Pain 127 (2007) 94–102

www.elsevier.com/locate/pain

Determining behavioural and physiological responses to pain

in infants at risk for neurological impairment
Bonnie Stevens a,b,*, Patrick McGrath c, Sharyn Gibbins d, Joseph Beyene e,
Lynn Breau f,g,h, Carol Camfield i, Allen Finley h, Linda Franck j, Alexandra Howlett k,
Celeste Johnston l, Patricia McKeever m, Karel O’Brien n, Arne Ohlsson n,
Janet Yamada b
a
Faculties of Nursing and Medicine, University of Toronto, Toronto, Ont., Canada
b
Research Institute, Hospital for Sick Children, Toronto, Ont., Canada
c
Psychology, Pediatrics, Psychiatry, Biomedical Engineering, Dalhousie University, IWK Health Centre, Halifax, NS, Canada
d
Interdisciplinary Research, Sunnybrook and Women’s College Health Sciences Centre, Toronto, Ont., Canada
e
Department of Public Health Sciences, University of Toronto and Biostatistics Methodology Unit, Research Institute, Toronto, Ont., Canada
f
Pediatric Pain Research Lab, IWK Health Centre, Halifax, NS, Canada
g
Faculties of Nursing and Pediatrics, Dalhousie University, Halifax, NS, Canada
h
Pediatric Pain Management, IWK Health Centre, Halifax, NS, Canada
i
Pediatrics, Dalhousie University and IWK Health Centre, Halifax, NS, Canada
j
UCL Institute of Child Health and Great Ormond Street Hospital for Children, London, UK
k
Department of Pediatrics, Dalhousie University, Halifax, NS, Canada
l
School of Nursing, McGill University, Montreal, Que., Canada
m
Faculty of Nursing, University of Toronto and Research Institute, Bloorview Kids Rehab, Toronto, Ont., Canada
n
Department of Pediatrics, Mount Sinai Hospital, Toronto, Ont., Canada

Received 30 August 2005; received in revised form 13 July 2006; accepted 3 August 2006

Abstract

Multiple researchers have validated indicators and measures of infant pain. However, infants at risk for neurologic impairment (NI)
have been under studied. Therefore, whether their pain responses are similar to those of other infants is unknown. Pain responses to heel
lance from 149 neonates (GA > 25–40 weeks) from 3 Canadian Neonatal Intensive Care units at high (Cohort A, n = 54), moderate
(Cohort B, n = 45) and low (Cohort C, n = 50) risk for NI were compared in a prospective observational cohort study. A significant
Cohort by Phase interaction for total facial action (F(6, 409) = 3.50, p = 0.0022) and 4 individual facial actions existed; with Cohort
C demonstrating the most facial action. A significant Phase effect existed for increased maximum Heart Rate (F(3, 431) = 58.1,
p = 0.001), minimum Heart Rate (F(3, 431) = 78.7, p = 0.001), maximum Oxygen saturation (F(3, 425) = 47.6, p = 0.001), and minimum
oxygen saturation (F(3, 425) = 12.2, p = 0.001) with no Cohort differences. Cohort B had significantly higher minimum (F(2, 79) = 3.71,
p = 0.029), and mean (F(2, 79) = 4.04, p = 0.021) fundamental cry frequencies. A significant Phase effect for low/high frequency Heart
Rate Variability (HRV) ratio (F(2, 216) = 4.97, p = 0.008) was found with the greatest decrease in Cohort A. Significant Cohort by Phase
interactions existed for low and high frequency HRV. All infants responded to the most painful phase of the heel lance; however, infants
at moderate and highest risk for NI exhibited decreased responses in some indicators.
 2006 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.

Keywords: Neonates; Pain responses; Procedural pain; Neurological impairment

*
Corresponding author. Tel.: +1 416 813 6595; fax: +1 416 813 8273.
E-mail address: [email protected] (B. Stevens).

0304-3959/$32.00  2006 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.pain.2006.08.012
 B. Stevens et al. / Pain 127 (2007) 94–102 95

1. Introduction age. Heroin or methadone addiction in the mother (defined

by either a history of active drug intake within 72 h before
Researchers have described the behavioural and delivery or a positive urine test from maternal urine) and phar-
physiologic responses of term and preterm newborn macologically induced paralysis in the infant precluded the
infant’s inclusion in the study.
infants to painful stimuli (e.g., heel lance) in precise
Infants who met the inclusion criteria were stratified into
detail (Anand et al., 1999, 2000; Coskun and Anand,
three cohorts. The development and validation of these
2000; Fitzgerald, 2000). Valid indicators of pain have cohorts has been previously established for this group of stud-
been established and a proliferation of pain assess- ies by Stevens and colleagues using a group of 10 neonatal
ment measures have been developed and validated experts (neonatologists, pediatric neurologists and neonatal
(Stevens et al., 2000; Duhn and Medves, 2004). These nurse practitioners). These experts reviewed all potential diag-
measures have been utilized to systematically evaluate noses of infants in the NICU in terms of risk for NI and inde-
interventions for relieving pain (Taddio et al., 1999; pendently categorized each diagnosis into the 3 cohorts. The
Ohlsson et al., 2000; Shah and Ohlsson, 2004; Stevens initial agreement between raters was high (approximately
et al., 2004) and to provide the evidence for the devel- 98%); discrepancies were easily resolved with discussion to
opment of professional guidelines and standards for achieve 100% consensus agreement. The cohorts were defined
as:
infant pain prevention and management (American
Academy of Pediatrics Committee, 2000; Anand,
• Cohort A: At high risk for NI (i.e., perinatal asphyxia, IVH
2001). However, infants at risk for neurologic impair- [Grade III or IV] or a syndrome or chromosomal anomaly);
ment (NI) generally have been excluded from or • Cohort B: At moderate risk for NI (i.e., acute disease pro-
under represented in this research. Therefore, there cesses such as persistent pulmonary hypertension of the
is little knowledge on how these infants respond to newborn, severe meconium aspiration, meningitis, hydro-
painful stimuli and if existing pain assessment indica- cephalus, necrotising enterocolitis);
tors and measures are reliable, valid and feasible for • Cohort C: At low risk for NI (i.e., respiratory distress
clinical and research use. requiring ventilation, sepsis).

