later cord
clamping compared with those allocated to receive earlier cord clamping.
Regarding maternal outcomes only 1 study reported postpartum haemorrhage and this found
no significant difference in the incidence of postpartum haemorrhage in women allocated to
the later cord clamping group compared with those allocated to the earlier cord clamping
group. The evidence was of very low quality. No other maternal outcomes were reported.
Subgroup analysis by use of uterotonic
Findings from a meta-analysis of the 15 RCTs (n=976) showed that the administration of a
uterotonic agent intravenously at birth did not change the direction of estimates of effects in
most of the reported outcomes reported in the main meta-analysis. However, 2 significant
results from the main analysis on the outcomes (reduction in intraventricular haemorrhage
[all grades] and reduction in the number of babies needing transfusion for anaemia) lost
significance at the subgroup analysis although the direction of the main effect is still
beneficial for the later clamping group. The evidence was of low and very low quality.
Subgroup analysis by strategy for increasing placental transfusion (cord milking)
Very low quality evidence from a sub-group analysis by strategy for increasing placental
transfusion from 2 RCTs (n=84) showed no difference in the 2 reported outcomes of infant
death and severe intraventricular haemorrhage between earlier cord clamping compared with
later cord clamping or earlier cord clamping compared with earlier cord clamping following
cord milking. Further meta-analyses found fewer babies who had either later cord clamping
or cord milking received a blood transfusion for anaemia compared with babies who had

The decision of when to clamp the cord has no or negligible resource implications, unless
there are secondary costs associated with adverse outcomes which are affected by the
timing of cord clamping. If a decision on timing improves outcomes then that decision is likely
to be cost neutral or cost saving and therefore if a certain timing is clinically optimal then it
will almost certainly be cost effective too.
No search for health economic evidence was undertaken for this question as the decision on
the timing of cord clamping was thought to have negligible resources implications.

327

15.7.1 Relative value places of the outcomes considered
The Guideline Development Committee prioritised both neonatal and maternal outcomes to
base the draft of their recommendations. The maternal outcomes agreed at the protocol
stage were mortality, clinical indicators (such as primary postpartum haemorrhage [PPH])
and the need for further intervention (such as blood transfusion, emergency anaesthesia),
length of hospital stay and women’s personal experience in terms of attachment and breast-
feeding.
In terms of the neonatal outcomes, the following were prioritised as the most important:
• respiratory disease
• brain injury
• treatment for hyperbilirubinaemia with exchange transfusion
• blood counts at 6 and 12 hours (haemoglobin or haematocrit).
The committee considered that severe intraventricular haemorrhage (IVH) (grade III/IV) may
be linked with significant long term neuro-developmental impairment, including cerebral palsy
(CP) in preterm infants, whereas many babies who have a mild IVH go on to develop
normally or might have only minimal disabilities associated with learning, such as reduced IQ
or poor concentration or mild behavioural difficulty. The committee noted that few studies
included in this review had looked at severe IVH (grade III/IV), although most of the studies
reported the rate of all grades of IVH. As a result, the committee focussed on both severe
IVH and all grades of IVH as proxies for poor neonatal neurodevelopment outcomes.
‘Transfused for anaemia’ was considered important in relation to reducing the risks
associated with blood products and donor exposure, which was a beneficial effect of later
cord clamping. It was felt that hyperbilirubinaemia may not be particularly relevant because
hyperbilirubinaemia is common, usually mild, and only rarely associated with serious
outcomes such as encephalopathy. Thus greater emphasis was placed on findings for
‘transfused for anaemia’ and IVH than hyperbilirubinaemia.
The committee felt that infant death was a less informative outcome here because in the
populations studied it was so rare that trials were unlikely to have adequate statistical power
to detect any difference. In the presence of very limited evidence for maternal outcomes,
committee decision-making was mainly driven by neonatal outcomes.

Very low quality evidence demonstrated that there was a significantly lower rate of all grades
of IVH and ‘transfused for anaemia’ and significantly higher haematocrit at 4 hours and 24
hours after birth in the delayed cord clamping group compared with the early group.
Furthermore, these benefits were confirmed in the subgroup analysis based on the use of
any uterotonic at birth. This was an important consideration for the Guideline Development
Committee when drafting the recommendations, because in their experience uterotonic use
is very common in clinical practice and is part of the active management of labour
recommended in the NICE guideline on intrapartum care.
However, no significant difference was found between the early and late cord clamping
groups for a number of neonatal outcomes such as hyperbilirubinaemia (treated), infant
death, anaemia of prematurity and respiratory distress syndrome.

328
The committee felt that the fact that there was no difference in Apgar scores at 5 minutes in
relation to the timing of cord clamping was reassuring; indeed, there was no evidence of any
other harm being associated with later cord clamping.
The committee noted that the majority of the studies defined delayed cord clamping as being
between 30 and 60 seconds after birth. In some studies the cord was clamped after a longer
interval (up to 180 seconds after birth). The committee felt that in clinical practice, delayed
cord clamping is generally conducted within the 30–60 second time limit and although they
felt the same benefits might be seen at other timings, they decided that the
recommendations should reflect the 30–60 second interval.
The committee noted that in nearly all studies the baby was kept below the level of the
placenta in order to facilitate blood flow. This practice was presumed to be beneficial, but the
committee noted that in a recent study in term babies it was found that the transfer of
placental blood when cord clamping was delayed was not reduced when the baby was
placed on the mother’s abdomen, above her uterus. This was also evident from 3 included
studies in this review where babies were held above the uterus.
The committee was aware that 2 out of 17 included studies examined the effect of ‘cord
milking’ compared with early cord clamping. Cord milking is a technique used to increase the
passage of blood along the cord to the baby so that the baby can be removed more quickly
for resuscitation or respiratory support. Results from these studies showed that cord milking
seems to provide a benefit similar to that of delayed cord clamping in terms of the rate of
‘transfused for anaemia’ and all grades of IVH.
The committee noted that all studies were carried out in resource rich countries where the
level of haemoglobin is generally higher than babies born in resource poor countries.
Hyperbilirubinaemia was more prevalent among babies where cord clamping was delayed
but this difference was not significant. The committee did not feel that there was evidence to
show that a slightly higher level of hyperbilirubinaemia in early life would lead to worse long-
term outcomes.
15.7.3 Consideration of health benefits and resource use

The decision of when to clamp the cord has no or negligible resource implications, unless
there are secondary costs associated with adverse outcomes which are affected by the
timing of cord clamping. If a decision on timing improves outcomes then that decision is likely
to be cost neutral or cost saving and therefore if a certain timing is clinically optimal then it
will almost certainly be cost effective too.

The quality of the evidence included in this section was low and very low. Two of the main
reasons for the studies being downgraded were risk of bias due to lack of blinding and
imprecision.
In particular, the Guideline Development Committee recognised some variations in the
methods and populations in the included trials in terms of gestational age, different
definitions of “early cord clamping” and “late cord clamping”, and the nature of the active
management of third stage, including the administration of a uterotonic. Timings of cord
clamping also varied between the studies and this variation was taken into consideration
during the committee’s discussion and interpretation of results.

329

The Guideline Development Committee noted that gestational age might make a difference
to the care strategy, and if a baby is very premature and needing immediate resuscitation
then early cord clamping might take precedence, yet these might be the babies with the most
to gain from a larger placental transfusion. The committee felt this decision would need to be
made on an individual case-by-case basis following the clinician’s judgement on the balance
between benefits and harms. Similarly, there might be other maternal reasons for separation
between the woman and baby straightaway after birth, for example in the case of severe
haemorrhage.
In addition, the committee was aware that a recent trial in term babies has cast doubt on the
assumption that the position of the baby in relation to the uterus is important, but noted that
this has not been tested in preterm babies, so no further conclusions can be made.

Given that there was limited evidence available in this area, the Guideline Development
Committee did not feel confident about making strong recommendations for practice
regarding the timing of cord clamping. They noted there is some evidence in favour of
delayed cord clamping and no evidence of harm is associated with it. The committee
identified that an advantage of delayed cord clamping, not addressed by any of the studies in
the evidence table, is that placental transfusion allows newborn infants to continue to receive
oxygen via the placenta as long as the cord is pulsing. In babies born with fetal distress, it is
believed that the passage of blood in the first minute can contribute to a better resuscitation.
Leaving the cord intact does not necessarily preclude other actions being taken for the
benefit of the baby simultaneously, for example giving oxygen. Given this, the committee
agreed that in most cases, clamping should not take place before 30 seconds after the birth
of the baby, and that in situations where speed is of the essence, cord milking should be
considered a reasonable alternative to delayed clamping.
updated guideline.

6. Is there any advantage to preterm babies from delayed
Research question versus early cord clamping, or cord milking?
Why this is needed
Importance to ‘patients’ or the Delay in cord clamping, or active cord milking, to ensure an
population adequate placental transfusion to the baby at the time of birth,
has been shown to be beneficial (see NICE guideline on
intrapartum care). Current evidence relates to term babies, but
it is possible that benefit would be greater for the preterm
baby, assisting transfer from the fetal circulation, and
improving haemoglobin and iron stores.
Relevance to NICE guidance Since the current guideline now recommends delayed cord
clamping or cord milking (providing mother and baby are
stable), the importance is low. The strength of the evidence
base for the recommendation needs to be enhanced.
Relevance to the NHS If a clear effect on improved neurodevelopmental outcome
can be demonstrated, the intervention of delaying cord
clamping, or cord milking, is likely to be highly cost-effective.

330
6. Is there any advantage to preterm babies from delayed

Research question versus early cord clamping, or cord milking?
prematurely
Current evidence base The current evidence relates to term babies and does not look
specifically at the preterm population.
Equality This group is defined only by gestational age at delivery.
Feasibility There have been previous studies in the area so further ones
would be feasible. Comparison of immediate with delayed
cord clamping should allow ascertainment of any improvement
in short-term stability or mortality, medium-term requirement
for blood transfusion, and long-term neurodevelopment of the
preterm infant.
The potential difficulty in setting up such a trial is that
researchers may feel the existing evidence on term babies is
sufficient to extrapolate to preterm babies.
Other comments Ascertainment of outcome could and should be masked from
knowledge of treatment group, but masking of the allocated
treatment from the healthcare professionals looking after the
mother and baby at delivery is not feasible.
16 Health economics
16.1 What is the clinical effectiveness of prophylactic
progesterone (vaginal or oral) in preventing preterm
labour in pregnant women considered to be at risk
of preterm labour and birth?
16.1.1 Review of the literature
A search was undertaken for health economic evidence on prophylactic progesterone to
prevent preterm labour in women considered to be at risk of preterm labour and birth. A total
of 149 studies were identified by the search. After reviewing titles and abstracts, 5 papers
were obtained. Three of these studies were excluded because they were not economic
evaluations, were reporting a conference abstract or lacked the relevant comparator. Two
studies were included in the literature review and are reported here (see Appendix H for
evidence tables).
A US study (Cahill 2010) used a decision analytic framework to evaluate the cost–utility of 4
strategies for the prevention of preterm labour (PTL) in women with threatened PTL.
• universal sonographic screening for cervical length and treatment with vaginal
progesterone
• cervical length screening for women with increased risk of preterm birth and treatment
with vaginal progesterone

331
• risk-based treatment with 17 α-hydroxprogesterone caproate (17-OHP-C) without

screening
• no screening or treatment.
The authors reported that the analysis was based on published evidence but the source of
the evidence is unclear. The authors concluded that universal sonographic screening for
cervical length and treatment with vaginal progesterone was the dominant strategy. They
further reported that Monte Carlo simulation showed this to be the dominant strategy 96.9%
of the time.
However, there are a number of quality concerns in the reporting of this analysis. Resource
use is not specifically reported, the source of model inputs is unclear, the perspective of the
analysis is not stated and the study has no reference list. Although the paper reports the
results of sensitivity analysis and a “worst case” scenario, the methods are barely described.
Furthermore, there is no attempt to quantify the uncertainty in any of the reported results.
A US study (Pizzi 2014) used a decision analytic approach to evaluate the cost effectiveness
of vaginal progesterone gel to a placebo for the prevention of PTL in women with threatened
PTL with a cervical length of 10–20 mm as measured by transvaginal ultrasound. The
population was based on women in a multicentre randomised controlled trial (RCT) who were
pregnant with singleton pregnancies. The decision analysis included efficacy and safety data
from the trial and used the cost per preterm birth averted as the measure of cost
effectiveness.
The author reported a cost year of 2011 and included costs incurred until the infant is
discharged from hospital. Based on a third party payer perspective, the author reported that
vaginal progesterone was dominant with vaginal progesterone producing cost savings of
USD 12,354 (US dollars) relative to placebo for an incremental benefit of 0.0426 births
averted. A probabilistic sensitivity analysis suggested that vaginal progesterone gel
dominated placebo in 79.2% of simulations.
The authors report a number of limitations with their study. In particular they note that
PREGNANT trial (Pizzi 2014) had a multi-country study design and that it might not be
appropriate to apply US costs to the services reported in the trial. The model had a relatively
short-time horizon but as the authors note, including long term morbidity would have
strengthened their reported conclusion.
1.2. What is the diagnostic accuracy of the following (alone or

in combination) in women with intact membranes to
identify preterm labour leading to preterm birth:
• clinical assessment (such as symptoms expressed by women, strength and
frequency of contractions, findings on vaginal examination)
• biochemical testing for markers for preterm labour namely cervicovaginal fetal
fibronectin and IGF-BP1 insulin-like growth factor binding protein-1
• cervical ultrasound features (such as cervical length and funnelling)?
16.1.2 Introduction
Preterm labour is a common occurrence in pregnancy and is associated with adverse
outcomes. However, there are interventions which can reduce the risks of adverse outcomes
(maternal corticosteroids or tocolysis for example). Diagnosis of preterm labour has the
potential to identify the women who would benefit from treatment whille providing
reassurance to the majority who are not in preterm labour.

332
The cost effectiveness of diagnosis cannot usually be considered in the absence of treatment
or management as it is the decisions that follow from a particular diagnosis that affect patient
outcomes. So, for example, diagnosis would not usually be cost effective if there was not an
effective treatment for the condition being diagnosed.
Therefore, although this analysis focuses on the diagnostic decision, it uses the output from
Section 16.4 on the effectiveness of tocolysis to quantify the benefit of diagnosing preterm
labour.
16.1.3 Methods
A cost–utility decision analytic model was developed in Microsoft Excel® to assess different
diagnostic strategies to identify preterm labour in women with suspected preterm labour and
intact membranes between gestational ages of 24+0 and 34+0 weeks. A range of alternative
diagnostic strategies to diagnose preterm labour within 48 hours were considered in the
clinical review, an outcome considered important because it is related to the decision-making
regarding the timing of steroid and magnesium sulfate administration. In addition, the
strategies of no diagnostic test/no treatment and treating all women without a diagnostic test
were included as alternatives.
The evidence on the diagnostic accuracy of the various diagnostic strategies was of
generally poor quality and often with serious limitations (see Chapter 9). Where there was
more than 1 study reporting the diagnostic accuracy of the test it was not thought appropriate
to synthesise these data and therefore data on the same diagnostic test were often
conflicting. For this reason, the evaluation took the form of a ‘what-if’ analysis. This involved
calculating the cost–utility for all combinations of sensitivity and specificity between 0% and
100% (10,201 combinations in total) and determining what the cost-effective strategy would
be for a diagnostic strategy with a certain cost at each of these different combinations; that
is, ‘if’ a diagnostic strategy had this particular diagnostic accuracy, then ‘what’ would the
cost-effective strategy be?
1. Treat based on the results of the diagnostic test.
2. Do not perform a diagnostic test but treat all women.
3. Do not perform a diagnostic test and do not treat.
Where a strategy involves treatment (1 and 2) then It was assumed that women were given
calcium channel blockers as a tocolytic, as that conforms with the committee
recommendation on tocolysis and what was assessed as the most cost-effective tocolytic in
Section 16.4. However, the model assumed that only true positives derive the benefit of
treatment. It was assumed that women not treated, either as a result of a negative test result
or the strategy itself, would be sent home.
The initial costs of diagnosis and treatment occur in the immediate term, but the model takes
a lifetime horizon of the baby in terms of both future costs and benefits as the outcomes
assessed have lifelong consequences. This reflects the outcomes included in the treatment
models – see Section 16.4.2.1.
The model did not consider the re-presentation of women with suspected preterm labour
following a previous negative test result. A schematic of the model is shown in Figure 20.
The model followed the approach of the model assessing the cost effectiveness of tocolytics
in women with suspected or diagnosed preterm labour by performing the analysis by
gestational age.

333
Figure 20: Schematic of model to assess the cost-effectiveness of alternative

diagnostic strategies for preterm labour

334
16.1.4 Diagnostic strategies, test accuracy and prevalence of actual preterm

labour
The various diagnostic strategies that were included in the clinical review are shown in Table
97. The diagnostic strategies evaluated based on clinical assessment (Bishop Score),
transvaginal ultrasound (cervical length), fetal fibronectin and insulin-like growth factor
binding protein-1 (pIGFBP-1) could be used alone or in combination.
Table 91: : List of diagnostic test strategies and their reported test sensitivity and
specificity
Study Year Diagnostic strategy Sensitivity Specificity
Treat all 100.0% 0.0%
No diagnosis 0.0% 100.0%
Schmitz 2008 Bishop score ≥4 94.0% 43.0%
Schmitz 2008 Bishop score ≥8 35.0% 97.0%
Schreyer 1989 Bishop score 4 to 6 69.2% 73.7%
Schmitz 2008 Bishop Score 4–7 and cervical length ≤20 mm 60.0% 64.0%
Gomez 2005 Cervical length <30 mm 88.2% 53.0%
Schmitz 2008 Cervical length <30 mm 88.0% 40.0%
Tsoi 2005 Cervical length ≤10 mm 81.0% 93.7%
Gomez 2005 Cervical length ≤15 mm 64.7% 90.4%
Bagga 2010 Cervical length ≤25 mm 62.5% 89.5%
Gomez 2005 Fetal fibronectin and cervical length <15 mm 41.2% 95.5%
Gomez 2005 Fetal fibronectin and cervical length <30 mm 58.8% 85.9%
LaShay 2000 Fetal fibronection test 75.0% 88.0%
Gomez 2005 Fetal fibronection test 58.8% 78.8%
Ting 2007 pIGFBP-1 100.0% 74.0%
Lembet 2002 pIGFBP-1 93.3% 81.0%
Brik 2010 pIGFBP-1 73.7% 64.9%
Kwek 2004 pIGFBP-1 66.7% 66.1%
Schmitz 2008 Selective test (TVUS + Bishop score) 88.0% 58.0%
TVUS transvaginal ultrasound
In order to determine the proportion of positives (true and false positives) who receive
treatment and the proportion of negatives who do not receive treatment (true and false
negatives), a prevalence of actual preterm labour of 10% was assumed based on the opinion
of the committee.
16.1.5 Costs
The approximate costs of the diagnostic tests are shown in Table 98. Although these values
are used to inform the ‘what-if’ analysis, they can also be varied themselves as part of a
sensitivity analysis to determine to what extent the cost of the diagnostic strategy is an
important driver of the optimal test/treat strategy.

335
Table 92: Diagnostic test costs

Unit
Test cost Notes
Clinical £62.50 Honest (2009), based on 37.5 minutes of midwife time; see also
assessment Table 115
Ultrasound £152 NHS Reference Costs 2013/14a
(TVUS)
pIGFBP-1 £40 Based on 10 minutes of midwife time, see also Table 115 and
£196.07 cost of HHH1206 Test kit partus (pack of 10)b
FFN £70 Based on approximate cost of £45 to reflect costs associated with
cassette and analyser, and 15 minutes of midwife time, see also
Table 115
FFN fetal fibronectin, TVUS transvaginal ultrasound
(a) Currency code MA36Z, Obstetrics outpatient procedure
(b) NHS supply chain catalogue February 2014
The costs of treatment following a positive diagnosis and the costs arising from adverse
outcomes are shown in Table 114 and Table 129 respectively. The lifelong costs of adverse
outcomes are discounted at an annual rate of 3.5%.
16.1.6 Baseline risk and treatment effectiveness

As noted earlier treatment effectiveness is based on the model used to assess the cost
effectiveness of tocolysis and the baseline data used in that model is described in Sections
16.4.2.2 and 16.4.2.3.
However, in this model treatment effectiveness is handled in a deterministic fashion
predominantly for ease of exposition in the ‘what-if’ analysis. This model assumes that
women would be given calcium channel blockers as their tocolytic treatment and this model
uses the mean relative treatment effect across the 100,000 iterations of the NMA to derive an
absolute risk for each of the 3 model outcomes with treatment by gestational age. These
absolute risks for these outcomes by gestational age are shown in Table 99 below.
Table 93: Absolute risk with calcium channel blockers by gestational age
Gestational RDS IVH
age Mortality (respiratory distress syndrome) (intraventricular haemorrhage)
24 weeks 0.507 0.612 0.080
25 weeks 0.375 0.819 0.099
26 weeks 0.253 0.852 0.099
27 weeks 0.197 0.755 0.050
28 weeks 0.131 0.563 0.012
29 weeks 0.135 0.537 0.010
30 weeks 0.075 0.468 0.006
31 weeks 0.067 0.306 0.006
32 weeks 0.045 0.229 0.003
33 weeks 0.036 0.280 0.000
34 weeks 0.023 0.112 0.000
16.1.7 Quality adjusted life years (QALY)

The benefits of diagnosis are based on providing the most cost-effective tocolytic treatment
to true positives using the analysis described in Section 16.4. Benefits from treatment in

336
terms of health related quality of life are derived from the potential of treatment to reduce the
adverse outcomes of neo-natal/perinatal mortality, respiratory distress syndrome and
intraventricular haemorrhage. The QALY loss associated with these adverse outcomes is
given in Table 130 and Table 131.
A 3.5% annual discount rate is applied to QALY losses occurring in the future.
16.1.8 Results
In the following analyses the cost of a diagnostic test was set to £152 which was based on
the cost of a transvaginal ultrasound (see Table 98). The cost-effective strategy for different
diagnostic test sensitivity and diagnostic test specificity by gestational age is summarised in
Figure 21 to Figure 33.
Figure 21: What-if analysis showing cost-effective strategies by test accuracy at a

gestational age of 24 weeks


337



338



339



340



341

Figure 32: Summary of cost-effective strategies by diagnostic accuracy for gestational

ages 24-31 weeks

342
Figure 33: Summary of cost-effective strategies by diagnostic accuracy for

gestational ages 32-34 weeks
Sensitivity analysis
i. Varying the cost of the diagnostic test
In this sensitivity analysis the ‘what-if’ thresholds for cost effectiveness were compared for a
diagnostic cost of £40 versus £152 for gestational ages of 24, 30 and 34 weeks.
Figure 34: Cost-effective strategies by diagnostic accuracy varying diagnostic test

cost at a gestational age of 24 weeks

343



344
ii. Varying the prevalence

In this sensitivity analysis the ‘what-if’ thresholds for cost effectiveness were compared for a
prevalence of 5%, 10% and 20% for gestational ages of 24, 30 and 34 weeks and a
diagnostic test cost of £152.
Figure 37: Cost-effective strategies by diagnostic accuracy varying prevalence at a



345

iii. Varying the cost of false negatives

In the previous analyses it was assumed that there was no cost with a false negative, other
than that associated with higher rates of adverse outcomes as a result of missing treatment.
In Figure 40 below the implication of changing from no cost of false negative to a cost of
£20,000 per false negative is shown for a woman of 24 weeks’ gestation. The figure of
£20,000 per false negative was used because such a high value was necessary to
demonstrate any effect of this input on cost-effectiveness thresholds at this gestational age.
Figure 40: Cost-effective strategies by diagnostic accuracy varying the cost of a

false negative at a gestational age of 24 weeks

346
Figure 41 shows how cost-effective strategies by diagnostic accuracy change when

assuming no cost of a false negative diagnosis, a £1,000 cost per false negative and a
£20,000 cost of a false negative for women of 30 weeks’ gestation.
Figure 41: Cost-effective strategies by diagnostic accuracy varying the cost of a false
negative at a gestational age of 30 weeks
Figure 42 shows the most cost-effective strategy by diagnostic accuracy when comparing no
false negative costs with cost per false negative diagnosis of £1,000.
Figure 42: Cost-effective strategies by diagnostic accuracy varying the cost of a false
negative at a gestational age of 34 weeks
iv. Varying the cost of treatment

It is possible that some units would not be able to treat all the women recommended for
treatment, particularly if ‘treat all’ was considered optimal, and this would necessitate the
transfer of a proportion of women to alternative units. These transport costs have not been

347
factored into the base-case analysis and therefore in this sensitivity analysis the impact of
adding a £300 transport cost for each treated woman was assessed. This £300 was based
on the ‘see, treat, convey’ ambulance cost (£231) given in the 2013–14 NHS Reference
costs and then rounded up to reflect any additional costs that may be involved in inter-
hospital transfers. The results for each gestational age are shown in Figure 43 to Figure 53.
Figure 43: Cost-effective strategies by diagnostic accuracy assuming a £300

transport cost per woman treated at a gestational age of 24 weeks
Figure 44: Cost-effective strategies by diagnostic accuracy assuming a £300 transport

cost per woman treated at a gestational age of 25 weeks

348



349



350



351



352

As is apparent from the results shown above the impact of this change (denoted by the
yellow area in the figures above) increases with gestational age up to 31 weeks before
diminishing. The limited impact at the lower gestational ages suggests a ‘treat all’ strategy
would remain cost effective at these ages although the analysis also reflects some of the
uncertainties around the precise gestational age at which the use of a diagnostic test would
become cost effective.

