Obesity and risk of respiratory tract infections: results of an infection-diary based cohort study

Background Respiratory tract infections (RTIs) are a major morbidity factor contributing largely to health care costs and individual quality of life. The aim of the study was to test whether obesity (BMI ≥ 30 kg/m2) is one of the risk factors underlying frequent RTIs in the German adult population. Methods We recruited 1455 individuals between 18 to 70 years from a cross-sectional survey on airway infections in Germany and invited them to self-report in diaries incident RTIs experienced during three consecutive winter/spring seasons. RTIs reported in these 18 months and summary measures adding-up individual RTIs were the outcomes of interest. Results Compared to individuals with normal weight, obese individuals reported a consistently higher frequency of upper and lower RTIs and predominantly fell in the upper 10% group of a diary sumscore adding-up 10 different RTI symptoms over time. Obesity was associated both with lower RTIs (adjustedOR = 2.02, 95%CI = 1.36–3.00) and upper RTIs (adjustedOR = 1.55, 95%CI = 1.22–1.96). Adjusting for demographic and lifestyle variables did only marginally affect ORs. Stratified analyses suggested a stronger association for women and effect modifications by sports activity and dietary habits. Conclusions We confirm the association of obesity with infection burden and present evidence for putative interaction with sports activity and dietary patterns. Electronic supplementary material The online version of this article (10.1186/s12889-018-5172-8) contains supplementary material, which is available to authorized users.


Background
Frequent and severe respiratory tract infections (RTIs) constitute an important morbidity factor in our society and a considerable cost burden in terms of medical treatment and time of work-loss [1,2]. RTIs are divided into upper RTIs (URTIs) including common cold, pharyngitis, otitis, sinusitis, laryngotracheitis, epiglottitis and lower RTIs (LRTIs) including bronchitis, pneumonia and bronchiolitis [3]. Individual exposure to infectious agents and host factors such as smoking [4,5] and vitamin D status [6,7] are believed to contribute to observed differences in RTI risk. In addition, the role of overweight (body mass index (BMI) = 25.0-29.9 kg/m 2 ) and in particular obesity (BMI ≥ 30 kg/m 2 ) in predisposition to RTIs is increasingly discussed [8][9][10][11][12][13]. This growing interest is driven by the rising number of overweight and obese individuals worldwide [14] and the emerging knowledge of notable immunological imbalances in association with obesity [15]. Most of the studies targeting adults explored the association of obesity with specific RTIs and their outcomes. Thus, obesity was associated with non-allergic rhinitis [8] and influenza like-illness [9]. Moreover, two population-based studies which investigated the role of obesity as risk factor for community acquired pneumonia (CAP) in the general population resulted in controversial findings [10,11]. Two recent Danish population-based studies reported an excess of a large spectrum of RTIs including pneumonia among obese people [12,13]. The overall aim of our study targeting the adult population in South Baden, Germany, is to identify risk factors for the susceptibility to RTIs. Here we present data on the role of obesity as contributing factor to a high RTI burden in the German society and explore effect modification by gender, sports activity and nutritional patterns.

Study population
Study participants (n = 1455) were recruited from the airway infection susceptibility (AWIS) cross sectional study querying RTI burden in an adult population in South-Baden, Germany [16]. The study protocol was approved by community officials and the Ethics Committee of the University of Freiburg (Ref. No. 258/11_120365). Based on the RTI history-score individuals of putative low, medium and high risk of future RTIs were invited to the actual sub-cohort. The RTI history score is summarizing information on the frequency and severity of RTIs and antibiotics use over the past two years, selfassessed RTI susceptibility, and occurrence of selected severe infections [16]. Study participants were requested to fill-in an additional questionnaire (baseline questionnaire) on lifestyle factors and co-morbidities and to complete monthly diaries registering the monthly occurrence and the duration (< 2 weeks, > 2 weeks) of RTIs, namely sinusitis, rhinitis, otitis media, pharyngitis/laryngitis, tonsillitis, influenza-like illness, bronchitis, pneumonia, pleurisy and other acute RTIs, from the beginning of November to the end of April of three seasons: 2012/13, 2013/14 and 2014/15. Furthermore, the intake of antibiotics, doctor visits, hospitalisation for RTIs and the impact of RTI symptoms on their daily activities were queried. Further recruitment details into the AWIS study and the present sub-cohort are presented under Additional files 1 and 2. Informed consent was obtained from all individual participants included in the study.