1.1. Pain in neonates at risk for neurologic impairment Due to the exploratory nature of this research, sample size
was calculated based on regression modeling and the number
of response indicators that were to be examined. Based on
Approximately 10% of neonates are admitted to the an estimate of 30 indicators (e.g., facial actions, cry and phys-
Neonatal Intensive Care Unit (NICU) (Joseph et al., iologic indicators) and using 5 subjects per variable, we deter-
1998) and 10% of these infants are at risk of NI (Robert- mined that 150 infants or 50 per risk group were required.
son et al., 1998). Neurologic impairment is the result of
a multiplicity of factors including; congenital syndrome/ 2.2. Data collection procedures
chromosomal abnormalities (e.g., congenital trisomies),
birth trauma (e.g., fractures, nerve injuries), extreme The study was approved by all Research Ethics Boards at
preterm birth (e.g., intraventricular haemorrhage the participating hospitals and associated universities. Eligible
[IVH]; necrotising enterocolitis [NEC]), and acquired ill- infants and their cohort assignment were determined by the
nesses with CNS involvement (Robertson et al., 1998). Research Coordinator at each of the 3 participating NICUs
In a retrospective cohort study of 194 neonates stratified and validated with the study steering group. After parents of
infants consented for their infant to participate in the study,
by risk for NI, Stevens et al. (2003) found that the high-
a research nurse who provided no clinical care for the partici-
est risk group for NI underwent significantly more pain-
pating infants abstracted the following data from the infant’s
ful procedures (>15/day) and were administered less chart: Birth weight, gestational age (GA) at birth, 1, 5 and
opioids than infants at minimal or no risk for NI. 10 min Apgar scores, severity of illness (SOI) (Neonatal Ther-
The purpose of this study was to compare behaviour- apeutic Interventions Scoring System; NTISS (Gray et al.,
al and physiological responses of neonates at high, mod- 1992)), neonatal mortality risk (Score for Neonatal Acute
erate and low risk for NI to a routinely performed Physiology: Perinatal Extension, SNAP:PE score (Richardson
painful procedure (i.e., heel lance) during hospitalization et al., 1993; Richardson and Escobar, 1998)), neurodevelop-
in the NICU. mental risk potential (Neurobiologic Risk Score; NBRS
(Brazy et al., 1993)), and the nature and frequency of painful
procedures, and analgesics and sedatives used throughout the
2. Methods first 7 days of life.

2.1. Study design and sample 2.2.1. Behavioural responses

Behavioural responses included facial actions and cries. The
In this prospective observational cohort study, neonates facial expression of pain is characterized by the mouth stretch-
were recruited from three tertiary level NICUs in Canada. Eli- ing vertically and horizontally, the eyes closed tightly, the brow
gible neonates were: (a) hospitalized in the NICU, (b) >25 bulging and the tongue cupped in a taut shape (Craig and
weeks gestational age at birth and (c) <6 weeks postnatal Grunau, 1993). To quantify facial expression in response to
96 B. Stevens et al. / Pain 127 (2007) 94–102

pain, the Neonatal Facial Coding System (NFCS; – Grunau (Nellcor Pulse Oximeter, Model N-3000; Hayward, CA) and
and Craig, 1987) was developed and validated (Grunau and the SATMASTER data collection system (EMG, Los Angeles,
Craig, 1990). Facial actions were collected by videotaping CA). Data were recorded second to second and sampled at
infant’s faces on an 8 mm camcorder (Sharp, Panasonic or 100 Hz. Signals were digitalized in the pulse oximeter, and
Sony) and audiotaping 4 standardized phases of the heel lance downloaded into a personal computer (PC). ECG segments
that included: baseline (3 min), warming (2 min), post-proce- of 128 s in the baseline/warming and immediate post-proce-
dure (heel stick, heel squeeze; 30 s blocks) and return to base- dure/return to baseline phases were edited, linearly detrended
line (5 min) of the heel lance procedure. Facial actions were and analyzed for power spectral density using HRView soft-
coded second-to-second according to the NFCS (Grunau ware (Boston Medical Technologies, Boston, MA). Standards
and Craig, 1990) by two trained research assistants who did defining specific frequency bandwidths commonly used to
not have any interactions with the infants. The videotapes were characterize and study power spectral analysis of heart rate
played back in real time on a Panasonic AG-1970 with stop- variability (HRV) in infants were followed. Normalized power
frame capability and clocked to the 4th decimal place. Each spectral values were calculated and reported for: high-frequen-
recording session was scored repeatedly for each of the facial cy power (0.15–1 Hz); low frequency power (0.04–0.15 Hz) and
actions, using laptop computer software developed in the lab the ratio of low-/high-frequency power (Saul et al., 1991; Task
written in BASIC that recorded the scores and allowed for Force of European Society of Cardiology, 1996; Berntson
information on artifacts to be included. A final score based et al., 1997).
on percentage of time the action was present was calculated
for the block of time of interest. 2.3. Data collection measures
Cries were audio recorded using a small Sennheiser unidi-
rectional microphone that was placed 10 cm from the infant’s Severity of illness is an important construct in infants in the
mouth. The microphone was connected to a Sony 500 high-fre- NICU but there is no gold standard measure for this construct
quency audiotape recorder with a tone generator to mark in infants. Therefore, two measures were utilized to tap into
events. A research assistant, blind to the study design, was different domains of SOI in the neonatal period. The Score
trained to conduct cry analyses using CSPEECH (Milenkovic, for Neonatal Acute Physiology: Perinatal Extension
1998). The first cry from the stick phase of each heelstick pro- (SNAP:PE, Richardson et al., 1993) captures predicted risk
cedure was digitized at 20 kHz using a 16-bit analog to digital for mortality and is a 26 item scale based on routine clinical
converter and low-pass filtered with a high-frequency cut-off of tests and vital signs during the first 12 h of life to predict out-
10 kHz to avoid aliasing. Cry analysis from Fast-Fourier come. Birth weight, Apgar score at 5 min, and small size for
transform spectrography was performed using a Pentium GA are added to the base SNAP score to form a newborn ill-
microcomputer with C-SPEECH SP that was modified to ness severity and mortality risk score, the SNAP:PE. This mea-
accommodate fundamental frequencies up to 4 kHz. One sure has been validated with over 27,000 infants (Richardson
research assistant conducted the cry analyses under the super- et al., 2001). The Neonatal Therapeutic Interventions Scoring
vision of a cry expert co-investigator (CJ) who was blind to System (NTISS, Gray et al., 1992), adapted from the Thera-
group assignment. The screen-mark function was used to iden- peutic Interventions Scoring System (TISS, Cullen et al.,
tify the voiced and unvoiced segments of each cry. Subsequent 1974), captures SOI through the frequency and intensity of
analysis only considered the pitch characteristics for voiced diagnostic procedures performed over time during the NICU
segments of the cries. The percent of voiced/voiceless data hospitalization. Validity was established with 1643 infants by
were reported but no other measurements were made for the comparing NTISS scores with mortality risk estimates
voiceless segments. The definition of voiceless and voiced is (r = 0.69, p < 0.0001), in hospital mortality rates (p < 0.05)
based on work by Michelsson (Michelsson, 1971). Cries were and a measure of nursing acuity (Medicus, r = 0.69,
analyzed from the auditory perspective of pitch which refers p < 0.0001). The Neurobiologic Risk Score (NBRS, Brazy
to how high or low the cry sounds and is precisely measured et al., 1993), used to assess infant neurobehavioural outcome,
as fundamental frequency (F0) and the temporal perspective is a 7-item scale scored at discharge to determine the degree of
of cry duration. These two cry characteristics are commonly hypoxemia, ischemia, metabolic aberration, and brain injury
assessed due to their saliency and usefulness to both lay and occurring during the course of the infant’s hospitalization.
professional care providers as well as researchers. Intra-rater The NBRS has demonstrated significant correlation with the
reliability was conducted every six months and was 98% for Bayley Mental Index (r = 0.43) and Psychomotor Develop-
pitch. ment (r = 0.51) and neurologic examination score (r = 0.65)
for up to 24 months corrected age.
2.2.2. Physiologic response
Disposable ECG electrodes and pulse oximetry probes were 2.4. Data management and statistical analyses
placed on the infants and ECG, respiratory rate and oxygen
saturation were continuously recorded using a cardiorespirato- All raw facial action, heart rate and oxygen saturation data
ry monitor and personal computer (1000 Hz sampling rate). were coded and reduced by trained coders in Dr. Stevens’s lab
Mean respiratory rate was recorded and an electronic event using methods developed by Stevens et al. (Stevens et al., 1999)
marker was used to mark each phase of the procedure. Simul- for each of the 4 phases of the heel lance and summarized for
taneous verbal and electronic event markers signaled each each event phase. Expert data coders were blinded to risk
phase of the heel lance procedure on the physiologic data col- group allocation status and to the purpose of the study. The
lection system (electronic) and videotaped data (verbal). Phys- coders typically simultaneously code data from multiple ongo-
iologic indicators were recorded using a pulse oximeter ing studies in a laboratory far removed from the clinical setting
 B. Stevens et al. / Pain 127 (2007) 94–102 97

