OBJECTIVES:

Excess adiposity upregulates proinflammatory adipokines in infancy that have also been implicated in the pathogenesis of bronchiolitis. The association between excess adiposity and severity of disease in bronchiolitis is unclear. We sought to examine the association between adiposity and length of hospitalization and risk of PICU transfer in children with bronchiolitis.

METHODS:

We conducted a retrospective cohort study examining infants 24 months and younger hospitalized at an academic children’s hospital with bronchiolitis, grouped by weight status (BMI z score and ponderal index). Data were extracted from the medical record, including the following relevant covariates: age, sex, race and/or ethnicity, and International Classification of Diseases, 10th Revision codes. Outcomes included length of stay (LOS) and PICU transfer. We used multiple regression to examine the association between each anthropometric measure and LOS and likelihood of PICU transfer.

RESULTS:

There were 765 children in the final sample, 599 without a significant comorbidity (eg, prematurity, congenital heart disease). The median LOS was 2.8 days (interquartile range 1.7–4.9 days). LOS increased with increasing ponderal index quartile (P = .001). After accounting for age and significant comorbidities, we used multivariable regression to identify a significant association between increasing ponderal index and LOS (P = .04) and no association between BMI and LOS. Logistic regression did not reveal an association between either anthropometric measure and PICU transfer.

CONCLUSIONS:

In this study, we identified an association between a measure of excess adiposity in infants and length of hospitalization for bronchiolitis. Further work is needed to confirm this association, examine potential mechanisms, and account for other potential confounders.

Excess adiposity in infancy has been associated with a chronic proinflammatory state, and there is also evidence of reprogramming of the response to infections.1  Overfeeding and rapid weight gain in infancy have been shown to activate proinflammatory pathways.1,2  One putative mechanism acts via adipokine-mediated modulation of T helper 1 cells, which play a key role in adaptive immunity to viral respiratory infection.3,4  Another potential pathway includes an adipokine-induced shift of T cells to a T helper 17 phenotype.5 

In models of bronchiolitis, the T helper 17 response has a key role. Interleukin 17 increases with severity of illness, and neutralization of interleukin 17 decreases mucous production and the inflammatory response.6  The innate immune system also plays a critical role in the initial recognition and response to respiratory viral pathogens, with a suboptimal interferon response to infections associated with increasing clinical severity.7  The aberrant systemic inflammatory response arising from excess adiposity in infancy has important overlap with the immunologic dysregulation seen in clinical bronchiolitis.

Multiple investigators have observed an increased risk of developing asthma later in childhood after early childhood obesity.810  The proposed pathophysiology underlying this association is in part attributed to proinflammatory adipokines, including higher levels of interleukin 6, leptin, and tumor necrosis factor α (TNF-α), which may influence the airway inflammatory response in the setting of obesity.3  Although the mechanisms underlying the development of asthma among a subset of children with obesity and the mechanisms by which excess adiposity in infancy may influence acute bronchiolitis pathology may be distinct, the underlying premise is the same. Namely, excess adiposity early in life has immunologic sequelae with potential consequences on the respiratory health of children.

Bronchiolitis is the most common reason that children <1 year of age are hospitalized.11  For bronchiolitis, the impact of infants’ adiposity, either on the length of hospitalization or likelihood of ICU transfer, has not been examined. Therefore, we sought to examine the association between anthropometric proxies for infant adiposity and both length of hospital stay and likelihood of PICU transfer.

We conducted a retrospective cohort study examining infants with bronchiolitis hospitalized at a large academic children’s hospital. Patients were grouped by weight status, as defined by BMI z score and ponderal index, 2 measures of weight status endorsed by the World Health Organization.12  The primary outcome was length of stay (LOS), and secondary outcomes were PICU transfer and PICU LOS.

Clinical data were abstracted from the electronic medical record system, Epic. The initial analytic sample included 898 patients admitted between October 2013 and June 2018. Institutional review board approval was obtained (STUDY00019212).

Data were extracted with relevant identifiers: age, sex, race, Hispanic ethnicity, weight (in kilograms) and height (in inches), primary International Classification of Diseases, 10th Revision (ICD-10) code, admission department, LOS (days), and discharge date and time. Inclusion criteria were limited to children with a diagnosis of bronchiolitis by ICD-10 code who were 0 to 24 months of age at the time of admission. Because all infants were <24 months of age, height was assumed to be measured while recumbent, as per hospital protocol, and was classified as a length. For patients without a measured weight or height during admission, measures taken within 1 month of admission were manually extracted from the chart. Those with missing height and weight data were excluded. For all patients with time in the PICU, admission and discharge departments were manually confirmed in the electronic medical record.

