Tropical Medicine and International Health doi:10.1111/tmi.

12557

volume 20 no 10 pp 1320–1328 october 2015

Stunting is associated with poor outcomes in childhood

pneumonia
Peter P. Moschovis1, Emmanuel O. D. Addo-Yobo2, Salem Banajeh3, Noel Chisaka4, David C. Christiani1,
Douglas Hayden1, Prakash Jeena5, William B. MacLeod6, Greta Mino7, Archana Patel8, Shamim Qazi9,
Mathuram Santosham10, Donald M. Thea6 and Patricia L. Hibberd1
1 Massachusetts General Hospital, Boston, MA, USA
2 Kwame Nkrumah University of Science & Technology, Komfo Anokye Teaching Hospital, Kumasi, Ghana
3 Al-Sabeen Hospital for Women & Children, Sana’a University, Sana’a, Yemen
4 World Bank, Washington, DC, USA
5 University of KwaZulu-Natal, Durban, South Africa
6 Boston University, Boston, MA, USA
7 Children’s Hospital Dr Francisco de Ycaza Bustamante, Guayaquil, Ecuador
8 Lata Medical Research Foundation, Indira Gandhi Government Medical College, Nagpur, India
9 World Health Organization, Geneva, Switzerland
10 Johns Hopkins University, Baltimore, MD, USA

Abstract objective Stunting affects 26.7% of children worldwide, and little is known about its effects on the
outcomes of childhood pneumonia. We evaluated the effect of stunting on the outcomes of
pneumonia among children enrolled in two large clinical trials.
methods We analysed data from two WHO and USAID-sponsored inpatient treatment trials, the
Severe Pneumonia Evaluation Antimicrobial Research study (n = 958) and the Amoxicillin Penicillin
Pneumonia International Study (n = 1702), which enrolled children aged 2–59 months across 16 sites
in LMICs. We assessed the effect of stunting (height-for-age Z score < 2) on treatment outcome
and time to resolution of hypoxaemic pneumonia.
results Among 2542 (96%) children with valid data for height, 28% were stunted and 12.8%
failed treatment by 5 days. The failure rate among stunted patients was 16.0% vs. 11.5% among
non-stunted patients [unadjusted RR = 1.24 (95% CI 1.08, 1.41); adjusted RR = 1.28 (95% CI 1.10,
1.48)]. An inverse relationship was observed between height and failure rates, even among non-
stunted children. Among 845 patients with hypoxaemic pneumonia, stunting was associated with a
lower probability of normalisation of respiratory rate [HR = 0.63 (95% CI 0.52, 0.75)] and oxygen
saturation [HR = 0.74 (95% CI 0.61, 0.89)].
conclusions Stunting increases the risk of treatment failure and is associated with a longer course
of recovery in children with pneumonia. Strategies to decrease stunting may decrease the burden of
adverse outcomes in childhood pneumonia in low-resource settings.

keywords pneumonia, malnutrition, treatment failure, global health

Chronic undernutrition is best reflected by stunting, or

Introduction
low height (or length)-for-age [7]. Stunting affects 162
Pneumonia is the leading cause of death among young million children (26.7% of all children) worldwide, 92%
children in developing nations, responsible for about of whom live in Asia or Africa [8]. For children with
1.2 million early childhood deaths annually [1, 2]. pneumonia, stunting may have harmful effects that are
Undernutrition is a known risk factor for developing distinct from those related to nutritional deficiency, as
and dying from pneumonia [3], but much recent height is the major determinant of lung size and pulmo-
research has focused on the effects of wasting or nary function [9, 10]. Stunting is independently associ-
underweight [4–6], which primarily reflects acute ated with an increased risk of parent-reported acute
undernutrition. respiratory infection [11], WHO-defined pneumonia [12],

1320 © 2015 John Wiley & Sons Ltd

Tropical Medicine and International Health volume 20 no 10 pp 1320–1328 october 2015