to keep them vague to the purpose of specific studies. To quan-

Met eligibility criteria and
tify facial expression, the Neonatal Facial Coding System approached (n=191)
(NFCS; Grunau and Craig, 1987) was used. The trained coders
have participated in facial coding over the past 5 years and
completed inter- and intra-rater reliability checks every 3
Refused
Refusedtoto
participate
participate
months (achieving levels > 95%). Cry was evaluated in terms (n=22)
of acoustic (F0) characteristics in Dr. Celeste Johnston’s lab (n=22)
and HRV indicators were analyzed in Dr. Linda Franck’s
lab by expert data coders who were blinded to risk group allo- Dropped out prior to
cation status and to the purpose of the study. study data collection
(n=20)
Descriptive statistics characterized the individual responses
to pain and demographic factors by infant NI risk group. Con-
tinuous measurements were summarized using means and Total participants N=149
standard deviations and frequencies as well as percentages to
describe categorical variables. Repeated measures analyses of Cohort A (n=54)
variance (RM ANOVA) with one between-subject factor (NI Cohort B (n=45)
Cohort) and one within-subject factor (Phase of heel lance pro- Cohort C (n=50)
cedure) were conducted to determine interaction and main
effects of NI Cohort or Phase. Repeated measures ANCOVA Fig. 1. Study flow diagram.
were conducted on the variables that had significant interac-
tions, controlling for measures of SOI (SNAP:PE and NTISS).
studies (3) or they did not want to participate in research
Simple effects analyses were also conducted to determine at
what phase of the heel lance differences were occurring and (3). Of 169 who agreed to participate, 20 were trans-
whether differences existed among cohort groups. Assump- ferred to another hospital or died prior to the start of
tions for using parametric statistics were assessed prior to the study. A total of 149 parents of neonates at high
the data analyses. A significance value of .05 was specified (Cohort A, n = 54), moderate (Cohort B, n = 45) and
for all statistical tests. SAS statistical software, version 8.2 low (Cohort C, n = 50) risk for NI participated in this
(SAS Institute Inc., Carey, NC) was employed for all data study (Fig. 1).
analyses.
3.2. Neonatal characteristics
3. Results
Characteristics of the neonates, comparative statistics
3.1. Participation rates and post hoc contrasts across cohorts are presented in
Table 1. Significant differences occurred in: birth weight,
A total of 191 parents of eligible infants were gestational age, Apgar score at 5 min and SNAP:PE
approached to participate in this study. Twenty-two scores. Post hoc tests indicated that infants in Cohort
parents refused to participate; stating that they were A had significantly higher birth weights than infants in
too overwhelmed (7), did not want their baby to be vid- Cohorts B and C, and significantly lower Apgar scores
eotaped (5), their infant had been through too much than infants in Cohort C. These indicators, although
while in the NICU (4), they were involved in too many statistically significant, were not thought to account

