Review

Tuberculosis as a cause or comorbidity of childhood

pneumonia in tuberculosis-endemic areas: a systematic review
Jacquie N Oliwa, Jamlick M Karumbi, Ben J Marais, Shabir A Madhi, Stephen M Graham

Pneumonia is a major cause of morbidity and mortality in infants and children worldwide, with most cases occurring Lancet Respir Med 2015
in tuberculosis-endemic settings. Studies have emphasised the potential importance of Mycobacterium tuberculosis in Published Online
acute severe pneumonia in children as a primary cause or underlying comorbidity, further emphasised by the January 29, 2015
http://dx.doi.org/10.1016/
changing aetiological range with rollout of bacterial conjugate vaccines in high mortality settings. We systematically
S2213-2600(15)00028-4
reviewed clinical and autopsy studies done in tuberculosis-endemic settings that enrolled at least 100 children aged
KEMRI Wellcome Trust Research
younger than 5 years with severe pneumonia, and that prospectively included a diagnostic approach to tuberculosis Programme, Department of
in all study participants. We noted substantial heterogeneity between studies in terms of study population and Public Health Research,
diagnostic methods. Of the 3644 patients who had culture of respiratory specimens for M tuberculosis undertaken, Nairobi, Kenya
(J N Oliwa MMed Paeds,
275 (7∙5%) were culture positive, and an acute presentation was common. Inpatient case-fatality rate for pneumonia
J M Karumbi BPharm); Marie
associated with tuberculosis ranged from 4% to 21% in the four clinical studies that reported pathogen-related Bashir Institute for Infectious
outcomes. Prospective studies are needed in high tuberculosis-burden settings to address whether tuberculosis is a Diseases and Biosecurity and
cause or comorbidity of childhood acute severe pneumonia. The Children’s Hospital at
Westmead, Sydney Medical
School, University of Sydney,
Introduction such as Pneumocystis jirovecii or cytomegalovirus. Sydney, NSW, Australia
Pneumonia is the leading cause of death in children Furthermore, most previous studies did not highlight the (B J Marais PhD); Medical
aged 1–59 months, accounting for an estimated 18% of potential importance of co-infections, as manifested by Research Council: Respiratory
and Meningeal Pathogens
under-5 mortality worldwide in 2011.1 In 2010, roughly a high prevalence of pneumococcal-respiratory viral
Research Unit
120 million episodes of pneumonia, 14 million severe co-infections (roughly 33%), which has since been (Prof S A Madhi PhD) and
pneumonia episodes, and 1·3 million deaths due to observed in children admitted to hospital with pneumonia Department of Science and
pneumonia in infants and children aged younger than in low-income, middle-income, and high-income Technology and National
Research Foundation: Vaccine
5 years were recorded.1–3 Most (81%) of these deaths settings.10,11 Furthermore, the studies were done before Preventable Diseases
occurred in the first 2 years of life. The epidemiology of the worldwide spread of the HIV epidemic. (S A Madhi), Faculty of Health
child pneumonia varies widely between different regions The HIV epidemic has had a major effect on the Sciences, University of the
of the world in terms of disease incidence, severity, and burden and mortality of pneumonia in children; Witwatersrand, Johannesburg,
South Africa; Centre for
associated mortality, and the contribution of causative bacterial pneumonia is more common and more severe International Child Health,
pathogens and prevalence of risk factors (table 1, in HIV-infected children compared with uninfected University of Melbourne
figure 1).4,5 Liu and colleagues2 report that most children.5 P jirovecii pneumonia (PCP) is frequently fatal Department of Paediatrics and
pneumonia episodes in children younger than 5 years in HIV-infected infants not receiving co-trimoxazole Murdoch Children’s Research
Institute, Royal Children’s
occurred in southeast Asia (39%) and Africa (26%), with preventive therapy and co-infections (concurrent Hospital, Melbourne, VIC,
sub-Saharan Africa accounting for 43% of pneumonia Australia
deaths, despite only constituting 19% of the world’s (Prof S M Graham PhD); and
under-5 population. International Union Against
Key messages
Tuberculosis and Lung Disease,
An understanding of the common causative pathogens • Tuberculosis is not often reported in young children Paris, France (S M Graham)
in high-burden settings is important to inform presenting with acute severe pneumonia in tuberculosis- Correspondence to:
case-management and potential preventive strategies, endemic settings Dr Jacquie Narotso Oliwa, KEMRI
such as vaccine development and delivery. Case- • Tuberculosis might be a direct cause of severe pneumonia Wellcome Trust Research
management and immunisation strategies have been Programme, Department of
or might be an underlying comorbidity that increases the Public Health Research,
informed by studies done in the 1980s which identified risk of secondary bacterial pneumonia 197 Lenana Place, Lenana Road,
Streptococcus pneumoniae and Haemophilus influenzae as • Clinical and autopsy studies have confirmed tuberculosis Nairobi, Kenya PO Box 43640-
the most common bacterial pathogens causing pneumonia in children that have died with severe pneumonia 00100
in children.6,7 These studies also showed that most [email protected]
• Restrictions of tuberculosis diagnostic techniques in
pneumonia-related deaths were due to bacterial rather children hinder estimation of actual burden and improved
than viral pneumonia, with the exception of measles. case detection
However, even in the case of measles-associated • Data on tuberculosis in children with acute severe
pneumonia deaths, 47–55% were associated with bacterial pneumonia are from a small number of studies in mainly
superinfection with S pneumoniae identified in 30–50% large urban-based hospitals with marked heterogeneity in
of confirmed bacterial co-infections.8 The diagnostic diagnostic approaches
techniques used in these studies restricted identification of • The non-specific clinical presentation of pulmonary
pathogens to bacteria and known common viruses.7,9 They tuberculosis in infants and young children highlights the
did not use diagnostics specific to the identification of urgent need for improved diagnostic instruments
M tuberculosis, atypical bacteria, or opportunistic pathogens

www.thelancet.com/respiratory Published online January 29, 2015 http://dx.doi.org/10.1016/S2213-2600(15)00028-4 1