Because there is no consensus on the best proportionality index proxy for adiposity in infants, we used the World Health Organization child growth standards to derive 2 age-standardized anthropometric measures of adiposity: BMI z score and ponderal index.13  Ponderal index is calculated as weight (kilograms)/length (meters cubed),12  compared with BMI, which is calculated as weight (kilograms)/length (meters squared). Ponderal index and BMI z score were examined both continuously and categorically: ponderal index by quartile within the sample and BMI z score by associated percentile (less than the fifth, fifth to 84.99th, 85th–94.99th, and >95th). BMI categories were labeled as underweight, healthy weight, overweight, and obesity, respectively.

The primary outcome for this analysis was LOS (in days), defined as the duration of admission to the inpatient unit, not including time in the emergency department. This was calculated by using the difference between discharge date and time and admission date and time. Secondary outcomes included PICU transfer from an inpatient ward and PICU LOS.

Because significant comorbidities may not only influence adiposity but also mark numerous coinciding and unknown clinical and social factors increasing LOS, the decision was made a priori to examine the associations of interest separately in this group. The broad categories of disease identified a priori were preterm birth, congenital respiratory disease, congenital heart disease, and genetic disorders. Two reviewers independently categorized ICD-10 codes as representing 1 of these 4 predetermined categories and a third reviewer acted as a tiebreaker, as necessary. Admission age, sex, and race and/or ethnicity, and insurance status (as a proxy for socioeconomic status) were also examined as potential confounders.

We conducted a power analysis to determine the number of patients needed to detect a clinically significant difference in LOS, which was set at 0.5 days. We used previous quality projects at our institution to inform our assumed LOS of 3 days for patients at a healthy weight, with an SD of 2 days. We set the α error at .05 and a power of 80%. To detect a significant difference using these parameters, we determined that we would need 252 subjects from each group (normal weight and overweight or obesity) or 504 subjects overall in unadjusted analyses. Assuming that the infants and children with overweight or obesity made up 30% of the sample14  and that we would need 252 subjects in the overweight or obesity group, we estimated needing an overall sample of 840 subjects.

For descriptive analyses, we first compared children with and without significant comorbidities by measures of adiposity and covariates of interest. We used Pearson χ2 tests for nominal variables, 2-sample t tests with a 2-sided α of .05 for continuous normally distributed data, and Kruskal-Wallis tests for continuous nonparametric data.

We then compared outcomes stratified by the presence of comorbid disease. We used Spearman’s rank correlation to examine the strength of the relationship between each anthropometric measure and the nonparametric outcomes of LOS and PICU LOS, we used Pearson χ2 tests to compare the risk of PICU transfer by stratum of each covariate, and we used Kruskal-Wallis tests to compare median LOS and PICU LOS by sex and race and/or ethnicity subgroups.

We used multivariate linear regression to determine the association between ponderal index and BMI z score on LOS as well as PICU LOS. LOS was log transformed to approximate normality. Age was considered both as a continuous variable and as a categorical variable, with age ≥12 months as the reference group. Race and/or ethnicity, sex, and age were considered individually as confounders by comparing crude and adjusted slope coefficient estimates to a threshold of 10% change in slope.15  Interaction term analysis was used to identify significant effect modifiers, with a threshold α level of .1 for the interaction term’s inclusion in the linear model for a continuous variable and a threshold α level of .25 for a categorical variable. Model diagnostics were performed on the final model, including residual analysis for violation of homoscedasticity, studentized residuals for identification of extreme cases, Cook’s distance for identification of influential points, and variation inflation factor analysis for examination of potential collinearity.

Multivariable logistic regression was used to examine the independent association between each weight-status measure and PICU transfer. Age and sex were conceptually identified as potential confounders and empirically included in the adjusted model. Race and/or ethnicity was evaluated for inclusion in the model on the basis of a ≥10% change in the odds ratio (OR). We examined the association by age through stratification.

There were 765 children in the final analytic sample, 599 without comorbidities and 166 with a significant comorbidity (prematurity, congenital respiratory disease, congenital heart disease, or a genetic disorder). The median LOS was 2.67 days for those without comorbidities (Table 1). There was an overall prevalence of ICU transfer of 8.4% among this subset of children, and of those transferred to the PICU, the median PICU stay was 1.7 days. Children with and without comorbidities significantly differed on the basis of BMI z score, age, race and/or ethnicity, and insurance type (Table 1). This validated the a priori decision to proceed with analysis of the subsets separately.