P. P. Moschovis et al. Stunting affects childhood pneumonia outcomes

radiographically confirmed pneumonia [13], with Mexico City, Mexico; Multan, Pakistan; Rawalpindi,
increased risk of RSV-related lower respiratory tract Pakistan; Sana’a, Yemen). Patients were enrolled between
infection (but not upper respiratory infection) [14], and May 1999 and May 2002 (APPIS) and August 2000 and
with increased risk of hospitalisation for pneumonia [15]. April 2004 (SPEAR). In both trials, children were recruited
Beyond increasing the risk of developing severe pneu- at the time of presentation to the hospital, assessed every
monia, the restricted lung growth associated with stun- 6 h through hospital discharge and evaluated at pre-
ting may make children less able to tolerate severe defined intervals after hospitalisation (for APPIS, days 5
pulmonary infection, more prone to develop hypoxaemia and 14; for SPEAR, days 5, 10–12 and 21–30).
and respiratory failure, and may impair recovery from
pneumonia [16, 17]. 1 To better understand the effect of
Participants
stunting on pneumonia outcomes, we evaluated the effect
of stunting on the risk of failing treatment and time to The two trials included children under age 5 years
recovery from pneumonia, using data from the previ- (3–59 months in APPIS, 2–59 months in SPEAR) with
ously completed Severe Pneumonia Evaluation Antimi- WHO-defined pneumonia. The APPIS trial recruited
crobial Research (SPEAR) study [18] and Amoxicillin children with the prior WHO definition of ‘severe pneu-
Penicillin Pneumonia International Study (APPIS) [19]. monia’ (cough or difficulty breathing with lower chest in-
These studies were multinational randomised clinical tri- drawing), while the SPEAR trial recruited children with
als of antibiotic regimens for pneumonia among children WHO-defined ‘very severe pneumonia’ (cough or diffi-
in low- and middle-income countries, settings with high culty breathing with either central cyanosis, severe respi-
rates of both stunting and pneumonia. We hypothesised ratory distress, or inability to drink) [20]. [Note that the
that (i) stunted children would have an increased risk of WHO severity criteria were revised in 2013 to simply
failing treatment of their episode of pneumonia and that ‘severe pneumonia’ vs. ‘pneumonia’ [21]; the APPIS and
(ii) stunted children would take longer to recover from SPEAR trials used the pre-2013 criteria for enrolment.]
pneumonia. The trials had similar exclusion criteria, which included
known history of asthma or wheezing, history of serious
adverse reaction to any of the study drugs, more than
Methods 24 h (>48 h in APPIS) of antibiotics already given for the
present illness, and other concomitant chronic illnesses or
Study design
severe infections (bacterial meningitis, cerebral malaria,
We analysed data from two large WHO and USAID- jaundice). APPIS excluded children with clinically evident
sponsored inpatient treatment trials in childhood HIV infection at the time of study enrolment. Two APPIS
pneumonia, the Amoxicillin Penicillin Pneumonia sites (Durban, South Africa, and Ndola, Zambia) tested
International Study (APPIS) (n = 1702) and the Severe participants for HIV using ELISA with Western blot
Pneumonia Evaluation Antimicrobial Research (SPEAR) confirmation in children older than 15 months and HIV-
study (n = 958). Details of the study design have been 1 DNA PCR in children younger than 15 months. Of the
published previously [18, 19], but briefly, these trials 523 children enrolled at these two sites, HIV status was
were randomised, open-label studies designed to evalu- known for 464 participants, of which 106 (23%) tested
ate the equivalence of oral amoxicillin vs. injectable positive for HIV. APPIS also excluded children with
penicillin for WHO-defined severe pneumonia (APPIS, severe acute malnutrition, defined as weight-for-age less
ISRCTN95821329) and to compare chloramphenicol than 3 SD or clinical kwashiorkor.
vs. ampicillin plus gentamicin for WHO-defined very
severe pneumonia (SPEAR, ISRCTN39543942).
Variables and measurement
Height (or length for children age <2 years) was
Setting
measured at the time of enrolment into the study by
The trials enrolled participants across 16 sites, all tertiary research personnel trained in standardised anthropomet-
hospitals in low- or middle-income countries (APPIS: ric techniques. We calculated Z scores (number of
Bogota, Columbia; Capetown, South Africa; Durban, standard deviations above or below the median) for
South Africa; Ho Chi Minh City, Vietnam; Islamabad, height-for-age based on the WHO Growth Standards
Pakistan; Kumasi, Ghana; Mexico City, Mexico; Nagpur, [22] using the WHO Anthro macros [23]. Stunting
India; Ndola, Zambia; SPEAR: Chandigarh, India; Dha- was defined as a height-for-age Z score (HAZ) < 2
ka, Bangladesh; Guayaquil, Ecuador; Lusaka, Zambia; [8].