Table 1
Neonatal characteristics of study sample by group
Cohort A Cohort B Cohort C F, p value
GA at birth (weeks) 34.00 (5.27)* n = 54 31.11 (5.34) n = 45 31.42 (3.93) n = 50 F(2, 146) = 5.41, p = 0.005
Birth weight (g) 2227.74 (1134.44)** n = 54 1784.73 (1097.09) n = 45 1656.88 (735.69) n = 50 F(2, 146) = 4.62, p = 0.011
Apgar (5 min) 7.23 (2.14)*** n = 53 7.69 (1.34) n = 45 8.26 (1.27) n = 50 F(2,145) = 5.03, p = 0.008
SNAP: PE scorea 20.85 (21.65) n = 53 24.24 (21.20)**** n = 45 13.14 (16.66) n = 50 F(2, 145) = 3.92, p = 0.022
NTISS scoreb 15.58 (8.82) n = 53 16.67 (7.70) n = 45 14.44 (5.21) n = 50 F(2, 145) = 1.07, p = 0.346
NBRS Scorec 5.28 (4.12)***** n = 54 4.40 (2.83) n = 45 1.98 (1.68) n = 50 F(2, 146) = 15.65, p = 0.001
All values expressed as means (standard deviation).
a
Score for Neonatal Acute Physiology: Perinatal Extension, SNAP:PE score (Richardson and Escobar, 1998, Richardson et al.).
b
Neonatal Therapeutic Interventions Scoring System; NTISS (Gray et al., 1992).
c
Neurobiologic Risk Score; NBRS (Brazy et al., 1993).
* Cohort A significantly higher than Cohort B + C; no difference between Cohorts B + C.
** Cohort A significantly higher than Cohort B + C; no difference between Cohorts B + C.
*** Cohort A significantly lower Apgar scores than Cohort C; no difference between Cohorts A + B and Cohorts B + C.
**** Cohort B significantly higher SNAP:PE scores than Cohort C; no difference between Cohorts A + B and Cohorts B + C.
***** Cohort A significantly higher than Cohort C; no difference between Cohort A + B and Cohort B + C.
98 B. Stevens et al. / Pain 127 (2007) 94–102

Table 2
Neonatal clinical outcomes by group
Cohort A Cohort B Cohort C F, p value
Days on ventilator 9.83 (15.37) n = 53 24.58 (27.97)*
n = 45 7.86 (17.01) n = 50 F(2, 145) = 9.31, p = 0.001
Days on CPAP 4.39 (7.92) n = 54 6.69 (12.80) n = 45 3.66 (7.00) n = 50 F(2, 146) = 1.32, p = 0.269
Days on low flow O2 air 14.94 (27.88) n = 53 28.67 (38.69)** n = 45 14.18 (28.90) n = 49 F(2, 144) = 3.05, p = 0.050
Days until no apnea, bradycardia 26.90 (35.43) n = 51 46.17 (44.67)** n = 42 28.58 (30.26) n = 50 F(2, 140) = 3.75, p = 0.026
and/or desaturation spells
Days to regain birth weight 11.49 (5.21) n = 45 10.67 (4.73) n = 36 13.26 (4.74) n = 42 F(2, 120) = 2.89, p = 0.059
Days to full enteral feeds 17.50 (16.85) n = 50 25.61 (22.48) n = 38 19.83 (19.51) n = 48 F(2, 133) = 1.92, p = 0.150
Days to full bottle/breast feeds 33.69 (36.65) n = 26 46.28 (45.77) n = 29 37.22 (28.35) n = 36 F(2, 88) = 0.870, p = 0.423
Days in hospital 48.46 (40.58) n = 52 58.91 (46.20) n = 44 45.51 (33.06) n = 49 F(2, 142) = 1.41, p = 0.247
All values expressed as means (standard deviation).
* Cohort B significantly higher than Cohorts A and C; no difference between Cohorts A + C.
** Cohort B significantly higher than Cohort A; no difference between Cohorts A + C and Cohorts B + C.

for a clinically important difference. Infants in Cohort B Cohort C during the post-procedure phase of the heel
had significantly higher SNAP:PE scores than infants in lance only (p < 0.001).
Cohort C, which was considered clinically significant Cry data were available for 82/149 infants. Of the
and therefore was controlled for in the analyses of pain remaining 67 infants, 28 did not cry, 28 were ventilat-
indicators. ed and could not produce an audible cry and 11 cry
samples were unable to be analyzed. Controlling for
3.3. Neonatal clinical outcomes SOI, there was a statistically significant difference
between Cohort groups (F(2, 77) = 3.41, p = 0.0381) on
Clinical outcomes, comparative statistics and post the minimum fundamental cry frequency. Bonferronni
hoc contrasts across cohorts are presented in Table 2 adjusted p-values for pair-wise comparisons indicated
(NBRS scores are reported in Table 1). Statistically sig- a statistically significant difference between Cohorts
nificant differences were noted in days on the ventilator, A and B only (p-value = 0.0355). Similar results were
days until no apnea episodes and NBRS scores. Post hoc found for the mean fundamental cry frequency. Con-
tests revealed that infants in Cohort B (those at moder- trolling for SOI, there remained a statistically signifi-
ate risk for NI and who had the highest SNAP:PE cant difference between Cohort groups (F(2, 77) = 3.82,
scores) required significantly more ventilatory support p = 0.0262). Bonferronni adjusted p-values for pair-
than infants in Cohorts A and C. Consistent with more wise comparisons indicated a statistically significant
severe illness, infants in Cohort B had significantly high- difference between Cohorts A and B only (p-value =
er mean number of days until they were free of apnea, 0.0229). Cohort B, therefore, had significantly higher
bradycardia and/or desaturation spells compared to minimum and mean fundamental cry frequencies com-
Cohort A, but not with Cohort C. Infants in Cohort pared to Cohort A (Fig. 3). There were no significant
A had significantly higher NBRS scores than infants in differences between Cohorts for the duration of first
Cohort C. cry (mean duration of 2.84 s in Cohort A; 2.48 s in

3.4. Neonatal pain responses

Raw Facial Activity
3.4.1. Behavioural indicators
Controlling for SOI, a significant interaction between Cohort A
450
Cohort and Phase, F(6, 409) = 3.50, p = 0.0022 existed for Cohort B
400
total facial action (Fig. 2). Post hoc Bonferonni adjusted 350 Cohort C
pairwise analysis revealed that Cohort C had a signifi- 300
250
cantly higher proportion of total facial expressions than
200
Cohorts A (p = 0.006) and B (p < 0.001) in the post-pro- 150
cedure phase of the heel lance. Individual facial actions 100
including eye squeeze F(6, 427) = 3.79, p = 0.0011, nasola- 50
bial furrow F(6, 421) = 3.18, p = 0.0046, vertical stretch 0
mouth F(6, 409) = 3.29, p = 0.0036, and taut tongue Baseline Warm *Post Return to
Procedure Baseline
F(6, 391) = 2.35, p = 0.0303 yielded similar results. Simple
effects analysis conditional on the phase of heel lance for Phases of Heel Lance
these statistically significant interactions and controlling Fig. 2. Summary of total facial activity by cohort by phase. *p = 0.002
for SOI showed these facial indicators to be different in Differences between Cohort groups at post-procedure.
 B. Stevens et al. / Pain 127 (2007) 94–102 99