 Review

bacterial, mycobacterial, fungal, or viral) are common in other causes of pneumonia, including measles vaccine,
HIV-infected children.12 Additionally, the HIV epidemic H influenzae type b (Hib), and pneumococcal conjugate
has substantially increased the incidence and trans- vaccines.17,20,21 Tuberculosis needs specific treatment (in
mission of tuberculosis in HIV endemic settings, contrast to many respiratory viruses) and treatment
particularly in young women, greatly increasing the risk outcomes in young children are usually excellent.
of tuberculosis in their infants.13 Reversal of the burden The contribution of tuberculosis to the burden of
of HIV in infants has been encouraging, with increasing pneumonia and death in childhood as a direct cause or
coverage of prevention of mother-to-child transmission underlying contributing factor is still poorly quantified.
of HIV, early antiretroviral therapy, and co-trimoxazole Cause-specific mortality estimates are usually modelled
prophylaxis for HIV-infected and HIV-exposed infants.14 from vital registration data with historical assumptions
and allow only the reporting of a single cause of death,
Potential contribution of tuberculosis to childhood which in the context of respiratory disease is not
pneumonia pathogen-specific. Additionally, the fact that children with
Although tuberculosis is a curable and preventable acute pneumonia symptoms might have microbiologically
disease, it is the second leading cause of death from an confirmed tuberculosis contradicts traditional teaching
infectious agent after HIV. In 2013, about 9∙0 million new and standard case management, in which tuberculosis is
cases of tuberculosis occurred, with 1∙5 million deaths only considered in children with prolonged persistent
worldwide, and most of the cases were from Asia and symptoms. The concept that tuberculosis might increase
Africa.15 Roughly 550 000 of the new cases were in children, susceptibility to secondary bacterial pneumonia in young
with 80 000 deaths in those who were HIV-uninfected.15,16 children is also not widely appreciated.
This number might be an underestimate owing to the To assess the association of tuberculosis with childhood
challenges of establishing the diagnosis of tuberculosis in pneumonia in tuberculosis-endemic areas, we did a
children. There is a growing awareness that children have systematic review of published literature reporting the
a high burden of tuberculosis-related disease that is often causes of severe pneumonia in infants and young
not reported as such.17 children that prospectively evaluated these children for
Previous studies of pneumonia in infants and young multiple infectious causes, including M tuberculosis. By
children might also have underestimated the contribution reviewing the available data on prevalence, clinical
of tuberculosis as a direct cause or comorbidity of acute presentation, diagnostic approaches, co-infection, and
community-acquired pneumonia in children because of outcome, we aimed to provide an overview of knowledge
the difficulties of microbiological confirmation in this age gaps and a resource for future research and advocacy.
group, especially in resource-restricted tuberculosis-
endemic settings.5 These settings are the ones that have Search strategy and selection criteria
the highest incidence of childhood pneumonia and We included studies of any design that were done in a
pneumonia-related mortality (figures 1, 2).3,4,16 Additionally, tuberculosis-endemic setting (country incidence ≥50 new
these settings have the highest prevalence of childhood cases per 100 000 people per year at the time of the study);
malnutrition and HIV infection worldwide, both common enrolled at least 100 children aged younger than 5 years
comorbidities that increase the risk and the mortality of who had a diagnosis of pneumonia or respiratory tract
tuberculosis and of pneumonia in young children.17–19 infection (defined as clinical evidence of severe or very
Furthermore, the relative interaction with tuberculosis as a severe pneumonia according to WHO criteria of acute
cause or contributor to childhood pneumonia in respiratory infection,22 or radiological evidence of lobar or
tuberculosis endemic areas is likely to be changed with patchy consolidation); and included a tuberculosis
increasing global uptake of vaccines that protect against diagnostic workup and described the diagnostic approach
in sufficient detail. We included studies that reported
additional comorbidities such as HIV or malnutrition as
Population aged Incidence per Total episodes Total deaths
<5 years (2010) child-year (×10⁶) (×10³) long as a lower respiratory tract infection provided the
main point of entry into the study. Case reports or series,
Africa 133 340 762 0·27 (0·14-0·63) 36·4 (18·2–84·4) 540·6 (43·8–627·3)
studies with older populations, and those done in high-
Americas 76 995 700 0·08 (0·04-0·18) 6·4 (3·3–14·5) 23·9 (22·6-35·6)
income countries not endemic for tuberculosis (incidence
Eastern 72 151 965 0·23 (0·11-0·53) 16·4 (8·2–38·0) 168·4 (147·3–217·1)
Mediterranean
<50 new cases per 100 000 people per year) were excluded.
Europe 54 605 243 0·03 (0·02-0·04) 1·6 (1·3–2·1) 18·1 (14·7–23·4)
The primary outcomes considered were the numbers and
proportions of tuberculosis cases diagnosed clinically or
Southeast Asia 179 956 087 0·26 (0·13–0·61) 47·4 (23·7–109·8) 443·8 (336·7–534·2)
culture-confirmed in children aged younger than 5 years
Western Pacific 116 411 580 0·11 (0·05-0·24) 12·2 (6·2–28·2) 61·9 (50·7–78·0)
with pneumonia.
World 633 461 337 0·19 (0·10–0·44) 120·4 (60·8–277·0) 1256·8 (1053·2–1482·9)
We recognised studies as potentially highly hetero-
The data in parentheses are uncertainty ranges. Adapted from Walker and colleagues.1 geneous but did not exclude any because of perceived
low quality (STROBE checklist).23 Heterogeneity included
Table 1: Pneumonia disease burden estimates by WHO region in children aged 0–4 years (2011)
study population, study setting and diagnostic methods

2 www.thelancet.com/respiratory Published online January 29, 2015 http://dx.doi.org/10.1016/S2213-2600(15)00028-4

 Review

Episodes per child-year

≤0·10
0·11–0·20
0·21–0·30
0·31–0·40
0·41–0·50

Figure 1: Incidence of clinical pneumonia in children less than 5 years of age (2012)4
Figure adapted from the World Health Organization (WHO) with permission. Small circles represent island populations.