© 2015 John Wiley & Sons Ltd 1321

Tropical Medicine and International Health volume 20 no 10 pp 1320–1328 october 2015

P. P. Moschovis et al. Stunting affects childhood pneumonia outcomes

Statistical methods
Box 1 Criteria for treatment failure
We performed all statistical analyses using SAS v. 9.3
• Persistent severe pneumonia at 5 days, defined as
(SAS Institute Inc., Cary, NC, USA). We compared the
any one of the following
baseline characteristics of stunted vs. non-stunted chil-
s Oxygen saturation < 90% or central cyanosis dren using Pearson’s v2 for categorical variables, Stu-
s Severe respiratory distress (e.g. grunting, very dent’s t-test for normally distributed continuous variables
severe chest indrawing) and the Wilcoxon rank sum test for continuous variables
s Signs of pneumonia with a general danger sign with a skewed distribution.
(inability to breastfeed or drink, lethargy/ We examined the effect of stunting on the risk of treat-
reduced level of consciousness, convulsions) ment failure initially as a single variable in a binary
• Development of any of the following: bacterial regression model and then using multivariate regression
meningitis, empyema, septic shock, renal failure to adjust for known confounders as main effects, and
or newly diagnosed comorbid conditions study site as a repeated effect assuming exchangeable cor-
• Serious adverse drug reaction relation within sites. Height-for-age was also evaluated as
• Modification of antibiotic treatment a continuous variable in a similar model.
• Voluntary withdrawal or absconding Among patients with hypoxaemic pneumonia (which
• Death would be classified as ‘severe pneumonia’ under the 2013
WHO guidelines [21]), we evaluated time to resolution
of hypoxaemia and tachypnea, using the SAS %CIF
macro to calculate the cumulative incidence function and
The primary outcome of the study was treatment fail-
plotting time to recovery using Kaplan–Meier method
ure at 5 days, a composite outcome used in both trials
[24]. Patients who died, left against medical advice or
that was updated to reflect the 2013 WHO guidelines for
were lost to follow-up were censored at the time at which
severe pneumonia (Box 1). [21] Secondary outcomes
they left the study. We also compared rates of resolution
included time to resolution of hypoxaemia [resolution
of hypoxaemia and tachypnea using a multivariate Cox
defined as oxygen saturation (SpO2) that
proportional-hazards model that included age, sex, treat-
remained ≥ 90% until discharge from the study] and tac-
ment group, initial oxygen saturation, initial respiratory
hypnea [resolution defined as respiratory rate (RR) that
rate, presence of danger signs (unable to eat or drink, dif-
remained normal until discharge from the study
ficult to arouse or convulsions), breastfeeding per WHO
(RR < 60 breaths per minute for children aged
recommendations, dehydration, immunisation status and
<2 months, RR < 50 breaths per minute for age 2–
study site as a random effect.
11 months and RR < 40 breaths per minute for age 12–
All analyses were conducted using intention-to-treat
59 months)]. All oxygen saturations found to be below
analysis. A P < 0.05 was considered significant for all
90% were confirmed with a repeat measurement. All
hypothesis testing.
respiratory rate assessments were performed twice; if the
two measurements differed by more than five breaths per
minute, a third measurement was performed. The mean Ethical considerations
of repeat measurements for respiratory rate and oxygen Families of patients who participated in both the APPIS
saturation was used for the analysis. and SPEAR trials provided informed consent after review-
We accounted for known confounders in the multivari- ing the risks and benefits of participation. The studies were
ate analyses, including age (measured in months), sex, approved by the institutional review boards of all hospitals
treatment group, initial oxygen saturation, initial respira- participating in the study, the WHO, the Boston University
tory rate, presence of danger signs (unable to eat or School of Public Health, and the Johns Hopkins University
drink, difficult to arouse, or convulsions), breastfeeding Bloomberg School of Public Health. The present analysis
per WHO recommendations, dehydration, immunisation has been evaluated and exempted by the Human Research
status and study site. In the two sites that tested children Committee (IRB) of Partners Healthcare and approved by
for HIV (Ndola and Durban), we performed a subgroup the APPIS and SPEAR Steering Committees.
analysis of the effect of stunting adjusting for HIV status,
age, sex, treatment group, initial oxygen saturation and
Role of the funding source
initial respiratory rate (there was insufficient variability
in these two sites to adjust for danger signs, dehydration The present analysis was supported by funding from the
or immunisation status). National Institutes of Health, which had no role in the