TABLE 1

Descriptive Analysis of the Sample Subset by Presence of Comorbidities

CharacteristicAll (N = 765)No Significant Comorbidities (n = 599)1+ Significant Comorbidity (n = 166)P
Ponderal index, mean (SD) 25.0 (4.6) 24.9 (4.5) 25.3 (4.8) .34a 
Ponderal index quartiles, n (%)    .96b 
 First quartile 192 (25) 149 (25) 43 (26) — 
 Second quartile 191 (25) 152 (25) 39 (24) — 
 Third quartile 191 (25) 150 (25) 41 (25) — 
 Fourth quartile 191 (25) 148 (25) 43 (26) — 
BMI z score, mean (SD) −0.30 (1.5) −0.24 (1.4) −0.51 (1.6) .04a 
BMI categories, n (%)    <.001b 
 Underweight 135 (18) 88 (15) 47 (28) — 
 Healthy wt 497 (65) 407 (68) 90 (54) — 
 Overweight 61 (8) 50 (8) 11 (6) — 
 Obesity 72 (9) 54 (9) 18 (11) — 
PICU transfer, n (%) 68 (9) 50 (8) 18 (11) .32b 
LOS, median (IQR) 2.83 (2–5) 2.67 (2–5) 3.64 (2–7) <.001c 
ICU LOS, median (IQR) 2.0 (1.1–5.0) 1.7 (1.0–4.1) 3.1 (1.5–6.9) .001c 
Age, mean (SD), median, wk 29 (26), 19 28 (26), 18 35 (28), 28 .003a 
Age <12 mo, n (%) 623 (81) 496 (83) 127 (77) .07b 
Female sex, n (%) 314 (41) 245 (41) 69 (42) .88b 
Race and/or ethnicity, n (%)    <.001b 
 Non-Hispanic white 471 (62) 365 (61) 106 (64) — 
 Hispanic 160 (21) 111 (19) 49 (30) — 
 Non-Hispanic African American 17 (2) 11 (2) 6 (4) — 
 Multiracial 43 (6) 41 (7) 2 (1) — 
 Other 74 (10) 71 (12) 3 (2) — 
Insurance type, n (%)    .01 
 Public 477 (62) 357 (60) 120 (72) — 
 Private 277 (36) 233 (39) 44 (27) — 
 Military, federal, or unspecified 10 (1) 8 (1) 2 (1) — 
CharacteristicAll (N = 765)No Significant Comorbidities (n = 599)1+ Significant Comorbidity (n = 166)P
Ponderal index, mean (SD) 25.0 (4.6) 24.9 (4.5) 25.3 (4.8) .34a 
Ponderal index quartiles, n (%)    .96b 
 First quartile 192 (25) 149 (25) 43 (26) — 
 Second quartile 191 (25) 152 (25) 39 (24) — 
 Third quartile 191 (25) 150 (25) 41 (25) — 
 Fourth quartile 191 (25) 148 (25) 43 (26) — 
BMI z score, mean (SD) −0.30 (1.5) −0.24 (1.4) −0.51 (1.6) .04a 
BMI categories, n (%)    <.001b 
 Underweight 135 (18) 88 (15) 47 (28) — 
 Healthy wt 497 (65) 407 (68) 90 (54) — 
 Overweight 61 (8) 50 (8) 11 (6) — 
 Obesity 72 (9) 54 (9) 18 (11) — 
PICU transfer, n (%) 68 (9) 50 (8) 18 (11) .32b 
LOS, median (IQR) 2.83 (2–5) 2.67 (2–5) 3.64 (2–7) <.001c 
ICU LOS, median (IQR) 2.0 (1.1–5.0) 1.7 (1.0–4.1) 3.1 (1.5–6.9) .001c 
Age, mean (SD), median, wk 29 (26), 19 28 (26), 18 35 (28), 28 .003a 
Age <12 mo, n (%) 623 (81) 496 (83) 127 (77) .07b 
Female sex, n (%) 314 (41) 245 (41) 69 (42) .88b 
Race and/or ethnicity, n (%)    <.001b 
 Non-Hispanic white 471 (62) 365 (61) 106 (64) — 
 Hispanic 160 (21) 111 (19) 49 (30) — 
 Non-Hispanic African American 17 (2) 11 (2) 6 (4) — 
 Multiracial 43 (6) 41 (7) 2 (1) — 
 Other 74 (10) 71 (12) 3 (2) — 
Insurance type, n (%)    .01 
 Public 477 (62) 357 (60) 120 (72) — 
 Private 277 (36) 233 (39) 44 (27) — 
 Military, federal, or unspecified 10 (1) 8 (1) 2 (1) — 

IQR, interquartile range; —, not applicable.

a

Two-sample t test.

b

Pearson χ2 test.

c

Kruskal-Wallis test.