Estimated new TB cases (all forms)

per 100 000 population per year
0–9·9
10–19
20–49
50–124
125–299
300–499
≥500
No data
Not applicable

Figure 2: WHO estimated tuberculosis incidence estimates per country (2013)16

Figure adapted from the World Health Organization (WHO) with permission. Small circles represent island populations.

used to clinically diagnose or microbiologically confirm across and within studies did not allow for pooled
tuberculosis, and all recognised factors that have a risk of estimates or meta-analysis of variables. Rather, the main
bias across studies for detection of the primary outcome characteristics for individual studies were listed. The
of the review and for mortality. Assessment of the risk of gold standard for tuberculosis diagnosis is culture
bias of individual studies identified potential sampling confirmation and so the principal summary measures
bias in many of the clinical studies (appendix). The that we aimed to report were the pooled numbers of See Online for appendix
substantial heterogeneity and recognised risk of bias culture-confirmed tuberculosis and as a proportion of

www.thelancet.com/respiratory Published online January 29, 2015 http://dx.doi.org/10.1016/S2213-2600(15)00028-4 3

 Review

rates of HIV infection.26,31–33,36,37 The other five were

949 records identified through database searching 4 records identified from other sources (experts in studies of children treated in hospital from a wide range
the field and bibliographies)
of high-burden tuberculosis settings (reported incidence
50–299 cases per 100 000 population per year) including
rural Africa, and low-HIV-prevalence settings in Asia,
953 records from all relevant sources namely Bangladesh and China.24,25,27,29,35 The three autopsy
studies were also from southern African countries with
825 records excluded after review of the title
very high tuberculosis incidence (South Africa, Zambia,
and Zimbabwe).28,30,34
128 abstracts reviewed
Tuberculosis diagnosis
Studies were done in hospitals with varying capabilities
90 abstracts excluded to microbiologically confirm tuberculosis, and not
30 reviews
18 case reports
all studies included mycobacterial culture for
20 foreign languages microbiological confirmation. Diagnostic approaches for
22 not relevant each study are summarised in table 2. Potential sampling
bias occurred within studies; four studies25,27,31,33 collected
38 full articles retrieved samples for culture for M tuberculosis in all study
participants, whereas four studies24,26,32,35 collected
7 full articles excluded samples for culture in a subset of participants, in which
5 foreign language criteria for selection were not clearly mentioned.
2 duplications (same study population)
One study used Xpert MTB/RIF (Cepheid, CA, USA)
additionally to culture in 214 children.27 Most of the 3644
31 full articles assessed for eligibility
samples taken for culture or Xpert MTB/RIF were
sputum samples obtained by induced sputum technique
17 articles based on inclusion criteria or gastric lavage, with an additional 94 samples from a
(low sample size, low tuberculosis
endemicity or high income settings) direct lung aspirate. One study reported multiplex PCR
results from a nasopharyngeal sample.29
Chest radiographs were done as part of the diagnostic
14 studies included in the final review
evaluation for pneumonia in all the studies but only
three of the studies27,30,31 compared radiological findings
Figure 3: PRISMA flow diagram of studies included in the Review
with those with a diagnosis of tuberculosis and those
those that had respiratory specimens cultured, with risk without. In a study of severely malnourished children in
of bias explicitly acknowledged. We deemed the risk of Bangladesh,27 no differences were noted except that one of
bias of individual studies for reporting of inpatient the 27 confirmed tuberculosis cases had a miliary pattern.
deaths negligible. In the South African autopsy study of HIV-infected
children,30 no differences were noted in radiological
Findings patterns between those with tuberculosis and those with
Study overview other HIV-related lung diseases.30 The clinical study by
Our search identified 14 articles that were eligible for the Zar and colleagues reported that hilar or mediastinal
final analysis: 11 prospective clinical studies and adenopathy was significantly more common in children
three autopsy studies (figure 3, table 2).24–37 We noted with tuberculosis than in those without (43% vs 12%).31
substantial heterogeneity between studies in terms of
factors that would potentially affect yield of tuberculosis Contribution of tuberculosis to pneumonia
diagnosis in children with pneumonia such as inclusion The proportion of pneumonia cases that were diagnosed
criteria, study setting, background tuberculosis incidence with tuberculosis ranged from 1% to 23%. The proportion
at the time of the study, prevalence of comorbidities such of culture-confirmed tuberculosis in children with
as severe malnutrition, and techniques used for sputum pneumonia also varied widely between the nine studies
collection and microbiological confirmation. The that included M tuberculosis culture, with five studies25–27,31,32
11 clinical studies of children treated in hospital included reporting culture-confirmed rates of 5–8%. One study
a total of 6504 infants and children with a clinical or reported culture-confirmed rates of 15%.33 Overall, of the
radiological diagnosis of pneumonia that were aged infants and young children with pneumonia who had
younger than 5 years. culture of respiratory specimens for M tuberculosis
Six of these studies were done in large urban-based performed, 7∙5% (275 of 3644) were culture positive. The
hospitals in South Africa and Malawi, which are very proportion of culture positive cases was higher in settings
high-burden tuberculosis settings (reported incidence with a very high tuberculosis burden at the time of the
≥300 cases per 100 000 population per year) with high study (incidence of ≥300 cases per 100 000 population per

4 www.thelancet.com/respiratory Published online January 29, 2015 http://dx.doi.org/10.1016/S2213-2600(15)00028-4