1322 © 2015 John Wiley & Sons Ltd

Tropical Medicine and International Health volume 20 no 10 pp 1320–1328 october 2015

P. P. Moschovis et al. Stunting affects childhood pneumonia outcomes

study design, data analysis, data interpretation or writing (13.6% vs. 12.8%, P for difference = 0.81), and patients
of the report. PPM had full access to the data and had who left the study or were lost to follow-up had similar
final responsibility for the decision to submit for publica- rates of stunting as those who remained in the study
tion. (34% vs. 28%, P for difference = 0.30).
The baseline characteristics of the study participants
are listed in Table 1. A majority of the participants
Results
(61.8%) were boys. Stunted children were on average
A total of 2660 patients participated in the two trials, older, less often breastfed appropriately, more often
and 2542 (96%) had valid data for height recorded dehydrated, and they exhibited danger signs more often.
(Figure 1). Between the two trials, 65 children (2.4%) left A slightly greater proportion of stunted children were
the study voluntarily or were lost to follow-up. The over- boys, although this difference was not statistically signifi-
all prevalence of stunting among patients in the study cant. On average, stunted children had a higher respira-
was 28% (26% of patients in APPIS, 31% of patients in tory rate than non-stunted children (despite being older),
SPEAR). Patients with missing height data (n = 118) had and this difference remained throughout the entire study
similar treatment failure rates as those with height data period (Figure 2).

APPIS trial: 1702 children with SPEAR trial: 958 children with very
severe pneumonia* across 9 sites severe pneumonia* across 8 sites

95 missing data for height 23 missing data for height

8 failed treatment 8 failed treatment
87 did not fail treatment 15 did not fail treatment

2542 children available for analysis

714 children stunted 1828 children not stunted

114 failed treatment 600 cured 211 failed treatment 1617 cured

* The APPIS and SPEAR studies were based on the pre-2013 WHO classification scheme for pneumonia. The
classification has been revised in the latest edition of the WHO guidelines.(21)

Figure 1 Study Population. *The APPIS and SPEAR studies were based on the pre-2013 WHO classification scheme for pneumonia.
The classification has been revised in the latest edition of the WHO guidelines [21].

© 2015 John Wiley & Sons Ltd 1323

Tropical Medicine and International Health volume 20 no 10 pp 1320–1328 october 2015

P. P. Moschovis et al. Stunting affects childhood pneumonia outcomes

Table 1 Baseline characteristics of study participants Overall, 341 patients (12.8%) met the primary end-
point of treatment failure by 5 days. Among stunted
Stunted Not stunted
patients, the failure rate was 16.0%, while among non-
(HAZ < 2) (HAZ ≥ 2) P for
n = 714 n = 1828 difference stunted patients, the failure rate was 11.5% (RR = 1.24
(95% CI 1.08, 1.41), P = 0.002). An inverse relationship
Age in months 9.0 (13.2) 6.9 (8.9) <0.0001 was observed between height-for-age Z score (HAZ) and
(median) failure rates, even among non-stunted children (see
Male sex 64.9% 60.8% 0.06
Figure 3). Stunting was associated with an increased risk
Appropriately 81.3% 81.6% 0.84
immunised of failure [RR = 1.28 (95% CI 1.10, 1.48), P = 0.001]
Breastfed per WHO 56.3% 66.1% <0.0001 after adjusting for age, sex, treatment group, initial oxy-
recommendations gen saturation, initial respiratory rate, presence of danger
Dehydrated 7.3% 4.2% 0.001 signs, breastfeeding per WHO recommendations, dehy-
Danger signs 37.8% 33.6% 0.04 dration, immunisation status and study site (see Table 2).
Oxygen saturation 92.0 (8.0) 92.0 (7.5) 0.50 For every one standard deviation increase in height-for-
% (median)
age (modelled as a continuous variable), the risk of
Respiratory rate 64.6 (12.4) 63.1 (12.5) 0.007
(mean) treatment failure decreased by 7% [RR = 0.93, (95% CI
Temperature in 37.6 (0.9) 37.6 (1.0) 0.27 0.89, 0.96), P = 0.0002]. Other variables that were
°C (mean) significantly associated with treatment outcome in the
multivariate model included sex, initial oxygen
HAZ, height-for-age Z score. For categorical variables, data repre-
saturation, dehydration and immunisation status.
sent number of patients (n, %), mean (SD), and median and inter-
quartile range (for age and oxygen saturation) as appropriate.
Because HIV is a known risk factor for stunting, we
P values are from Pearson v2 for categorical variables, Student’s performed a subgroup analysis of treatment failure by
t-test for normally distributed continuous variables and the 5 days among all children for whom HIV testing results
Wilcoxon rank sum test for skewed variables (age and oxygen were available (testing was only performed at two sites,
saturation). Ndola and Durban). There remained a significant associa-
tion between stunting and treatment failure by 5 days