353
16.1.9 Discussion
The results presented in Section 16.2.3 do provide a cost-effective rationale for adopting a
different diagnostic strategy by gestational age as summarised in Figure 32 and Figure 33. At
the lower gestational age, tests require good diagnostic accuracy and particular sensitivity in
order to be preferred to a strategy of ‘treat all’, which for analytical purposes can be
considered as having 100% sensitivity and 0% specificity. This is because the absolute
benefits of treatment are much higher at the lower gestational ages and therefore there are
greater implications of missing false negative in terms of benefits foregone. As the threshold
in Figure 21 shows, any percentage point reduction in sensitivity has to be compensated for
by a much larger percentage point increase in specificity to be considered of equivalent cost
effectiveness. In other words, a relatively large reduction in false positives with their
associated costs is necessary to compensate for any increase in false negatives.
However, as gestational age increases the issue of false positives becomes more important
as a determinant of cost effectiveness. Less benefits are foregone by false negatives and
therefore smaller reductions in false positives are needed to maintain cost effectiveness. In
Figure 29 for example, the threshold occurs where a percentage point reduction in sensitivity
can be traded approximately for a percentage point increase in specificity. As a corollary of
false positives becoming more important at higher gestational ages relative to false
negatives, then less good diagnostic accuracy is required for treatment based on a
diagnostic test to be cost effective. This reflects the different trade-off between sensitivity and
specificity at higher gestational ages. Compared with treat all, treating based on a diagnostic
test can lead to a very large reduction in costs associated with false positives which can
mean that the additional benefits of treating all, which are smaller in absolute terms, can no
longer be achieved at acceptable cost.
Indeed at 33–34 weeks’ gestation in the base-case analysis, treat all was never a cost-
effective strategy but rather with increasing gestational age better diagnostic accuracy was
necessary in order to justify treatment. The rationale for this is that at 33–34 weeks treat all is
less cost effective than no diagnosis and treatment. However, treatment can still be cost
effective if the diagnostic test can identify sufficient true positives without too large cost in
terms of false positives.
Between 32 and 33 weeks there is a ‘tipping point’ between treat all and no treatment at
lower levels of diagnostic accuracy (see Figure 29 and Figure 30). As diagnostic accuracy
declines there is a smaller chance that a diagnostic test is preferred relative to either treat all
or no treat. When the absolute benefits of treatment are relatively large then a treat all
strategy is more likely to be cost effective. As absolute benefits of treatment decline then a
no treat strategy is more likely to be cost effective. In the model this change occurs at a
gestational age of 33 weeks.
However, as Figure 39 suggests, the finding that no diagnosis and no treatment is more cost
effective than treat all is very sensitive to the prevalence of preterm birth. At 33 and 34
weeks’ gestation, treat all was preferable to no diagnosis and no treatment at a prevalence of
11% and 19% respectively.
Sensitivity analysis suggested that the cost of the diagnostic test (within plausible ranges)
was not an important driver of cost-effective thresholds for treat all, treat based on diagnostic
test and no diagnosis and no treatment. This was especially the case at the lower gestational
age where assuming an almost 4-fold increase in the cost of the diagnostic test had a
negligible impact (see Figure 34). Again, this is because the cost of diagnosis is relatively
insignificant when compared with the losses in health related quality of life and ‘downstream’
costs associated with false negatives. Nevertheless, the analysis did demonstrate, other
things being equal, that less good diagnostic accuracy would be required for a cheaper test

354
to be cost effective compared with treat all (Figure 34 and Figure 35) and no treat and no
diagnosis (Figure 36).
As noted earlier, the sensitivity analysis did indicate that the cost-effective strategy by
diagnostic accuracy was sensitive to changes in the prevalence of the preterm model,
especially at the older gestational ages (Figure 37, Figure 38 and Figure 39). This is
important because there is some uncertainty as to the precise prevalence in this population.
The reason that the model is sensitive to assumptions about prevalence is because the
importance of false negatives increases and the importance of false positives diminishes with
increasing prevalence.
The base-case analysis assumed that there were no additional costs with false negatives
other than those arising from the downstream costs of adverse outcomes. However, the
sensitivity analysis did not suggest that the results were sensitive to this assumption (Figure
40, Figure 41 and Figure 42). This is demonstrated by the fact that a false negative cost of
£20,000 was used in order to demonstrate an impact, but this is a figure way in excess of
what could be considered plausible. Although it is reasonable to assume that there are some
costs associated with false negatives, such as additional appointments with healthcare
professionals for ongoing symptoms, for example, and extra investigations as part of
differential diagnosis, these would typically be at least an order of magnitude less than
£20,000 per patient.
As noted, this analysis does provide a rationale for adopting a different approach according
to gestational age. Figure 21 suggests that a test must have a sensitivity greater than or
equal to 87% (higher if specificity is 100% or less) at a gestational age of 24 weeks to be
preferred to a strategy of treat all. There are a number of studies that report sensitivity of
87% or more (see Table 97) but they are either contradicted by another study of a similar test
or don’t have high enough specificity to make them a cost-effective option.
However, the model very clearly suggests that treat all does not remain a cost-effective
option for all gestational ages. Given the limitations and quality of the diagnostic studies
included it is not a straightforward matter to determine precisely the gestational age when the
approach should change. Indeed, the precise gestational age at which it becomes cost
effective to use a diagnostic test may vary by unit, especially where transfer costs out of unit
are being incurred. Nevertheless, there is some evidence, subject to uncertainty and
limitations, from the clinical review that some tests might achieve the diagnostic accuracy to
be considered cost effective relative to treat all at 30 weeks.
Figure 27 shows that the following combinations of sensitivity and specificity (or better) make
treating based on a diagnostic test more cost effective than treat all at a gestational age of 30
weeks (see Table 100). Both studies of transvaginal ultrasound using a cervical length of 15
mm or less fall have diagnostic accuracy figures that are sufficient to make treatment based
on a diagnostic test be considered cost effective relative to treat all. Using transvaginal
ultrasound and cervical length of 10 mm or less also has diagnostic accuracy figures that
would support a recommendation when compared with treat all, but this is only based on a
single study. In addition the model suggested that treating based on a transvaginal
ultrasound and cervical length of 15 mm or less could be considered cost effective relative to
a strategy of no test and no treat at the higher gestational ages.
Table 94: Test accuracy threshold for cost effectiveness of using a diagnostic test at a
Sensitivity Specificity
100% 17%
99% 19%
98% 21%

355
Sensitivity Specificity
97% 22%
96% 24%
95% 26%
94% 28%
93% 29%
92% 31%
91% 33%
90% 35%
89% 36%
88% 38%
87% 40%
86% 42%
85% 43%
84% 45%
83% 47%
82% 49%
81% 51%
80% 52%
79% 54%
78% 56%
77% 58%
76% 59%
75% 61%
74% 63%
There are a number of considerations that need to be taken into account when the results of
the above health economic analysis are interpreted. First, although the analysis took a
standard incremental approach with respect to broad categories of treat all, treat based on a
diagnostic test, or do not treat and do not diagnose, it did not do so with respect to different
combinations of sensitivity and specificity that would be considered cost effective relative to
treat all and do not diagnose or treat. Nor within the context of the what-if analysis did it do
this for specific diagnostic accuracy data provided by the included studies. This was because
it was thought that this diagnostic accuracy data had such severe limitations that the broader
what-if approach would be more useful. However, the uncertainty inherent in the reported
diagnostic accuracy evidence adds another level of uncertainty with respect to what may
considered the most cost-effective diagnostic test or combinations of test.
Secondly, the analysis departed from the NICE reference case by not including a
probabilistic sensitivity analysis. However, it is important to note that this model took into
account the interdependence of treatments and diagnosis in determining cost effectiveness
and the treatment effect size was derived from the mean relative treatment effect of a
treatment found to be cost effective in a probabilistic sensitivity analysis. It would have been
possible to sample the relative treatment effect from the iterations produced for the NMA but
if treatment was considered cost effective based on the mean relative treatment effect it is
almost certain that a probabilistic sensitivity analysis would confirm this finding, albeit with
some quantification of the uncertainty. Furthermore, this what-if analysis is by construction
deterministic with respect to sensitivity and specificity and yet the uncertainty surrounding the

356
diagnostic accuracy of the various strategies is more difficult to quantify than the uncertainty
surrounding treatment. Extensive 1-way sensitivity analysis was undertaken to investigate
the important drivers of cost effectiveness (including sensitivity, specificity, gestational age,
prevalence, diagnostic and cost of false negatives).
Finally, in assessing the benefits of diagnosis the model assumes that false negatives miss
the benefits of treatment and experience the baseline risk of various outcomes. However,
this reflects a worst case scenario and at least a proportion of women sent home as
negatives are likely to re-present in sufficient time to still benefit from treatment.
16.1.10 Conclusion
This what-if analysis provides strong evidence that treatment is cost effective even when
considering the costs of identifying the women suitable for treatment. It also provides
evidence that the most cost-effective diagnostic strategy varies with gestational age. At lower
gestational ages when the absolute risks are high then treating all women with suspected
preterm labour and intact membranes can be cost effective even when allowing for the fact
that 90% of those treated might not derive any treatment benefit.
The model also suggest that treatment can remain cost effective at higher gestational ages
when absolute risks are lower, providing a diagnostic test can be applied with sufficiently
good diagnostic accuracy.
Although a change in diagnostic strategy according to gestational age is indicated by this
analysis, the gestational age at which this change should take place is difficult to precisely
identify given the uncertainty with respect to the precise diagnostic accuracy of the various
tests. Nevertheless, Figure 27 and Figure 38 suggest that 30 weeks and above may be a
reasonable gestational age at which to require treatment to be guided by a positive
diagnostic test, and thereby reduce inconvenience to women and costs to the health service
when absolute risks are relatively low. There is some suggestion from the diagnostic studies
reviewed that at 30 weeks transvaginal ultrasound using a cervical length of 15 mm or less
could have sufficient diagnostic accuracy to be considered cost effective relative to treat all,
do not diagnose and do not treat, or other diagnostic tests or combinations of tests which do
not have a cost-effective sensitivity/specificity combination.
16.2 What is the clinical and cost effectiveness of

magnesium sulfate given to women at high risk of
giving birth preterm(defined as those suspected to
be in preterm labour or diagnosed as being in
preterm labour and those having planned preterm
birth) for preventing cerebral palsy and other
neurological disorders in babies born at different
preterm gestations?
16.2.1 Introduction
Preterm labour carries a high risk of neonatal mortality and morbidity which can result in
large ongoing costs for the healthcare system, as a result of cerebral palsy for example.
There has been increasing research interest in recent years in magnesium sulfate as a
treatment that may offer some degree of neuroprotection for preterm birth. Clearly, there is
potential for any treatment that reduces the adverse consequences of preterm birth to be
cost effective given the substantial losses in health related quality of life and healthcare costs

357
resulting from these adverse consequences. A small published literature (see Section 16.3.2)
suggested that magnesium sulfate is a cost-effective treatment for neuroprotection, but these
analyses were performed outside a UK setting and therefore it was thought that it would be
useful to develop a new model utilising the clinical review that was undertaken for this
guideline.
16.2.2 Methods
A decision analytic model was developed in Microsoft Excel® to assess the cost
effectiveness of magnesium sulfate given to women for neuroprotection between 24+0 and
32+0 weeks of pregnancy and at high risk of preterm birth. A schematic of the model is
shown below in Figure 54.
Figure 54: Schematic of the model for the use of magnesium sulfate
for neuroprotection

358
16.2.3 Model probabilities and treatment effect size

A clinical review undertaken for the guideline assessed the following outcomes:
• stillbirth
• neonatal mortality before discharge
• neonatal/paediatric mortality between discharge and follow-up
• total perinatal, neonatal and paediatric mortality*
• grade III or IV intracranial haemorrhage (ICH)*
• periventricular leukomalacia (PVL)*
• any cerebral palsy
• moderate or severe cerebral palsy at 2 years*
• gross motor dysfunction at 2 years
• any developmental delay at 2 years
• cognitive dysfunction at 2 years
• blindness at 2 years
• deafness at 2 years
• maternal death
• any maternal adverse effects
• maternal adverse effects leading to stopping of infusion
• maternal cardiac or respiratory arrest
• drop in maternal blood pressure of more than 15 mmHg
• maternal hypotension.
Only the outcomes marked with an asterisk were included in the model. Priority was given to
outcomes that have the most important impact on health related quality of life and/or where a
substantial saving to the NHS would result from an averted case. With respect to the
outcomes included in the model, a lifetime approach was considered with respect to costs
and health related quality of life.
Clearly there would be double counting if more than 1 neonatal/paediatric mortality outcome
was included and therefore the most comprehensive measure of mortality was chosen.
Moderate or severe palsy was chosen over any cerebral palsy as mild cerebral palsy, which
will explain most of the difference, has a much smaller impact on future health service costs
and health related quality of life.
The outcomes relating to motor function, cognitive development and any developmental
delay were not included because, although important, it was thought their impact and costs
would occur predominantly outside the health domain (such as in the education sector).
Furthermore, their exclusion from the model was unlikely to have an important bearing on the
overall cost effectiveness as treatment effects were small and not statistically significant.
Outcomes relating to deafness, blindness, maternal mortality, maternal cardiac or respiratory
arrest and maternal hypotension were excluded as the number of events was negligible and
any differences fell a long way short of statistical significance.
The effect of treatment on maternal blood pressure of more than 15 mmHg was not included
in the analysis as it was felt to be a more intermediate marker of outcomes of interest and
unlikely to have an important independent effect on health related quality of life and costs.

359
Broad measures of maternal adverse effects were not included as it was thought they most
likely would contribute to treatment failure and/or would have only a very short-term effect on
health related quality of life.
Meta analyses undertaken for the clinical review were used to estimate the baseline risk and
the treatment effect size as shown in Table 101 and Table 102 respectively. In addition to the
point estimate used in deterministic analysis, the tables also show the parameters that are
used for probabilistic sensitivity analysis.
Table 95: Baseline risks for model of magnesium sulfate for neuroprotection
Alpha
Outcome Risks Distribution a Betaa Source
Neonatal/paediatric 10.8% Beta 242 2001 Guideline meta-analysis
mortality
Cerebral palsy 3.4% Beta 59 1656 Guideline meta-analysis
ICH 5.0% Beta 90 1709 Guideline meta-analysis
PVL 2.7% Beta 48 1751 Guideline meta-analysis
ICH intracranial haemorrhage, PVL periventricular leukomalacia
a. The Alpha parameter is given by the number of events in the controls in the meta- analysis. The Beta
parameter is the number without events in the controls
Table 96: Treatment effect size for model of magnesium sulfate for neuroprotection
Relative
Outcome risk Distribution Mua Sigmaa Source
Neonatal/paediatric 0.95 Log-normal −0.044 0.088 Guideline meta-analysis
mortality
Cerebral palsy 0.61 Log-normal −0.490 0.211 Guideline meta-analysis
ICH 0.81 Log-normal −0.189 0.154 Guideline meta-analysis
PVL 0.94 Log-normal −0.053 0.206 Guideline meta-analysis
a. Mu is calculated as the natural log of the relative risk and sigma is calculated as the standard error of the log of
the relative risk
16.2.4 Costing and resource use

In accordance with the NICE Guidelines Manual (NICE 2012) costing was undertaken from
the perspective of the NHS and personal social services. Costs are based on 2015 prices
unless otherwise stated.
Discounting of costs was not exactly as per the NICE Reference Case. No discounting is
necessary for treatment as this all occurs at the start of the intervention, but the outcomes
evaluated are often lifetime in their impact and their costs should be discounted. However,
the paper used to estimate these costs used a discount rate of 5% rather than the 3.5%
discount rate suggested by NICE (Kruse 2009). It was not possible to recalculate the costs
using a 3.5% discount rate as a temporal breakdown of the overall lifetime cost was not
provided.
16.2.4.1 Treatment cost

The cost of magnesium sulfate are based on an initial 4 g intravenous (IV) bolus followed by
1 g per hour IV infusion thereafter for a period of 24 hours. The pharmaceutical component of
treatment costs are shown in Table 103.

360
Table 97: Magnesium sulfate costs

Dose Per unit cost Quantity Total Source
20 ml (4 g) amp £16.98 1 £16.98 BNF (March 2015)
2 ml (1 g) amp £1.15 24 £27.60 BNF (March 2015)
Total - - £44.58
It addition it was assumed that the woman would require ante-natal monitoring for the whole
24 hour period for which she was on treatment and this was based on NHS Reference costs,
see Table 104. So the total treatment cost was £1,081
Table 98: Antenatal observation

Currency Code Currency description Cost Source
NZ16Z Ante-natal routine observation £1,036 NHS Reference Costs 2013/14
16.2.4.2 Downstream costs

In addition to the costs of the intervention, it is important to compare the alternatives in terms
of their impact on costs that arise subsequent to the intervention decision which are
attributable either to the intervention itself, such as adverse events, or to preterm birth. Table
105 details the costs associated with model outcomes.
Table 99: Outcome related costs

Outcome Cost Distribution SE Source
Perinatal, neonatal or £1,480 Normal £159 NHS Reference Costs 2013/14a
paediatric mortality
Cerebral palsy £74,608b Deterministic - Kruse 2009
ICH £22,382c Deterministic - Kruse 2009
PVL £74,608d Deterministic - Kruse 2009
a. XB03Z Paediatric critical care, advanced critical care 3
b. Kruse 2009 estimated in year 2000 prices that the lifetime health care costs for cerebral palsy using an
annual discount rate of 5% was €66,155 for men and €65,288 for women. The mid-point of this estimate was
used and converted into GBP using an exchange rate of £0.737 = €1 (http://www.exchangerates.org.uk/ -
accessed 03/04/2015). It was then converted into 2013/14 prices using the HCHS (The Hospital & Community
Health Services) Index
c. It was assumed that Grade III and Grade IV ICH would be similar in cost to cerebral palsy. Alvarez 1994
suggest that 30% of ICH is of severity Grade III and Grade IV and therefore the cost of ICH was estimated as 0.3
x £74,608
d. It was assumed that the costs of PVL were the same as the costs for cerebral palsy
16.2.5 QALYs
A lifetime QALY loss was assigned to each of the 4 outcomes assessed in the model as
shown in Table 106. In the case of a neonatal or paediatric death it was assumed that this
would result in a loss of 80 years of life based on current life expectancy in the UK. It was
assumed that all 80 years would have be lived with a health state utility of 0.82 based on UK
population norms (Kind 1983). An annual discount rate of 3.5% was applied for each year
lived in full health in accordance with NICE methods.
The health state utility for moderate to severe cerebral palsy was taken from the literature
(Cahill 2011) with a value of 0.55. A life expectancy of 60 years was assumed and the total
discounted QALY for that life expectancy calculated assuming there were no additional co-
morbidities. This was then subtracted from the discounted QALY of an individual who lived
80 years with a health state utility of 0.82 in order to estimate the overall lifetime QALY loss
associated with moderate to severe cerebral palsy.

361
It was assumed that the QALY loss from PVL would be the same as for moderate to severe
cerebral palsy and that the QALY loss from ICH would be one-third of that from cerebral
palsy.
Table 100: QALY losses associated with adverse model outcomes

Outcome QALY loss
Perinatal, neonatal or paediatric mortality 22.70
Cerebral palsy 8.50
Intracranial haemorrhage 2.80
Periventricular leukomalacia 8.50
16.2.6 Sensitivity analysis

Probabilistic sensitivity analysis, and one-way and multi-way sensitivity analyses were
undertaken to assess the robustness of the model results given uncertainty surrounding
various model inputs.
16.2.7 Results
Deterministic base case results are presented in Table 107, Figure 55, Figure 56 and Figure
57. This analysis suggests that magnesium sulfate for neuroprotection is dominant, being
cheaper and more effective than no magnesium sulfate for neuroprotection. It is cheaper
because the savings from a reduction in adverse outcomes more than offset the cost of
treatment.
Table 101: Incremental costs and QALYs of magnesium sulfate for neuroprotection
Outcome Incremental costs Incremental QALYs
Treatment £1,081 N/A
Neonatal/paediatric mortality −£7 0.11
Cerebral palsy −£995 0.11
Intercranial haemorrhage −£193 0.02
PVL −£102 0.01
Total −£215 0.26

362
Figure 55: Deterministic base-case analysis of magnesium sulfate for neuroprotection

shown on a cost-effectiveness plane
Figure 56: Breakdown of base case analysis of costs for magnesium sulfate for
neuroprotection versus no magnesium sulfate for neuroprotection

363
Figure 57: Breakdown of base case analysis of QALY losses for magnesium
Figure 58 shows results from the probabilistic sensitivity analysis of 10,000 Monte Carlo
simulations plotted on a cost-effectiveness plane. This suggests that the probability
magnesium sulfate is cost effective for neuroprotection compared with not giving magnesium
sulfate for neuroprotection is 85.9% using a £20,000 willingness to pay for a QALY decision
threshold. Across the 10,000 simulations magnesium sulfate for neuroprotection had a net
mean benefit of £4,900 when compared with a strategy of no magnesium sulfate for
neuroprotection.