 Review

Country Tuberculosis Participants Duration of Inclusion Tuberculosis Tuberculosis cases Case-fatality rate and HIV prevalence
(setting) population (age) symptoms on criteria diagnosis (% of enrolled) other characteristics (number of HIV-
incidence presentation of tuberculosis cases infected over
per 100 000 for tuberculosis number of
per year 38* cases tuberculosis cases
tested for HIV)
Very-high-burden settings (tuberculosis incidence ≥300 cases per 100 000 people per year)
Graham37 Malawi (urban 328 288 of under-5’s Not reported WHO severe or Clinical† 5 (1·7%) 20% (1 of 5) died aged 40% (2 of 5)
(2005–06) and peri- (median very severe 5 months
urban) 5 months pneumonia
[range 2–59])
Moore26 South Africa 406 2439 77% of cases had Admission to Culture of sputum 421 (17%) of 2439 376 first and 64% (241 of 376)
(1998– (urban) (3–59 months) cough for hospital for lower in 1334 children enrolled; 90 (7%) of 45 recurrent episodes;
2006) <10 days respiratory tract when tuberculosis 1334 sputum 4% (4 of 90) of
duration infection clinically samples culture culture-confirmed
suspected confirmed cases died in hospital;
49% (206 of 421)
cases discharged
following response to
empirical antibiotics
and not initiated on
tuberculosis treatment
McNally33 South Africa 780 358 (median 85% of cases had WHO severe or Culture of sputum 53 (15%), all 64% (34 of 53) of 72% (38 of 53)
(2001–02) (urban) 4·8 months symptoms for very severe culture confirmed cases were aged
[IQR 2·7–13]) <2 weeks pneumonia <1 year; 11 (21%) died;
maternal tuberculosis
associated with poor
outcomes
Zar31 (1998) South Africa 406 250 (median Enrolment WHO severe or Culture of sputum 20 (8%), all culture 15% (3 of 20) died 55% (11 of 19)
(urban) 6 months criteria: cough very severe confirmed
[IQR 3–16]) <14 days pneumonia
duration
Madhi32 South Africa 406* 1215 Enrolment WHO severe or Culture of sputum 69 (6%); 69 (8%) 58 (84%) aged 52% (36 of 69)
(1997–98) (urban) (2–59 months) criteria: cough very severe in 858 children of 858 culture <2 years; 7 (10%) also
for <14 days pneumonia when tuberculosis confirmed had bacteraemia
duration clinically
suspected
Graham36 Malawi (urban 479 150 (median Not reported WHO severe or Clinical† 9 (6%) All cases had close 89% (8 of 9)
(1996) and 5 months [IQR very severe tuberculosis contact
peri-urban) 2–59]) pneumonia and poor response to
antibiotics
High-burden settings (tuberculosis incidence 50–299 cases per 100 000 people per year)
Nantongo25 Uganda 193 231 (median 37% of cases had WHO severe or Clinical†; culture of 37 (16%) cases; 24 (65%) aged 28% (14 of 51)
(2011) (urban) 15 months [IQR cough for very severe sputum 12 (5%) culture- <2 years; young age
7–36]) <2 weeks pneumonia confirmed (<1 year) and contact
history associated with
confirmed tuberculosis
Chisti27 Bangladesh 225 385 (median Median duration Severely Clinical†; culture 8 (23%); 27 (7% ) 4 (5%) died within Not tested; low HIV
(2011–12) (urban) 10 months [IQR of cough for malnourished; (n=385) and culture or Xpert‡ 3 months prevalence setting
2-59 months]) cases: 7 days (IQR radiological Xpert‡ (n=214) of confirmed
4–8) consolidation sputum
Hammitt24 Kenya (rural) 298 810 Not reported WHO severe or Clinical†; culture 5 (0·6%); 2 (2%) 108 investigated for Not reported for
(2010) (1–59 months) very severe of sputum of 108 sputum tuberculosis were tuberculosis cases;
pneumonia (n=108) sampled culture selected from 10% for severe
confirmed 810 severe pneumonia pneumonia cases
cases
Wang29 China (urban) 92 100 (mean Not reported Radiological Multiplex PCR of 1 (1 %) S pneumoniae and Not tested; low HIV
(2004–05) 15·7 months) evidence of nasopharyngeal M tuberculosis prevalence setting
pneumonia specimens; identified in same
culture not done specimen
Adegbola35 The Gambia 189 278 Not reported WHO severe or Culture of lung 5 (1·8%); 2 (2 %) All 5 cases were 2% (3 of 155)
(1990–92) (urban and (3–58 months) very severe aspirate (n=94) or of 120 sampled severely malnourished; malnourished
peri-urban) pneumonia; induced sputum culture confirmed 2 cases also had subgroup
radiological (n=26) bacteria cultured from
consolidation lung aspirate
(Table 2 continues on next page)

www.thelancet.com/respiratory Published online January 29, 2015 http://dx.doi.org/10.1016/S2213-2600(15)00028-4 5

 Review

Country, Tuberculosis Participants Duration of Inclusion criteria Tuberculosis Tuberculosis cases Case-fatality rate and HIV prevalence
(setting) population (age) symptoms on diagnosis (% of enrolled) other characteristics (number HIV-
incidence presentation of tuberculosis cases infected over
per 100 000 for tuberculosis number of
per year 38* cases tuberculosis cases
tested for HIV)
(Continued from previous page)
Autopsy studies
Chintu34 Zambia (urban 645 264 Not reported Death from Histopathology 54 (20%); 35 (65%) aged 59% (32 of 54)
(1997–2000) and peri-urban (1–192 months) respiratory including Ziehl- pulmonary <18 months; 12 cases
disease in Neelsen stain tuberculosis in had concurrent
hospital 42 cases and miliary pyogenic pneumonia
tuberculosis in
12 cases
Rennert30 South Africa 406 93 (mean No case had HIV-related death Histopathology 4 (4%) 3 (13% ) of 23 deaths All HIV-infected
(1998–99) (urban) 10·5 months cough for with antemortem including Ziehl- in children of 1 year or
[range 1·5–69·8]) >1 week lung disease Neelsen stain and older were
culture tuberculosis cases
Ikeogu28 Zimbabwe 362 184 (mean Not reported Dead on arrival or Microscopy and 8 (4%); All severely 75% (6 of 8)
(1992–93) (urban and 11·1 months shortly thereafter culture of lung 4 disseminated and malnourished; 6 cases
peri-urban) [range 1–55]) tissue 4 pulmonary had concurrent
tuberculosis pyogenic pneumonia

*Tuberculosis incidence per 100 000 population at time of study from World Bank Estimates.38 †“Clinical” included history of contact, response to antibiotics, chest radiograph, and tuberculin skin test. ‡Xpert
MTB/RIF (Cepheid, CA, USA).