Mean respiratory rate over study period

70

65
Mean respiratory rate (breaths per minute)

60

55

50

45

40
Stunted (HAZ < –2)
35
Not Stunted (HAZ ≥ –2)
30
Day 1

Day 2

Day 3

Day 5

Day 14
Baseline

Figure 2 Mean respiratory rate by

stunting.

1324 © 2015 John Wiley & Sons Ltd

Tropical Medicine and International Health volume 20 no 10 pp 1320–1328 october 2015

P. P. Moschovis et al. Stunting affects childhood pneumonia outcomes

Rate of treatment failure by

height-for-Age Z score
25%

20%

15%

10%

5%

0%
< –3 –3 to –2 –2 to –1 –1 to 0 0 to +1 +1 to +2 +2 to +3 >+3

Figure 3 Rate of treatment failure by n = 377 n = 337 n = 566 n = 561 n = 414 n = 201 n = 56 n = 30
height-for-age Z score. P for trend
(Cochran-Armitage) <0.0001. Error bars P for trend (Cochran-Armitage) <0.0001
represent 95% confidence intervals Error bars represent 95% confidence intervals

Table 2 Multivariate model of predictors of treatment failure at [RR = 1.72 (95% CI 1.09, 2.74)]. Stunted children had a
5 days non-significant increased risk of death [RR = 2.32 (95%
CI 0.99, 5.45)].
n Relative risk 95% CI P value We then examined the effect of stunting on the rate of
Stunting (HAZ < 2) 1.28 (1.10, 1.48) 0.001 recovery among the 845 patients with hypoxaemic pneu-
Age in months 0.99 (0.97, 1.01) 0.53 monia (‘severe pneumonia’ per the updated 2013 WHO
Female sex 1.25 (1.09, 1.43) 0.002 definition) [21], where recovery was defined as time to
Initial oxygen 0.71 (0.64, 0.78) <0.0001 resolution of tachypnea and hypoxaemia. Median time to
saturation (per 10%) normalisation of respiratory rate was significantly longer
Initial respiratory 1.08 (0.99, 1.19) 0.10
among stunted compared to non-stunted children
rate (per 10
breaths/min)
(5.25 days vs. 3.75 days; see Figure 4). After adjusting
Dehydration 0.54 (0.43, 0.68) <0.0001 for age, sex, treatment group, initial oxygen saturation,
Danger signs 0.90 (0.78, 1.04) 0.15 initial respiratory rate, presence of danger signs, breast-
Immunised appropriately 0.74 (0.60, 0.91) 0.005 feeding per WHO recommendations, dehydration, immu-
Breastfed per 1.03 (0.88, 1.21) 0.67 nisation status and study site in a multivariate Cox
WHO standards regression model, we found that the probability of nor-
HAZ, height-for-age Z score. malisation of respiratory rate was significantly lower
Area under ROC Curve: 0.786. among those who were stunted vs. those who were not
[HR = 0.63 (95% CI 0.52, 0.75), P < 0.0001].
[RR = 1.62 (95% CI 1.44, 1.81), P < 0.0001], even after We found a similar effect on time to recovery from
adjusting for HIV status. There was no evidence of inter- hypoxaemia, with a longer median time to normalisation
action between HIV and stunting (Pinteraction = 0.67). of oxygenation among stunted patients vs. non-stunted
When we evaluated differences in types of treatment patients (3.0 days vs. 2.75 days; see Figure 5). In the
failure among stunted vs. non-stunted children, we found multivariate Cox regression model, we found that the
that stunted children were more likely to have persistent probability of recovery from hypoxaemia was
hypoxaemia at 5 days [RR = 1.35 (95% 1.16, 1.58)] or significantly lower among stunted vs. non-stunted chil-
to be diagnosed with new comorbid conditions dren [HR = 0.74 (95% CI 0.61, 0.89), P = 0.001].