Among those without significant comorbidities, LOS increased with increasing ponderal index quartile (P = .001, nonparametric test of trend). Similarly, Spearman’s rank correlation between ponderal index as a continuous variable and LOS revealed a weak positive correlation (r = 0.14; P < .001; Table 2). The proportion of children with a PICU transfer did not significantly differ by ponderal index (P = .52; Table 2) or BMI z score category (P = .52; Table 2). BMI z score had no significant association with ICU LOS (r = 0.05; P = .72; Table 2), and ponderal index revealed a weak association with ICU LOS (r = 0.27; P = .06; Table 2).

TABLE 2

Clinical Outcomes Examined by Adiposity Variables and Demographic Characteristics: Children Without Comorbidities

CharacteristicLOS, Median (IQR), dPPICU Transfer, n (%)PPICU LOS, Median (IQR), dP
Ponderal index r = 0.14 <.001a — — r = 0.27 .06a 
Ponderal index quartiles — .004b — .52c — .56b 
 First quartile 2.10 (1.42–3.29) — 9 (6) — 1.40 (1.32–2.85) — 
 Second quartile 2.69 (1.54–4.91) — 12 (8) — 2.59 (1.35–5.45) — 
 Third quartile 2.86 (1.61–4.44) — 13 (9) — 2.20 (1.50–4.80) — 
 Fourth quartile 2.87 (1.60–5.30) — 16 (11) — 3.55 (1.75–6.20) — 
BMI z score r = −0.02 .56a — — r = 0.05 .72a 
BMI categories — .86b — .52c — .33b 
 Underweight 2.57 (1.65–4.84) — 5 (6) — 2.40 (1.32–2.58) — 
 Healthy wt 2.64 (1.51–4.25) — 33 (8) — 2.60 (1.40–4.80) — 
 Overweight 2.91 (1.73–4.59) — 6 (12) — 1.70 (1.40–3.80) — 
 Obesity 2.72 (1.69–4.43) — 6 (11) — 6.40 (3.8–6.60) — 
Age on admission r = −0.18 <.001a — — r = −0.16 .27a 
Age categories, mo — .001b — .53c — .09b 
 <12 2.72 (1.60–4.78) — 43 (9) — 3.00 (1.50–5.70) — 
 ≥12 2.00 (1.23–3.21) — 7 (7) — 1.40 (1.30–1.50) — 
Sex — .32b — .64c — .97b 
 Female 2.72 (1.69–4.41) — 22 (9) — 1.90 (1.32–6.20) — 
 Male 2.63 (1.46–4.54) — 28 (8) — 2.72 (1.50–4.30) — 
Race and/or ethnicity — .30b — .44c — .55b 
 Non-Hispanic white 2.60 (1.53–4.40) — 30 (8) — 2.95 (1.40–6.00) — 
 Hispanic 2.83 (1.62–4.93) — 10 (9) — 2.30 (0.90–3.40) — 
 Non-Hispanic African American 3.02 (1.70–3.54) — 1 (9) — 1.20 (1.20–1.20) — 
 Multiracial 2.84 (1.77–6.09) — 6 (15) — 2.72 (1.80–3.70) — 
 Other 2.44 (1.34–4.00) — 3 (4) — 3.80 (1.40–6.60) — 
Insurance type — .19b — .45d — .37b 
 Public 2.77 (1.61–4.70) — 34 (10) — 3.2 (1.50–5.70) — 
 Private 2.28 (1.51–4.09) — 16 (7) — 2 (1.33–3.80) — 
 Military, federal, or unspecified 2.57 (1.64–4.43) — 0 (0) — — — 
CharacteristicLOS, Median (IQR), dPPICU Transfer, n (%)PPICU LOS, Median (IQR), dP
Ponderal index r = 0.14 <.001a — — r = 0.27 .06a 
Ponderal index quartiles — .004b — .52c — .56b 
 First quartile 2.10 (1.42–3.29) — 9 (6) — 1.40 (1.32–2.85) — 
 Second quartile 2.69 (1.54–4.91) — 12 (8) — 2.59 (1.35–5.45) — 
 Third quartile 2.86 (1.61–4.44) — 13 (9) — 2.20 (1.50–4.80) — 
 Fourth quartile 2.87 (1.60–5.30) — 16 (11) — 3.55 (1.75–6.20) — 
BMI z score r = −0.02 .56a — — r = 0.05 .72a 
BMI categories — .86b — .52c — .33b 
 Underweight 2.57 (1.65–4.84) — 5 (6) — 2.40 (1.32–2.58) — 
 Healthy wt 2.64 (1.51–4.25) — 33 (8) — 2.60 (1.40–4.80) — 
 Overweight 2.91 (1.73–4.59) — 6 (12) — 1.70 (1.40–3.80) — 
 Obesity 2.72 (1.69–4.43) — 6 (11) — 6.40 (3.8–6.60) — 
Age on admission r = −0.18 <.001a — — r = −0.16 .27a 
Age categories, mo — .001b — .53c — .09b 
 <12 2.72 (1.60–4.78) — 43 (9) — 3.00 (1.50–5.70) — 
 ≥12 2.00 (1.23–3.21) — 7 (7) — 1.40 (1.30–1.50) — 
Sex — .32b — .64c — .97b 
 Female 2.72 (1.69–4.41) — 22 (9) — 1.90 (1.32–6.20) — 
 Male 2.63 (1.46–4.54) — 28 (8) — 2.72 (1.50–4.30) — 
Race and/or ethnicity — .30b — .44c — .55b 
 Non-Hispanic white 2.60 (1.53–4.40) — 30 (8) — 2.95 (1.40–6.00) — 
 Hispanic 2.83 (1.62–4.93) — 10 (9) — 2.30 (0.90–3.40) — 
 Non-Hispanic African American 3.02 (1.70–3.54) — 1 (9) — 1.20 (1.20–1.20) — 
 Multiracial 2.84 (1.77–6.09) — 6 (15) — 2.72 (1.80–3.70) — 
 Other 2.44 (1.34–4.00) — 3 (4) — 3.80 (1.40–6.60) — 
Insurance type — .19b — .45d — .37b 
 Public 2.77 (1.61–4.70) — 34 (10) — 3.2 (1.50–5.70) — 
 Private 2.28 (1.51–4.09) — 16 (7) — 2 (1.33–3.80) — 
 Military, federal, or unspecified 2.57 (1.64–4.43) — 0 (0) — — — 