Summary of Cry Frequencies by infants (e.g., non-nutritive sucking). However, the con-
Cohort verse could also be true. This preliminary comparison
Cohort A of responses across cohorts is the precursor for generat-
1600 ing behavioural and physiological indicators of pain and
Cohort B*
1400
to determine if existing infant pain measures are valid
Cohort C for assessing pain in this vulnerable infant population.
1200 Assuming that pain is a multidimensional phenome-
1000 non, the best validated infant pain measures are multi-
Hz item scoring systems comprised of multidimensional
800
behavioural indicators or a composite of behavioural
600 and physiological indicators (Carbajal et al., 1997,
400 2003). A few of these measures include context and indi-
vidual infant characteristics (Stevens et al., 1996; van
200 Dijk et al., 2000) but none have incorporated risk for
0 NI. Infants with NI may show fewer and less clear emo-
Mean Maximum Minimum tional responses than healthy controls (Fraiberg, 1979;
Fundamental Frequency (F0) Yoder, 1987) making pain assessment more challenging.
All infants responded to heel lances performed in the
Fig. 3. Summary of cry frequencies by cohort. *Note: Cohort B is
significant higher for mean and minimum fundamental frequency. NICU, although response magnitude differed across
cohort groups. The meaning of response differences is
complicated as no direct ‘‘gold standard’’ measure of
Cohort B; 2.23 s in Cohort C); F(2, 79) = 1.35, pain exists in neonates. To better understand the mean-
p = 0.266. ing of response differences, we standardized the method
of conducting the painful procedure and collecting
3.4.2. Physiologic indicators response data over four distinct phases of the procedure
A significant main Phase effect was found across using a cadre of indicators that currently possess the
groups with increased maximum heart rate in the return best evidence for pain in this age group. Yet, it is a dis-
to baseline phase (F(3, 431) = 58.1, p = 0.001), increased tinct possibility that a broader range of indicators is
minimum heart rate in the post-procedure phase, required. From the physiologic domain, Slater et al.
(F(3, 431) = 78.7, p = 0.001), and lowest maximum oxy- (2006) and Bartocci et al. (2006) have both used Near
gen in the post-procedure phase (F(3, 425) = 47.6, Infrared Spectroscopy (NIRS) to evaluate cortical
p = 0.001). There were no Cohort by Phase interactions responses of infants to heel lance and intravenous starts,
and no Cohort group main effects with these physiologic respectively. From the behavioural perspective, War-
indicators. nock and Sandrin (2004) used an ethological approach
Controlling for SOI, there were significant Cohort by to identify 11 classes of behaviours within 5 phases of
Phase interactions for high-frequency HRV circumcision including head and upper body move-
(F(4, 214) = 2.74, p = 0.0294) and low frequency HRV ments, cry, hand movements, breath holding and facial
(F(4, 214) = 2.69, p = 0.0322) in the post-procedure phase grimacing. These approaches demonstrate the benefits
immediately following the heel lance. However, post hoc of systematic observation to articulate additional indica-
Bonferonni pair-wise comparisons, adjusted for SOI, tors to complement those that have already been
indicated no significant cohort differences. There was a validated.
significant Phase effect for the low frequency/high fre- To date, facial expression has been reported as the
quency HRV ratio (F(2, 216) = 4.97, p = 0.008) with the most consistent, sensitive and reliable indicator of pain
greatest decrease at post-procedure. in infants. However, there is speculation that differences
in facial musculature, hypotonia, and aberrant neural
4. Discussion information programming may affect facial pain
responses in this population (Ciccihetti and Sroufe,
Infants at risk for NI may suffer more pain than 1978; Ciccihetti and Beegley, 1990) but there has been
healthy infants due to; (a) discounting or denial of sig- no empirical validation during an acute painful event.
nals of acute distress by health professionals and parents If we hypothesize that facial expression is the most valid
(Walco et al., 1994), (b) misconceptions about the and consistent means of determining pain in infants with
capacity and sensitivity for pain, (c) effects of underlying NI, then, we may conclude, based on selected physiolog-
disease on the nociceptive system, (d) inadequate knowl- ical and behavioural indicators, that either the cohorts
edge to systematically study the problem (Oberlander differed in their pain experience or that the ability of
et al., 1999; Oberlander and O’Donnell, 2001) and (e) infants in the different cohorts to express or display pain
the inability to use analgesic methods used in other was the reason for the variation observed. Further
100 B. Stevens et al. / Pain 127 (2007) 94–102