© 2015 John Wiley & Sons Ltd 1325

Tropical Medicine and International Health volume 20 no 10 pp 1320–1328 october 2015

P. P. Moschovis et al. Stunting affects childhood pneumonia outcomes

Time to normal respiratory rate

1.0

0.8
Proportion of patients

0.6

0.4

0.2

Not stunted (HAZ >= –2)

Stunted (HAZ < –2)
0.0
Figure 4 Time to normal respiratory rate
0 5 10 15 20 25 30
among stunted vs. non-stunted children
Days with severe pneumonia.

Time to normal oxygen saturation

1.0

0.8
Proportion of patients

0.6

0.4

0.2

Not stunted (HAZ >= –2)

Stunted (HAZ < –2)
0.0
Figure 5 Time to normal oxygen
0 5 10 15 20 25 30 saturation among stunted vs. non-stunted
Days children with severe pneumonia.

this relationship was also observed among children with

Discussion
height that would be considered normal (HAZ ≥ 2).
Among 2542 young children with pneumonia enrolled in Among the 845 children with hypoxaemic pneumonia in
two large clinical trials across 16 sites in 11 low- and mid- the study, stunting was also associated with a longer course
dle-income countries, stunting was associated with an of recovery from pneumonia, as reflected by time to nor-
increased risk of treatment failure at 5 days. An inverse malisation of respiratory rate and oxygen saturation.
relationship between height and risk of treatment failure Stunting may have both acute and long-term effects on
was most pronounced among children with stunting, but the outcomes of pneumonia. During the acute course of

1326 © 2015 John Wiley & Sons Ltd

Tropical Medicine and International Health volume 20 no 10 pp 1320–1328 october 2015

P. P. Moschovis et al. Stunting affects childhood pneumonia outcomes

pneumonia treatment and recovery, stunted children may anthropometric data on each child would more accu-
have less ventilatory reserve than children with greater rately reflect the exposure risk.
height and thus a greater risk of poor outcomes. As venti- Our study highlights the importance of stunting among
lation is a function of tidal volume and respiratory rate, children being treated for pneumonia. Recognition of this
and tidal volume in young children is a function of risk factor may help identify children at greater risk of
height/length [25], the higher respiratory rate we poor outcome, enabling providers to triage resources
observed in stunted patients may be a compensatory appropriately. More importantly, our findings highlight
mechanism to ensure adequate ventilation. the need for further research into prevention and treat-
Chronic undernutrition also has long-term conse- ment of stunting, as well as the effects of stunting on
quences on lung development and growth. Alveolarisa- lung development and adult pulmonary function. While
tion, which begins in the antenatal period, continues new interventions and strategies are being developed,
through childhood [26], and poor growth early in life is global health efforts must focus on proven strategies to
associated with reduced lung function in adulthood [27]. combat stunting, including promotion of breastfeeding,
In fact, much of the global ethnic and racial variability in prevention and treatment of micronutrient deficiency and
pulmonary function may be attributed to nutritional dif- protein-energy malnutrition, and introduction of
ferences [28]. In addition, the injury to the developing appropriate complementary feeding [32].
lung caused by pneumonia can lead to long-term sequelae
and is associated with reductions in adult pulmonary
Acknowledgements
function [27, 29] and in adult height [30]. It remains
unknown whether stunting affects the growth of both air- The SPEAR and APPIS studies were funded by the
ways and parenchyma, or whether it disproportionately Department of Child and Adolescent Health and Devel-
affects the airways (i.e. causing dysanaptic lung growth). opment, WHO; Center for International Health and
Clearly, the relationship between undernutrition, severe Development, Boston University; and Johns Hopkins
infections (including pneumonia) and growth is complex, Bloomberg School of Public Health, Baltimore. This ana-
as undernutrition increases the risk of severe infection, lysis was supported by the National Heart, Lung, and
and recurrent infections can contribute to poor growth. Blood Institute of the National Institutes of Health under
Given the high global burden of stunting, and more than award number F32HL124951. Statistical support was
120 million cases of childhood pneumonia annually [1], provided by Harvard Catalyst | The Harvard Clinical and
the overlap between these two diseases – and the vicious Translational Science Center (National Center for
cycle of infection and poor growth – merits further study. Research Resources and the National Center for Advan-
Our findings must be interpreted in the context of the cing Translational Sciences, National Institutes of Health
study design, and this highlights several limitations of our Award UL1 TR001102) and financial contributions from
study. The outcomes of pneumonia are affected by many Harvard University and its affiliated academic healthcare
host and pathogen factors; we attempted to control for centers. The content is solely the responsibility of the
several major predictors of outcome in our multivariate authors and does not necessarily represent the official
models, but the trials did not collect data on all possible views of the World Health Organization or National
confounders. Air pollution, for example, is associated Institutes of Health. The results of this study were pre-
with both stunting and a higher risk of pneumonia [31], sented in part at the American Thoracic Society Interna-
but no data on this or other environmental exposures tional Conference, San Diego, USA, 16–21 May 2014.
were available for children participating in the trials.
HIV is a risk factor for both stunting and treatment fail-
ure, but HIV testing was not performed in most of the References
sites. (We did perform a subgroup analysis with the avail- 1. Walker CL, Rudan I, Liu L et al. Global burden of childhood
able data, and adjusting for HIV status did not change pneumonia and diarrhoea. Lancet 2013: 381: 1405–1416.
the findings of our study). Second, because stunting is the 2. UN Inter-agency Group for Child Mortality Estimation.
final common outcome of many disease processes (includ- Levels and Trends in Child Mortality Report 2013. 2013.
ing undernutrition and chronic infection), grouping all 3. Black RE, Victora CG, Walker SP et al. Maternal and child
stunted children together fails to account for important undernutrition and overweight in low-income and middle-
income countries. Lancet 2013: 382: 427–451.
differences in comorbidities. Third, because stunting
4. Chisti MJ, Tebruegge M, La Vincente S, Graham SM, Duke
reflects chronic exposures, a single data point (height at
T. Pneumonia in severely malnourished children in develop-
the time of study enrolment) may not fully capture the ing countries - mortality risk, aetiology and validity of
duration or degree of undernutrition. Longitudinal