Table 2: Studies assessing the contribution of tuberculosis to pneumonia in children aged younger than 5 years in tuberculosis-endemic areas

year; 232 (8%) of 2800 pneumonia cases in which samples with antituberculosis treatment started later once culture
were available for culture) than in studies done in high results became available.26
tuberculosis burden settings (incidence of 50–299 cases
per 100 000 people per year; 43 (5%) of 844 pneumonia Association with HIV infection
cases in which samples were available for culture). HIV co-infection was common (28–89%) in children
diagnosed with tuberculosis in the HIV-endemic settings
Relation to vaccine coverage of eastern and southern Africa.25,26,32,33,36,37,39 One study32
The national immunisation programme in six of the study reported a 23-fold (95% CI 13–48) higher incidence of
sites included Hib conjugate vaccine in early infancy.24–27,33,37 admission to hospital with culture-confirmed
The only study26 that included children who received a tuberculosis presenting as acute severe pneumonia in
pneumococcal conjugate vaccine reported follow up of a HIV-infected children aged younger than 2 years (1470
randomised placebo-controlled trial of the nine-valent cases per 100 000 per year) than in HIV-uninfected
pneumococcal conjugate vaccine in South African infants. children (65 per 100 000 per year). However, the
The main aim of the study was to assess protective efficacy proportion of patients that were culture positive for
against invasive pneumococcal disease and all-cause M tuberculosis was similar between HIV-infected and
radiological confirmed pneumonia during the first 2 years HIV-uninfected children treated in hospital for acute
of life. A post-hoc vaccine-probe analysis from this study26 pneumonia in studies in HIV-endemic settings.25,26,32,33,39
estimated that 43–47% of treatment in hospital for culture-
confirmed tuberculosis in HIV-infected and HIV- Mortality
uninfected children in this setting could be due to Mortality in children with pneumonia and diagnosis of
superimposed pneumococcal co-infection. tuberculosis was not consistently reported. In those
studies that reported inpatient deaths in tuberculosis
Symptoms associated with tuberculosis cases from HIV-endemic African settings, case-fatality
The duration of respiratory symptoms such as cough rates ranged from 4% to 21%.26,31,33,37 The study of severely
before admission was acute in most patients with malnourished Bangladeshi children27 followed all children
tuberculosis when this feature was reported in the study until 12 weeks after discharge and reported deaths in four
(table 2),26,27,31–33 with the exception of the Ugandan study25 (four of 86 [5%] patients with tuberculosis that were
that reported persistent cough of more than 2 weeks’ discharged; note, one patient died of tuberculosis in
duration was more common in pneumonia cases with hospital) of the patients with tuberculosis. Autopsy
tuberculosis compared with those without. One study studies provide additional data on the contribution of
noted that 49% of patients with tuberculosis responded tuberculosis to pneumonia-related deaths in children.
to first-line empirical antibiotic treatment for community- This contribution ranged from 4% to 20% in children
acquired pneumonia and were well enough to discharge who died from respiratory disease in three settings with

6 www.thelancet.com/respiratory Published online January 29, 2015 http://dx.doi.org/10.1016/S2213-2600(15)00028-4

 Review

very high tuberculosis incidence rates.28,30,34 These autopsy have been reported, with or without respiratory disease,
studies were also done at the peak of the HIV epidemic and clinical diagnosis is especially challenging in this
and before the rollout of preventive measures, such as group.43 Third, many of the studies were in HIV-endemic
co-trimoxazole preventive therapy and universal settings before the rollout of interventions that have
antiretroviral therapy for HIV-infected children. The substantially reduced HIV prevalence in young children in
selection criteria in these studies were highly variable and those settings and reduced the susceptibility to tuberculosis
only one study30 provided antemortem clinical data of children that are living with HIV. Although we noted the
(table 2). Disseminated tuberculosis was common, as prevalence of tuberculosis in patients with pneumonia
were co-infections of tuberculosis with pyogenic being similar between HIV-infected and HIV-uninfected
pneumonia in children (most were younger than 5 years children,25,26,31–33 the risk of tuberculosis was increased in
of age) dying from respiratory disease. Polymicrobial HIV-infected children not receiving antiretroviral
infections were also noted to be common and associated therapy.32,44 Finally, the two studies from Malawi relied on
with a worse outcome in one of the clinical studies,33 with clinical suspicion, such as a positive contact history and
M tuberculosis identified in 18% of HIV-infected and 29% poor response to antibiotics, and reported the lowest
of HIV-uninfected infants with acute pneumonia who prevalence of tuberculosis of studies from the highly
failed empirical first-line antibiotic therapy. endemic countries.36,37 Relying solely on clinical criteria for
the diagnosis of tuberculosis might overestimate rather
Discussion than underestimate the prevalence of the disease;45
Pneumonia is a major cause of under-5 mortality however, this issue might not be the case in children with
worldwide, and tuberculosis is a treatable and preventable tuberculosis who present to hospital when they have an
disease in young children that most often presents as a acute bacterial pneumonia because they might respond to
lower respiratory tract disease. This Review provides antibiotics and the underlying tuberculosis might be
evidence of the prevalence of tuberculosis in infants and missed, as noted in the study that followed the
young children admitted to hospital with predominantly pneumococcal conjugate vaccine study cohort.26
acute pneumonia in a range of tuberculosis-endemic Case-management guidelines often advise health
settings. The findings of this Review should, however, be workers to consider the diagnosis of tuberculosis in
interpreted with caution because of the heterogeneity of infants and children with chronic cough. Tuberculosis is
study populations and diagnostic approaches between known to be common in studies of children with
studies, the sampling bias for diagnosis within some persistent cough in tuberculosis-endemic settings.46,47
studies, and the acknowledged difficulties of diagnosis of However, in this Review we noted that many of the
tuberculosis in children. In view of the poor specificity of confirmed tuberculosis cases presented with acute
clinical features of tuberculosis in young children,40 the cough.25–27,31–33 Furthermore, many of the study participants
most robust data are provided by studies that sought were infants. Although tuberculosis can directly cause
culture confirmation. An important finding was that 275 of severe pneumonia and disseminated disease, especially
3644 (7∙5%) of patients with severe pneumonia in whom in infants, many of these children are likely to present to
respiratory specimens were collected for M tuberculosis hospital with a bacterial pneumonia complicating
culture had culture-confirmed disease (especially because underlying pulmonary tuberculosis. Many of the studies
culture has low diagnostic sensitivity in young children reported bacterial–tuberculosis co-infection.28,29,32–35
with intrathoracic tuberculosis at about 30–60%, One study reported that 10% of culture-confirmed
dependent on the specific disease manifestation).41,42 Our tuberculosis cases also had bacteria isolated from blood
findings also show that tuberculosis might be an important culture, despite this being a test of low sensitivity (5–15%)
contributor to pneumonia-related deaths in young children for bacterial pneumonia.32 Furthermore, tuberculosis
because of underdiagnosis or comorbidity predisposing to cases improved with antibiotics for community-acquired
bacterial co-infection.11,26,31,33,34 pneumonia, and admission to hospital with tuberculosis
The findings from these studies are not likely to be was significantly less common in children who had
representative of the epidemiology of childhood received the pneumococcal conjugate vaccine compared
pneumonia in tuberculosis-endemic areas in general. with placebo.26 A seasonal correlation between invasive
First, four of the studies were from large urban hospitals pneumococcal disease and tuberculosis cases that is
in South Africa,26,31–33 a country that is highly endemic for particularly pronounced in HIV-infected individuals has
tuberculosis and HIV, with routine access to conjugated been reported by the same group in Johannesburg.48 On
Hib and pneumococcal vaccines at the time of the studies. the basis of the results from the pneumococcal conjugate
Second, the studies from Bangladesh and The Gambia vaccine-probe design and clinical response to empirical
focused on tuberculosis diagnosis in malnourished antibiotic treatment against bacterial pneumonia, the
children with respiratory symptoms.27,35 Although the scarce evidence shows that almost half the children with
bidirectional association between tuberculosis and culture-confirmed tuberculosis were admitted to hospital
malnutrition is well recognised, surprisingly few data on because of bacterial (and particularly pneumococcal)
the prevalence of tuberculosis in malnourished children pneumonia.26