364
Figure 58: Base case probabilistic sensitivity analysis of magnesium sulfate for
neuroprotection versus no magnesium sulfate for neuroprotection
The cost-effectiveness acceptability curve for the base-case probabilistic sensitivity analysis
is shown in Figure 59. Across all willingness to pay thresholds magnesium sulfate has the
highest probability of being cost effective. This is so even when the decision maker is not
willing to pay anything for a QALY and this is because magnesium sulfate for neuroprotection
was the cheapest strategy for approximately 60% of the simulations as a result of savings
from reduced adverse outcomes more than offsetting treatment outcomes.

365
Figure 59: Cost-effectiveness acceptability curve for base case probabilistic

sensitivity analysis
16.2.8 Sensitivity analysis

The base-case analysis strongly suggests that magnesium sulfate for neuroprotection is cost
effective. Below are presented a number of sensitivity analyses to test how robust that
conclusion is with respect to changes in model parameters. Clearly, lowering the treatment
cost and/or increasing the QALY associated with adverse outcomes would only reinforce the
base-case conclusion. Therefore, these sensitivity analyses are intended to explore the
thresholds at which magnesium sulfate for neuroprotection would cease to be cost effective.
If those input values required to achieve those thresholds are not thought plausible then
confidence in the base-case result can be strengthened.
16.2.8.1 Increasing the treatment costs

The treatment cost would have to be increased to £6,413 for magnesium sulfate for
neuroprotection to be no longer considered cost effective at a £20,000 willingness to pay for
a QALY according to the deterministic analysis. In a probabilistic sensitivity analysis of
10,000 Monte Carlo simulations at this treatment cost and using a £20,000 willingness to pay
threshold the results indicated that magnesium sulfate would have a 47.0% probability of
being the most cost-effective treatment option and had a net mean benefit of −£513 across
the 10,000 simulations. This result is displayed in Figure 60 and the associated cost-
effectiveness acceptability curve in Figure 61.

366
Figure 60: Monte Carlo simulation of magnesium sulfate for neuroprotection assuming
a treatment cost of £6,413 and a willingness to pay of £20,000 per QALY
Figure 61: Cost-effectiveness acceptability curve for magnesium sulfate for

neuroprotection assuming a treatment cost of £6,413

367
If a £30,000 per QALY threshold was used to assess cost effectiveness then the treatment
cost would have to exceed £8,971. A probabilistic sensitivity analysis of 10,000 Monte Carlo
simulations at this treatment cost and £30,000 willingness to pay threshold suggested that
magnesium sulfate would have a 47.1% probability of being the most cost-effective treatment
option and had a net mean benefit of −£698 across the 10,000 simulations. This result is
displayed in Figure 62 and the associated cost-effectiveness acceptability curve in Figure 63.
a treatment cost of £8,971 and a willingness to pay of £30,000 per QALY

368
Figure 63: Cost-effectiveness acceptability curve for magnesium sulfate for

neuroprotection assuming a treatment cost of £8,971
16.2.8.2 Reducing the QALY loss from adverse outcomes

There is considerable uncertainty around the health state utility loss associated with adverse
outcomes and they are treated deterministically even in the probabilistic sensitivity analysis.
In this analysis we restrict the QALY loss to that arising from mortality and assume in this
analysis that this is 11 QALYs, about half what was used in the base-case analysis.
A probabilistic sensitivity analysis of 10,000 Monte Carlo simulation found that there was a
70.9% probability of magnesium sulfate for neuroprotection being cost effective with these
assumptions about reduced QALY loss from adverse outcomes while keeping all other model
inputs at their base-case values. Across the 10,000 simulations the net mean benefit of
magnesium sulfate was £1,075. The plot of these 10,000 simulations is shown in Figure 64
and the associated cost-effectiveness analysis acceptability curve is shown in Figure 65.

369
that a QALY loss only arises from mortality and that this loss is only half the
value assumed in the base-case analysis
Figure 65: Cost-effectiveness acceptability curve magnesium sulfate for

neuroprotection assuming that a QALY loss only arises from mortality and
that this loss is only half the value assumed in the base-case analysis

370
16.2.8.3 Reducing the cost of adverse outcomes

In this sensitivity analysis the cost of all adverse outcomes are removed and the treatment
cost is increased until magnesium sulfate is no longer cost effective in the deterministic
analysis at a willingness to pay of £20,000 and £30,000 per QALY. This occurs at treatment
costs of £5,117 and £7,675 respectively. Probabilistic sensitivity analysis for these 2
scenarios is shown in Figure 66 and Figure 68 with their respective cost-effectiveness
acceptability curves depicted in Figure 67 and Figure 69.
The probability of magnesium sulfate being cost effective in the Monte Carlo simulation with
the lower treatment cost and willingness to pay for a QALY was 48% with a net mean benefit
of −£346. In the Monte Carlo simulation with a higher treatment cost and willingness to pay
for a QALY the probability of magnesium sulfate given for neuroprotection being cost
effective was also 48% with a net mean benefit of −£597 across the 10,000 simulations.
no costs from adverse outcomes and a treatment cost of £5,117 and a
willingness to pay of £20,000 per QALY

371

neuroprotection no costs from adverse outcomes and a treatment cost of
£5,117
no costs from adverse outcomes and a treatment cost of £7,675 and a
willingness to pay of £30,000 per QALY

372

neuroprotection no costs from adverse outcomes and a treatment cost of
£7,675
16.2.8.4 Two-way sensitivity analysis varying the treatment cost and the QALY loss
from mortality
In this sensitivity analysis 2 model inputs are varies across a wide range of values to
estimate a combined cost-effectiveness threshold. It also indicates the trade-off necessary in
order across these 2 variables necessary for cost effectiveness.
Treatment costs are varied between £1,000 and £20,000 and the QALY loss from mortality is
varied between 5 and 25 QALYs, and the results of this sensitivity analysis are shown in
Figure 79 for a £20,000 willingness to pay for a QALY. The base-case analysis falls a long
way from the threshold with a treatment cost of just over £1,000 and a 22.70 QALY loss from
mortality.

373
Figure 70: Graph to show the two-way relationship between QALY loss from mortality
and treatment cost holding all other values constant at their base case value
16.2.8.5 Two-way sensitivity analysis varying the cost of cerebral palsy and the
QALY loss from cerebral palsy
In this sensitivity analysis the cost and QALY loss from cerebral palsy are varied between
wide levels, including much lower values assumed in the base case analysis. However, the
conclusion that magnesium sulfate given for neuroprotection is cost effective remains the
case for all the cost QALY combinations assessed in this analysis as displayed in Figure 71.

374
Figure 71: Graph to show the two-way relationship between QALY loss from cerebral
palsy and cerebral palsy cost holding all other values constant at their base
case value
16.2.8.6 Treatment at a gestional age of 34 weeks

The risks in the base case analysis are based on the control group in the meta-analysis
undertaken as part of the review of the clinical evidence for this guideline (see Table 101).
These controls would reflect the range of gestational ages included in the clinical trials but in
practice the risk of adverse events will fall with increasing gestational age which is likely to
make treatment relatively less cost effective. Therefore, in this sensitivity analysis we explore
whether treatment at a gestational age of 34 weeks is likely to be cost effective.
In this sensitivity analysis we used a baseline risk of cerebral palsy of 0.7% (Marret 2007) – a
much lower level than the 3.4% risk used in the base-case analysis (see Table 101) – but
assumed that the relative treatment effect would be unaltered (see Table 102). We
additionally made the conservative assumption that there would be no treatment effect for
other outcomes considered in the health economic model.
The cost in the base case analysis of £1,081 is based on magnesium sulfate being a
standalone intervention and includes a cost routine antenatal observation as part of an
inpatient admission. However, the guideline has recommended tocolysis for women up to 34
weeks and therefore it can be reasonably argued that these costs are not incremental to the
magnesium sulfate intervention and therefore in this sensitivity analysis the costs are limited
to the drug costs of £45.
Deterministic results are presented for this sensitivity analysis in Table 108 and Figure 72.
This analysis suggests that magnesium sulfate for neuroprotection remains dominant, albeit
much less so than in the base-case analysis.

375
Table 102: Incremental costs and QALYs of magnesium sulfate for neuroprotection at
a gestational age of 34 weeks
Outcome Incremental costs Incremental QALYs
Treatment £45 N/A
Neonatal/paediatric mortality £0 0.00
Cerebral palsy −£139 0.02
Intercranial haemorrhage £0 0.00
PVL £0 0.00
Total −£94 0.02
Figure 72: Deterministic sensitivity analysis of magnesium sulfate for neuroprotection

at a gestational age of 34 weeks
To reflect that magnesium sulfate might require more intensive monitoring than would be
required for tocolytic treatment alone, the treatment cost was increased to determine the
threshold for cost effectiveness at a willingness to pay of £20,000 per QALY and £30,000 per
QALY respectively. This suggested that magnesium sulfate would remain cost effective for
neuroprotection at a gestational age at 34 weeks providing that the treatment cost was £455
or lower at a £20,000 per QALY willingness to pay threshold or that the treatment cost is
£614 or lower for a higher £30,000 per QALY willingness to pay threshold.
16.2.9 Discussion
This model supports other published economic evaluations in finding magnesium sulfate to
be a cost-effective intervention for neuroprotection in preterm birth (Cahill 2011, Bickford
2013). In our analysis, magnesium sulfate was found to be dominant when compared with an
alternative of no magnesium sulfate.
Probabilistic sensitivity analysis took into account treatment uncertainty where magnesium
sulfate was only found to offer a statistically significant benefit for 1 of the 4 outcomes, but
nevertheless found that magnesium sulfate had an 86% probability of being the most cost-
effective treatment using base-case inputs.

376
The finding that magnesium sulfate was cost effective was generally robust to changes in
model inputs that were made to favour the alternative of no magnesium sulfate for
neuroprotection. A sensitivity analysis demonstrated a treatment cost threshold at which
magnesium sulfate for neuroprotection would no longer be cost effective but this treatment
cost had to be more than 5 times the treatment cost used in the model, which included a cost
of hospitalisation for all women.
The assumptions made with respect to cerebral palsy were potentially important to the
results of the analysis as this is where the clinical review for our guideline found the greatest
evidence of treatment benefit. However, sensitivity analyses assuming a much lower cost of
cerebral palsy and a much lower QALY gain from an averted case did not alter the finding
that magnesium sulfate was cost effective.
A sensitivity analysis suggested that treatment up to a gestational age of up to 34 weeks
could be cost effective, providing that the relative treatment effect for cerebral palsy is
maintained even though the absolute risk would be much lower. However, not unexpectedly,
this sensitivity analysis did indicate that the relative cost effectiveness of magnesium sulfate
would be lower at higher gestational ages when there is not much evidence of benefit of such
interventions.
16.2.10 Conclusion
The model produced for this guideline provides strong support for the cost effectiveness of
magnesium sulfate given for neuroprotection and this is reflected in the recommendations
made by the Guideline Development Committee.
16.3 What is the clinical and cost effectiveness of

tocolyticsgiven to women with suspected or
diagnosed preterm labour to improve outcomes:
• progesterone/progestogens
• beta-sympathomimetics
• oxytocin receptor antagonists
• cyclo-oxygenase enzyme inhibitors
• non-steroidal anti-inflammatory drugs
• magnesium sulfate?
16.3.1 Introduction
Preterm birth can be costly for the health services and accounts for a disproportionate
amount of infant death and morbidity. Indeed, very preterm birth, which accounts for just 1%
of UK births, is implicated in more than half of infant deaths (RCOG Guideline on Tocolysis
for Women in Preterm Labour). Therefore, prevention of preterm birth is important in order to
improve child outcomes. There are a range of medications which have been proposed as
having a tocolytic function and there is considerable variation in their cost. Cost-effectiveness
analysis can therefore be potentially helpful in making decisions between the use of these
particular alternatives or none.

377
16.3.2 Methods
A decision analytic model was developed in Microsoft Excel® to assess the cost
effectiveness of drugs given to women with suspected or diagnosed preterm labour in order
to delay birth and by so doing improve neonatal outcomes. A simplified schematic of the
model is shown in Figure 73.
Figure 73: Schematic of decision tree model to assess the cost-

effectiveness of drugs given to women with suspected or
diagnosed preterm labour
The model has been developed so that cost effectiveness can be assessed by gestational
age in weekly increments from 24 weeks to 34 weeks. In total 6 different drug classes can be
compared in addition to standard care, although the model can be set to run with 1 or more
treatment alternatives excluded. The various treatment alternatives compared within the
model are as follows:
• standard care
• betamimetics

378

• magnesium sulfate
• oxytocin receptor blockers
• prostaglandin inhibitors
• nitrates.
Treatment is expected to work by delaying birth and to reflect this in the schematic (see
Figure 73) it might have seemed more intuitive if the first chance nodes following treatment
had been ‘delayed birth’ and ‘not delayed’. However, it was not possible from the clinical data
to determine different probabilities for adverse events according to whether labour was
delayed or not. Furthermore, in calculating overall health gain from a particular treatment it is
the risk of the outcome by treatment which is required which, if the data existed, would
simply be a weighted average of the risks in the delayed/not delayed group.
16.3.2.1 Clinical outcomes

As part of the protocol for the clinical review, the Guideline Development Committee
prioritised the outcomes listed in Table 109.
Table 103: Committee prioritised outcomes for clinical review of tocolysis

Outcome Included in the NMA
Maternal outcomes
Maternal mortality
Adverse events – discontinuation of treatment Included
Maternal infection
Neonatal outcomes
Perinatal mortality Included
Neonatal mortality Includeda
Delay of birth by more than 48 hours Included
Mean gestational age at birth Included
Respiratory distress syndrome (RDS) Includeda
Chronic lung disease/bronchopulmonary dysplasia
Intraventricular haemorrhage (IVH) Includeda
White matter injury/periventricular leucomalacia
Neonatal infection/sepsis Included
Neurodevelopmental disability
a. Used as an outcome in the health economic model
The health economic model was restricted to outcomes for which a network meta-analysis
(NMA) was undertaken in order to ensure all the drug treatment classes could be compared
in a consistent manner. However, if it was thought evidence from pairwise comparison of
outcomes might have a bearing on cost effectiveness, the committee would be able to use
that information additionally in making their recommendations.
Of the 8 outcomes included in the NMA, 3 outcomes were chosen for inclusion in the model:
neonatal mortality; respiratory distress syndrome (RDS); and intraventricular haemorrhage
(IVH). Although delay of birth is the objective of treatment, that was considered to be an
intermediate marker for the real end-points of interest, namely improved neonatal and

379
maternal outcomes. Similarly, discontinuation of treatment and mean gestational age were
considered to be proxies for other outcomes influencing health related quality of life.
Neonatal infection/sepsis was not included because prognosis is often good unless it leads
to a death or neurodevelopmental problems in which case it would usually be captured within
the included outcomes. Neonatal and perinatal mortality are both important outcomes but is
thought that there would be issues of double counting if both outcomes were included and
therefore it was felt that neonatal mortality was the more useful outcome of the two.
16.3.2.2 Baseline data

The NMA generates a measure of treatment effect for each drug class relative to placebo. It
was assumed that placebo could be used to represent a standard care or no drug treatment
to delay birth option. However, many of the studies in the NMA were quite dated and
therefore the committee considered that the outcomes in the placebo arms in those trials
were unlikely to represent the current risk of standard care for the outcomes included in the
health economic analysis.
Therefore, rather than use the placebo risk in trials as the baseline risks for mortality, RDS
and IVH were estimated from different sources. However, this is not without its own
limitations as it is likely that the data that is used to inform the baseline risk includes women
with suspected or diagnosed preterm labour who will often have been given some drug
treatment in order to delay birth. If a single drug class dominated current practice then it
would have been possible to use this as the baseline treatment and measure the other
treatments effectiveness relative to this drug class. However, the committee considered there
was too much variation in current practice to do this. Therefore, if drug treatment to delay
birth is effective the baseline risks used in this analysis may under-estimate the risk
associated with standard care or no drug treatment.
Although the health economic model used data from the NMA on neonatal mortality, a more
useful outcome would have been all deaths including stillbirths, but, as noted in Section
16.4.2.1, the trial data did not record the data in a manner that allowed it to be analysed in
this way. The baseline data presented in Table 110 gives the risk of all perinatal (including
stillbirth) and neonatal death and it is assumed that the relative treatment effect derived from
the NMA on neonatal death will be the same when applied across all deaths. This is a strong
assumption, but at least the NMA did not find statistically significant differences in perinatal
mortality for the drug treatment classes in this analysis.
Table 104: Baseline death rate with no drug treatment to delay birth by gestational
agea
Gestational age (weeks) Births Deathsb Mortality rate
24 759 473 0.623
25 801 386 0.482
26 887 299 0.337
27 1009 268 0.266
28 1178 211 0.179
29 1339 210 0.184
30 1623 169 0.104
31 2140 197 0.092
32 3094 191 0.062
33 4141 209 0.050
34 6975 226 0.032
a. Source: Live births, stillbirths and infant deaths by gestational age at birth, 2011 birth cohort (ONS, 2011)

380
b. Includes stillbirths, early perinatal deaths and late perinatal deaths
Published data was used in order to estimate the baseline risk of RDS and IVH (Ross, 2014)
and is shown in Table 111 and Table 112 respectively. In order to sample the baseline risk
for probabilistic sensitivity analysis it was useful to estimate an actual number of events and
this was done by multiplying the births in Table 110 by the mortality rate reported in the
paper.
Table 105: Baseline respiratory distress syndrome rate with no drug treatment to
delay birth by gestational age
Gestational age (weeks) Births RDS casesa RDS rate
24 759 531 0.700
25 801 720 0.899
26 887 824 0.929
27 1009 847 0.839
28 1178 765 0.649
29 1339 709 0.622
30 1623 892 0.555
31 2140 791 0.370
32 3094 866 0.280
33 4141 1,407 0.340
34 6975 976 0.140
a. Estimated
Table 106: Baseline intraventricular haemorrhage rate with no drug treatment to delay
birth by gestational age
Gestational age (weeks) Births IVH casesa IVH rate
24 759 189 0.249
25 801 240 0.300
26 887 266 0.300
27 1009 161 0.160
28 1178 47 0.040
29 1339 40 0.035
30 1623 32 0.020
31 2140 42 0.020
32 3094 30 0.010
33 4141 0 0.000
34 6975 0 0.000
a. Estimated
The baseline risks for all 3 model outcomes by gestational age are depicted graphically in
Figure 74.

381
Figure 74: Baseline risks by gestational age
li i k
100%
90%
80%
70%
60%
50%
40%
16.3.2.3 Mortality due to RDS and IVH

Trials included in the NMA will have counted RDS and IVH cases even where those cases
resulted in death. This double counting is accounted for in the model by estimating the
mortality associated with RDS and IVH. This enables the model to estimate the proportion of
babies with RDS and IVH that survive, which is important when calculating the QALYs (see
Section 16.4.2.6).
It is assumed that the mortality rate from RDS and IVH does not vary by gestational age.
RDS mortality was estimated using published US data (American Lung Association Lung
Disease Data 2008). This data suggested that RDS affected 16,268 babies born in the USA
in 2005, with 875 of those cases resulting in death. The mortality rate of IVH has been
estimated at between 27% and 50% from severe (high-grade) IVH and at 5% for from low-
grade haemorrhage. Table 113 summarises the RDS and IVH mortality risk used in the
model’s base-case analysis.
Table 107: RDS and IVH mortality rates by gestational age

Gestational age (weeks) RDS mortality rate IVH mortality rate
24 0.054 0.300
25 0.054 0.300
26 0.054 0.300
27 0.054 0.300
28 0.054 0.300
29 0.054 0.300
30 0.054 0.300
31 0.054 0.300
32 0.054 0.300
33 0.054 0.300
34 0.054 0.300

382
16.3.2.4 Treatment effectiveness

Section 16.4.2.2 outlines how the baseline risk of 3 model outcomes has been estimated.
This baseline is assumed to represent the risk when no drugs are given to delay preterm
birth. The NMA estimates a treatment effect size for each of the drug classes in this analysis
relative to this baseline risk. A baseline risk and a relative treatment effect allow the absolute
risk or probability of the outcome to be calculated for each treatment class.
Absolute risk for Betamimetics = baseline risk × relative risk
This absolute risk can then be used to generate the weighted QALYs and costs associated
with different drug classes.
The model assumes that the relative treatment effect, derived from the NMA, will be the
same across gestational age. Women with pregnancies less than 26 weeks were not
included in the studies that made up the NMA but the committee considered it reasonable to
assume that treatment would be equally effective in these women. However, the absolute
treatment effect varies with gestational age, reflecting the different baseline risks at different
gestational ages.
16.3.2.5 Costs
Costs were based on an NHS and Personal Social Services perspective as outlined in the
NICE reference case in Developing NICE guidelines: the manual. Costs were based on a
2015 price year or as close to 2015 prices as could be estimated when sourced from an
earlier price year.
16.3.2.6 Treatment costs

For the NMA that provides the estimates of treatment effectiveness, the committee agreed
that it was reasonable to construct the network by treatment class as opposed to individual
drugs. The rationale for this is that they expected there to be little variation in treatment
effectiveness by class. Where treatments are assumed to be equally effective it follows that
the most cost-effective treatment among them will be the cheapest. Therefore, the committee
agreed that the costing for each treatment class would be based on the cheapest drug in
class that they would be willing to recommend if this drug was shown to be cost effective.
The costs used for treatment costs are shown in Table 114. How these costs were derived is
described below.
Table 108: Treatment costs

Drug class Cost Source
Betamimetics £81 BNF 2015
Calcium channel blockers £14 BNF 2015
Magnesium sulfate £169 BNF 2015
Oxytocin receptor blockers £517 BNF 2015
Prostaglandin inhibitors £14 BNF 2015
Nitrates £16 BNF 2015
In addition to the pharmaceutical costs of the treatment it is assumed that some staff time is
required in order to administrate. For each treatment the staffing time is estimated and a total
cost of staff time is calculated based on the unit costs of staff time shown in Table 115.