© 2015 John Wiley & Sons Ltd 1327

Tropical Medicine and International Health volume 20 no 10 pp 1320–1328 october 2015

P. P. Moschovis et al. Stunting affects childhood pneumonia outcomes

WHO clinical signs: a systematic review. Trop Med Int severe pneumonia among children aged 2-59 months in low
Health 2009: 14: 1173–1189. resource settings: multicentre randomised controlled trial
5. Preidis GA, McCollum ED, Mwansambo C, Kazembe PN, (SPEAR study). BMJ 2008: 336: 80–84.
Schutze GE, Kline MW. Pneumonia and malnutrition are 19. Addo-Yobo E, Chisaka N, Hassan M et al. Oral amoxicillin
highly predictive of mortality among African children hospi- versus injectable penicillin for severe pneumonia in children
talized with human immunodeficiency virus infection or aged 3 to 59 months: a randomised multicentre equivalency
exposure in the era of antiretroviral therapy. J Pediatr 2011: study. Lancet 2004: 364: 1141–1148.
159: 484–489. 20. World Health Organization. Pocket Book of Hospital Care
6. Chisti MJ, Salam MA, Ashraf H et al. Clinical risk factors for Children: Guidelines for the Management of Common
of death from pneumonia in children with severe acute mal- Illnesses with Limited Resources. WHO Press: Geneva,
nutrition in an urban critical care ward of Bangladesh. PLoS 2005.
ONE 2013: 8: e73728. 21. World Health Organization. Pocket Book of Hospital Care
7. Waterlow JC. Classification and definition of protein-calorie for Children: Guidelines for the Management of Common
malnutrition. Br Med J 1972: 3: 566–569. Illnesses with Limited Resources (2nd edn), WHO Press:
8. United Nations Children’s Fund, World Health Organiza- Geneva, 2013.
tion, The World Bank. UNICEF-WHO-World Bank Joint 22. WHO Multicentre Growth Reference Study Group. WHO
Child Malnutrition Estimates. 2012. Child Growth Standards: Length/Height-for-Age, Weight-
9. Quanjer PH, Stanojevic S, Cole TJ et al. Multi-ethnic reference for-Age, Weight-for-Length, Weight-for-Height and Body
values for spirometry for the 3-95-yr age range: the global lung Mass Index-for-Age: Methods and Development. World
function 2012 equations. Eur Respir J 2012: 40: 1324–1343. Health Organization: Geneva, 2006.
10. Zverev Y. Prediction of peak expiratory flow rates in stunted 23. World Health Organization. WHO Anthro Macro for SAS.
children. Cent Afr J Med 2001: 47: 74–78. 2011;3.2.2 [1 Sep 2014]
11. Mwiru R, Spiegelman D, Hertzmark E et al. Nutritional 24. Putter H, Fiocco M, Geskus RB. Tutorial in biostatistics:
predictors of acute respiratory infections among children competing risks and multi-state models. Stat Med 2007: 26:
born to HIV-infected women in Tanzania. J Trop Pediatr 2389–2430.
2013: 59: 203–208. 25. Nguyen TT, Hoo AF, Lum S, Wade A, Thia LP, Stocks J.
12. le Roux DM, Myer L, Nicol MP, Zar HJ. Incidence and New reference equations to improve interpretation of infant