r, Spearman’s correlation coefficient; —, not applicable.

a

Spearman’s test of independence.

b

Kruskal-Wallis test.

c

Pearson χ2 test.

d

Fisher’s exact test.

For children with significant comorbidities (n = 166), there was no significant correlation between ponderal index as a continuous variable and LOS (r = 0.14; P = .08), and although there was an apparent increase in LOS with increasing ponderal index quartile, this was nonsignificant (P = .09, nonparametric test of trend). Similarly, BMI z score for children with ≥1 comorbidity was not significantly associated with LOS in bivariate analyses, and there was no significant trend detected.

Age on admission revealed a weak negative correlation with LOS (r = −0.18; P < .001; Table 2). Further examination of age revealed a significant association between ponderal index and age on admission (P < .001, Kruskal-Wallis test). Sex, race and/or ethnicity, and insurance type were also examined, and none were significantly associated with LOS or risk of PICU transfer (Table 2).

LOS was not normally distributed (P < .001, Shapiro-Wilk test). The data best approximated a normal distribution after log transformation. A residual analysis of the final model did not indicate a violation of homoscedasticity. Forty observations were identified as extreme cases (absolute value of studentized residuals >2). However, none were identified as influential points. No evidence of collinearity was identified.

Model building demonstrated that ponderal index and log LOS had a significant association in the linear model. On the basis of the aforementioned criteria for confounding of a 10% change in the adjusted slope coefficient estimate for the weight-status variable, only the categorical variable of age, with >12 months as the reference group, qualified as a confounder. The inclusion of this covariate reduced the explanatory power of the model by 32%. The final adjusted association model predicted a 1.62% change in predicted LOS (days) per unit change of ponderal index (Table 3).

TABLE 3

Predicted LOS by Ponderal Index: Children Without Comorbidities

PredictorβSEP% Change in LOSa
Crude model     
 Ponderal index .01 0.003 .001 2.37 
 Constant .16 0.08 .04 — 
Adjusted for age     
 Ponderal index .01 0.003 .04 1.62 
 Age <12 mob .08 0.04 .04 — 
 Constant .17 0.08 .025 — 
PredictorβSEP% Change in LOSa
Crude model     
 Ponderal index .01 0.003 .001 2.37 
 Constant .16 0.08 .04 — 
Adjusted for age     
 Ponderal index .01 0.003 .04 1.62 
 Age <12 mob .08 0.04 .04 — 
 Constant .17 0.08 .025 — 

The variables that were found to be insignificant and were therefore left out of the final model include sex and race and/or ethnicity. —, not applicable.

a

Per unit change in ponderal index.

b

Age >12 mo is the reference group.