exploration into factors that may influence this expres- a reliable index of central NI related to neurological
sion is required. These differences were sustained when damage, increased risk of poor developmental outcome,
the confounding effects of SOI were removed. However, and death in infants (Porges, 1995). The HRV analysis
this conclusion is only valid in somuch as the selected may be useful as an index of pain-related stress during
range of indicators employed. noxious procedures such as heel stick (Gunnar and
Crying is an innate behaviour that gains the attention Porter, 1995; Lindh et al., 1997; Oberlander et al.,
of caregivers. Certain diagnostic categories, specifically 2000; Grunau et al., 2001; Morison et al., 2001) and cir-
hyperbilirubinemia (Wasz-Hockert et al., 1971), cumcision (Porter et al., 1997). The interpretation of
asphyxia (Michelsson, 1971), crie du chat syndrome similarity in response across all the physiological param-
(Vuorenkosk et al., 1966), Down’s syndrome eters could, on the one hand, raise doubts about the
(Wasz-Hockert et al., 1968) and prematurity (Michels- added value of measuring HRV. This indicator may
son et al., 1983; Lester, 1987) were identified by Scandi- have particular importance in assessing pain in infants
navian researchers several decades ago using cry at risk for NI, because HRV responses indicated interac-
characteristics, in particular, F0, as well as melody and tive differences amongst the risk groups and particular
bi-phonation. In response to pain, the cries of infants phases of the procedure whereas HR and oxygen satura-
with Down syndrome or asphyxia (Fisichelli et al., tion did not. However, when SOI was controlled, cohort
1966; Lind et al., 1970; Michelsson, 1971; Michelsson differences were not sustained. Further research is need-
et al., 1977; Michelsson et al., 1983) were less frequent, ed to explore the complex interrelationships with other
less variable in intensity and F0 (pitch) and of longer reactivity systems within the individual. Challenges in
latency from the painful stimulus than cries in infants terms of feasibility and clinical utility of such an indica-
with no disabilities (Fisichelli et al., 1966). Conversely, tor will also need to be considered.
Robb found that infants ‘‘at risk’’ were no different from As pain has both immediate- and long-term
those of infants not at risk on cry characteristics (Robb consequences, we must be able to assess the efficacy of
and Goberman, 1997). This inconsistency may be attrib- pain-relieving interventions, to prevent or minimize con-
uted to the different methods of cry analyses and dura- sequences and enhance infant outcomes in all infants.
tion of the temporal domain. Oberlander et al. (2002) examined responses in preterm
In this study, F0 was significantly higher in infants in infants at 32 weeks PCA, who had parenchymal brain
Cohort B, who also had the highest SOI. When SOI was injury (PBI) compared to a matched group of infants
controlled in the analyses, the group at moderate risk without neurologic injury. No significant differences in
for NI continued to demonstrate higher minimum and facial expression (except to tongue protrusion), finger
mean fundamental frequency than other groups, splay, or physiologic indicators (HR, HRV) (Oberlander
suggesting that other factors may be involved either et al., 2002) were found. Although most of the infants
independently or in interaction with NI risk status to with PBI went on to develop cerebral palsy, in the neo-
account for differences. Evaluation of additional cry natal period, pain responses remained intact. These
characteristics including peak spectral energy, phona- results differ somewhat from the results of the present
tion, jitter and other temporal characteristics (latency, study, where differences in selected behavioural vari-
expiration, pause, inspiration, and rhythmicity) should ables existed between the cohort groups. Differences in
also be evaluated in future cry analyses. these two studies warrant further investigation to
Consistent with previous research (Craig et al., 1993; expand and validate indicators for pain assessment in
McIntosh et al., 1993; Stevens and Johnston, 1994; infants with a broad range of neurologic impairments
Stevens et al., 1994; Stevens et al., 1997; Abu-Saad and developmental ages.
et al., 1998; Stevens, 1998), heart rate increased and oxy- Establishment of the psychometric properties and
gen saturation decreased during the most painful phases feasibility of existing infant pain measures provides a
of the procedure. However, there was no impact of risk solid basis for understanding pain in infants with NI.
for NI on these variables. Because of the potentially Although the applicability of existing measures is not
altered physiological arousal and responsiveness to pain fully understood, until we achieve a more comprehen-
and other sensory stimuli in the neurologically impaired sive understanding of the impact of NI on infant pain
infant, new or additional physiological indicators may responses (and revised or novel indicators, measures or
be required to more accurately assess pain in these measurement approaches), both clinicians and research-
infants. Data from the present study challenge this ers need encouragement to use existing validated mea-
assumption to some extent; however, the rationale for sures. All these infants responded to pain in the most
the lack of response differences is not clear. painful phase of the heel lance procedure even though
Heart rate variability has been hypothesized to be a the response magnitude differed according to the NI risk
more sensitive indicator of parasympathetic and sympa- status in some indicators. Differences were mostly
thetic reactivity and autonomic function in relation to observed between infants at least risk for NI and either
acute and chronic stressors, such as pain. HRV has been those at moderate to severe risk (where few differences
 B. Stevens et al. / Pain 127 (2007) 94–102 101