383
Table 109: Staff unit costs

Staff Unit cost per hour Source
Nurse £120a Curtis (2014)
Doctor £41b Curtis (2014)
a. Based on per hour of patient contact for the full-time equivalent basic salary for Agenda for Change band
6. This cost per hour is based on observations about the ratio of direct to indirect time on face to face contact. It is
assumed that 41% of a nurses time is spent on direct patient with the remaining 59% spent on non-patient
activities such as administration and paperwork
b. Based on a Foundation House officer 2 on a 48 hour week
Betamimetics
Two betamimetics were compared to obtain the lowest cost in class:
i. Terbutaline sulfate
A summary of the doses and mode of administration used in the included studies is shown
below:
• How: oral 5 to 10 mg every 4 to 6 hours
• Laohapojanart: 10 microgram per minute with an increase of 5 microgram per minute
every 10 minutes if required until 25 microgram per minute reached
• Mavaldi: 0.25 mg of loading dose subcutaneously; same dose repeated every 45 minutes
• Motazadian: 250 microgram subcutaneous followed by the same dose every 45 minutes
For the purposes of this costing dosing was based on the following
(http://www.fpnotebook.com/ob/Pharm/Trbtln.htm):
Intravenous:
1.Start: 10 microgram/minute
2.Increase rate by 5 microgram per minute every 10 minutes
3.Maximum: 25 microgram per minute
4.Once controlled, decrease dose 5 microgram every 30 minutes
5.Titrate dose down to lowest effective dose
It was assumed that a total of 50.55 mg is administered over a period of 48 hours.
• 1–10 minutes = 10×10 microgram = 0.1 mg
• 31 minutes to 24 hours = 1410×25 microgram = 35.25 mg
• 24 hours to 24 hours 30 minutes = 30×20 microgram = 0.6 mg
• 24 hours 30 minutes to 25 hours = 30×15 microgram = 0.45 mg
• 25 hours to 48 hours = 1380×10 microgram = 13.8 mg
From the BNF (accessed 30 March 2015), the price of 1 ml ampule of terbutaline sulfate 500
microgram/ml for injection was £0.43. The treatment cost was calculated for terbutaline
sulfate as shown in Table 116.
Table 110: Calculation of treatment cost of terbutaline sulfate

Unit cost Unit cost per mg
Injection (0.5 mg) £0.43 £0.86
Total infused 50.55 mg £43.37
Staff Minutes Cost

384

Nurse 15 £30.00
Doctor 15 £10.25
Total Cost £83.72
ii. Salbutamol
below:
• Jannet: dilution of 2.5 mg in a 500 ml 5% weight per volume glucose solution, with an
initial flow rate of 30 ml/hour, 0.15 mg/hour
• Jannet: IV infusion; initial dose 12 microgram per minute. The dose was increased by 6
microgram per minute at 10 minute intervals up to maximum of 50 microgram per minute
until the desired effect was achieved
• Motazadian: IV bolus 0.1 mg followed by the same boluses every 5 minutes
It was assumed that a total of 79 mg is administered over a period of 48 hours:

• 71 minutes to 24 hours = 1370×45 microgram = 61.65 mg
• 24–30 hours = 360×22.5 microgram = 8.1 mg
From the BNF (accessed 30 March, 2015), the price of a 5 ml ampule of 1 mg/1 ml
salbutamol (as sulfate) solution for intravenous infusion was £2.48. The treatment cost for
salbutamol was calculated as shown in Table 117.
Table 111: Calculation of treatment cost of salbutamol

Solution for IV infusion £2.48 £0.50
Drug Maximum Total drug cost
IV infusion 79 mg £39.18
Staff Minutes Cost
Nurse 15 £30.00
Doctor 15 £10.25
Total cost £79.43
Calcium channel blockers

Two calcium channel blockers were compared to obtain the lowest cost in class:
i. Nifedipine
below:

385
• Al-quattan: 30 mg of oral loading dose followed by 20 mg orally after 120 minutes

• Al-omari: 10 mg orally, by chewing every 15 minutes, maximum dose 40 mg in the first
hour then 10 mg every 4–6 hours
• Haghighi: 10 mg capsule given sublingually repeated every 20 minutes (up to maximum
of 40 mg during the first hour of treatment)
• Kashanian: 10 mg orally up to maximum of 4 doses. If contractions subsided then 20 mg
every 6 hours for the first 24 hours; 20 mg every 8 hours for the second 24 hours; 10 mg
every 8 hours for the next 24 hours. (If the contractions contined or blood pressure was
below 90/50 mmHg, administration of nifedipine discontinued.)
• Klauser: 30 mg of oral loading dose followed by 20–30 mg orally every 4 to 6 hours
• Koks: 30 mg of oral loading dose followed by 20 mg 2 to 4 times daily
• Laohapojanart: 10 mg crushed and swallowed, 10 mg every 20 minutes with maximum
40 mg in the first hours. After that 20 mg every 4 hours up to 72 hours.
• Mavaldi: 30 mg of oral loading dose followed by 20 mg orally after 90 minutes
• Papatsonis: 10 to 40 mg of oral loading dose at the first hour followed by maintenance
dose of 60–160 mg of slow releasing daily until 34 weeks
• Taherian: Start with 10 mg orally repeated every 20 minutes (maximum dose 40 mg in
the first hour)
For the purpose of this costing it was assumed that 40 mg is given as a loading dose
followed by 20 mg every 6 hours for a period of 72 hours (240 mg). From the BNF (accessed
30 March 2015), the price of a 90 capsule pack of 10 mg nifedipine was £7.30. The treatment
cost for nifedipine was calculated as shown in Table 118.
Table 112: Calculation of treatment cost of nifedipine

90 capsules £7.30 £0.01
Drug Dose Cost
Loading dose 40 £0.32
Subsequent dose 240 £1.95
Staff Minutes Cost
Nurse 5 £10.00
Doctor 5 £3.42
Total cost £15.69
ii. Nicardipine
below:
• Jannet: dilution of 50 mg in a 500 ml 5% weight per volume glucose solution with an

initial flow rate of 30 ml/hour, 3 mg/hour
• Larmon: 40 mg oral dose, after 2 hours if needed further 20 mg oral dose
For the purposes of this costing it was assumed that an initial dose of 40 mg is followed by a
subsequent dose of 20 mg. From the BNF (accessed 30 March 2015), the price of a 56
capsule pack of 20 mg nicardipine was £6.00. The treatment cost for nicardipine was
calculated as shown in Table 119.

386
Table 113: Calculation of treatment cost of nicardipine

56 capsules £6.00 £0.005
Drug Dose Cost
Capsules 60 mg £0.32
Staff Minutes Cost
Nurse 5 £10.00
Doctor 5 £3.42
Total cost £13.74
Magnesium sulfate
A summary of the doses and mode of administration that was used in the included studies is
shown below:
• Borna: loading dose 4 to 6 g in 20% solution followed by an continuous infusion of 2 to 4

g/hour
• Cotton: loading dose 4 g IV over 15 minutes then followed by a continuous IV infusion of
2 g/hour
• El-Sayed: 4 g IV bolus, then at the rate of 2–4 g/hour
• Haghighi; loading dose 6 g IV followed by an infusion of 2 g/hour increase to maximum 4
g/hour
• Klauser: loading dose 6 g IV over 20 minutes followed by an infusion of 4-6 g/hour
• Larmon: loading dose 6 g IV (2 g/hour) increase to max 4 g/hour
• Mcwhorter: loading dose 4 to 6 g in 20% solution followed by an continuous infusion at
the rate of 2 to 4 g/hour
• Taherian: loading dose 4 g IV over 15 minutes followed by an IV infusion of 2–3 g/hour
• Wilkins: loading dose 4 g IV over 15 minutes followed by an IV infusion of 2 g/hour
For the purpose of this costing it was assumed that 4 g is given as a loading dose followed
by 2 g/hour by intravenous infusion for a period of 48 hours (96 g). From the BNF (accessed
30 March 2015), the price of a 20 ml (4 g) ampule of magnesium sulfate for injection was
£16.98 and the cost of a 2 ml (1 g) ampule of magnesium sulfate was £1.18. The treatment
cost for magnesium sulfate was calculated as shown in Table 120.
Table 114: Calculation of treatment cost of magnesium sulfate

20 ml (4 g) ampule £16.98 £0.004
2 ml (1 g) ampule £1.15 £0.001
Drug Dose Cost
Loading dose 4000 mg £16.98
Continuous infusion 96,000 mg £110.40
Staff Minutes Cost
Nurse 15 £30.00
Doctor 15 £10.25
Total cost £167.63

387
Oxytocin receptor blockers

below:
• Al-omari: bolus 6.7 mg IV over 1 minute followed by an IV infusion of 18 mg/hour for 3

hours followed by 6 mg/hour for 24–48 hours
It was assumed that the dosing was as indicated by the BNF (accessed 30 March 2015),
intravenous injection, initially 6.75 mg over 1 minute, then by intravenous infusion 18
mg/hour for 3 hours, then 6 mg/hour for 45 hours. From the BNF (accessed 30 March 2015),
the price of a 0.9 ml (6.75 mg) vial of atosiban (as acetate) for injection was £18.41 and the
cost of a 5 ml vial (7.5 mg/ml) for intravenous infusion was £52.82. The treatment cost for
atosiban was calculated as shown in Table 121.
Table 115: Calculation of treatment cost of oxytocin receptor blockers

Injection vial (6.75 mg) £18.41 £2.73
IV vial (37.5 mg) £52.82 £1.41
Drug Dose Cost
Intravenous injection 6.75 £18.41
IV first infusion first 3 hours 54 £76.06
IV infusion next 45 hours 270 £380.30
Staff Minutes Cost
Nurse 15 £30.00
Doctor 15 £10.25
Total cost £515.02
Nitrates
below:
• El-sayed: 100 microgram IV bolus, then at a rate of 1 to 10 microgram/kg/minute

• Smith: transdermal patch
For the purpose of this costing it was assumed that women would be given 4 ‘10’ patches
releasing 10 mg/24 hours of glyceryl trinitrate over the course of 48 hours. From the BNF
(accessed 30 March 2015) the price of a 28 pack ‘10’ patch of glyceryl trinitrate is £14.06.
The treatment cost for nitrates was calculated as shown in Table 122.
Table 116: Calculation of treatment cost of nitrates

‘10’ patch £14.06 £0.05
Drug Dose Cost
4 patches 40 mg £2.01
Staff Minutes Cost

388

Nurse 5 £10.00
Doctor 5 £3.42
Total cost £15.43
Prostaglandin inhibitors
Five prostaglandin inhibitors were compared to obtain the lowest cost in class:
i. Indomethacin
below:
• Besinger: 50 mg orally initially then 25 to 50 mg orally every 4 hours until the contraction
ceased then 25 mg maintenance therapy orally every 4–6 hours
• Kashanian: 100 mg suppository, a repeat administrated 1 hour later with the same dose;
maximum dose 200 mg daily
• Klauser: 100 mg suppository, a repeat administrated 2 hours later with the same dose,
followed by 50 mg orally every 6 hours for 12 hours
• Spearing: 100 mg suppository, a repeat administrated 12 hours later with the same dose,
followed by 25 mg orally every 6 hours for 48 hours
For the purposes of this costing it was assumed that a 100 mg suppository is administered
followed by a repeat administration with the same dose. This is then followed by 25 mg orally
every 6 hours for 48 hours. From the BNF (accessed 30 March 2015) the price of a 10 pack
of 100 mg indomethacin suppositories is £17.61 and the price of a 28 capsule pack of 25 mg
indomethacin is £1.32. The treatment cost for indomethacin was calculated as shown in
Table 123.
Table 117: Calculation of treatment cost of indomethacin

28 capsules £1.32 £0.002
Suppositories £17.61 £0.18
Drug Dose Cost
Suppositories 200 mg £35.22
Staff Minutes Cost
Nurse 5 £10.00
Doctor 5 £3.42
Total £49.01
ii. Celecoxib
below:
• Borna: 100 mg orally twice daily

For the purposes of this costing it was assumed that the women receive 100 mg orally twice
daily over 48 hours (400 mg in total). From the BNF (accessed 30 March 2015) the price of a

389
60 capsule pack of 100 mg celecoxib is £21.55. The treatment cost for celecoxib was
calculated as shown in Table 124.
Table 118: Calculation of treatment cost of celecoxib

60 capsules £21.55 £0.004
Drug Dose Cost
Staff Minutes Cost
Nurse 5 £10.00
Doctor 5 £3.42
Total cost £14.85
iii. Sulindac
For the purposes of this costing it was assumed that the dose is 200 mg every 12 hours for
48 hours (800 mg in total). From the BNF (accessed 30 March 2015) the price of a 56 tablet
pack of 200 mg sulindac is £38.29. The treatment cost for sulindac was calculated as shown
in Table 125.
Table 119: Calculation of treatment cost of sulindac

56 tablets £38.29 £0.003
Drug Dose Cost
Tablets 800 mg £2.74
Staff Minutes Cost
Nurse 5 £10.00
Doctor 5 £3.42
Total cost £16.15
iv. Ketorolac
For the purposes of this costing it was assumed that ketorolac is administered
intramuscularly as a 60 mg loading dose followed by 30 mg every 6 hours for 48 hours. From
the BNF (accessed 30 March 2015) the price of a 1 ml ampule of 30 mg/ml ketorolac
trometamol for injection was £1.07. The treatment cost for ketoreloc was calculated as
shown in Table 126.
Table 120: Calculation of treatment cost of ketorolac

Injection (1 ml ampule) £1.07 £0.036
Drug Dose Cost
Loading dose 60 mg £2.14
Repeat doses 240 mg £8.56
Staff Minutes Cost

390

Nurse 15 £30.00
Doctor 15 £10.25
Total cost £50.95
v. Mefenamic acid
A study (Mital 1992) reported using Mefenamic acid 500 mg 3 times daily, although the
duration was not specified in the abstract. For the purposes of this costing this dosing was
used and it was assumed this was administered over 48 hours. From the BNF (accessed 30
March 2015) the price of a 28 tablet pack of 500 mg mefenamic acid is £2.62. The treatment
cost for mefenamic acid was calculated as shown in Table 127.
Table 121: Calculation of treatment cost of mefenemic acid

28 tablets £2.62 £0.0002
Drug Dose Cost
Tablets 3000 mg £0.56
Staff Minutes Cost
Nurse 5 £10.00
Doctor 5 £3.42
Total cost £13.98
A summary of all the costs of the different tocolytic by class is given in Table 128 and
indicates the lowest cost option for use in the model.
Table 122: Summary of tocolytic treatment costs

Magnesium sulfate
Mefenemic acid
Nitroglycerine
Indomethacin
Lowest cost
Nicardipine
Salbutamol
Terbutaline
Nifedipine
Celecoxib
Ketorolac
Atosiban
Sulindac
Betamimetics £85 £79 – – – – – – – – – – £79

Calcium channel – – £16 £14 – – – – – – – – £14
blockers
Magnesium sulfate – – – – £16 – – – – – – – £168
8
Nitrates – – – – – £15 – – – – – – £15
Oxytocin receptor – – – – – – £515 – – – – – £515
blockers
Prostaglandin – – – – – – – £49 £15 £16 £51 £14 £14
inhibitors
Costs rounded to nearest pound

391
16.3.2.7 Costs relating to adverse outcomes

In addition to the costs of treatment it is important that downstream costs are also factored
into the analysis as more effective treatments are likely to lead to lower costs arising from
adverse outcomes. The costs relating to adverse outcomes used in the base-case analysis
are shown in Table 129.
Table 123: Costs of adverse outcomes

Outcome Cost Standard error Distribution
Deatha £0 £106 Deterministic
RDSb £7,000 £36 Deterministic
IVHc £23,700 - Deterministic
a. NHS Reference Costs 2012/13, XB03Z Paediatric critical care, intensive care, advanced
b. NHS Reference Costs 2011/12, XB01Z Paediatric critical care, intensive care ECMO/ECLS
c. It was assumed that IVH would have the same cost as ICH. It was additionally assumed that Grade III
and Grade IV ICH would be similar in cost to cerebral palsy. A European paper (Kruse 2009) estimated in year
2000 prices that the lifetime healthcare costs for cerebral palsy using an annual discount rate of 5% was
€66,155 for men and €65,288. The mid-point of this estimate was used and converted into GBP using an
exchange rate of £0.83 = €1 (http://www.exchangerates.org.uk/ - accessed 26/03/2014). It was then converted
into 2011/12 prices using the HCHS (The Hospital & Community Health Services) Index. One study (Alvarez
1994) suggested that 30% of ICH is of severity Grade III and Grade IV and therefore the cost of ICH was
estimated as 0.3 x £79,000
16.3.2.8 Quality adjusted life years (QALYs)

A lifetime QALY loss was assigned to each of the 3 outcomes included in the model
(mortality, RDS and IVH). To calculate the QALY loss from mortality it was assumed that the
maximum QALY loss associated with neonatal/perinatal death would be 22.7 QALYs. This is
based on the current life expectancy in England and Wales of 80 years and the simplifying
assumption that each year would be lived with a health state utility of 0.82, a value based on
a UK population norm (Kind 1982) and with an annual discount rate of 3.5% applied to that
health state utility. However, in practice a lower QALY gain is likely to arise from averting
mortality at earlier gestational ages as births at these ages are associated with shorter life
expectancy and greater morbidity, other than that already captured by RDS and IVH.
To proxy this effect, data on 1 year survival was used to estimate the proportion of averted
neonatal/perinatal deaths where the baby would not survive the first year of life. It was
assumed that deaths within the first year would not have any QALY associated with them but
that all those surviving the first year would receive the maximum 22.7 QALYs. Therefore, the
overall QALY gain from averting a neonatal/perinatal loss is the weighted average of the
QALY gain experienced by those surviving the first year (22.7 QALYs) and the zero QALY
from deaths occurring within that first year. The proportions surviving the first year varies with
gestational age as shown in Figure 75 and the QALY loss assigned to neonatal/perinatal
mortality by gestational age is shown in Table 130.
Table 124: QALY loss from neonatal/perinatal mortality by gestational age

Gestational age
(weeks) Survive 1st Year Death 1st year Maximum QALY Weighted QALY
24 0.878 0.122 22.7 19.92
25 0.920 0.080 22.7 20.89
26 0.937 0.063 22.7 21.27
27 0.955 0.045 22.7 21.69
28 0.977 0.022 22.7 22.18
29 0.988 0.012 22.7 22.44

392
Gestational age
(weeks) Survive 1st Year Death 1st year Maximum QALY Weighted QALY
30 0.996 0.004 22.7 22.61
31 0.992 0.008 22.7 22.52
32 0.992 0.008 22.7 22.53
33 0.995 0.005 22.7 22.58
34 0.996 0.004 22.7 22.61
Figure 75: Mortality rate in first year by gestational age at birth
Source: ONS 2011
The QALY losses from RDS and IVH not leading to neonatal/perinatal mortality are as shown
in Table 131.
Table 125: QALY loss from RDS and IVH

Outcome QALY loss
RDS 3.85
IVH 4.5
The QALY loss from RDS values was essentially a ‘dummy’ value reflecting the highly
variable prognosis. Therefore it is reasonable to posit a significant QALY loss to capture the
proportion having poor long-term outcomes but a proportion would also have good long-term

393
outcomes and therefore the QALY loss can expected to be considerably less than that
arising from mortality. The QALY loss from IVH was based on the value we used for ICH in
the model looking at the cost effectiveness of magnesium sulfate for neuroprotection
In the base-case analysis it is assumed that the NHS has a willingness to pay of £20,000 per
QALY.
16.3.2.9 Probabilistic sensitivity analysis

It is usual and good practice when reporting a relative treatment effect to provide confidence
intervals in addition to the point estimate. This is because there is uncertainty around the
point estimate due to sampling error. As the sample size is increased this uncertainty is
reduced which, everything else being equal, is reflected in narrower confidence intervals.
Similarly, in health economic analysis it is important to take into account the uncertainty
around model inputs. This can sometimes be achieved through 1-way sensitivity analysis,
where 1 input value at a time is varied in order to assess what change that input has on the
model’s results. However, although that can often provide useful insights into what inputs are
driving the model’s results, it is inadequate to address the uncertainty which exists
simultaneously across all model inputs.
Probabilistic sensitivity analysis, using Monte Carlo simulation techniques, allows for
uncertainty across all model inputs to be addressed. Simulation involves running the model
many times. In each simulation, rather than taking an input’s point estimate it is sampled
from its probability distribution. For inputs that are based on a large sample the probability
distribution will be relatively narrow and the sampled inputs will reflect that. Conversely, input
values derived from a small sample will have a relatively wide probability distribution.
For example, take the baseline risk of death in this model at a gestational age of 30 weeks.
From Table 110 it can be seen that the point estimate is 0.104. In the probabilistic sensitivity
analysis this is sampled from a beta distribution which is constrained to lie between 0 and 1
as a mortality rate must do. Two parameters, alpha and beta, are needed to describe a
particular beta distribution. Alpha is simply the number of events, in this case 169 deaths.
Beta is the number of non-events or surviving babies and is 1454 at 30 weeks’ gestation.
Using a random number generator, computer software (Microsoft Excel® for example) can
be used to sample from the beta probability distribution for this combination of alpha and
beta. Below, 10 samples from this distribution are shown.
Table 126: Example of Monte Carlo simulation of baseline mortality risk

Sample number Sampled value
1 0.101
2 0.114
3 0.102
4 0.096
5 0.103
6 0.107
7 0.099
8 0.112
9 0.095
10 0.095

394
If these values were being sampled as part of a Monte Carlo simulation, then these are the
values that would be used for baseline mortality in the first 10 simulations of the model. The
average of these 10 sampled values is 0.103 which is close to the point estimate. This is to
be expected and consequently the larger the number of simulations, the less role for
sampling error to affect the result.
The NMA was used to simulate 100,000 log odds ratios for each treatment class relative to
baseline from the posterior distribution of the probability of event (IVH, RDS or neonatal
mortality). For each outcome, these values are a random sample from the joint distribution of
the probabilities and therefore maintain any correlation between them. Using a mathematical
transformation these log odds ratios are then converted into absolute probabilities. For the
RDS outcome, nitrates were not included in the NMA and therefore it was conservatively
assumed that they would have the same risk of RDS as baseline.
When the user runs the probabilistic sensitivity analysis, they can stipulate the number of
simulations to run up to a maximum of 100,000. If a number less than a 100,000 is chosen
the model will run a sequence of the 100,000 previously generated simulations but starting at
a random point.
16.3.2.10 Results
Table 133 shows the results of the probabilistic sensitivity analysis based on 10,000
simulations for a gestational age of 24 weeks for the base-case analysis.
Table 127: Base case PSA result based on 10,000 simulations for a gestational age of
24 weeks
Mean Mean net Probability
Treatment Mean cost QALY benefit cost effective ICER
Nitrates −£962 1.136 £23,685 0.36 n/a
Calcium channel −£916 2.467 £50,246 0.34 £35 per
blockers QALY
Prostaglandin inhibitors −£277 −0.39 −£7,525 0.04 Dominated
Magnesium sulfate −£221 −1.516 −£30,101 0.01 Dominated
Betamimetics −£206 0.011 £425 0.01 Dominated
Standard care £0 0 £0 0.04 Dominated
Oxytocin receptor £270 1.613 £31,985 0.20 Dominated
blockers
a. Mean costs and QALYs are calculated relative to standard care
Using the mean net benefit would suggest that calcium channel blockers are the most cost-
effective treatment; however, nitrates have a marginally higher probability of being cost
effective.
Figure 76 shows a plot of 1000 simulations of the base-case analysis on the cost-
effectiveness plane with incremental costs and QALYs shown relative to standard care
(origin) in women at 24 weeks of pregnancy. Figure 77 is a similar plot of the same 1000
simulations but restricted to the 3 most cost-effective treatments as assessed by net mean
benefit.
The cost-effectiveness acceptability curve is shown in Figure 78. This shows the probability
of different treatments being cost effective as the willingness to pay for a QALY is varied and
therefore at a willingness to pay of £20,000 per QALY, the probabilities shown in Figure 78
correspond to the probabilities in Table 133.

