Underweight, overweight, and obesity as independent risk factors for hospitalization in adults and children from influenza and other respiratory viruses

Background The relationship between obesity and risk of complications described during the 2009 influenza pandemic is poorly defined for seasonal influenza and other viral causes of influenza‐like illness (ILI). Methods An observational cohort of hospitalized and outpatient participants with ILI was conducted in six hospitals in Mexico. Nasopharyngeal swabs were tested for influenza and other common respiratory pathogens. Results A total of 4778 participants were enrolled in this study and had complete data. A total of 2053 (43.0%) had severe ILI. Seven hundred and seventy‐eight (16.3%) were positive for influenza, 2636 (55.2%) were positive for other viral respiratory pathogens, and 1364 (28.5%) had no respiratory virus isolated. Adults with influenza were more likely to be hospitalized if they were underweight (OR: 5.20), obese (OR: 3.18), or morbidly obese (OR: 18.40) compared to normal‐weight adults. Obese adults with H1N1 had a sixfold increase in odds of hospitalization over H3N2 and B (obese OR: 8.96 vs 1.35, morbidly obese OR: 35.13 vs 5.58, respectively) compared to normal‐weight adults. In adults with coronavirus, metapneumovirus, parainfluenza, and rhinovirus, participants that were underweight (OR: 4.07) and morbidly obese (OR: 2.78) were more likely to be hospitalized as compared to normal‐weight adults. All‐cause influenza‐like illness had a similar but less pronounced association between underweight or morbidly obesity and hospitalization. Conclusions There is an increased risk of being hospitalized in adult participants that are underweight or morbidly obese, regardless of their viral pathogen status. Having influenza, however, significantly increases the odds of hospitalization in those who are underweight or morbidly obese.


| BACKG ROU N D
Until the onset of the 2009 influenza A/H1N1 pandemic, body mass index (BMI) was not widely appreciated as an independent risk factor for influenza. While there was no increase in the rates of ILI being reported in obese patients, 1 early reports during the pandemic noted an association between severity of illness and obesity. [2][3][4][5][6][7][8][9][10][11] Several studies in adults found an association between obesity and death due to pandemic influenza. [11][12][13][14] Another study, however, found no association between obesity and mortality in intensive care unit (ICU) participants. 15 Far fewer studies examined syndromic influenza-like illness (rather than restricted to influenza). Two studies found a link between obesity and hospitalization from influenza-like illness, 16,17 and another showed higher rates of outpatient visits for influenza-like illness in obese patients. 18 One additional study evaluated rates of obesity in the community and demonstrated that communities with a greater prevalence of obesity were more likely to have high influenza-related hospitalization rates. 19 None of these studies, however, determined the etiology of the participants illness, and none examined low body mass index as a risk factor.
Given the limited data with syndromic ILI severity and obesity, we examined the association between body mass and severity of influenza-like illness (ILI) during four consecutive years (2010)(2011)(2012)(2013)(2014) in both pediatric and adult populations. We hypothesized that those individuals categorized as overweight, obese, and morbidly obese would be more likely to be hospitalized with ILI from influenza, but also from other respiratory viruses, as compared to those participants with a normal body mass.

| Study design and sites
Beginning in April 2010, participants were enrolled in ILI 002, an observational cohort study conducted by the Mexican Emerging Infectious Disease Clinical Research Network, Mexico (La Red). The study was conducted at five centers in Mexico City, Mexico, and one in San Luis Potosí, Mexico. The study sites are located in urban environments and include two general hospitals, two tertiary care hospitals (one that serves those with respiratory problems and one that serves those with metabolic disorders), and two tertiary care pediatric centers.

| Case definition and study population
The study population included participants of any age who presented with an influenza-like illness (ILI). Influenza-like illness was defined by the presence of at least one respiratory symptom (eg, shortness of breath, cough) and either fever (≥38°C or subjective feverishness) or one or more non-respiratory symptoms (eg, malaise, headache). The participants included were those who sought medical attention at a study center and agreed to participate in the study.

| Virology
Nasopharyngeal swab and nasal aspirate samples were sent to a cen-  The outcome measure of severity of disease was defined as having severe disease if participants were hospitalized within 14 days of enrollment. Participants that were in the emergency room >24 hours yet never admitted were excluded as these cannot be clearly categorized as outpatients or hospitalized.

| Statistical analyses
All analyses were performed separately for pediatric and adult participants. Descriptive statistics include presentation of proportion for categorical variables and means and standard deviations for continuous variables. Logistic regression models were used to examine the relationship between severity of illness and BMI. Sex, chronic medical condition (present/absent), and age (as a linear term) were included in all models to adjust for imbalances.
Interactions between BMI and other covariates (adjustment variables and pathogen categories) were examined. When interaction terms were significant, logistic regression analyses were performed for each level of the covariate. We calculated the odds ratio (OR) of being hospitalized based on body mass category, where normal was the reference category. We estimated 95% confidence intervals (95% CI) and P-values of the odds ratios using the Firth method to account for separation that may occur in logistic regression models due to small sample sizes. 23 SAS version 9 (SAS Institute Inc., Cary, NC, USA) and R 3.1.0 (R Foundation for Statistical Computing, Vienna, Austria) were used to complete all analysis.

| Regulatory aspects
The Institutional Review Board (IRB) at each site approved this study. The study was conducted following the principles of the International Conference on Harmonization's Good Clinical Practice, Declaration of Helsinki, and the Mexican General Health Law. All participants provided informed consent. The project was registered on clinicaltrials.gov (NCT01418287).   (Table 3) In pediatric participants, there were no significant interactions between BMI and the pathogen groups; therefore, all of the participants were analyzed together. Only obese pediatric participants were statistically significantly more likely to develop severe ILI compared to their counterparts with a normal z-score (OR: 2.2, P = 0.002; Table 4).

| D ISCUSS I ON
We have demonstrated that adults that are underweight or morbidly obese are more likely to be hospitalized from an influenzalike illness, regardless of the causative agent of the illness, than normal-weight adults. The risk follows a "U"-shaped curve, where individuals at both extremes (ie, those that are underweight or morbidly obese) were more likely to develop severe ILI when compared with normal-weight, overweight, or obese individuals. The increased risk of underweight and morbidly obese adults was stronger for influenza-positive adults compared with those positive for other respiratory viruses or negative for any respiratory viruses.
Analysis of the risk of specific respiratory viral pathogens or strains of influenza was difficult given the small number of participants in each of these groups. Despite the large confidence intervals, the data seem to indicate that the association between morbid obesity and risk of severe ILI is not just statistically significant in adults with influenza A/H1N1 pdm 2009. A statistically significant OR was also found in participants with influenza H3N2 and B. While the association was also present for these strains of influenza, the odds ratio was six times higher in influenza H1N1 participants (OR: 35.13 vs 5.58), which may explain why individuals with H1N1 were the first in which the association between body mass and disease severity was postulated. However, influenza H3N2 and influenza B still have twice the risk of hospitalization as the other most common respiratory viruses. Preclinical data suggest reduced vaccine efficacy in obesity, 24,25 and while the increased risk does not seem to be due to reduced antibody titers, 26,27 it may be from reduced cellular immunity. 28 However, this would not explain the differential risk of H1N1 over H3N2, nor of all influenza over other respiratory viruses.
The risk associated with low body mass is a finding that has previously been described as it relates to all-cause mortality, 29

CO N FLI C T O F I NTE R E S T
The authors have no conflict of interest to declare. Age, gender, and presence of chronic conditions were also taken into account when running the multiple logistic regression model. Statistically significant ORs and their corresponding P-values are bolded.