severity of childhood pneumonia in the first year of life in a lung function. Pediatr Pulmonol 2013: 48: 370–380.
South African birth cohort: the Drakenstein Child Health 26. Narayanan M, Owers-Bradley J, Beardsmore CS et al. Alve-
Study. Lancet Glob Health 2015: 3: e95–e103. olarization continues during childhood and adolescence: new
13. Coles CL, Fraser D, Givon-Lavi N et al. Nutritional status evidence from helium-3 magnetic resonance. Am J Respir
and diarrheal illness as independent risk factors for alveolar Crit Care Med 2012: 185: 186–191.
pneumonia. Am J Epidemiol 2005: 162: 999–1007. 27. Tennant PW, Gibson GJ, Pearce MS. Lifecourse predictors
14. Okiro EA, Ngama M, Bett A, Cane PA, Medley GF, James of adult respiratory function: results from the Newcastle
Nokes D. Factors associated with increased risk of progression Thousand Families Study. Thorax 2008: 63: 823–830.
to respiratory syncytial virus-associated pneumonia in young 28. Mukhopadhvay S, Macleod KA, Ong TJ, Ogston SA. “Eth-
Kenyan children. Trop Med Int Health 2008: 13: 914–926. nic” variation in childhood lung function may relate to pre-
15. Dharmage SC, Rajapaksa LC, Fernando DN. Risk factors of ventable nutritional deficiency. Acta Paediatr 2001: 90:
acute lower respiratory tract infections in children under five 1299–1303.
years of age. Southeast Asian J Trop Med Public Health 29. Barker DJ, Godfrey KM, Fall C, Osmond C, Winter PD,
1996: 27: 107–110. Shaheen SO. Relation of birth weight and childhood respira-
16. Villamor E, Misegades L, Fataki MR, Mbise RL, Fawzi tory infection to adult lung function and death from chronic
WW. Child mortality in relation to HIV infection, nutri- obstructive airways disease. BMJ 1991: 303: 671–675.
tional status, and socio-economic background. Int J Epidem- 30. Bozzoli C, Deaton A, Quintana-Domeque C. Adult height
iol 2005: 34: 61–68. and childhood disease. Demography 2009: 46: 647–669.
17. Sempertegui F, Estrella B, Rodriguez O et al. Zinc as an 31. Harris AM, Sempertegui F, Estrella B et al. Air pollution
adjunct to the treatment of severe pneumonia in Ecuadorian and anemia as risk factors for pneumonia in Ecuadorian
children: a randomized controlled trial. Am J Clin Nutr children: a retrospective cohort analysis. Environ Health
2014: 99: 497–505. 2011: 3: 93.
18. Asghar R, Banajeh S, Egas J et al. Chloramphenicol versus 32. Bhutta ZA, Salam RA. Global nutrition epidemiology and
ampicillin plus gentamicin for community acquired very trends. Ann Nutr Metab 2012: 61(Suppl 1): 19–27.

Corresponding Author Peter P. Moschovis, Massachusetts General Hospital, Harvard Medical School, Divisions of Pulmonary/
Critical Care Medicine and Global Health, 125 Nashua Street, Suite 840, Boston, MA 02114 USA. Tel.: +1 617 643 9687;
Fax +1 617 724 9948, E-mail: pmoschovis@mgh.harvard.edu

1328 © 2015 John Wiley & Sons Ltd