395
Figure 76: Cost-effectiveness plane for the base case analysis showing all treatments
in the analysis for women at 24 weeks’ gestation
Figure 77: Cost-effectiveness plane for the base case analysis, restricted to the 3 most
cost-effective treatments for women at 24 weeks’ gestation
Figure 79 and Figure 80 respectively show how the net mean benefit and the probability that
each treatment is cost effective varies with gestational age. This shows that although
treatment becomes relatively less cost effective with increasing length of gestation, as shown
by declining mean net benefit, that calcium channel blockers continues to be a cost-effective
treatment for all diagnosed and suspected preterm births between 24 and 34 weeks at a
willingness to pay threshold of £20,000 per QALY.

396
Figure 78: Cost-effectiveness acceptability curve for base-case analysis and at a

Figure 79: Chart to show net mean benefit by treatment and gestational age in the
base case analysis

397
Figure 80: Chart to show probability a treatment is cost effective (at a willingness to
pay of £20,000 per QALY) by gestational age in the base case analysis
16.3.2.11 Sensitivity analysis

A number of sensitivity analyses are presented below to illustrate how sensitive the model’s
results are to some key assumptions and input parameters.
16.3.2.12 Estimating the effect of treatment on mortality using the NMA undertaken
on perinatal mortality
In the base-case analysis the effect of treatment on mortality was estimated using the NMA
on neonatal mortality. However, an NMA was also undertaken on the outcome of perinatal
mortality, although for reasons outlined earlier it was not possible to combine neonatal and
perinatal mortality outcomes. In this sensitivity analysis the effect of treatment on mortality is
estimated from the relative treatment effects derived from the NMA on perinatal mortality and
is based on 10,000 simulations.
Table 128: PSA result based on 10,000 simulations for a gestational age of 24 weeks
with treatment effect on mortality estimated using data on relative treatment
effect size from the NMA on perinatal mortality
Mean Mean Mean net Probability
Treatment cost QALY benefit cost-effective ICER
Nitrates −£945 8.672 £174,381 0.89 Dominates
Calcium channel blockers −£905 1.579 £32,491 0.03 Dominated
Prostaglandin inhibitors −£250 1.742 £35,086 0.04 Dominated
Magnesium sulfate −£205 −0.386 −£7,521 0.01 Dominated
Betamimetics −£198 0.117 £2,354 0.00 Dominated
Standard care £0 0 £0 0.00 Dominated
Oxytocin receptor blockers £283 0.965 £19,018 0.02 Dominated

398
Table 134 reports the results of this sensitivity analysis in detail for women at 24 weeks’
gestation with the cost-effectiveness acceptability curve for this analysis shown in Figure 81.
This sensitivity analysis suggests that nitrates are the most cost-effective treatment
dominating all alternatives, with the lowest mean costs and highest mean QALYs across the
simulation. Allowing for uncertainty in model inputs, and treatment effect size in particular,
they also have a very high probability of being the most cost-effective treatment alternative.
Figure 82 shows a plot of 1000 simulations of the base-case analysis on the cost-
effectiveness plane with incremental costs and QALYs shown relative to standard care
(origin) in women at 24 weeks’s gestation. Figure 83 is a similar plot of the same 1000
simulations but restricted to the 4 most cost-effective treatments as assessed by net mean
benefit.
Figure 84 and Figure 85 show, respectively, how the net mean benefit and the probability
that each treatment is cost effective varies with gestational age. These suggest that although
the relative cost effectiveness of treatment diminishes with increasing gestational age,
nitrates continue to have a very high probability of being the most cost-effective treatment
across all gestational ages considered in the model.
Figure 81: Cost-effectiveness acceptability curve using the NMA on neonatal mortality
to estimate the relative treatment effect on mortality at a gestational age of
24 weeks

399
Figure 82: Cost-effectiveness plane with the relative treatment effect on mortality
based on the NMA for perinatal mortality and showing all treatments in the
analysis for women at 24 weeks’ gestation

400
Figure 83: Cost-effectiveness plane with the relative treatment effect on mortality
based on the NMA for perinatal mortality, restricted to the 4 most cost-
effective treatments for women at 24 weeks’ gestation
Figure 84: Chart to show net mean benefit by treatment and gestational age using the
NMA on neonatal mortality to estimate the relative treatment effect on
mortality

401
Figure 85: Chart to show probability a treatment is cost-effective (at a willingness to

pay of £20,000 per QALY) by gestational age using the NMA on neonatal
mortality to estimate the relative treatment effect on mortality
16.3.2.13 Varying the QALY loss from IVH and RDS

In this sensitivity analysis the QALY loss from IVH and RDS were both reduced to 0.5
QALYs. Calcium channel blockers remained the most cost-effective treatment at 24 weeks
and 34 weeks respectively with the highest net mean benefit. At 24 weeks the ICER for
calcium channel blockers was £20 per QALY relative to nitrates which dominated other
treatment alternatives at that gestational age. At 34 weeks calcium channel blockers had an
ICER of £25 per QALY relative to standard care while dominating all other treatments.
In addition, another sensitivity analysis was performed in which the QALY loss from RDS and
IVH was increased to 10 QALYs. At 24 weeks the ICER for calcium channel blockers was
£24 per QALY relative to nitrates which dominated other treatment alternatives at that
gestational age. At 34 weeks calcium channel blockers had an ICER of £24 per QALY
relative to standard care while dominating all other treatments.
16.3.3 Discussion
One of the advantages of an economic analysis of this type is that it allows benefits across
different outcome measures to be synthesised into a single measure of effect. In the base-
case analysis calcium channel blockers were found to be the most cost-effective intervention
with the highest net mean benefit across 10,000 Monte Carlo simulations and a very low
ICER relative to non-dominated treatment alternatives. Although the model showed that the
cost effectiveness declined with increasing gestational age, calcium channel blockers
remained the most cost-effective treatment across all gestational ages (see Figure 76).
Interestingly, although calcium channel blockers had a similar probability of being the most
cost effective compared with nitrates, it was the latter which had the highest probability of
being the most cost-effective treatment. This apparent discrepancy reflected the wide
confidence intervals reported for the nitrates treatment log odds ratios in the NMA, as can be
seen in Figure 76 and Figure 77.
The base-case model did not address the possible costs of diagnosis which can cause the
costs of achieving particular outcomes to be underestimated. Nor did it include the costs of

402
hospitalisation because, with the possible exception of standard care, this cost would be
identical across the different treatment alternatives. However, this model was used to inform
the model that did consider the diagnosis of preterm labour in women with suspected
preterm labour and intact membranes (see Section 16.2 ). That model found that treatment
remained cost effective even when including diagnostic costs, hospitalisation costs and the
treatment of false positives. This finding is consistent with the net mean benefits observed for
calcium channel blockers (see Figure 79).
Sensitivity analysis suggests that the model is not sensitive to the QALY values used for
RDS and IVH outcomes. Unlike mortality the model does not vary the QALY loss from RDS
and IVH with gestational age, partly because of the difficulty with estimating how these QALY
losses would vary with gestational age. Nevertheless, from a clinical perspective, the
prognosis from these outcomes is related to gestational age, with lower gestational age at
birth being associated with worse prognosis. However, this sensitivity analysis suggests that
an approach which varied QALY loss by age for these outcomes would have a negligible
impact on results.
It is also apparent from the differences in net mean costs that the model’s results would not
be particularly sensitive to differences in treatment costs, especially at the lower gestational
ages, as the treatment cost represents only a small proportion of the difference in net mean
benefit between the various treatment alternatives.
One sensitivity analysis which did have a huge bearing on the model’s results was basing the
relative treatment effect on mortality on the perinatal mortality NMA rather than the neonatal
mortality NMA. This difference was driven by the NMA on perinatal mortality which found a
very high probability that nitrates were the most effective treatment for this outcome. This
was then reflected in this sensitivity analysis where nitrates were estimated to have an 89%
probability of being the most cost-effective treatment. However, the Guideline Development
Committee discussed that this benefit of nitrates needs to be balanced against the potential
harm to the fetus. They noted that the number of trials including nitrates was small and that
therefore the results needed be interpreted with caution. In addition, there was no NMA data
for nitrates for the outcome of respiratory distress syndrome.
In terms of modelling the cost effectiveness of treatment, mortality across the entire
perinatal/neonatal mortality period is the real outcome of interest and that is reflected in how
the baseline mortality risk was calculated. However, due to overlaps in definition it was not
possible to generate a single NMA for the relative treatment effect covering the broader
period due to issues of double counting. The recommendations of the committee reflect that
neonatal mortality was selected as the more important outcome measure. Nevertheless, the
fact that the 2 analyses produced different results should be considered as a limitation of the
analysis and may reduce confidence in the result.
The cost effectiveness results were driven largely by the outcomes of the NMA. In the
methods section the choice of outcomes was explained as more than 3 outcomes were
evaluated with network meta-analyses. One NMA that was not included was delay in preterm
birth. In some respects this might be considered the best measure of tocolysis as the
benefits of tocolysis are predicated on them achieving such a delay. In that NMA
prostaglandin inhibitors came out as the most likely to be the effective treatment, had the
largest point estimate of relative treatment effect and was significantly better than placebo for
this outcome. However, it wasn’t included in the economic model as it wasn’t considered a
‘hard’ outcome and it would be anticipated that a treatment that was more successful in
delaying birth would also be the most successful at reducing adverse outcomes.
Nevertheless, it is worth noting that even for the outcome of delay in preterm birth,
prostaglandin inhibitors were not significantly better than calcium channel blockers, which
were also significantly better than placebo for delay in preterm labour.

403
There are some limitations with this analysis. Baseline risks are based on populations who
may be in receipt of treatment, including tocolysis to improve preterm outcomes, and
therefore may underestimate the risk of not providing treatment and may therefore similarly
underestimate the absolute treatment effect which would tend to cause cost effectiveness to
be underestimated.
The studies included in the network analyses did not look at pregnancies of less than 26
weeks but the committee considered it would be reasonable to extrapolate relative treatment
effects to women with gestations of 24–26 weeks. The model did not make a similar
extrapolation for women with pregnancies of over 34 weeks as the committee did not think
that any delay after this age would be considered sufficiently worthwhile. Furthermore,
although not considered explicitly in the model considering the diagnosis of preterm labour,
the evidence for gestations of 33 and 34 weeks was that treatment was less obviously cost
effective.
16.3.4 Conclusion
This model provides reasonably strong evidence that calcium channel blockers can be
considered as a cost-effective tocolytic treatment for women with diagnosed or suspected
preterm labour between 24 and 34 weeks’ gestation. They are additionally a relatively cheap
tocolytic and the Guideline Development Committee thought that they were often a first-line
treatment already.

404
Glossary and abbreviations
17 Glossary and abbreviations

17.1 Key terms
Term Definition
Symptoms of preterm A woman has presented before 37+0 weeks of pregnancy reporting
labour symptoms that might be indicative of preterm labour (such as
abdominal pain), but no clinical assessment (including speculum or
digital vaginal examination) has taken place.
Suspected preterm labour A woman is in suspected preterm labour if she has reported symptoms
of preterm labour and has had a clinical assessment (including a
speculum or digital vaginal examination) that confirms the possibility of
preterm labour but rules out established labour.
Diagnosed preterm labour A woman is in diagnosed preterm labour if she is in suspected preterm
labour and has had a positive diagnostic test for preterm labour.
Established preterm A woman is in established preterm labour if she has progressive
labour cervical dilatation from 4 cm with regular contractions (see the
definition of the established first stage of labour in the NICE guideline
on intrapartum care).
Preterm prelabour rupture A woman is described as having P-PROM if she has ruptured
of membranes (P-PROM) membranes before 37+0 weeks of pregnancy but is not in established
labour.
'Rescue' cervical cerclage Cervical cerclage performed as an emergency procedure in a woman
with premature cervical dilatation and often with exposed fetal
membranes.
17.2 Glossary
Term Definition
Abstract Summary of a study, which may be published alone or as an
introduction to a full scientific paper.
Accelerations (fetal An abrupt increase in fetal heart rate above baseline withtime from
monitoring) onset to peak of the acceleration less than 30 seconds and the total
lasting less than 2 minutes.
Acidosis An increased acidity in the blood and other body tissue.
Active management of A package of care comprising the following components:
the third stage routine use of drugs to cause contraction of the uterus
clamping and cutting of the cord
controlled cord traction after signs of separation of the placenta.
Allocation concealment The process used to prevent advance knowledge of group assignment
in a radnomised controlled trial (RCT). The allocation process should
be impervious to any influence by the individual making the allocation,
by being administered by someone who is not responsible for recruiting
participants.
Amniotic fluid The protective liquid surrounding the baby within the amniotic sac of a
pregnant woman.
Antenatal antibiotic The use of antibiotics to prevent infections in antenatal care.
prophylaxis

405
Term Definition
Antepartum haemorrhage Bleeding from or into the genital tract, occurring from 24+0 weeks of
pregnancy and prior to the birth of the baby.
Apgar score A measure of the physical condition of a newborn infant.
Applicability How well the results of a study or NICE evidence review can answer a
clinical question or be applied to the population being considered.
Arm (of a clinical study) Subsection of individuals within a study who receive one particular
intervention, for example placebo arm.
Association Statistical relationship between 2 or more events, characteristics or
other variables. The relationship may or may not be causal.
Attrition bias Systematic differences between comparison groups in withdrawals or
exclusion of participants from a study.
Available case analysis Analysis of data that is available for participants at the end of follow-up.
Before-and-after study A study that investigates the effects of an intervention by measuring
particular characteristics of a population both before and after taking
the intervention, and assessing any change that occurs.
Baseline The initial set of measurements at the beginning of a study (after run-in
period where applicable) with which subsequent results are compared.
Baseline variability (fetal Fluctuations in the fetal heart rate of more than 2 cycles per minute.
monitoring)
Bias Influences on a study that can make the results look better or worse
than they really are. (Bias can even make it look as if a treatment
works when it does not.) Bias can occur by chance, deliberately or as a
result of systematic errors in the design and execution of a study. It can
also occur at different stages in the research process, for example
during the collection, analysis, interpretation, publication or review of
research data. For examples see selection bias, performance bias,
information bias, confounding factor and publication bias.
Bishop score A prelabour scoring system based on clinical examination of the cervix,
to assist in predicting whether induction of labour will be required.
Bradycardia (fetal Slow heart rate; for the term fetus, this is defined as a heart rate of less
monitoring) than 110 beats per minute.
Breech (presentation) A baby which is so positioned in the womb that the buttocks or feet are
delivered first.
Bronchopulmonary A chronic lung disorder of infants and children.
dysplasia
Bulging membranes Amniotic membranes bulging through the opening of the cervix.
Caesarean section A surgical operation for delivering a baby by cutting through the wall of
the mother's abdomen. This may be an elective (planned) or
emergency procedure.
Cardiotocography Electronic recording of the fetal heart rate using either a Doppler
ultrasound transducer strapped to the woman’s abdomen, or an
electrode attached to the fetal scalp, plus a second toco transducer
strapped to the woman’s abdomen to record uterine contractions.
Carer (caregiver) Someone who looks after family, partners or friends in need of help
because they are ill, frail or have a disability.
Case-control study A study to find out the cause(s) of a disease or condition. This is done
by comparing a group of patients who have the disease or condition
(cases) with a group of people who do not have it (controls) but who
are otherwise as similar as possible (in characteristics thought to be
unrelated to the causes of the disease or condition). This means the

406
Term Definition
researcher can look for aspects of their lives that differ to see if they
may cause the condition. For example, a group of people with lung
cancer might be compared with a group of people the same age that
do not have lung cancer. The researcher could compare how long both
groups had been exposed to tobacco smoke. Such studies are
retrospective because they look back in time from the outcome to the
possible causes of a disease or condition.
Case series Report of a number of cases of a given disease, usually covering the
course of the disease and the response to treatment. There is no
comparison (control) group of patients.
Cephalic (presentation) A baby so positioned in the womb that the head is delivered first.
Cerebral palsy The general term for a number of neurological conditions that affect
movement and co-ordination.
Cervical cerclage A surgical treatment for cervical incompetence or insufficiency.
Cervical shortening A condition in pregnant women where the cervix becomes softer and
weaker than normal.
Cervical trauma Physical injury to the cervix including surgery; for example previous
cone biopsy (cold knife or laser), large loop excision of the
transformation zone (LLETZ) – any number) or radical diathermy.
Chorioamnionitis An inflammation of the fetal membranes (amnion and chorion) due to a
bacterial infection.
Chronic lung disease A general term for long-term respiratory problems in premature babies.
Clinical audit A systematic process for setting and monitoring standards of clinical
care. Whereas ‘guidelines’ define what the best clinical practice should
be, ‘audit’ investigates whether best practice is being carried out.
Clinical audit can be described as a cycle or spiral. Within the cycle
there are stages that follow a systematic process of establishing best
practice, measuring care against specific criteria, taking action to
improve care and monitoring to sustain improvement. The spiral
suggests that as the process continues, each cycle aspires to a higher
level of quality.
Clinical efficacy The extent to which an intervention is active when studied under
controlled research conditions.
Clinical effectiveness How well a specific test or treatment works when used in the 'real
world' (for example when used by a doctor with a patient at home),
rather than in a carefully controlled clinical trial. Trials that assess
clinical effectiveness are sometimes called management trials. Clinical
effectiveness is not the same as efficacy.
Clinician A healthcare professional who provides patient care; for example a
doctor, nurse or physiotherapist.
Cochrane Review The Cochrane Library consists of a regularly updated collection of
evidence based medicine databases including the Cochrane Database
of Systematic Reviews (reviews of randomised controlled trials
prepared by the Cochrane Collaboration).
Cognitive dysfunction The loss of intellectual functions such as thinking, remembering and
reasoning of sufficient severity to interfere with daily functioning.
Cohort study A study with 2 or more groups of people – cohorts – with similar
characteristics. One group receives a treatment, is exposed to a risk
factor or has a particular symptom and the other group does not. The
study follows their progress over time and records what happens.

407
Term Definition
Comorbidity A disease or condition that someone has in addition to the health
problem being studied or treated.
Concealment of allocation The process used to ensure that the person deciding to enter a
participant into a randomised controlled trial does not know the
comparison group into which that individual will be allocated. This is
distinct from blinding and is aimed at preventing selection bias. Some
attempts at concealing allocation are more prone to manipulation than
others, and the method of allocation concealment is used as an
assessment of the quality of a trial.
Confidence interval (CI) There is always some uncertainty in research. This is because a small
group of patients is studied to predict the effects of a treatment on the
wider population. The confidence interval is a way of expressing how
certain we are about the findings from a study, using statistics. It gives
a range of results that is likely to include the 'true' value for the
population. The CI is usually stated as '95% CI', which means that the
range of values has a 95 in 100 chance of including the 'true' value. For
example, a study may state that “based on our sample findings, we are
95% certain that the 'true' population blood pressure is not higher than
150 and not lower than 110”. In such a case the 95% CI would be 110
to 150. A wide confidence interval indicates a lack of certainty about
the true effect of the test or treatment – often because a small group of
patients has been studied. A narrow confidence interval indicates a
more precise estimate (for example if a large number of patients have
been studied).
Confounding factor Something that influences a study and can result in misleading findings
if it is not understood or appropriately dealt with. For example, a study
of heart disease may look at a group of people who exercise regularly
and a group who do not exercise. If the ages of the people in the 2
groups are different, then any difference in heart disease rates
between the 2 groups could be because of age rather than exercise.
Therefore age is a confounding factor.
Consensus methods Techniques used to reach agreement on a particular issue. Consensus
methods may be used to develop NICE guidance if there is not enough
good quality research evidence to give a clear answer to a question.
Formal consensus methods include Delphi and nominal group
techniques.
Continuous outcome Data with a potentially infinite number of possible values within a given
range. Height, weight and blood pressure are examples of continuous
variables.
Control group A group of people in a study who do not receive the treatment or test
being studied. Instead, they may receive the standard treatment
(sometimes called 'usual care') or a dummy treatment (placebo). The
results for the control group are compared with those for a group
receiving the treatment being tested. The aim is to check for any
differences. Ideally, the people in the control group should be as similar
as possible to those in the treatment group, to make it as easy as
possible to detect any effects due to the treatment.
Cord milking After delivery, the caregiver holds the umbilical cord and squeezes
blood down the cord into the baby.
Cord prolapse When the umbilical cord comes out of the uterus with or before the
presenting part of the fetus.
Corticosteroids Anti-inflammatory medicines.