www.thelancet.com/respiratory Published online January 29, 2015 http://dx.doi.org/10.1016/S2213-2600(15)00028-4 7

 Review

The first step is for clinicians managing infants and young

Search strategy and selection criteria children with severe pneumonia in tuberculosis-endemic
We searched PubMed, clinical trials register, Cochrane reviews, DARE, Embase, and CINHAL countries to be aware that tuberculosis might be a cause or
for articles published in English between April 30, 1949 and May 27, 2014 with the MeSH contributor. At present, this recognition is not the case and
search terms “child*”, “infan*”, “paediatric”, “pediatric”, “tuberc*”, “TB”, “MTb”, “mycobact*” an aim of this Review is to improve awareness. Improved
(restrict to pulmonary TB), “pneum*”, “ARI” (acute respiratory tract infections), “LRTI” (lower diagnosis will probably depend on the future development
respiratory tract infections), “mortality”, “death”, “morbidity”, “infect*”, “inciden*”, “preval*”. of a point-of-care test that does not rely on sputum
Search terms used included “tuberculosis OR mycobacter* AND pneumonia AND child OR sampling. This test is now an important focus of
children OR paediatric OR pediatric” and “tuberculosis [MeSH Terms] OR tuberculosis [All research.52,53 Studies also need to be done in a wider range
Fields] AND pneumonia [MeSH Terms] OR pneumonia [All Fields] AND child [MeSH Terms] of settings than has been the case so far, such as rural-
OR child [All Fields] OR child* [MeSH Terms] OR child* [All Fields] OR children [All Fields]) OR based, secondary-level care settings that include a sufficient
paediatrics [MeSH Terms] OR paediatrics [All Fields] OR paediatric [All Fields] OR paediatrics period of follow up after discharge to appropriately manage
[MeSH Terms] OR pediatrics [All Fields] OR pediatric [All Fields]”. Additionally, we reviewed suspected or culture-confirmed tuberculosis.
the reference lists of original studies and review papers and contacted experts in the field of In conclusion, this Review suggests that tuberculosis is
child tuberculosis in an attempt to include all relevant publications. important in the pathogenesis of acute childhood
The key information abstracted from included papers were study characteristics including pneumonia in countries with a high incidence of
year of study, country and setting, tuberculosis incidence at the time of the study, study tuberculosis, either as a direct cause or as an underlying
sample size and age range, inclusion criteria, diagnostic approach, reported outcomes, risk factor that increases susceptibility to bacterial
and HIV prevalence in tuberculosis cases (when available). pneumonia. Interpretation of findings from previous
studies is restricted by recognised diagnostic challenges
The mortality associated with a diagnosis of tuberculosis and substantial heterogeneity between studies with risk
is a concern. In view of the high prevalence of of bias. Prospective studies from several epidemiological
comorbidities and co-infections in these children, settings that use optimum diagnostic techniques are
differentiation between children that died with underlying needed to better understand the contribution of
tuberculosis, which was complicated by concurrent or tuberculosis to child pneumonia and to improve clinical
superimposed infections, from those that died directly management.
because of tuberculosis is not possible. The autopsy Contributors
studies suggest that bacterial co-infections are important All authors contributed to the concept and plan for this Review. JNO and
causes of death in children with tuberculosis.28,30,34 The JMK did the literature review and analysis with input from BJM and
SMG. JNO, BJM, and SMG developed the first draft and all authors
autopsy studies included in this Review all had large provided major contributions to the final manuscript. JNO abstracted
sample sizes but did not necessarily represent the full the data and JMK, BJM, SAM, and SMG verified accuracy. JNO and JMK
range of children dying with pneumonia, because consent independently screened the titles and abstracts of all papers identified by
for autopsy was less than 25% in the studies in Zambia the search and applied the predefined study selection criteria to identify
eligible studies.
and Zimbabwe,28,34 whereas the South African study only
included HIV-infected children.30 Only one study reported Declaration of interests
SAM has received honoraria, but not linked to this work, from
antemortem clinical data.30 Small autopsy studies were GlaxoSmithKline, Pfizer, Novartis, and Sanofi Pasteur. The other authors
not included in this Review, such as a study from Botswana declare no competing interests.
that reported tuberculosis prevalence of 21% in children Acknowledgments
dying with respiratory disease in Francistown.49 We thank Mike English (KEMRI Wellcome Trust, Kenya, and Oxford
Additionally, an autopsy study from Bangladesh reported University, Oxford, UK) for his helpful review of the manuscript.
tuberculosis prevalence of 3% in children who died from References
pneumonia.50 A review of autopsy studies of deaths due to 1 Walker CL, Rudan I, Liu L, et al. Global burden of childhood
pneumonia and diarrhoea. Lancet 2013; 381: 1405–16.
respiratory infections in Africa reported that about 8·5% 2 Liu L, Johnson HL, Cousens S, et al. Global, regional, and national
of all cases in children had pulmonary tuberculosis, which causes of child mortality: an updated systematic analysis for 2010
was statistically more prevalent in HIV-uninfected cases.51 with time trends since 2000. Lancet 2012; 379: 2151–61.
This Review shows the need for improved diagnostics 3 Zar HJ, Ferkol TW. The global burden of respiratory disease-impact
on child health. Pediatr Pulmonol 2014; 49: 430–34.
for tuberculosis in infants and young children, including 4 Rudan I, Boschi-Pinto C, Biloglav Z, Mulholland K, Campbell H.
those presenting with acute severe pneumonia. The Epidemiology and aetiology of childhood pneumonia.
clinical presentation and radiological features frequently Bull World Health Organ 2008; 86: 408–16.
5 Graham SM. Child pneumonia: current status, future prospects.
overlap and even a clinical response to antibiotics does not Int J Tuberc Lung Dis 2010; 14: 1357–61.
necessarily exclude a diagnosis of tuberculosis, which 6 Shann F. The management of pneumonia in children in developing
might be the underlying cause predisposing to countries. ClinInfect Dis 1995; 21 (suppl 3): S218–25.
susceptibility for acute bacterial pneumonia treatment in 7 Selwyn BJ. The epidemiology of acute respiratory tract infection in
young children: comparison of findings from several developing
hospital. If an infant develops tuberculosis, infection of the countries. Coordinated Data Group of BOSTID Researchers.
infant by a close contact who has tuberculosis (which Rev Infect Dis 1990; 12 (suppl 8): S870–88.
might or might not have been diagnosed) is very likely, and 8 Duke T, Mgone CS. Measles: not just another viral exanthem.
Lancet 2003; 361: 763–73.
so this information should always be carefully sought after.