existed). These results are important for health care pro- Berntson GG, Bigger Jr JT, Eckberg DL, Grossman P, Kaufmann PG,
viders who need to more carefully observe for these Malik M, et al. Heart rate variability: origins, methods, and
interpretive caveats. Psychophysiology 1997;34:623–48.
responses in this group of infants. This increased vigi- Brazy JE, Goldstein RF, Oehler JM, Gustafson KE, Thompson Jr
lance combined with further examination of indicators RJ. Nursery neurobiologic risk score: levels of risk and
not currently included in existing pain measures (e.g., relationships with nonmedical factors. J Dev Behav Pediatr
HRV, acoustic cry characteristics such as F0 or spectral 1993;14:375–80.
energy, imaging (Bartocci et al., 2006; Slater et al., 2006) Carbajal R, Paupe A, Hoenn E, Lenclen R, Olivier-Martin M. DAN:
evaluation behavioral scale of acute pain in newborn infants. Arch
and behaviours such as finger splay or a gaping mouth Pediatr 1997;4:623–8.
(Als et al., 1994; Holst et al., 2004), head and torso Carbajal R, Veerapen S, Couderc S, Jugie M, Ville Y. Analgesic effect
movements (Warnock and Sandrin, 2004)) will be cru- of breast feeding in term neonates: randomised controlled trial.
cial for establishing valid, reliable and feasible measures BMJ 2003;326:13–5.
for pain assessment. Additionally, attributing greater Ciccihetti D, Beegley M. Children with Down syndrome: a develop-
mental perspective. New York: Cambridge University Press; 1990.
attention to the under management of pain in this pop- Ciccihetti D, Sroufe LA. An organizational view of affect: illustration
ulation is crucial. from the study of Down syndrome infants. In: Lewis M, Rosenb-
lum L, editors. The development of affect. New York: Plenum;
1978. p. 309–50.
Coskun V, Anand KJS. Development of supraspinal pain processing.
Acknowledgements In: Anand KJS, Stevens BJ, McGrath PJ, editors. Pain in neonates.
2nd revised and enlarged edition. Amsterdam: Elsevier Science
Funding is acknowledged from the Canadian Insti- B.V.; 2000. p. 23–54.
tutes of Health Research (MOP-37884) and the Bloor- Craig KD, Grunau RVE. Neonatal pain perception and behavioral
measurement. In: Anand KJS, Stevens BJ, McGrath PJ, editors.
view Childrens Hospital Foundation. We would also
Pain in neonates. Amsterdam: Elsevier; 1993. p. 67–105.
like to acknowledge financial support from the Ontar- Craig KD, Grunau RVE, Linton J, Hadjistavropoulos H. Pain in the
io Ministry of Health for the Career Scientist Award preterm neonate: behavioral and physiological indices. Pain
and the Signy Hildur Eaton Chair in Paediatric Nurs- 1993;52:287–300.
ing Research at Sick Kids to B. Stevens and a Canada Cullen DJ, Civetta JM, Briggs BA, Ferrara LC. Therapeutic inter-
vention scoring system: a method for quantitative comparison of
Research Chair award to Patrick McGrath. We
patient care. Crit Care Med 1974;2:57–60.
acknowledge the research nurses who participated in Duhn L, Medves J. A systematic integrative review of infant pain
the collection and management of study data includ- assessment tools. Adv Neonatal Care 2004;4:126–40.
ing, Anne Jack, Marie Bagg, Mary Anne Fagan, Janet Fisichelli VR, Haber A, Davis J. Audible characteristics of the cries of
Narciso, Kim Caddell, Janet Chee, Salena Moham- normal infants and those with Down’s syndrome. Percept Mot
Skills 1966;23:744–6.
med-Breault, Sandy Lin and Karolina Kupczyk.
Fitzgerald M. Development of the peripheral and spinal pain system.
Finally, we are grateful for the willingness of the fam- In: Anand KJS, Stevens BJ, McGrath PJ, editors. Pain in neonates.
ilies of infants who agreed to participate in this re- 2nd revised and enlarged edition. Amsterdam: Elsevier Science
search study. B.V.; 2000. p. 9–21.
Fraiberg S. Blind infants and their mothers: an examination of the
system. In: Bullowa M, editor. Before speech. New York: Cam-
bridge University Press; 1979. p. 149–69.
References Gray JE, Richardsohn DK, McCormick MC, Workman-Daniels K,
Goldman DA. Neonatal therapeutic intervention scoring system: a
Abu-Saad H, Bours G, Stevens B, Jammars J. Assessment of pain in therapy-based severity-of-illness index. Pediatrics 1992;90:561–7.
the neonate. Semin Perinatol 1998;22:402–16. Grunau RVE, Craig KD. Facial activity as a measure of neonatal pain
Als H, Lawhon G, Duffy FH, McAnulty GB, Gibes-Grossman R, expression. In: Tyler DC, Krane EJ, editors. Advances in pain
Blickman JG. Individualized developmental care for the very low- research and therapy, pediatric pain. NewYork: Raven Press;
birth-weight preterm infant. Medical and neurofunctional effects. 1990. p. 147–53.
JAMA 1994;272:853–8. Grunau RVE, Craig KD. Pain expression in neonates: facial action
American Academy of Pediatrics Committee on Fetus and Newborn and cry. Pain 1987;28:395–410.
Committee on Drugs Section on Anesthesiology. Section on Grunau RVE, Oberlander TF, Whitfield MF, Fitzgerald C, Lee SK.
Surgery Canadian Paediatric Society Fetus and Newborn Com- Demographic and therapeutic determinants of pain reactivity in
mittee. Prevention and management of pain and stress in the very low birth weight neonates at 32 weeks’ postconceptual age.
neonate. Pediatrics 2000;105:454–61. Pediatrics 2001;107:105–12.
Anand KJ. Consensus statement for the prevention and management Gunnar MR, Porter FL. Neonatal stress reactivity: predictions to later
of pain in the newborn. Arch Pediatr Adolesc Med emotional temperament. Child Dev 1995;66:1–13.
2001;155:173–80. Heart rate variability. Standards of measurement, physiological
Anand KJ, Coskun V, Thrivikraman KV, Nemeroff CB, Plotsky PM. interpretation and clinical use. Task force of the European society
Long-term behavioral effects of repetitive pain in neonatal rat pups. of cardiology and the North American society of pacing and
Physiol Behav 1999;66:627–37. electrophysiology. Eur Heart J 1996;17: 354–81.
Anand KJS, Stevens BJ, McGrath PJ. Pain in neonates. 2nd revised Holst L, Grunau R, Oberlander T, Whitfield M. Specific newborn
and enlarged edition. Amsterdam: Elsevier Science B.V.; 2000. individualized developmental care and assessment program move-
Bartocci M, Bergqvist LL, Lagercrantz H, Anand KJ. Pain activates ments are associated with acute pain in preterm infants in the
cortical areas in the preterm newborn brain. Pain 2006;122:109–17. neonatal intensive care unit. Pediatrics 2004;114:65–72.
102 B. Stevens et al. / Pain 127 (2007) 94–102