408
Term Definition
Cost–benefit analysis Cost-benefit analysis is one of the tools used to carry out an economic
(CBA) evaluation. The costs and benefits are measured using the same
monetary units (for example UK pounds) to see whether the benefits
exceed the costs.
Cost–consequence Cost–consequence analysis is one of the tools used to carry out an
analysis (CCA) economic evaluation. This compares the costs (such as treatment and
hospital care) and the consequences (such as health outcomes) of a
test or treatment with a suitable alternative. Unlike cost–benefit
analysis or cost-effectiveness analysis, it does not attempt to
summarise outcomes in a single measure (like the quality adjusted life
year) or in financial terms. Instead, outcomes are shown in their natural
units (some of which may be monetary) and it is left to decision-makers
to determine whether, overall, the treatment is worth carrying out
Cost-effectiveness Cost-effectiveness analysis is one of the tools used to carry out an
analysis (CEA) economic evaluation. The benefits are expressed in non-monetary
terms related to health, such as symptom-free days, heart attacks
avoided, deaths avoided or life years gained (that is, the number of
years by which life is extended as a result of the intervention).
Cost-effectiveness model An explicit mathematical framework, which is used to represent clinical
decision problems and incorporate evidence from a variety of sources
in order to estimate the costs and health outcomes.
Cost–utility analysis Cost–utility analysis is one of the tools used to carry out an economic
(CUA) evaluation. The benefits are assessed in terms of both quality and
duration of life, and expressed as quality adjusted life years (QALYs).
See also Utility.
COX proportional hazard In survival analysis, a statistical model that asserts that the effect of the
model study factors (for example the intervention of interest) on the hazard
rate (the risk of occurrence of an event) in the study population is
multiplicative and does not change over time.
Credible interval (CrI) The Bayesian equivalent of a confidence interval.
Decelerations (fetal A decrease in the fetal heart rate below the baseline rate.
monitoring)
Decision analysis An explicit quantitative approach to decision-making under uncertainty
based on evidence from research. This evidence is translated into
probabilities, and then into diagrams or decision trees which direct the
clinician through a succession of possible scenarios, actions and
outcomes.
Decision to delivery The time taken between the decision to expedite a birth and the birth.
interval
Delayed cord clamping A birth practice where the umbilical cord is not clamped or cut until
after pulsations have ceased, or until after the placenta is delivered.
Diagnosed preterm labour A woman is in diagnosed preterm labour if she is in suspected preterm
labour and has had a positive diagnostic test for preterm labour.
Dichotomous outcomes Outcome that can take 1 of 2 possible values, such as dead/alive,
smoker/non-smoker, present/not present (also called binary data).
Dilated cervix Open cervix.
Discounting Costs and perhaps benefits incurred today have a higher value than
costs and benefits occurring in the future. Discounting health benefits
reflects individual preference for benefits to be experienced in the
present rather than the future. Discounting costs reflects individual

409
Term Definition
preference for costs to be experienced in the future rather than the
present.
Dominance A health economics term. When comparing tests or treatments, an
option that is both less effective and costs more is said to be
'dominated' by the alternative.
Drop-out A participant who withdraws from a trial before the end.
Economic evaluation An economic evaluation is used to assess the cost effectiveness of
healthcare interventions (that is, to compare the costs and benefits of a
healthcare intervention to assess whether it is worth doing). The aim of
an economic evaluation is to maximise the level of benefits – health
effects – relative to the resources available. It should be used to inform
and support the decision-making process; it is not supposed to replace
the judgement of healthcare professionals. There are several types of
economic evaluation: cost–benefit analysis, cost consequence
analysis, cost-effectiveness analysis, cost-minimisation analysis and
cost–utility analysis. They use similar methods to define and evaluate
costs, but differ in the way they estimate the benefits of a particular
drug, programme or intervention.
Early cord clamping Clamping carried out in the first 60 seconds after birth.
Effect (as in effect A measure that shows the magnitude of the outcome in 1 group
measure, treatment compared with that in a control group. For example, if the absolute risk
effect, estimate of effect, reduction is shown to be 5% and it is the outcome of interest, the effect
effect size) size is 5%. The effect size is usually tested, using statistics, to find out
how likely it is that the effect is a result of the treatment and has not
just happended incidentally.
Effectiveness How beneficial a test or treatment is under usual or everyday
conditions, compared with doing nothing or opting for another type of
care.
Efficacy How beneficial a test, treatment or public health intervention is under
ideal conditions (for example in a laboratory), compared with doing
nothing or opting for another type of care.
Established labour Labour is established when there is both of:
• regular painful contractions, and
• progressive cervical effacement (thinning) and dilatation
beyond 4 cm.
Established preterm A woman is in established preterm labour if she has progressive
labour cervical dilatation from 4 cm with regular contractions (see the
definition of established preterm labour in the NICE guideline on
intrapartum care).
Epidemiological study The study of a disease within a population, defining its incidence and
prevalence and examining the roles of external influences (for example
infection, diet) and interventions.
EQ-5D (EuroQol 5 A standardised instrument used to measure health-related quality-of-
dimensions) life. It provides a single index value for health status.
Equivalence study A trial designed to determine whether the response to 2 or more
treatments differs by an amount that is clinically unimportant. This is
usually demonstrated by showing that the true treatment difference is
likely to lie between a lower and an upper equivalence level of clinically
acceptable differences.
Evidence Information on which a decision or guidance is based. Evidence is
obtained from a range of sources including randomised controlled

410
Term Definition
trials, observational studies, expert opinion (of clinical professionals or
patients).
Exclusion criteria Explicit standards used to decide which studies should be excluded
(literature review) from consideration as potential sources of evidence.
Exclusion criteria (clinical Criteria that define who is not eligible to participate in a clinical study.
study)
Extended dominance If Option A is both more clinically effective than Option B and has a
lower cost per unit of effect when both are compared with a do-nothing
alternative, then Option A is said to have extended dominance over
Option B. Option A is therefore more cost effective and should be
preferred, other things remaining equal.
Extrapolation An assumption that the results of studies of a specific population will
also hold true for another population with similar characteristics.
Expectant management Waiting for events to take their natural course. This would usually
include observation of the woman and/or baby’s condition.
Exposed membranes When the cervix opens, the membranes are exposed.
Fetal blood sampling A technique to measure the level of acid–base status of the baby’s
blood. A sample of blood is taken from the baby’s scalp and either the
pH or lactate value is measured. It is used as an adjunct to
cardiotocography to help to clarify whether the baby is developing an
acidosis when may cause additional interventions to be required.
Fetal monitoring Method used to monitor the fetal heartbeat during labour.
Fetal fibronectin A fibronectin protein produced by fetal cells.
Fetal growth restriction A condition where growth of the fetus slows or ceases when it is in the
uterus.
Fixed-effect model In meta-analysis, a model that calculates a pooled effect estimate
using the assumption that all observed variation between studies is
caused by the play of chance. Studies are assumed to be measuring
the same
overall effect.
Follow-up Observation over a period of time of an individual, group or initially
defined population whose appropriate characteristics have been
assessed in order to observe changes in health status or health-related
variables.
Forest plot A graphical representation of the individual results of each study
included in a meta-analysis together with the combined meta-analysis
result. The plot also allows readers to see the heterogeneity among the
results of the studies. The results of individual studies are shown as
squares centred on each study’s point estimate. A horizontal line runs
through each square to show each study’s confidence interval. The
overall estimate from the meta-analysis and its confidence interval are
shown at the bottom, represented as a diamond. The centre of the
diamond represents the pooled point estimate, and its horizontal tips
represent the confidence interval.
Gestation The period of development in the uterus from conception until birth.
Gestational age A term used during pregnancy to describe how far along the pregnancy
is, measured in weeks.
Generalisability The extent to which the results of a study hold true for groups that did
not participate in the research. See also external validity.
Gold standard A method, procedure or measurement that is widely accepted as being
the best available to test for or treat a disease.

411
Term Definition
GRADE, GRADE profile A system developed by the GRADE Working Group to address the
shortcomings of present grading systems in healthcare. The GRADE
system uses a common, sensible and transparent approach to grading
the quality of evidence. The results of applying the GRADE system to
clinical trial data are displayed in a table known as a GRADE profile.
Gross motor dysfunction Dysfunction in the movement of the large muscles of the body.
Harms Adverse effects of an intervention.
Hazard ratio A hazard is the rate at which events happen, so that the probability of
an event happening in a short time interval is the length of time
multiplied by the hazard. Although the hazard may vary with time, the
assumption in proportional hazard models for survival analysis is that
the hazard in one group is a constant proportion of the hazard in the
other group. This proportion is the hazard ratio.
Health economics Study or analysis of the cost of using and distributing healthcare
resources.
Health related quality of A measure of the effects of an illness to see how it affects someone's
life (HRQoL) day-to-day life.
Heterogeneity The term is used in meta-analyses and systematic reviews to describe
when the results of a test or treatment (or estimates of its effect) differ.
Imprecision Results are imprecise when studies include relatively few patients and
few events and thus have wide confidence intervals around the
estimate of effect.
Inclusion criteria Explicit criteria used to decide which studies should be considered as
(literature review) potential sources of evidence.
Incremental cost The extra cost linked to using one test or treatment rather than another.
Or the additional cost of doing a test or providing a treatment more
frequently.
Incremental cost The difference in the mean costs in the population of interest divided by
effectiveness ratio (ICER) the differences in the mean outcomes in the population of interest for
one treatment compared with another
Incremental net benefit The value (usually in monetary terms) of an intervention net of its cost
(INB) compared with a comparator intervention. The INB can be calculated
for a given cost-effectiveness (willingness to pay) threshold. If the
threshold is £20,000 per QALY gained then the INB is calculated as:
(£20,000×QALYs gained) minus incremental cost.
Indirectness The available evidence is different to the review question being
addressed, in terms of population, intervention, comparison and
outcome (PICO).
Induction of labour A procedure where the midwife or doctor starts labour artificially by
using a membrane sweep, pessary or hormone drip.
Infant death The death of a child aged less than 1 year.
Insulin-like growth factor A protein that in humans is encoded by the IGFBP1 gene.
binding protein-1
Instrumental birth Birth in which the use of instruments is required.
Intellectual delay A disability characterised by significant limitations both in intellectual
functioning (reasoning, learning, problem solving) and in adaptive
behaviour, which covers a range of everyday social and practical skills.
Intention-to-treat analysis An assessment of the people taking part in a clinical trial, based on the
(ITT) group they were initially (and randomly) allocated to. This is regardless
of whether or not they dropped out, fully complied with the treatment or

412
Term Definition
switched to an alternative treatment. Intention-to-treat analyses are
often used to assess clinical effectiveness because they mirror actual
practice: that is, not everyone complies with treatment and the
treatment people receive may be changed according to how they
respond to it.
Intermittent auscultation Intermittent measurement of the fetal heart rate using a Doppler
ultrasound or a Pinard stethoscope.
Intervention In medical terms this could be a drug treatment, surgical procedure,
diagnostic or psychological therapy. Examples of public health
interventions could include action to help someone to be physically
active or to eat a more healthy diet.
Intracranial haemorrhage Bleeding within the skull cavity or brain.
Intraventricular Bleeding into the brain's ventricular system, where the cerebrospinal
haemorrhage fluid is produced and circulates.
Haematocrit The volume percentage of red blood cells in blood.
Hyperbilirubinaemia A condition in which there is too much bilirubin in the blood.
Hypoxia A condition in which the body or a region of the body is deprived of
adequate oxygen supply.
Kappa statistic A statistical measure of inter-rater agreement that takes into account
the agreement occurring by chance.
Labour The process of delivering a baby and the placenta, membranes and
umbilical cord from the uterus to the vagina to the outside world.
Length of stay The total number of days a participant stays in hospital.
Licence See ‘Product licence’.
Life years gained Mean average years of life gained per person as a result of the
intervention compared with an alternative intervention.
Likelihood ratio The likelihood ratio combines information about the sensitivity and
specificity. It tells you how much a positive or negative result changes
the likelihood that a patient would have the disease. The likelihood ratio
of a positive test result (LR+) is sensitivity divided by (1 minus
specificity).
Liquor The protective liquid contained by the amniotic sac of a pregnant
woman.
Long-term infant The rate of illness and disease in children.
morbidity
Loss to follow-up Patients who have withdrawn from the clinical trial at the point of follow-
up.
Low birth weight A birth weight of a live born infant of less than 2500 g (5 pounds 8
ounces).
Magnesium sulfate An inorganic salt containing magnesium, sulfur and oxygen, with the
formula MgSO4.
Markov model A method for estimating long-term costs and effects for recurrent or
chronic conditions, based on health states and the probability of
transition between them within a given time period (cycle).
McDonald suture A purse-string stitch used to cinch the cervix shut.
Mean An average value, calculated by adding all the observations and
dividing by the number of observations.
Mean difference In meta-analysis, a method used to combine measures on continuous
scales (such as weight), where the mean, standard deviation and

413
Term Definition
sample size in each group are known. The weight given to the
difference in means from each study (for example how much influence
each study has on the overall results of the meta-analysis) is
determined by the precision of its estimate of effect.
Mechanical ventilation A technique in which gas is moved toward and from the lungs through
an external device connected directly to the patient.
Median The value of the observation that comes half-way when the
observations are ranked in order.
Meta-analysis A method often used in systematic reviews. Results from several
studies of the same test or treatment are combined to estimate the
overall effect of the treatment.
Mid-trimester loss The death of a fetus in the second trimester (3–6 months of
pregnancy).
Minimal important Thresholds for clinical importance, which represent minimal important
difference (MID) differences for benefit or for harm; for example the threshold at which
drug A is less effective than drug B by an amount that is clinically
important to patients.
Multiple pregnancy A pregnancy in which there is more than 1 fetus.
Multivariate model A statistical model for analysis of the relationship between 2 or more
predictor (independent) variables and the outcome (dependent)
variable.
Necrotising enterocolitis A medical condition primarily seen in preterm infants, where portions of
the bowel undergo necrosis.
Neonatal death The death of a baby within the first 28 days of life.
Neonatal intensive care Intensive care for ill or preterm newborn infants.
Net monetary benefit The value (usually in monetary terms) of an intervention net of its cost.
(NMB) The NMB can be calculated for a given cost-effectiveness (willingness
to pay) threshold. If the threshold is £20,000 per QALY gained then the
NMB is calculated as: (£20,000×QALYs gained) minus cost.
Network meta-analysis Meta-analysis in which multiple treatments (that is, 3 or more) are
being compared using both direct comparisons of interventions within
randomised controlled trials and indirect comparisons across trials
based on a common comparator.
Neurodevelopmental Disabilities in the functioning of the brain that affect a child's behaviour,
delay memory or ability to learn.
Neonatal morbidity Health disorders in neonates occurring the first 4 weeks of life.
Nitrazine A pH indicator.
Number needed to treat The average number of patients who need to be treated to get a
(NNT) positive outcome. For example, if the NNT is 4, then 4 patients would
have to be treated to ensure 1 of them gets better. The closer the NNT
is to 1, the better the treatment. For example, if you give a stroke
prevention drug to 20 people before 1 stroke is prevented, the number
needed to treat is 20.
Observational study Individuals or groups are observed or certain factors are measured. No
attempt is made to affect the outcome. For example, an observational
study of a disease or treatment would allow 'nature' or usual medical
care to take its course. Changes or differences in 1 characteristic (for
example whether or not people received a specific treatment or
intervention) are studied without intervening. There is a greater risk of
selection bias than in experimental studies.

414
Term Definition
Odds ratio (OR) Odds are a way to represent how likely it is that something will happen
(the probability). An odds ratio compares the probability of something in
1 group with the probability of the same thing in another. An odds ratio
of 1 between 2 groups would show that the probability of the event (for
example a person developing a disease or a treatment working) is the
same for both. An odds ratio greater than 1 means the event is more
likely in the first group. An odds ratio less than 1 means that the event
is less likely in the first group. Sometimes probability can be compared
across more than 2 groups - in this case, 1 of the groups is chosen as
the 'reference category', and the odds ratio is calculated for each group
compared with the reference category. For example, to compare the
risk of dying from lung cancer for non-smokers, occasional smokers
and regular smokers, non-smokers could be used as the reference
category. Odds ratios would be worked out for occasional smokers
compared with non-smokers and for regular smokers compared with
non-smokers. See also confidence interval, relative risk, risk ratio.
Odds ratio (OR) Odds are a way to represent how likely it is that something will happen
(the probability). An odds ratio compares the probability of something in
1 group with the probability of the same thing in another. An odds ratio
of 1 between 2 groups would show that the probability of the event (for
example a person developing a disease, or a treatment working) is the
same for both. An odds ratio greater than 1 means the event is more
likely in the first group. An odds ratio less than 1 means that the event
is less likely in the first group. Sometimes probability can be compared
across more than 2 groups – in this case, 1 of the groups is chosen as
the 'reference category' and the odds ratio is calculated for each group
compared with the reference category. For example, to compare the
risk of dying from lung cancer for non-smokers, occasional smokers
and regular smokers, non-smokers could be used as the reference
category. Odds ratios would be worked out for occasional smokers
compared with non-smokers and for regular smokers compared with
non-smokers.
See Confidence interval, Relative risk, Risk ratio.
Opportunity cost The loss of other healthcare programmes displaced by investment in or
introduction of another intervention. This may be best measured by the
health benefits that could have been achieved had the money been
spent on the next best alternative healthcare intervention.
Outcome The impact that a test, treatment, policy, programme or other
intervention has on a person, group or population. Outcomes from
interventions to improve the public's health could include changes in
knowledge and behaviour related to health, societal changes (for
example a reduction in crime rates) and a change in people's health
and wellbeing or health status. In clinical terms, outcomes could
include the number of patients who fully recover from an illness or the
number of hospital admissions, and an improvement or deterioration in
someone's health, functional ability, symptoms or situation.
Researchers should decide what outcomes to measure before a study
begins.
P value The p value is a statistical measure that indicates whether or not an
effect is statistically significant. For example, if a study comparing 2
treatments found that 1 seems more effective than the other, the p
value is the probability of obtaining these results by chance. By
convention, if the p value is below 0.05 (that is, there is less than a 5%
probability that the results occurred by chance) it is considered that
there probably is a real difference between treatments. If the p value is

415
Term Definition
0.001 or less (less than a 1% probability that the results occurred by
chance), the result is seen as highly significant. If the p value shows
that there is likely to be a difference between treatments, the
confidence interval describes how big the difference in effect might be.
Performance bias Systematic differences between intervention groups in care provided
apart from the intervention being evaluated. Blinding of study
participants (both the recipients and providers of care) is used to
protect against performance bias.
Perinatal death Death occurring after 24 completed weeks of pregnancy and within 7
days after birth.
Periventricular A form of white-matter brain injury, characterised by the necrosis of
leucomalacia white matter near the lateral ventricles.
Physiological A package of care comprising the following components:
management of the third • no routine use of uterotonic drugs
stage
• no clamping of the cord until pulsation has stopped
• delivery of the placenta by maternal effort.
Preterm prelabour rupture A woman is described as having P-PROM if she has ruptured
of membranes (P-PROM) membranes before 37+0 weeks of pregnancy but is not in established
labour.
Placebo A fake (or dummy) treatment given to participants in the control group
of a clinical trial. It is indistinguishable from the actual treatment (which
is given to participants in the experimental group). The aim is to
determine what effect the experimental treatment has had – over and
above any placebo effect caused because someone has received (or
thinks they have received) care or attention.
Placebo effect A beneficial (or adverse) effect produced by a placebo and not due to
any property of the placebo itself.
Placental abruption A complication of pregnancy where the placenta has separated from
the uterus of the mother.
Placental alpha A human protein that was first isolated in amniotic fluid.
microglobulin
Planned preterm birth The planned birth of an infant before 37 weeks of pregnancy due to
medical complications.
Post-hoc analysis Statistical analyses that are not specified in the trial protocol and are
generally suggested by the data.
Postpartum haemorrhage Blood loss over 500 ml from the vagina following labour.
Power (statistical) The ability to demonstrate an association when one exists. Power is
related to sample size; the larger the sample size, the greater the
power and the lower the risk that a possible association could be
missed.
Preterm birth The birth of an infant before 37 weeks of pregnancy
Preterm labour Regular contractions of the uterus resulting in changes in the cervix
that start before 37 weeks of pregnancy.
Preterm prelabour rupture Rupture of the membranes before 37 weeks of pregnancy, occurring
of membranes before the onset of labour.
Pre-eclampsia A disorder of pregnancy characterised by high blood pressure and a
large amount of protein in the urine.
Primary care Healthcare delivered outside hospitals. It includes a range of services
provided by GPs, nurses, health visitors, midwives and other

416
Term Definition
healthcare professionals and allied health professionals such as
dentists, pharmacists and opticians.
Primary outcome The outcome of greatest importance, usually the one in a study that the
power calculation is based on.
Product licence An authorisation from the MHRA to market a medicinal product.
Progesterone A steroid hormone released by the corpus luteum that stimulates the
uterus to prepare for pregnancy.
Prognosis A probable course or outcome of a disease. Prognostic factors are
patient or disease characteristics that influence the course. Good
prognosis is associated with low rate of undesirable outcomes; poor
prognosis is associated with a high rate of undesirable outcomes.
Prophylactic antibiotics Antibiotics used for the prevention of infection complications.
Prophylactic cervical A treatment for cervical weakness (also termed cervical incompetence
cerclage or insufficiency) to prevent preterm birth and miscarriage.
Prophylactic Progesterone by vaginal suppository to reduce the incidence of
progesterone spontaneous preterm birth.
Prospective study A research study in which the health or other characteristic of
participants is monitored (or 'followed up') for a period of time, with
events recorded as they happen. This contrasts with retrospective
studies.
Publication bias Publication bias occurs when researchers publish the results of studies
showing that a treatment works well and don't publish those showing it
did not have any effect. If this happens, analysis of the published
results will not give an accurate idea of how well the treatment works.
This type of bias can be assessed by a funnel plot.
Puerperal sepsis Serious infection affecting the mother after giving birth.
Pyrexia A fever.
‘Rescue’ cervical cerclage Cervical cerclage performed as an emergency procedure in a woman
with premature cervical dilatation and often with exposed fetal
membranes.
Respiratory distress A syndrome in premature infants caused by developmental
syndrome insufficiency of surfactant production and structural immaturity in the
lungs.
Sepsis A whole-body inflammation caused by an infection.
Shirodkar suture A non-absorbable stitch that is inserted and put around the cervix to
hold it closed.
Special care baby unit A unit taking premature and term babies who do not require intensive
care, but are unable to be cared for on a normal ward.
Speculum examination A method for visualising the cervix (the opening of the uterus) and the
interior walls of the vagina, using an instrument.
Spontaneous preterm Regular contractions of the uterus resulting in changes in the cervix
labour that start before 37 weeks of pregnancy that occur with no intervention.
Stakeholder An organisation with an interest in a topic that NICE is developing a
clinical guideline or piece of public health guidance on. Organisations
that register as stakeholders can comment on the draft scope and the
draft guidance. Stakeholders may be:
• manufacturers of drugs or equipment
• national patient and carer organisations
• NHS organisations

417
Term Definition
• organisations representing healthcare professionals.
Standard deviation (SD) A measure of the spread or dispersion of a set of observations,
calculated as the average difference from the mean value in the
sample.
Stillbirth The death of a baby after 24 weeks of pregnancy but before birth.
Subgroup analysis An analysis in which the intervention effect is evaluated in a defined
subset of the participants in a trial, or in complementary subsets.
Suspected preterm labour A woman is in suspected preterm labour if she has reported symptoms
of preterm labour and has had a clinical assessment (including a
speculum or digital vaginal examination) that confirms the possibility of
preterm labour but rules out established labour.
Symptoms of preterm A woman has presented before 37+0 weeks of pregnancy reporting
labour symptoms that might be indicative of preterm labour (such as
abdominal pain), but no clinical assessment (including speculum or
digital vaginal examination) has taken place.
Systematic review (SR) A review in which evidence from scientific studies has been identified,
appraised and synthesised in a methodical way according to
predetermined criteria. It may include a meta-analysis.
Tachycardia (fetal Rapid heart rate; for the term fetus, this is defined as a heart rate of
monitoring) over 160 beats per minute.
Third stage of labour The interval from the birth of the baby to the expulsion of the placenta
and membranes.
Time horizon The time span over which costs and health outcomes are considered in
adecision analysis or economic evaluation.
Tocolytic A drug used to prevent or lessen uterine contractions.
Transfer This term indicates where responsibility for the woman’s care passes
from one healthcare professional to another. This may or may not also
involve a physical transfer of the woman from one birth setting to
another.
Transvaginal ultrasound An internal ultrasound scan to look at a women’s reproductive system.
Treatment allocation Assigning a participant to a particular arm of a trial.
Univariate Analysis which separately explores each variable in a data set.
Upper uterine segment It is the portion of the uterus above the bladder edge. The lower
segment is the portion of the uterus normally covered anteriorly by the
bladder; the lower segment is not well formed until the last trimester.
Uterotonic A drug used to induce uterine contractions.
Utility In health economics, a 'utility' is the measure of the preference or value
that an individual or society places upon a particular health state. It is
generally a number between 0 (representing death) and 1 (perfect
health). The most widely used measure of benefit in cost–utility
analysis is the quality adjusted life year (QALY), but other measures
include disability adjusted life years (DALYs) and healthy year
equivalents (HYEs).
Vaginal birth The birth of a baby through the vagina.