8 www.thelancet.com/respiratory Published online January 29, 2015 http://dx.doi.org/10.1016/S2213-2600(15)00028-4

 Review

9 Scott JA. The global epidemiology of childhood pneumonia 20 years 32 Madhi SA, Petersen K, Madhi A, Khoosal M, Klugman KP.
on. Bull World Health Organ 2008; 86: 494–96. Increased disease burden and antibiotic resistance of bacteria
10 Michelow IC, Olsen K, Lozano J, et al. Epidemiology and clinical causing severe community-acquired lower respiratory tract
characteristics of community-acquired pneumonia in hospitalized infections in human immunodeficiency virus type 1-infected
children. Pediatrics 2004; 113: 701–07. children. Clin Infect Dis 2000; 31: 170–76.
11 Madhi SA, Klugman KP. A role for Streptococcus pneumoniae in 33 McNally LM, Jeena PM, Gajee K, et al. Effect of age, polymicrobial
virus-associated pneumonia. Nat Med 2004; 10: 811–13. disease, and maternal HIV status on treatment response and cause
12 Graham SM. Non-tuberculosis opportunistic infections and other of severe pneumonia in South African children: a prospective
lung diseases in HIV-infected infants and children. descriptive study. Lancet 2007; 369: 1440–51.
Int J Tuberc Lung Dis 2005; 9: 592–602. 34 Chintu C, Mudenda V, Lucas S, et al. Lung diseases at necropsy in
13 Marais BJ, Gupta A, Starke JR, El Sony A. Tuberculosis in women African children dying from respiratory illnesses: a descriptive
and children. Lancet 2010; 375: 2057–59. necropsy study. Lancet 2002; 360: 985–90.
14 UNAIDS JUNPoHA. Global report: UNAIDS report on the global 35 Adegbola RA, Falade AG, Sam BE, et al. The etiology of pneumonia
AIDS epidemic 2013. http://www.unaids.org/sites/default/files/en/ in malnourished and well-nourished Gambian children.
media/unaids/contentassets/documents/epidemiology/2013/gr2013/ Pediatr Infect Dis J 1994; 13: 975–82.
UNAIDS_Global_Report_2013_en.pdf (accessed Dec 2, 2014). 36 Graham SM, Mtitimila EI, Kamanga HS, Walsh AL, Hart CA,
15 Seddon JA, Shingadia D. Epidemiology and disease burden of Molyneux ME. Clinical presentation and outcome of Pneumocystis
tuberculosis in children: a global perspective. Infect Drug Resist carinii pneumonia in Malawian children. Lancet 2000; 355: 369–73.
2014; 7: 153–65. 37 Graham SM, Mankhambo L, Phiri A, et al. Impact of human
16 WHO. Global Tuberculosis Report 2014. Geneva: World Health immunodeficiency virus infection on the etiology and outcome of
Organization, 2014. severe pneumonia in Malawian children. Pediatr Infect Dis J 2011;
30: 33–38.
17 Graham SM, Sismanidis C, Menzies HJ, Marais BJ, Detjen AK,
Black RE. Importance of tuberculosis control to address child 38 The World Bank Group. The World Bank Incidence of Tuberculosis
survival. Lancet 2014; 383: 1605–07. (per 100, 000 people). 2014. http://data.worldbank.org/indicator/
SH.TBS.INCD?page=3 (accessed July 18, 2014).
18 Chisti MJ, Tebruegge M, La Vincente S, Graham SM, Duke T.
Pneumonia in severely malnourished children in developing 39 Zar HJ, Tannenbaum E, Apolles P, Roux P, Hanslo D, Hussey G.
countries—mortality risk, aetiology, and validity of WHO clinical Sputum induction for the diagnosis of pulmonary tuberculosis in
signs: a systematic review. Trop Med Int Health 2009; 14: 1173–89. infants and young children in an urban setting in South Africa.
Arch Dis Child 2000; 82: 305–08.
19 Chisti MJ, Ahmed T, Pietroni MA, et al. Pulmonary tuberculosis in
severely-malnourished or HIV-infected children with pneumonia: a 40 Marais BJ, Gie RP, Obihara CC, Hesseling AC, Schaaf HS,
review. J Health Popul Nutr 2013; 31: 308–13. Beyers N. Well defined symptoms are of value in the diagnosis of
childhood pulmonary tuberculosis. Arch Dis Child 2005; 90: 1162–65.
20 Gilani Z, Kwong YD, Levine OS, et al. A literature review and
survey of childhood pneumonia etiology studies: 2000–2010. 41 Swaminathan S, Rekha B. Pediatric tuberculosis: global overview
Clin Infect Dis 2012; 54 (suppl 2): S102–08. and challenges. Clin Infect Dis 2010; 50 (suppl 3): S184–94.
21 Scott JA, English M. What are the implications for childhood 42 Marais BJ, Hesseling AC, Gie RP, Schaaf HS, Enarson DA,
pneumonia of successfully introducing Hib and pneumococcal Beyers N. The bacteriologic yield in children with intrathoracic