Among children without significant comorbidities, logistic regression did not demonstrate a significant difference in odds of PICU transfer by ponderal index (OR 1.05; 95% confidence interval 0.98–1.11) or BMI z score (OR 1.11; 95% confidence interval 0.91–1.37). Odds of PICU transfer also did not significantly differ by adiposity when stratified by age of admission.

We examined whether trends in ICU transfer, LOS, or PICU LOS changed over time. We found no significant difference in PICU transfer by admission year (P = .12, Pearson χ2 test), no significant trend in odds of ICU transfer by admission year (P = .18, Cochran-Armitage test of trend), and no significant difference in mean LOS (0.13, Kruskal-Wallis test) or PICU LOS (0.31, Kruskal-Wallis test) based on admission year.

With this study, we are the first, to our knowledge, to examine the association of infant adiposity with length of hospitalization and risk of PICU transfer. We found that among infants ≤24 months of age admitted for treatment of bronchiolitis without significant comorbid disease, LOS increased with increasing ponderal index. This association persisted after accounting for age. The only other study in which the association between adiposity and bronchiolitis was explored was a small case series of 13 Japanese children; the authors did not identify an association between 3 serum adipokines (leptin, adiponectin, and TNF-α) and BMI and their clinical outcome of a pulmonary index score.16 

Examining this association in another population of infants would be an important first step to either confirm or refute this association. The clinical relevance of the observed association, with 1.6% increase in LOS per unit change of ponderal index, will also likely be questioned. As an example, if 2 male infants aged 3 months were hospitalized for bronchiolitis, 1 being at the median weight for age and length and the other being at the median length but the 98th percentile weight for age, their ponderal indexes would differ by 7. Thus, there would be an 11% difference in their expected LOS. When we use the median LOS in our entire sample (2.8 days), this results in a 0.3-day difference. This may be meaningful at scale when we consider the prevalence of obesity and bronchiolitis.

There is no consensus standard of obesity in infants, and the best proportionality index proxy for adiposity in infants is unclear.14  Studies to attempt to correlate anthropometric measures of obesity with actual adiposity have been inconsistent.17,18  More recent studies in older children have revealed that an index with height cubed (ponderal index) compared with BMI is a more accurate assessment of adiposity.19  We examined both ponderal index and BMI z score. Although ponderal index revealed an association with LOS in both the bivariate analysis and linear regression, which may have clinical relevance, it is still only a proxy measure for the actual adiposity and its associated proinflammatory milieu.

Just as the anthropometric measures used to explore this potential association are proxies for the true exposure of interest, LOS is a surrogate measure for disease severity, which may limit the strength of the observed association. Although LOS is a commonly used proxy for disease severity in bronchiolitis20,21  and is a patient-centered outcome, it leaves many confounders unmeasured. For example, children with obesity and bronchiolitis may have more difficult intravenous access, and therefore potentially longer LOS, secondary to difficulties with fluid management, rather than a secondary effect of adipokines on the respiratory system. They may also have confounding social or logistic barriers to timely discharge owing to associated differences in insurance status, transportation, or social factors, which were not represented in the model.

A possible pathophysiology underlying an association between adiposity and bronchiolitis disease severity is through adipokines such as leptin, adiponectin, and TNF-α, cytokines that are upregulated in the context of obesity. In previous small studies,16  researchers have failed to identify an association between these adipokines and disease severity in bronchiolitis; however, their relationship to lower airway obstructive diseases has been examined in studies of older children and adults,2224  and in small studies of infants with asthma infected with respiratory syncytial virus, has been suggested that higher leptin levels may play a role in disease course.25  Future work would benefit from correlating adiposity with serum adipokines in a large cohort of infants with bronchiolitis.

Although ponderal index revealed an association with length of hospitalization, neither ponderal index nor BMI z score was associated with increased odds of PICU transfer, either in bivariate analyses or multivariate logistic regression. This could suggest that there is no increased risk of acute decompensation based on adiposity, although we also must consider that this study was not powered to examine this outcome. Because LOS was calculated by time on the inpatient ward, we considered whether the results might be biased toward the null by differential risk of direct admission to the PICU by adiposity; however, after stratification for comorbid disease, the adiposity distribution did not significantly differ between those admitted directly to the PICU and those admitted to the regular floor.