418
17.3 Abbreviations
Abbreviation Definition
17OHP-C 17 α-hydroxyprogesterone caproate
ARD Absolute risk difference
AYC Area under the curve
bpm Beats per minute
CI Confidence interval
CLD Chronic lung disease
CP Cerebral palsy
CrI Credible interval
CRP C-reactive protein
CS Caesarean section
CTG Cardiotocography
DIC Deviance information criteria
EFM Electronic fetal heart monitoring
EGA Estimated gestational age
FBS Fetal blood sampling
FEM Fetal electronic monitoring
fFN Fetal fibronectin
FHR Fetal heart rate
GRADE Grading of Recommendations, Assessment, Development and
Evaluations
HCG Human chorionic gonadotropin
HEED Health Economic Evaluations Database
HTA Health Technology Assessment
IA Intermittent auscultation
ICER Incremental cost-effectiveness ratio
ICH Intracranial haemorrhage
IGFBP-1 Insulin-like growth factor binding protein-1
IPD Individual patient data
IV Intravenous
IVH Intraventricular haemorrhage
LLETZ Large loop excision of the transformation zone
LR+ Likelihood ratio of a positive test result
LR− Likelihood ratio of a minus test result
MgSO4 Magnisium sulfate
MID Minimally important difference
MTC Mixed treatment comparison
NCC-WCH National Collaborating Centre for Women’s and Children’s Health
NHS EED NHS Economic Evaluation Database
NICE National Institute for Care Excellence
NICU Neonatal intensive care unit
NMA Network meta-analysis
NR Not reportable

419
Abbreviation Definition
OR Odds ratio
p Probability
PAMG-1 Placental alpha microglobulin-1
PBLNQ Preterm Birth Learning Needs Questionnaire
PICO Patient, intervention, comparison, outcome
pIGFBP-1 Phosphorylated insulin-like growth factor binding protein-1
PPH Primary postpartum haemorrhage
P-PROM Preterm premature rupture of membranes
PTL Preterm labour
PTLB Preterm labour and birth
PVL Periventricular leukomalacia
QALY Quality of life year
QUADAS Quality Assessment of Diagnostic Accuracy Studies
RCT Randomised controlled trial
RDS Respiratory distress syndrome
RR Risk ratio
SD Standard deviation
SE Standard error
SMD Standardised mean differences
SR Systematic review
TSU Technical support unit
TVUS Transvaginal ultrasound

420
References
18 References
Alfirevic 2012
Alfirevic,Z., Stampalija,T., Roberts,D., Jorgensen,A.L., Cervical stitch (cerclage) for
preventing preterm birth in singleton pregnancy, Cochrane Database of Systematic Reviews,
4, CD008991-, 2012
Alfirevic 2013
Alfirevic,Zarko, Milan,Stephen J., Livio,Stefania, Caesarean section versus vaginal delivery
for preterm birth in singletons, Cochrane Database of Systematic Reviews, -, 2013
Althaus 2005
Althaus,J.E., Petersen,S.M., Fox,H.E., Holcroft,C.J., Graham,E.M., Can electronic fetal
monitoring identify preterm neonates with cerebral white matter injury?, Obstetrics and
Gynecology, 105, 458-465, 2005
Althuisius 2003
Althuisius,S.M., Dekker,G.A., Hummel,P., van Geijn,H.P., Cervical inc, Cervical
incompetence prevention randomized cerclage trial: emergency cerclage with bed rest
versus bed rest alone, American Journal of Obstetrics and Gynecology, 189, 907-910, 2003
Aoki 2014
Aoki,S., Ohnuma,E., Kurasawa,K., Okuda,M., Takahashi,T., Hirahara,F., Emergency
cerclage versus expectant management for prolapsed fetal membranes: a retrospective,
comparative study, Journal of Obstetrics and Gynaecology Research, 40, 381-386, 2014
Asztalos 2013
Asztalos, E. V., Murphy, K. E., Willan, A. R., Matthews, S. G., Ohlsson, A., Saigal, S.,
Armson, B. A., Kelly, E. N., Delisle, M. F., Gafni, A., Lee, S. K., Sananes, R., Rovet, J.,
Guselle, P., Amankwah, K., Saleem, M., Sanchez, J., Macs- Collaborative Group, Multiple
courses of antenatal corticosteroids for preterm birth study: outcomes in children at 5 years
of age (MACS-5), JAMA Pediatrics, 167, 1102-10, 2013
Atarod 2014
Atarod,Z., Taghipour,M., Roohanizadeh,H., Fadavi,S., Taghavipour,M., Effects of single
course and multicourse betamethasone prior to birth in the prognosis of the preterm
neonates: A randomized, double-blind placebo-control clinical trial study, Journal of
Research in Medical Sciences, 19, 715-719, 2014
Azlin 2010
Azlin,M.I., Bang,H.K., An,L.J., Mohamad,S.N., Mansor,N.A., Yee,B.S., Zulkifli,N.H.,
Tamil,A.M., Role of phIGFBP-1 and ultrasound cervical length in predicting pre-term labour,
Journal of Obstetrics and Gynaecology, 30, 456-459, 2010
Bagga 2010

421
References
Bagga,R., Takhtani,M., Suri,V., Adhikari,K., Arora,S., Bhardwaj,S., Cervical length and

cervicovaginal HCG for prediction of pre-term birth in women with signs and symptoms of
pre-term labour, Journal of Obstetrics and Gynaecology, 30, 451-455, 2010
Bartnicki 1996
Bartnicki,J., Casal,D., Kreaden,U.S., Saling,E., Vetter,K., Fetal fibronectin in vaginal
specimens predicts preterm delivery and very-low-birth-weight infants, American Journal of
Obstetrics and Gynecology, 174, 971-974, 1996
Benattar 1997
Benattar,C., Taieb,J., Fernandez,H., Lindendaum,A., Frydman,R., Ville,Y., Rapid fetal
fibronectin swab-test in preterm labor patients treated by betamimetics, European Journal of
Obstetrics, Gynecology, and Reproductive Biology, 72, 131-135, 1997
Berghella 2011
Berghella,V., Rafael,T.J., Szychowski,J.M., Rust,O.A., Owen,J., Cerclage for short cervix on
ultrasonography in women with singleton gestations and previous preterm birth: a meta-
analysis, Obstetrics and Gynecology, 117, 663-671, 2011
Bickford 2013
Bickford,C.D., Magee,L.A., Mitton,C., Kruse,M., Synnes,A.R., Sawchuck,D., Basso,M.,
Senikas,V.M., von,Dadelszen P., MAG-CP Working Group., Magnesium sulphate for fetal
neuroprotection: a cost-effectiveness analysis, BMC Health Services Research, 13, 527-,
2013
Blencowe 2013
Blencowe H, Lee AC, Cousens S, Bahalim A, Narwal R, Zhong N, Chou D, Say L, Modi N,
Katz J, Vos T, Marlow N, Lawn JE. Preterm birth-associated neurodevelopmental impairment
estimates at regional and global levels for 2010. Pediatric research. 2013 Dec;74 Suppl 1:17-
34.
Botsis 2006
Botsis,D., Makrakis,E., Papagianni,V., Kouskouni,E., Grigoriou,O., Dendrinos,S.,
Creatsas,G., The value of cervical length and plasma proMMP-9 levels for the prediction of
preterm delivery in pregnant women presenting with threatened preterm labor, European
Journal of Obstetrics, Gynecology, and Reproductive Biology, 128, 108-112, 2006
Bowes 1980
Bowes,W.A.,Jr., Gabre,S.G., Bowes,C., Fetal heart rate monitoring in premature infants
weighing 1,500 grams or less, American Journal of Obstetrics and Gynecology, 137, 791-
796, 1980
Braithwaite 1986
Braithwaite,N.D.J., Milligan,J.E., Shennan,A.T., Fetal heart rate monitoring and neonatal
mortality in the very preterm infant, American Journal of Obstetrics and Gynecology, 154,
250-254, 1986
Brik 2010

422
References
Brik,M., Hernandez,A.I., Pedraz,C.C., Perales,A., Phosphorylated insulin-like growth factor

binding protein-1 and cervical measurement in women with threatening preterm birth, Acta
Obstetricia et Gynecologica Scandinavica, 89, 268-274, 2010
British Association of Perinatal Medicine 2008
British Association of Perinatal Medicine, The Management of Babies born Extremely
Preterm at less than 26 weeks of gestation, A Framework for Clinical Practice at the time of
Birth, 2008
Report of a Working Group
Burrus 1994
Burrus,D.R., O'Shea,T.M.,Jr., Veille,J.C., Mueller-Heubach,E., The predictive value of
intrapartum fetal heart rate abnormalities in the extremely premature infant, American
Journal of Obstetrics and Gynecology, 171, 1128-1132, 1994
Burwick 2011
Burwick,R.M., Zork,N.M., Lee,G.T., Ross,M.G., Kjos,S.L., Cervilenz assessment of cervical
length compared to fetal fibronectin in the prediction of preterm delivery in women with
threatened preterm labor, Journal of Maternal-Fetal and Neonatal Medicine, 24, 127-131,
2011
Cahill 2010
Cahill,A.G., Odibo,A.O., Caughey,A.B., Stamilio,D.M., Hassan,S.S., Macones,G.A.,
Romero,R., Universal cervical length screening and treatment with vaginal progesterone to
prevent preterm birth: a decision and economic analysis, American Journal of Obstetrics and
Gynecology, 202, 548-548, 2010
Cahill 2011
Cahill,A.G., Odibo,A.O., Stout,M.J., Grobman,W.A., Macones,G.A., Caughey,A.B.,
Magnesium sulfate therapy for the prevention of cerebral palsy in preterm infants: a decision-
analytic and economic analysis, American Journal of Obstetrics and Gynecology, 205, 542-
547, 2011
Carroll 1995
Carroll,S.G., Papaioannou,S., Nicolaides,K.H., Assessment of fetal activity and amniotic fluid
volume in the prediction of intrauterine infection in preterm prelabor amniorrhexis, American
Cazan-London 2005
Cazan-London,G., Mozurkewich,E.L., Xu,X., Ransom,S.B., Willingness or unwillingness to
perform cesarean section for impending preterm delivery at 24 weeks' gestation: a cost-
effectiveness analysis, American Journal of Obstetrics and Gynecology, 193, 1187-1192,
2005
Colbourn 2007
Colbourn,T., Asseburg,C., Bojke,L., Philips,Z., Claxton,K., Ades,A.E., Gilbert,R.E., Prenatal
screening and treatment strategies to prevent group B streptococcal and other bacterial
infections in early infancy: Cost-effectiveness and expected value of information analyses,
Health Technology Assessment, 11, 21-108, 2007
Costeloe 2012

423
References
Costeloe KL, Hennessy EM, Haider S, Stacey F, Marlow N, Draper ES. Short term outcomes
after extreme preterm birth in England: comparison of 2 birth cohorts in 1995 and 2006 (the
EPICure studies). BMJ (Clinical research ed. 2012;345:e7976.
MacKay 2010
MacKay DF, Smith GC, Dobbie R, Pell JP. Gestational age at delivery and special
educational need: retrospective cohort study of 407,503 schoolchildren. PLoS medicine.
2010;7(6):e1000289.
Crowther 2003
Crowther,C.A., Hiller,J.E., Doyle,L.W., Haslam,R.R., Australasian Collaborative Trial of
Magnesium Sulphate (ACTOMg SO, Effect of magnesium sulfate given for neuroprotection
before preterm birth: a randomized controlled trial, JAMA, 290, 2669-2676, 2003
Crowther 2013
Crowther,Caroline A., McKinlay,JD Christopher, Middleton,Philippa, Harding,Jane E., Repeat
doses of prenatal corticosteroids for women at risk of preterm birth for improving neonatal
health outcomes, Cochrane Database of Systematic Reviews, -, 2013
Curti 2012
Curti,A., Simonazzi,G., Farina,A., Mehmeti,H., Facchinetti,F., Rizzo,N., Exam-indicated
cerclage in patients with fetal membranes at or beyond external os: a retrospective
evaluation, Journal of Obstetrics and Gynaecology Research, 38, 1352-1357, 2012
Danti 2011
Danti,L., Prefumo,F., Lojacono,A., Corini,S., Testori,A., Frusca,T., The combination of short
cervical length and phIGFBP-1 in the prediction of preterm delivery in symptomatic women,
Journal of Maternal-Fetal and Neonatal Medicine, 24, 1262-1266, 2011
Daskalakis 2006
Daskalakis,G., Papantoniou,N., Mesogitis,S., Antsaklis,A., Management of cervical
insufficiency and bulging fetal membranes, Obstetrics and Gynecology, 107, 221-226, 2006
Del Valle 1992
Del Valle,G.O., Joffe,G.M., Izquierdo,L.A., Smith,J.F., Gilson,G.J., Curet,L.B., The
biophysical profile and the nonstress test: poor predictors of chorioamnionitis and fetal
infection in prolonged preterm premature rupture of membranes, Obstetrics and Gynecology,
80, 106-110, 1992
Demirci 2011
Demirci,O., Unal,A., Demirci,E., Sozen,H., Akdemir,Y., Boybek,E., Ertekin,A., Sonographic
measurement of cervical length and risk of preterm delivery, Journal of Obstetrics and
Gynaecology Research, 37, 809-814, 2011
Diaz 2009
Diaz,J., Chedraui,P., Hidalgo,L., Medina,M., The clinical utility of fetal fibronectin in the
prediction of pre-term birth in a low socio-economic setting hospital in Ecuador, Journal of
Maternal-Fetal and Neonatal Medicine, 22, 89-93, 2009
Dodd 2013

424
References
Dodd,Jodie M., Jones,Leanne, Flenady,Vicki, Cincotta,Robert, Crowther,Caroline A.,

Prenatal administration of progesterone for preventing preterm birth in women considered to
be at risk of preterm birth, Cochrane Database of Systematic Reviews, -, 2013
Douvas 1984
Douvas,S.G., Meeks,G.R., Graves,G., Intrapartum fetal heart rate monitoring as a predictor
of fetal distress and immediate neonatal condition in low-birth weight (<1,800 grams) infants,
American Journal of Obstetrics and Gynecology, 148, 300-302, 1984
Doyle 2014
Doyle,L.W., Anderson,P.J., Haslam,R., Lee,K.J., Crowther,C., Australasian Collaborative
Trial of Magnesium Sulphate (ACTOMgSO, School-age outcomes of very preterm infants
after antenatal treatment with magnesium sulfate vs placebo, JAMA, 312, 1105-1113, 2014
Eleje 2014
Eleje,G.U., Ezugwu,E.C., Ogunyemi,D., Eleje,L.I., Ikechebelu,J.I., Igwegbe,A.O.,
Okonkwo,J.E., Ikpeze,O.C., Udigwe,G.O., Onah,H.E., Nwosu,B.O., Ezeama,C.O.,
Ezenkwele,E.P., Accuracy and cost-analysis of placental alpha-microglobulin-1 test in the
diagnosis of premature rupture of fetal membranes in resource-limited community settings,
Journal of Obstetrics and Gynaecology Research, 41, 29-38, 2014
Elimian 2014
Elimian,A., Goodman,J., Escobedo,M., Nightingale,L., Knudtson,E., Williams,M., Immediate
compared with delayed cord clamping in the preterm neonate: a randomized controlled trial,
Eroglu 2007
Eroglu,D., Yanik,F., Oktem,M., Zeyneloglu,H.B., Kuscu,E., Prediction of preterm delivery
among women with threatened preterm labor, Gynecologic and Obstetric Investigation, 64,
109-116, 2007
Farb 1983
Farb,H.F., Arnesen,M., Geistler,P., Knox,G.E., C-reactive protein with premature rupture of
membranes and premature labor, Obstetrics and Gynecology, 62, 49-51, 1983
Fisk 1987
Fisk,N.M., Fysh,J., Child,A.G., Gatenby,P.A., Jeffery,H., Bradfield,A.H., Is C-reactive protein
really useful in preterm premature rupture of the membranes?, British Journal of Obstetrics
and Gynaecology, 94, 1159-1164, 1987
Fleming 2004
Fleming,A., Bonebrake,R., Istwan,N., Rhea,D., Coleman,S., Stanziano,G., Pregnancy and
economic outcomes in patients treated for recurrent preterm labor, Journal of Perinatology,
24, 223-227, 2004
Garite & Freeman 1982
Garite,T.J., Freeman,R.K., Chorioamnionitis in the preterm gestation, Obstetrics and
Gynecology, 59, 539-545, 1982
Gaucher & Payot 2011

425
References
Gaucher,N., Payot,A., From powerlessness to empowerment: Mothers expect more than

information from the prenatal consultation for preterm labour, Paediatrics and Child Health,
16, 638-642, 2011
Giles 2000
Giles,W., Bisits,A., Knox,M., Madsen,G., Smith,R., The effect of fetal fibronectin testing on
admissions to a tertiary maternal-fetal medicine unit and cost savings, American Journal of
Gomez 2005
Gomez,R., Romero,R., Medina,L., Nien,J.K., Chaiworapongsa,T., Carstens,M., Gonzalez,R.,
Espinoza,J., Iams,J.D., Edwin,S., Rojas,I., Cervicovaginal fibronectin improves the prediction
of preterm delivery based on sonographic cervical length in patients with preterm uterine
contractions and intact membranes.[Erratum appears in Am J Obstet Gynecol. 2005
Jul;193(1):308-9], American Journal of Obstetrics and Gynecology, 192, 350-359, 2005
Gramellini 2007
Gramellini,D., Fieni,S., Kaihura,C., Modena,A.B., Cervical length as a predictor of preterm
delivery: gestational age-related percentiles vs fixed cut-offs, Acta Bio-Medica de l Ateneo
Parmense, 78, 220-224, 2007
Griffin 1997
Griffin,T., Kavanaugh,K., Soto,C.F., White,M., Parental evaluation of a tour of the neonatal
intensive care unit during a high-risk pregnancy, JOGNN - Journal of Obstetric, Gynecologic,
and Neonatal Nursing, 26, 59-65, 1997
Gupton & Heaman 1994
Gupton,A., Heaman,M., Learning needs of hospitalized women at risk for preterm birth,
Applied Nursing Research, 7, 118-124, 1994
Haas 2012
Haas,D.M., Caldwell,D.M., Kirkpatrick,P., McIntosh,J.J., Welton,N.J., Tocolytic therapy for
preterm delivery: systematic review and network meta-analysis, BMJ, 345, e6226-, 2012
Hawrylyshyn 1983
Hawrylyshyn,P., Bernstein,P., Milligan,J.E., Soldin,S., Pollard,A., Papsin,F.R., Premature
rupture of membranes: the role of C-reactive protein in the prediction of chorioamnionitis,
American Journal of Obstetrics and Gynecology, 147, 240-246, 1983
Hodnett 2010
Hodnett,Ellen D., Fredericks,Suzanne, Weston,Julie, Support during pregnancy for women at
increased risk of low birthweight babies, Cochrane Database of Systematic Reviews, -, 2010
Holmes 2001
Holmes,P., Oppenheimer,L.W., Gravelle,A., Walker,M., Blayney,M., The effect of variable
heart rate decelerations on intraventricular hemorrhage and other perinatal outcomes in
preterm infants, Journal of Maternal-Fetal Medicine, 10, 264-268, 2001
Holst 2006

426
References
Holst,R.M., Jacobsson,B., Hagberg,H., Wennerholm,U.B., Cervical length in women in

preterm labor with intact membranes: relationship to intra-amniotic inflammation/microbial
invasion, cervical inflammation and preterm delivery, Ultrasound in Obstetrics and
Gynecology, 28, 768-774, 2006
HoutzagerBA 2006
Houtzager BA,Hogendoorn SM,Papatsonis DN,Samsom JF,van Geijn HP,Bleker OP,van
Wassenaer AG, Long-term follow up of children exposed in utero to nifedipine or ritodrine for
the management of preterm labour, BJOG : an international journal of obstetrics and
gynaecology, -, 2006
Iams 1995
Iams,J.D., Casal,D., McGregor,J.A., Goodwin,T.M., Kreaden,U.S., Lowensohn,R.,
Lockitch,G., Fetal fibronectin improves the accuracy of diagnosis of preterm labor, American
Aina-Mumuney 2007
Aina-Mumuney,A.J., Althaus,J.E., Henderson,J.L., Blakemore,M.C., Johnson,E.A.,
Graham,E.M., Intrapartum electronic fetal monitoring and the identification of systemic fetal
inflammation, Journal of Reproductive Medicine, 52, 762-768, 2007
Ismail 1985
Ismail,M.A., Zinaman,M.J., Lowensohn,R.I., Moawad,A.H., The significance of C-reactive
protein levels in women with premature rupture of membranes, American Journal of
Jain & Morris 1998
Jain,K., Morris,P.G., A clinical study to evaluate the usefulness of the MAST test in
diagnosing prelabour rupture of membranes, Journal of Obstetrics and Gynaecology, 18, 33-
36, 1998
Jaju & Dhabadi 2011
Jaju,P.B., Dhabadi,V.B., Nifedipine versus ritodrine for suppression of preterm labor and
analysis of side effects, Journal of Obstetrics and Gynaecology of India, 61, 534-537, 2011
Kariniemi 1984
Kariniemi,V., Jarvenpaa,A.L., Teramo,K., Fetal heart rate patterns and perinatal outcome of
very-low-birthweight infants, British Journal of Obstetrics and Gynaecology, 91, 18-22, 1984
Kashanian 2014
Kashanian,M., Zamen,Z., Sheikhansari,N., Comparison between nitroglycerin dermal patch
and nifedipine for treatment of preterm labor: a randomized clinical trial, Journal of
Perinatology, 34, 683-687, 2014
Kenyon 2008
Kenyon,S., Pike,K., Jones,D.R., Brocklehurst,P., Marlow,N., Salt,A., Taylor,D.J., Childhood
outcomes after prescription of antibiotics to pregnant women with preterm rupture of the
membranes: 7-year follow-up of the ORACLE I trial, Lancet, 372, 1310-1318, 2008
Kenyon 2013