vaccines in developing countries? PLoS Med 2008; 5: e86. tuberculosis. Clin Infect Dis 2006; 42: e69–71.
22 WHO. Pocket Book of Hospital care for children; Guidelines for the 43 Jaganath D, Mupere E. Childhood tuberculosis and malnutrition.
management of common childhood illnesses. Geneva: World J Infecti Dis 2012; 206: 1809–15.
Health Organisation, 2005. 44 Hesseling AC, Cotton MF, Jennings T, et al. High incidence of
23 von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, tuberculosis among HIV-infected infants: evidence from a South
Vandenbroucke JP. The Strengthening the Reporting of Observational African population-based study highlights the need for improved
Studies in Epidemiology (STROBE) statement: guidelines for tuberculosis control strategies. Clin Infect Dis 2009; 48: 108–14.
reporting observational studies. Prev Med 2007; 45: 247–51. 45 Osborne CM. The challenge of diagnosing childhood tuberculosis
24 Hammitt LL, Kazungu S, Morpeth SC, et al. A preliminary study in a developing country. Arch Dis Child 1995; 72: 369–74.
of pneumonia etiology among hospitalised children in Kenya. 46 Kumar M, Biswal N, Bhuvaneswari V, Srinivasan S. Persistent
Clin Infect Dis 2012; 54 (suppl 2): S190–99. pneumonia: underlying cause and outcome. Indian J Pediatr 2009;
25 Nantongo JM, Wobudeya E, Mupere E, et al. High incidence of 76: 1223–26.
pulmonary tuberculosis in children admitted with severe 47 Gokdemir Y, Cakir E, Kut A, et al. Bronchoscopic evaluation of
pneumonia in Uganda. BMC Pediatr 2013; 13: 16. unexplained recurrent and persistent pneumonia in children.
26 Moore DP, Klugman KP, Madhi SA. Role of Streptococcus J Paediatr Child Health 2013; 49: E204–07.
pneumoniae in hospitalisation for acute community-acquired 48 Dangor Z, Izu A, Moore DP, et al. Temporal association in
pneumonia associated with culture-confirmed hospitalisations for tuberculosis, invasive pneumococcal disease
Mycobacterium tuberculosis in children: a pneumococcal conjugate and influenza virus illness in South African children. PLoS One
vaccine probe study. Pediatr Infect Dis J 2010; 29: 1099–104. 2014; 9: e91464.
27 Chisti MJ, Graham SM, Duke T, et al. A prospective study of the 49 Ansari NA, Kombe AH, Kenyon TA, et al. Pathology and causes of
prevalence of tuberculosis and bacteraemia in Bangladeshi children death in a series of human immunodeficiency virus-positive and
with severe malnutrition and pneumonia including an evaluation of -negative pediatric referral hospital admissions in Botswana.
Xpert MTB/RIF assay. PloS One 2014; 9: e93776. Pediatr Infect Dis J 2003; 22: 43–47.
28 Ikeogu MO, Wolf B, Mathe S. Pulmonary manifestations in HIV 50 Tomashefski JF Jr, Butler T, Islam M. Histopathology and aetiology
seropositivity and malnutrition in Zimbabwe. Arch Dis Child 1997; of childhood pneumonia: an autopsy study of 93 patients in
76: 124–28. Bangladesh. Pathology 1989; 21: 71–78.
29 Wang Y, Kong F, Yang Y, Gilbert GL. A multiplex PCR-based reverse 51 Bates M, Mudenda V, Mwaba P, Zumla A. Deaths due to respiratory
line blot hybridization (mPCR/RLB) assay for detection of bacterial tract infections in Africa: a review of autopsy studies.
respiratory pathogens in children with pneumonia. Curr Opin Pulm Med 2013; 19: 229–37.
Pediatr Pulmonol 2008; 43: 150–59. 52 Portevin D, Moukambi F, Clowes P, et al. Assessment of the novel
30 Rennert WP, Kilner D, Hale M, Stevens G, Stevens W, T-cell activation marker-tuberculosis assay for diagnosis of active
Crewe-Brown H. Tuberculosis in children dying with HIV-related tuberculosis in children: a prospective proof-of-concept study.
lung disease: clinical-pathological correlations. Int J Tuberc Lung Dis Lancet Infect Dis 2014; 14: 931–38.
2002; 6: 806–13. 53 Anderson ST, Kaforou M, Brent AJ, et al. Diagnosis of childhood
31 Zar HJ, Hanslo D, Tannenbaum E, et al. Aetiology and outcome of tuberculosis and host RNA expression in Africa. N Engl J Med 2014;
pneumonia in human immunodeficiency virus-infected children 370: 1712–23.
hospitalized in South Africa. Acta Paediatr 2001; 90: 119–25.

www.thelancet.com/respiratory Published online January 29, 2015 http://dx.doi.org/10.1016/S2213-2600(15)00028-4 9