For the subgroup with ≥1 significant comorbid disease, we did not find a significant difference in hospital LOS, odds of PICU transfer, or PICU LOS. This may be a true finding, or it may be attributable to the small sample size because the study was not powered to identify a significant association in this subgroup.

In this retrospective cohort study, limitations include data availability and correct abstraction. We took measures to confirm data validity during exploratory analyses and to revisit patient charts individually when necessary. For example, all admission and discharge departments for patients with reported PICU time were manually confirmed. Another important limitation is the lack of reliable information on breastfeeding, both historically for the child and during the hospitalization. Formula feeding has a known association with rapid infant weight gain26  and, potentially, with risk of hospitalization for bronchiolitis,27  and breast milk has antiinflammatory properties28  that could have a clinical impact in children with bronchiolitis.

Infants without significant comorbidities who are hospitalized for bronchiolitis may have longer LOSs with increasing ponderal index. Future research should be used to examine whether this association is confirmed in other populations and to better define the association between severity of disease and adipokines to elucidate the potential biological underpinnings of this relationship.

We thank Jose Rodriguez for his assistance with data extraction.

Ms Haag performed data cleaning, abstracted information from the medical record as needed, conducted data analysis, and drafted the manuscript; Dr Goldfarb performed data cleaning, abstracted information from the medical record as needed, and edited the manuscript; Dr Austin provided input on the methodology and edited and critically examined the manuscript; Dr Noelck helped design the data collection and methods and edited and critically examined the manuscript; Dr Foster conceived and designed the initial study, conducted data analysis, edited the manuscript, and critically examined the manuscript for important intellectual content; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

FUNDING: Dr Foster was supported by National Institutes of Health grant K23DK109199. Funded by the National Institutes of Health (NIH).