427
References
Kenyon,Sara, Boulvain,Michel, Neilson,James P., Antibiotics for preterm rupture of

membranes, Cochrane Database of Systematic Reviews, -, 2013
Klauser 2012
Klauser,C.K., Briery,C.M., Keiser,S.D., Martin,R.W., Kosek,M.A., Morrison,J.C., Effect of
antenatal tocolysis on neonatal outcomes, Journal of Maternal-Fetal and Neonatal Medicine,
25, 2778-2781, 2012
Klauser 2014
Klauser,C.K., Briery,C.M., Martin,R.W., Langston,L., Magann,E.F., Morrison,J.C., A
comparison of three tocolytics for preterm labor: a randomized clinical trial, Journal of
Maternal-Fetal and Neonatal Medicine, 27, 801-806, 2014
Kurki 1990
Kurki,T., Teramo,K., Ylikorkala,O., Paavonen,J., C-reactive protein in preterm premature
rupture of the membranes.[Erratum appears in Arch Gynecol Obstet 1990;247(2):106],
Archives of Gynecology and Obstetrics, 247, 31-37, 1990
Kwek 2004
Kwek,K., Khi,C., Ting,H.S., Yeo,G.S., Evaluation of a bedside test for phosphorylated insulin-
like growth factor binding protein-1 in preterm labour, Annals of the Academy of Medicine,
Singapore, 33, 780-783, 2004
Lam 2003
Lam,F., Istwan,N.B., Jacques,D., Coleman,S.K., Stanziano,G.J., Managing perinatal
outcomes: the clinical benefit and cost-effectiveness of pharmacologic treatment of recurrent
preterm labor, Managed Care, 12, 39-46, 2003
LaShay 2000
LaShay,N., Gilson,G., Joffe,G., Qualls,C., Curet,L., Will cervicovaginal interleukin-6
combined with fetal fibronectin testing improve the prediction of preterm delivery?, Journal of
Maternal-Fetal Medicine, 9, 336-341, 2000
Lembet 2002
Lembet,A., Eroglu,D., Ergin,T., Kuscu,E., Zeyneloglu,H., Batioglu,S., Haberal,A., New rapid
bed-side test to predict preterm delivery: phosphorylated insulin-like growth factor binding
protein-1 in cervical secretions, Acta Obstetricia et Gynecologica Scandinavica, 81, 706-712,
2002
Lewis 1999
Lewis,D.F., Adair,C.D., Weeks,J.W., Barrilleaux,P.S., Edwards,M.S., Garite,T.J., A
randomized clinical trial of daily nonstress testing versus biophysical profile in the
management of preterm premature rupture of membranes, American Journal of Obstetrics
and Gynecology, 181, 1495-1499, 1999
Lukes 1997
Lukes,A.S., Thorp,J.M.,Jr., Eucker,B., Pahel-Short,L., Predictors of positivity for fetal
fibronectin in patients with symptoms of preterm labor, American Journal of Obstetrics and
Gynecology, 176, 639-641, 1997
Luthy 1987

428
References
Luthy,D.A., Shy,K.K., van,Belle G., Larson,E.B., Hughes,J.P., Benedetti,T.J., Brown,Z.A.,

Effer,S., King,J.F., Stenchever,M.A., A randomized trial of electronic fetal monitoring in
preterm labor, Obstetrics and Gynecology, 69, 687-695, 1987
Malak 1996
Malak,T.M., Sizmur,F., Bell,S.C., Taylor,D.J., Fetal fibronectin in cervicovaginal secretions as
a predictor of preterm birth, British Journal of Obstetrics and Gynaecology, 103, 648-653,
1996
March 2011
March,M., De,VecianaM, Parson,A., The efficacy of umbilical cord milking on the reduction of
red blood cell transfusion rates in infants born between 24 and 28 6/7 weeks gestation - A
randomized controlled trial, American Journal of Obstetrics and Gynecology, 204, S204-,
2011
Marret 2007
Marret,S., Marpeau,L., Zupan-Simunek,V., Eurin,D., Leveque,C., Hellot,M.F., Benichou,J.,
PREMAG trial group., Magnesium sulphate given before very-preterm birth to protect infant
brain: the randomised controlled PREMAG trial*, BJOG: An International Journal of
Obstetrics and Gynaecology, 114, 310-318, 2007
Marret 2008
Marret,S., Marpeau,L., Benichou,J., Benefit of magnesium sulfate given before very preterm
birth to protect infant brain, Pediatrics, 121, 225-226, 2008
Martin 1974
Martin,Jr, Siassi,B., Hon,E.H., Fetal heart rate patterns and neonatal death in low birthweight
infants, Obstetrics and Gynecology, 44, 503-510, 1974
Matsuda 2003
Matsuda,Y., Maeda,T., Kouno,S., The critical period of non-reassuring fetal heart rate
patterns in preterm gestation, European Journal of Obstetrics, Gynecology, and
Reproductive Biology, 106, 36-39, 2003
McKenna 1999
McKenna,D.S., Chung,K., Iams,J.D., Effect of digital cervical examination on the expression
of fetal fibronectin, Journal of Reproductive Medicine, 44, 796-800, 1999
Mercer 2012
Mercer,B., Antibiotics in the management of PROM and preterm labor, Obstetrics and
Gynecology Clinics of North America, 39, 65-76, 2012
Mittendorf 2002
Mittendorf,R., Dambrosia,J., Pryde,P.G., Lee,K.S., Gianopoulos,J.G., Besinger,R.E.,
Tomich,P.G., Association between the use of antenatal magnesium sulfate in preterm labor
and adverse health outcomes in infants, American Journal of Obstetrics and Gynecology,
186, 1111-1118, 2002
Mugford 1991

429
References
Mugford,M., Piercy,J., Chalmers,I., Cost implications of different approaches to the

prevention of respiratory distress syndrome, Archives of Disease in Childhood, 66, 757-764,
1991
Nankali 2014
Nankali,A., Jamshidi,P.K., Rezaei,M., The Effects of Glyceryl Trinitrate Patch on the
Treatment of Preterm Labor: A Single-blind Randomized Clinical Trial, Journal of
Reproduction and Infertility, 15, 71-77, 2014
Nikbakht 2014
Nikbakht,R., Taheri,Moghadam M., Ghane'ee,H., Nifedipine compared to magnesium sulfate
for treating preterm labor: A randomized clinical trial, Iranian Journal of Reproductive
Medicine, 12, 145-150, 2014
Nisenblat 2006
Nisenblat,V., Alon,E., Barak,S., Gonen,R., Bader,D., Ohel,G., Fetal heart rate patterns and
neurodevelopmental outcome in very low birth weight infants, Acta Obstetricia et
Gynecologica Scandinavica, 85, 792-796, 2006
Olatunbosun 1995
Olatunbosun,O.A., al-Nuaim,L., Turnell,R.W., Emergency cerclage compared with bed rest
for advanced cervical dilatation in pregnancy, International Surgery, 80, 170-174, 1995
Palacio 2007
Palacio,M., Sanin-Blair,J., Sanchez,M., Crispi,F., Gomez,O., Carreras,E., Coll,O.,
Cararach,V., Gratacos,E., The use of a variable cut-off value of cervical length in women
admitted for preterm labor before and after 32 weeks, Ultrasound in Obstetrics and
Gynecology, 29, 421-426, 2007
Perrone 2012
Perrone,G., Anceschi,M.M., Capri,O., Galoppi,P., Pizzulo,S., Buccheri,M., Pascone,R.,
Nofroni,I., Brunelli,R., Maternal C-reactive protein at hospital admission is a simple predictor
of funisitis in preterm premature rupture of membranes, Gynecologic and Obstetric
Investigation, 74, 95-99, 2012
Pizzi 2014
Pizzi,L.T., Seligman,N.S., Baxter,J.K., Jutkowitz,E., Berghella,V., Cost and cost effectiveness
of vaginal progesterone gel in reducing preterm birth: an economic analysis of the
PREGNANT trial, Pharmacoeconomics, 32, 467-478, 2014
POPPY Steering Group 2009
Family-centred care in neonatal units. A summary of research results and recommendations
from the POPPY Parents of premature babies) project. London: NCT; 2009
Porto 2011
Porto,A.M., Coutinho,I.C., Correia,J.B., Amorim,M.M., Effectiveness of antenatal
corticosteroids in reducing respiratory disorders in late preterm infants: randomised clinical
trial, BMJ (Clinical research ed.), Vol.342, pp.d1696, 2011., -
Rabe 2012

430
References
Rabe,H., az-Rossello,J.L., Duley,L., Dowswell,T., Effect of timing of umbilical cord clamping

and other strategies to influence placental transfusion at preterm birth on maternal and infant
outcomes. [Update of Cochrane Database Syst Rev. 2004;(4):CD003248; PMID: 15495045],
Cochrane Database of Systematic Reviews, 8, CD003248-, 2012
Ranjit 2015
Ranjit,T., Nesargi,S., Rao,P.N., Sahoo,J.P., Ashok,C., Chandrakala,B.S., Bhat,S., Effect of
early versus delayed cord clamping on hematological status of preterm infants at 6 wk of
age, Indian Journal of Pediatrics, 82, 29-34, 2015
RCOG 2011
Cervical Cerclage, RCOG Green-top Guideline No. 60
RCOG 2015
RCOG, Clamping of the Umbilical Cord and Placental Transfusion (Scientific Impact Paper
No. 14), 2015
Rayburn 1987
Rayburn,W.F., Johnson,M.Z., Hoffman,K.L., Donn,S.M., Nelson,R.M.,Jr., Intrapartum fetal
heart rate patterns and neonatal intraventricular hemorrhage, American Journal of
Perinatology, 4, 98-101, 1987
Roberts & Dalziel 2013
Roberts,Devender, Dalziel,Stuart R., Antenatal corticosteroids for accelerating fetal lung
maturation for women at risk of preterm birth, Cochrane Database of Systematic Reviews, -,
2013
Romem & Artal 1984
Romem,Y., Artal,R., C-reactive protein as a predictor for chorioamnionitis in cases of
premature rupture of the membranes, American Journal of Obstetrics and Gynecology, 150,
546-550, 1984
Romero 2012
Romero,R., Nicolaides,K., Conde-Agudelo,A., Tabor,A., O'Brien,J.M., Cetingoz,E.,
Da,Fonseca E., Creasy,G.W., Klein,K., Rode,L., Soma-Pillay,P., Fusey,S., Cam,C.,
Alfirevic,Z., Hassan,S.S., Vaginal progesterone in women with an asymptomatic sonographic
short cervix in the midtrimester decreases preterm delivery and neonatal morbidity: a
systematic review and metaanalysis of individual patient data, American Journal of
Rouse 2008
Rouse,D.J., Hirtz,D.G., Thom,E., Varner,M.W., Spong,C.Y., Mercer,B.M., Iams,J.D.,
Wapner,R.J., Sorokin,Y., Alexander,J.M., Harper,M., Thorp,J.M.,Jr., Ramin,S.M.,
Malone,F.D., Carpenter,M., Miodovnik,M., Moawad,A., O'Sullivan,M.J., Peaceman,A.M.,
Hankins,G.D., Langer,O., Caritis,S.N., Roberts,J.M., Eunice Kennedy Shriver NICHD
Maternal-Fetal Medicine Units Network., A randomized, controlled trial of magnesium sulfate
for the prevention of cerebral palsy, New England Journal of Medicine, 359, 895-905, 2008

431
References
Sakai 2003
Sakai,M., Sasaki,Y., Yamagishi,N., Tanebe,K., Yoneda,S., Saito,S., The preterm labor index
and fetal fibronectin for prediction of preterm delivery with intact membranes, Obstetrics and
Gynecology, 101, 123-128, 2003
Salim 2012
Salim,R., Garmi,G., Nachum,Z., Zafran,N., Baram,S., Shalev,E., Nifedipine compared with
atosiban for treating preterm labor: a randomized controlled trial, Obstetrics and Gynecology,
120, 1323-1331, 2012
Sawyer 2013
Sawyer,A., Rabe,H., Abbott,J., Gyte,G., Duley,L., Ayers,S., Parents' experiences and
satisfaction with care during the birth of their very preterm baby: A qualitative study, BJOG:
An International Journal of Obstetrics and Gynaecology, 120, 637-643, 2013
Schmitz 2006
Schmitz,T., Maillard,F., Bessard-Bacquaert,S., Kayem,G., Fulla,Y., Cabrol,D., Goffinet,F.,
Selective use of fetal fibronectin detection after cervical length measurement to predict
spontaneous preterm delivery in women with preterm labor, American Journal of Obstetrics
and Gynecology, 194, 138-143, 2006
Schmitz 2008
Schmitz,T., Kayem,G., Maillard,F., Lebret,M.T., Cabrol,D., Goffinet,F., Selective use of
sonographic cervical length measurement for predicting imminent preterm delivery in women
with preterm labor and intact membranes, Ultrasound in Obstetrics and Gynecology, 31, 421-
426, 2008
Schreyer 1989
Schreyer,P., Caspi,E., Bar,NatanN, Tal,E., Weinraub,Z., The predictive value of fetal
breathing movement and Bishop score in the diagnosis of 'true' preterm labor, American
Senden & Owen 1996
Senden,I.P., Owen,P., Comparison of cervical assessment, fetal fibronectin and fetal
breathing in the diagnosis of preterm labour, Clinical and Experimental Obstetrics and
Gynecology, 23, 5-9, 1996
Shy 1988
Shy,K.K., Olshan,A.F., Hickok,D.E., Luthy,D.A., Electronic fetal monitoring during premature
labor and the occurrence of perinatal mortality in very low birthweight infants, Birth, 15, 14-
18, 1988
Siassakos 2009
Siassakos,D., O'Brien,K., Draycott,T., Healthcare evaluation of the use of atosiban and
fibronectin for the management of pre-term labour, Journal of Obstetrics and Gynaecology,
29, 507-511, 2009

432
References
Skoll 2006
Skoll,A., St,Louis P., Amiri,N., Delisle,M.F., Lalji,S., The evaluation of the fetal fibronectin test
for prediction of preterm delivery in symptomatic patients, Journal of Obstetrics and
Gynaecology Canada: JOGC, 28, 206-213, 2006
Smith 2012
Smith,E.J., Muller,C.L., Sartorius,J.A., White,D.R., Maslow,A.S., C-reactive protein as a
predictor of chorioamnionitis, Journal of the American Osteopathic Association, 112, 660-
664, 2012
Sotiriadis 2010
Sotiriadis,A., Kavvadias,A., Papatheodorou,S., Paraskevaidis,E., Makrydimas,G., The value
of serial cervical length measurements for the prediction of threatened preterm labour,
European Journal of Obstetrics, Gynecology, and Reproductive Biology, 148, 17-20, 2010
Stupin 2008
Stupin,J.H., David,M., Siedentopf,J.P., Dudenhausen,J.W., Emergency cerclage versus bed
rest for amniotic sac prolapse before 27 gestational weeks. A retrospective, comparative
study of 161 women, European Journal of Obstetrics, Gynecology, and Reproductive
Biology, 139, 32-37, 2008
Swamy 2005
Swamy,G.K., Simhan,H.N., Gammill,H.S., Heine,R.P., Clinical utility of fetal fibronectin for
predicting preterm birth, Journal of Reproductive Medicine, 50, 851-856, 2005
Tagore & Kwek 2010
Tagore,S., Kwek,K., Comparative analysis of insulin-like growth factor binding protein-1
(IGFBP-1), placental alpha-microglobulin-1 (PAMG-1) and nitrazine test to diagnose
premature rupture of membranes in pregnancy, Journal of Perinatal Medicine, 38, 609-612,
2010
Tanir 2008
Tanir,H.M., Sener,T., Yildiz,Z., Cervicovaginal fetal fibronectin (FFN) for prediction of preterm
delivery in symptomatic cases: a prospective study, Clinical and Experimental Obstetrics and
Gynecology, 35, 61-64, 2008
Tanir 2009
Tanir,H.M., Sener,T., Yildiz,Z., Cervical phosphorylated insulin-like growth factor binding
protein-1 for the prediction of preterm delivery in symptomatic cases with intact membranes,
Journal of Obstetrics and Gynaecology Research, 35, 66-72, 2009
Tekesin 2005
Tekesin,I., Marek,S., Hellmeyer,L., Reitz,D., Schmidt,S., Assessment of rapid fetal
fibronectin in predicting preterm delivery, Obstetrics and Gynecology, 105, 280-284, 2005
Ting 2007
Ting,H.S., Chin,P.S., Yeo,G.S., Kwek,K., Comparison of bedside test kits for prediction of
preterm delivery: phosphorylated insulin-like growth factor binding protein-1 (pIGFBP-1) test
and fetal fibronectin test, Annals of the Academy of Medicine, Singapore, 36, 399-402, 2007

433
References
Tsoi 2005
Tsoi,E., Fuchs,I.B., Rane,S., Geerts,L., Nicolaides,K.H., Sonographic measurement of
cervical length in threatened preterm labor in singleton pregnancies with intact membranes,
Ultrasound in Obstetrics and Gynecology, 25, 353-356, 2005
Tsoi 2006
Tsoi,E., Akmal,S., Geerts,L., Jeffery,B., Nicolaides,K.H., Sonographic measurement of
cervical length and fetal fibronectin testing in threatened preterm labor, Ultrasound in
Valdes 2012
Valdes,E., Salinas,H., Toledo,V., Lattes,K., Cuellar,E., Perucca,E., Diaz,R., Montecinos,F.,
Reyes,A., Nifedipine versus fenoterol in the management of preterm labor: a randomized,
multicenter clinical study, Gynecologic and Obstetric Investigation, 74, 109-115, 2012
van Baaren 2014
van Baaren,G.J., Vis,J.Y., Wilms,F.F., Oudijk,M.A., Kwee,A., Porath,M.M., Oei,G.,
Scheepers,H.C., Spaanderman,M.E., Bloemenkamp,K.W., Haak,M.C., Bolte,A.C., Bax,C.J.,
Cornette,J.M., Duvekot,J.J., Nij Bijvanck,B.W., van,Eyck J., Franssen,M.T., Sollie,K.M.,
Vandenbussche,F.P., Woiski,M., Grobman,W.A., van der Post,J.A., Bossuyt,P.M.,
Opmeer,B.C., Mol,B.W., Predictive value of cervical length measurement and fibronectin
testing in threatened preterm labor, Obstetrics and Gynecology, 123, 1185-1192, 2014
Wex 2009
Wex,J., Connolly,M., Rath,W., Atosiban versus betamimetics in the treatment of preterm
labour in Germany: an economic evaluation, BMC pregnancy and childbirth, 9, 23-, 2009
Wex 2011
Wex,J., bou-Setta,A.M., Clerici,G., Di Renzo,G.C., Atosiban versus betamimetics in the
treatment of preterm labour in Italy: clinical and economic importance of side-effects,
European Journal of Obstetrics, Gynecology, and Reproductive Biology, 157, 128-135, 2011
Yoon 1996
Yoon,B.H., Jun,J.K., Park,K.H., Syn,H.C., Gomez,R., Romero,R., Serum C-reactive protein,
white blood cell count, and amniotic fluid white blood cell count in women with preterm
premature rupture of membranes, Obstetrics and Gynecology, 88, 1034-1040, 1996
Young 2012
Young,E., Tsai,E., O'Riordan,A., A qualitative study of predelivery counselling for extreme
prematurity, Paediatrics and Child Health, 17, 432-436, 2012

434
Appendices
Appendices
All appendices are contained in separate files:
Appendix A: Scope
Appendix B: Stakeholders
Appendix C: Declarations of interest
Appendix D: Review protocols
Appendix E: Search stratagies
Appendix F: PRISMA flow diagrams
Appendix G: Excluded studies
Appendix H: Evidence tables
Appendix I: Forest plots
Appendix J: Network meta-analysis of tocolytics

435

Full Guideline PDF 2176838029

Uploaded by

Copyright:

Available Formats

Full Guideline PDF 2176838029

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Full Guideline PDF 2176838029

Uploaded by

Copyright:

Available Formats

NCC-WCH

Preterm labour and birth

2.3.2 Undertaking new health economic analysis.............................................. 47

5.4 Evidence profile ................................................................................................... 94

7.7.4 Quality of evidence ................................................................................ 126

9.7.4 Quality of evidence ................................................................................ 170

11.2.4 Evidence statements .............................................................................. 252

13.3.2 Review question..................................................................................... 291

14.7.5 Other considerations .............................................................................. 314

16.2.4 Costing and resource use ...................................................................... 360

Jane Hawdon Consultant Neonatologist, Barts Health NHS Trust

Jane Plumb Lay member

Farrah Pradhan Lay member

Marianne Rowntree Midwife, Plymouth Hospitals NHS Trust

Meekai To Consultant in Fetal Medicine and Obstetrics, Kings College Hospital

1.1.3 NCC-WCH staff

Zosia Beckles Information Scientist (from October 2014)

Liz Bickerdike Research Assistant (until September 2013)

Shona Burman-Roy Senior Research Fellow

Anne Carty Project Manager (from March 2015)

Maryam Gholitabar Research Fellow

Paul Jacklin Senior Research Fellow – Health Economist

Juliet Kenny Project Manager (until March 2015)

Rosalind Lai Information Scientist (until October 2014)

Hugo Pedder Statistician (from September 2014)

Amy Wang Research Fellow (from January 2015 to June 2015)

1.2 Care algorithm

1.4 Key research recommendations

1.4.2 Identifying infection in women with preterm prelabour rupture of

1.4.3 Effectiveness of ‘rescue’ cerclage

1.4.4 Magnesium sulfate for neuroprotection: bolus plus infusion

1.5 Research recommendations

1.6 Other versions of the guideline

1.7 Schedule for updating the guideline

2 Guideline development methodology

2.1.3 Who developed this guideline?

2.1.4 What this guideline covers

2.1.4.1 Groups that will be covered

• women diagnosed to be in spontaneous preterm labour

2.1.4.2 Key clinical issues that will be covered

• prophylactic use of vaginal progesterone for women considered to be at risk of

2.1.5 What this guideline does not cover

2.1.5.1 Groups that will not be covered

2.1.5.2 Clinical issues that will not be covered

2.1.6 Relationships between this guideline and other NICE guidance

2.1.6.1 Related NICE guidance

2.2.1 Developing the review questions and outcomes

A total of 18 review questions were identified.

Table 1: Review questions

Chapte Type of review Review questions Outcomes

Chapte Type of review Review questions Outcomes

Chapte Type of review Review questions Outcomes

Chapte Type of review Review questions Outcomes

Chapte Type of review Review questions Outcomes

Chapte Type of review Review questions Outcomes

Chapte Type of review Review questions Outcomes

15 Interventional In preterm birth, Maternal outcomes

2.2.2 Searching for evidence

2.2.2.1 Clinical literature search

2.2.2.2 Health economic literature search