Outcome measures
In order to describe the association between obesity and RTIs, different outcome indicators were considered: outcomes at the level of each month ["any RTI", "any URTI" (sinusitis, rhinitis, otitis media, pharyngitis/laryngitis and tonsillitis), "any LRTI" (bronchitis, pneumonia and pleurisy), "≥3 RTIs", "any long lasting infection" (> 2 weeks)]; at the level of each winter season ("≥4 months with infections", "≥3 long lasting infections"); and at the individual level (i.e. are defined once per individual and covering the overall study period). The ten specific RTI symptom categories were considered with the binary symptom indicators "infection reported" or "no infection reported" for each month.
When counting the episodes for the outcome indicator "≥3 long lasting infections", different infection symptoms were counted as separate episodes, even if they overlapped in time. However, within one symptom category at least one month without this specific infection was required to call it a new episode. We also calculated a monthly diary RTI score, averaging the ten RTI symptom categories with the coding "0" for "no infection reported", "1" for "reported infection with duration < 2 weeks", and "2" for "reported infection present with duration >2 weeks". Missing values for individual infection items were treated as zero. If an individual RTI symptom was reported, but information on duration was missing, it was counted as "reported infection with duration < 2 weeks". If all items were missing, no diary score was computed. The diary RTI score at the monthly level was expanded to a score at the seasonal level by averaging over the six months (November-April) of each season, and to an overall score at the individual level by averaging over all available months. The respective upper 10% of these diary scores within each month, season and overall served as additional outcome indicators.
Further variables considered in the study were age, gender, self-reported weight and height for BMI calculation (BMI was categorized as < 30 (non-obese), 25 ≤ BMI < 30 (overweight) and ≥30 (obese)), educational level, contact with children, comorbidities, removed immunological organs, smoking status, sports activity and dietary intake patterns. Details on these variables are described in the Additional file 1 and supplementary information on dietary intake patterns is presented in Additional file 3.

Statistical analysis
Statistical analysis was performed using Stata (version 14 STATSCorp, USA). Descriptive statistics: Monthly prevalences of individual RTI symptoms were computed by taking the average over all subjects available at each month and then averaging over all 18 months covered. Prevalences at the seasonal level were computed accordingly averaging over all three seasons covered. The corresponding confidence intervals (CIs) and p-values are based on a generalised linear model with identity link and binomial type variance together with robust variance estimates. The frequency of long lasting infections among all months with infections was analysed accordingly. However, due to the limited number of cases for tonsillitis and otitis media we determined the monthly frequency of long-lasting infections by pooling the data over all seasons and for pneumonia by pooling all indicated months.

Odds ratios (ORs) for outcome variables and adjustments
At the monthly level ORs were computed using a logistic regression model with a random intercept applied to the individual data for each month taking the 18 months as a categorical covariate into account in addition to the obesity status indicator. Due to its small prevalence, pleurisy was not considered as single outcome in these analyses. Outcomes at the seasonal level were analysed accordingly with the individual data for each winter season and taking into account the three seasons as a categorical covariate. Outcomes at the individual level were analysed using a logistic regression model. Results are ORs and 95% CIs. Adjusted ORs are based on including age groups and education as simultaneous categorical covariates. Furthermore, in order to study the stability of the obesity-RTI association with respect to potential confounders, ORs were adjusted by respective variables. Subjects with incomplete covariate data were excluded from multivariate analyses.

Subgroup analysis
Effect modification by a binary variable was assessed by fitting an overall model with the corresponding interactions parametrized so that we could directly read off the two subgroup-specific ORs. Effect modification by sports activity and nutrition patterns was explored among those representing the lower and upper third of respective scores.

Characteristics of the study population
The study population comprised 1455 individuals (931 female and 524 male) with a median age of 51.08 years. Based on BMI calculated from self-reported weight and height, 2.1% of the population was underweight (BMI < 18.5 kg/m 2 ), 54% had a normal weight (18.5 kg/m 2 ≤ BMI < 25 kg/m 2 ), 31.1% was overweight, and 12.8% was considered obese (Table 1). In women, the distribution was 2.8%, 60.21%, 25.0%, and 12.1% and in men 0.76%, 43.1%, 41.8%, and 14.3%, respectively. The study participants were mainly of medium and high educational level, non-or ex-smokers, moderately affected by selected co-morbidities and they reported rather infrequent contact to small children. Further information on the study population and completed diaries is reported in Table 1 and Additional file 4.

Reported RTIs over 18 months covering three winter seasons
Missing rates of single items in the returned diaries were limited and ranged from 1.2% for rhinitis and pharyngitis/laryngitis to 2.6% for other acute respiratory infections. Study participants reported most frequently rhinitis (26.6%), followed by influenza-like illness (11.4%) and pharyngitis/laryngitis (10.5%), whereas pleurisy (0.10%) was rarely experienced. Any URTI (31.5%) was more frequent than any LRTI (7.9%). Apart from the LRTIs bronchitis, pneumonia and pleurisy, which more men than women reported, all other RTIs were more prevalent among women (Table 2). Seasonal patterns of reported infections show a February peak for two of the three assessed infection seasons (2012/13 and 2014/15, see Additional file 5). Respiratory infections with a high fraction of long duration were almost exclusively LRTIs, namely pneumonia (59%), followed by bronchitis (48.2%). Men were overrepresented among those with long-lasting RTIs ( Table 2).