1
Singer
K
,
Lumeng
CN
.
The initiation of metabolic inflammation in childhood obesity
.
J Clin Invest
.
2017
;
127
(
1
):
65
73
2
Skilton
MR
,
Marks
GB
,
Ayer
JG
, et al
.
Weight gain in infancy and vascular risk factors in later childhood
.
Pediatrics
.
2013
;
131
(
6
)
e1821
e1828
3
Carpaij
OA
,
van den Berge
M
.
The asthma-obesity relationship: underlying mechanisms and treatment implications
.
Curr Opin Pulm Med
.
2018
;
24
(
1
):
42
49
4
Moldoveanu
B
,
Otmishi
P
,
Jani
P
, et al
.
Inflammatory mechanisms in the lung
.
J Inflamm Res
.
2009
;
2
:
1
11
5
Winer
S
,
Paltser
G
,
Chan
Y
, et al
.
Obesity predisposes to Th17 bias
.
Eur J Immunol
.
2009
;
39
(
9
):
2629
2635
6
Mangodt
TC
,
Van Herck
MA
,
Nullens
S
, et al
.
The role of Th17 and Treg responses in the pathogenesis of RSV infection
.
Pediatr Res
.
2015
;
78
(
5
):
483
491
7
Heinonen
S
,
Rodriguez-Fernandez
R
,
Diaz
A
,
Oliva Rodriguez-Pastor
S
,
Ramilo
O
,
Mejias
A
.
Infant immune response to respiratory viral infections
.
Immunol Allergy Clin North Am
.
2019
;
39
(
3
):
361
376
8
McCallister
M
,
Medrano
R
,
Wojcicki
J
.
Early life obesity increases the risk for asthma in San Francisco born Latina girls
.
Allergy Asthma Proc
.
2018
;
39
(
4
):
273
280
9
Kelly
B
,
West
J
,
Yang
TC
,
Mason
D
,
Hasan
T
,
Wright
J
.
The association between body mass index, primary healthcare use and morbidity in early childhood: findings from the Born in Bradford cohort study
.
Public Health
.
2019
;
167
:
21
27
10
Brüske
I
,
Flexeder
C
,
Heinrich
J
.
Body mass index and the incidence of asthma in children
.
Curr Opin Allergy Clin Immunol
.
2014
;
14
(
2
):
155
160
11
Shay
DK
,
Holman
RC
,
Newman
RD
,
Liu
LL
,
Stout
JW
,
Anderson
LJ
.
Bronchiolitis-associated hospitalizations among US children, 1980-1996
.
JAMA
.
1999
;
282
(
15
):
1440
1446
12
Zaniqueli
D
,
Oliosa
PR
,
Neves
FS
, et al
.
Ponderal index classifies obesity in children and adolescents more accurately than body mass index z-scores
.
Pediatr Res
.
2019
;
86
(
1
):
128
133
13
WHO Multicentre Growth Reference Study Group
.
WHO Child Growth Standards based on length/height, weight and age
.
Acta Paediatr Suppl
.
2006
;
450
:
76
85
14
Ogden
CL
,
Carroll
MD
,
Kit
BK
,
Flegal
KM
.
Prevalence of childhood and adult obesity in the United States, 2011-2012
.
JAMA
.
2014
;
311
(
8
):
806
814
15
Kleinbaum
D
,
Kupper
L
,
Nizam
A
,
Rosenberg
E
.
Applied Regression Analysis and Other Multivariable Methods
. 5th ed.
Boston, MA
:
Cengage Learning
;
2014
16
Kawamata
R
,
Gunji
Y
,
Ozaki
A
, et al
.
Relationships between adipokine profiles, physique index, and severity of bronchiolitis in infancy
.
Fetal Pediatr Pathol
.
2017
;
36
(
5
):
347
356
17
De Cunto
A
,
Paviotti
G
,
Ronfani
L
, et al
.
Can body mass index accurately predict adiposity in newborns?
Arch Dis Child Fetal Neonatal Ed
.
2014
;
99
(
3
):
F238
F239
18
Bell
KA
,
Wagner
CL
,
Perng
W
,
Feldman
HA
,
Shypailo
RJ
,
Belfort
MB
.
Validity of body mass index as a measure of adiposity in infancy
.
J Pediatr
.
2018
;
196
:
168
174.e1
19
Peterson
CM
,
Su
H
,
Thomas
DM
, et al
.
Tri-ponderal mass index vs body mass index in estimating body fat during adolescence
.
JAMA Pediatr
.
2017
;
171
(
7
):
629
636
20
Ralston
SL
,
Lieberthal
AS
,
Meissner
HC
, et al
;
American Academy of Pediatrics
.
Clinical practice guideline: the diagnosis, management, and prevention of bronchiolitis [published correction appears in Pediatrics. 2015;136(4):782]
.
Pediatrics
.
2014
;
134
(
5
).
21
Zhang
L
,
Mendoza-Sassi
RA
,
Wainwright
C
,
Klassen
TP
.
Nebulised hypertonic saline solution for acute bronchiolitis in infants
.
Cochrane Database Syst Rev
.
2017
;(
12
):
CD006458
22
Shore
SA
,
Fredberg
JJ
.
Obesity, smooth muscle, and airway hyperresponsiveness
.
J Allergy Clin Immunol
.
2005
;
115
(
5
):
925
927
23
Yuksel
H
,
Sogut
A
,
Yilmaz
O
,
Onur
E
,
Dinc
G
.
Role of adipokines and hormones of obesity in childhood asthma
.
Allergy Asthma Immunol Res
.
2012
;
4
(
2
):
98
103
24
Morishita
R
,
Franco
MC
,
Suano-Souza
FI
,
Solé
D
,
Puccini
RF
,
Strufaldi
MW
.
Body mass index, adipokines and insulin resistance in asthmatic children and adolescents
.
J Asthma
.
2016
;
53
(
5
):
478
484
25
Chen
ZG
,
Li
M
,
Ji
JZ
, et al
.
The role of serum leptin in infants with wheezing after respiratory syncytial virus infected [in Chinese]
.
Zhonghua Shi Yan He Lin Chuang Bing Du Xue Za Zhi
.
2009
;
23
(
1
):
38
40
26
Appleton
J
,
Russell
CG
,
Laws
R
,
Fowler
C
,
Campbell
K
,
Denney-Wilson
E
.
Infant formula feeding practices associated with rapid weight gain: a systematic review
.
Matern Child Nutr
.
2018
;
14
(
3
):
e12602
27
Lanari
M
,
Prinelli
F
,
Adorni
F
, et al
;
Italian Neonatology Study Group on RSV Infections
.
Maternal milk protects infants against bronchiolitis during the first year of life. Results from an Italian cohort of newborns
.
Early Hum Dev
.
2013
;
89
(
suppl 1
):
S51
S57
28
Andreas
NJ
,
Kampmann
B
,
Mehring Le-Doare
K
.
Human breast milk: a review on its composition and bioactivity
.
Early Hum Dev
.
2015
;
91
(
11
):
629
635

Competing Interests

POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.

FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.