Joseph KS, Kramer MS, Marcoux S, Ohlsson A, Wen SW, Allen A, Saul JP, Berger RD, Albrecht P, Stein SP, Chen MH, Cohen RJ.
et al. Determinants of preterm birth rates in Canada from 1981 Transfer function analysis of the circulation: unique insights into
through 1983 and from 1992 through 1994. N Engl J Med cardiovascular regulation. Am J Physiol 1991;261:H1231–45.
1998;339:1434–9. Shah V, Ohlsson A. Venepuncture versus heel lance for blood
Lester BM. Developmental outcome prediction from acoustic cry sampling in term neonates. The Cochrane Database of Systematic
analysis in term and preterm infants. Pediatrics 1987;80:529–34. Reviews 2004, Issue 4. Art. No.: CD001452.pub2. doi:10.1002/
Lind J, Vuorenkoski V, Rosberg G, Partanen TJ, Wasz-Hockert O. 14651858.CD001452.pub2.
Spectrographic analysis of vocal response to pain stimuli in infants Slater R, Cantarella A, Gallella S, Worley A, Boyd S, Meek J, et al.
with Down’s syndrome. Dev Med Child Neurol 1970;12:478–86. Cortical pain responses in human infants. J Neurosci
Lindh V, Wiklund U, Sandman PO, Hakansson S. Assessment of 2006;26:3662–6.
acute pain in preterm infants by evaluation of facial expression and Stevens B. Composite measures of pain. In: Finley GA, McGrath PJ,
frequency domain analysis of heart rate variability. Early Hum Dev editors. Measurement of pain in infants and children. Progress in
1997;48:131–42. pain research and measurement 1998;vol. 10. Seattle: IASP Press;
McIntosh N, Van Veen L, Brameyer H. The pain of heelprick and its 1998. p. 161–78.
measurement in preterm infants. Pain 1993;52:71–4. Stevens B, Johnston C. Physiological responses of premature infants to
Michelsson K. Cry analyses of symptomless low birth weight neonates a painful stimulus. Nurs Res 1994;43:226–31.
and of asphyxiated newborn infants. Acta Paediatr Scand Suppl Stevens B, Johnston C, Franck L, Petryshen P, Jack A, Foster G. The
1971;216:1–45. efficacy of developmentally sensitive interventions and sucrose for
Michelsson K, Jarvenpaa AL, Rinne A. Sound spectrographic analysis reliving procedural pain in very low birth weight neonates. Nurs
of pain cry in preterm infants. Early Hum Dev 1983;8:141–9. Res 1999;48:35–43.
Michelsson K, Sirvio P, Wasz-Hockert O. Pain cry in full-term Stevens B, Johnston C, Gibbins S. Pain assessment in neonates. In:
asphyxiated newborn infants correlated with late findings. Acta Anand KJS, Stevens BJ, McGrath PJ, editors. Pain in neonates.
Paediatr Scand 1977;66:611–6. 2nd revised and enlarged edition. Amsterdam: Elsevier Science
Milenkovic P. CSPEECH. Madison: University of Wisconsin; 1998. B.V.; 2000. p. 101–34.
Morison SJ, Grunau RE, Oberlander TF, Whitfield MF. Relations Stevens B, Johnston C, Horton L. Factors that influence the infant’s
between behavioral and cardiac autonomic reactivity to acute pain behavioural pain responses. Pain 1994;59:101–9.
in preterm neonates. Clin J Pain 2001;17:350–8. Stevens B, Johnston C, Petryshen P, Taddio A. The premature infant
Oberlander TF, Grunau RE, Fitzgerald C, Whitfield MF. Does pain profile: development and initial validation. Clin J Pain
parenchymal brain injury affect biobehavioral pain responses in 1996;12:13–22.
very low birth weight infants at 32 weeks’ postconceptional age? Stevens B, McGrath P, Gibbins S, Beyene J, Breau L, Camfield C,
Pediatrics 2002;110:570–6. et al. Procedural pain in newborns at risk for neurologic impair-
Oberlander TF, Grunau RVE, Whitfield MF, Fitzgerald C, Pitfield S, ment. Pain 2003;105:27–35.
Saul JP. Biobehavioral pain responses in former extremely low Stevens B, Taddio A, Ohlsson A, Einarson T. The efficacy of sucrose
birth weight infants at four months’ corrected age. Pediatrics for relieving procedural pain in neonates – a systematic review and
2000;105:e6. meta-analysis. Acta Paediatr 1997;86:837–42.
Oberlander TF, O’Donnell ME. Beliefs about pain among profession- Stevens B, Yamada J, Ohlsson A. Sucrose for analgesia in newborn
als working with children with significant neurologic impairment. infants undergoing painful procedures. The Cochrane Database of
Dev Med Child Neurol 2001;43:138–40. Systematic Reviews 2004, Issue 3. Art. No.: CD001069.pub2.
Oberlander TF, O’Donnell ME, Montgomery CJ. Pain in children doi:10.1002/14651858.CD001069.pub2.
with significant neurological impairment. J Dev Behav Pediatr Taddio A, Ohlsson K, Ohlsson A. Lidocaine–prilocaine cream for
1999;20:235–43. analgesia during circumcision in newborn boys. The Cochrane
Ohlsson A, Taddio A, Jadad AR, Stevens B. Evidence-based decision Database of Systematic Reviews 1999, Issue 3. Art. No.:
making, systematic reviews and the Cochrane Collaboration: CD000496. doi:10.1002/14651858.CD000496.
implications for neonatal analgesia. In: Anand KJS, Stevens BJ, van Dijk M, de Boer JB, Koot HM, Tibboel D, Passchier J,
McGrath PJ, editors. Pain in neonates. 2nd revised and enlarged Duivenvoorden HJ. The reliability and validity of the COMFORT
edition. Amsterdam: Elsevier Science B.V.; 2000. p. 251–68. scale as a postoperative pain instrument in 0 to 3-year-old infants.
Porges SW. Cardiac vagal tone: a physiological index of stress. Pain 2000;84:367–77.
Neurosci Biobehav Rev 1995;19:225–33. Vuorenkosk V, Lind J, Partanen TJ, Lejeune J, Laourcade J, Wasz-
Porter FL, Wolf CM, Gold J, Lotsoff D, Miller JP. Pain and pain Hockert O. Spectrograph analysis of cries from children with
management in newborn infants: a survey of physicians and nurses. maladie du cri du chat. Annales Pasediatrisa Fenniae
Pediatrics 1997;100:626–32. 1966;12:174–80.
Richardson DK, Corcoran JD, Escobar GJ, Lee SK. SNAP-11 and Walco GA, Cassidy RC, Schechter NL. Pain, hurt, and harm. The
SNAPPE-11: simplified newborn illness severity and mortality risk ethics of pain control in infants and children. N Engl J Med
scores. J Pediatr 2001;138:92–100. 1994;331:541–4.
Richardson DK, Escobar GJ. Simplified and revalidated score for Warnock F, Sandrin D. Comprehensive description of newborn
neonatal acute physiology (SNAP-11) maintains excellent predic- distress behavior in response to acute pain (newborn male
tive performance. Pediatr Res 1998;43:227A. circumcision). Pain 2004;107:242–55.
Richardson DK, Gray JE, McCormick MC, Workman K, Goldmann Wasz-Hockert O, Koivisto M, Vuorenkoski V, Lind J. Spectrographic
DA. Score for neonatal acute physiology: a physiologic severity analysis of pain cry in hyperbilirubinemia. Biol Neonate
index for neonatal intensive care. Pediatrics 1993;91:617–23. 1971;17:260–71.
Robb MP, Goberman AM. Application of an acoustic cry template to Wasz-Hockert O, Lind J, Vuorenkoski V, Partanen T, Valanne T. The
evaluate at risk newborns: preliminary findings. Biol Neonate infant pain cry: a spectrographic and auditory analysis. Lon-
1997;71:131–6. don: Spastics International; 1968.
Robertson CM, Svenson LW, Joffres MR. Prevalence of cerebral palsy Yoder PJ. Relationship between degree of infant handicap and clarity
in Alberta. Can J Neurol Sci 1998;25:117–22. of infant cues. Am J Ment Defic 1987;91:639–41.