Association between obesity and reported RTIs
Compared to normal weight individuals, overweight and obese people consistently had a higher prevalence (Table 3) for the single RTIs, URTIs, LRTIs, as well as the other outcome parameters we looked at with other acute infections and pneumonia as the exceptions. For pneumonia, only obese subjects had a higher prevalence. The overweight group was typically falling in between the groups with normal weight and obesity ( Table 3). The strongest association was seen for pneumonia and bronchitis, and accordingly, any LRTI was more strongly associated with obesity than any URTI. Long-lasting RTIs, frequent RTIs and high diary scores were also more strongly associated with obesity than the individual symptoms. Adjustments by age and education did only marginally change these estimates. Among subjects with an infection, long lasting infections were again associated with obesity, reaching significance for any RTI, rhinitis, pharyngitis/laryngitis, influenza-like illness, and bronchitis ( Table 3).

Robustness of associations to confounding
For a better understanding of the robustness of the relationship between RTI burden and obesity, the effect of adjusting for putative confounders was explored (Additional file 6). The studied demographic and lifestyle variables (age, gender, education level, smoking status, contact to children, asthma, sports activity, dietary patterns and previous removal of immune organs) did only marginally affect ORs. However, adjustment for asthma, chronic obstructive pulmonary disease (COPD) or a summary score covering all queried co-morbidities weakened the relationship between obesity and all outcomes considerably. Adjustment for vitamin D levels among those for which serum was available (n = 508), had only a slight effect on the magnitude of the association between obesity and RTI outcomes.

Effect modification by gender, sports activity and nutritional pattern
The association between obesity and RTI outcomes was more prominent for women than for men and reached statistical significance only for the former (Table 4). For most outcomes this interaction was not significant, with the individual level diary score as an exception. When looking at sports activity, for most outcomes the association with obesity was confined to those physically more active and not seen for those reporting little sports activity (Table 5). For all outcomes the association was less pronounced in the latter group (compare the ratios of ORs in Table 5), a difference that reached significance for all outcomes except those with low prevalence. Typically the prevalence of an outcome was only increased in the small group of people with obesity and higher sports activity whereas all other groups presented rather similar patterns. Similarly, the prevalence of outcomes was increased among people with obesity and a more favourable nutritional pattern, but comparable among the other groups ( Table 6). The interaction reaches significance for the majority of outcomes.

Discussion
RTIs constitute an important morbidity factor considering the high health care costs, the time lost from work, and the impaired quality of life among those recurrently affected [1,2,17]. Obesity belongs to one of the host risk factors for RTI and has possibly an emerging role due to the dramatically increasing prevalence of obesity worldwide. In the present study, we report on the association of obesity with individual RTIs as well as with a diary score summarising different incident RTI symptoms over a period of 18 months. Our investigation could demonstrate an association between obesity and RTIs confirming previous findings on influenza-like illness [9], bronchitis [18] and pneumonia [10,12]. We also saw an association between obesity and rhinitis, sinusitis and pharyngitis/laryngitis. An elevated risk for sinusitis among obese was also reported in a populationbased cohort of Danish women [13]. None of the two Danish population-based studies [12,13] used ORs of monthly prevalence, but hazard ratios (HRs), as they could identify events on a daily basis. The HR of 1.6 [12] for the association with RTIs and the HR of 1.48 [13] for the association with URTIs are, however, of similar magnitude to the risk estimates which we observed. Mechanistically, excess adiposity might weigh down host defence as several mouse as well as human studies have suggested [19,20]. The here observed associations were more prominent for LRTIs compared to URTIs, but evident for both, and more pronounced when considering long lasting or frequent RTIs compared to single symptoms. Based on the infection diary data, we generated a RTI diary score summing-up all ten symptoms and allowing to average per month, per whole season or over the whole period of three years. Considering the upper ten percentile of the distribution of such scores as an outcome, associations were typically stronger than when considering single symptoms, and interactions were more pronounced. Moreover, the results of the seasonal score were very similar or even stronger than those of the three-years score, arguing for the adequacy to query six months infectious events in future studies to identify the infection-prone sub-group of the population. Lifestyle habits seem to contribute to an individual's risk for RTI. Among them, cigarette smoking has been reported as a major environmental risk factor for recurrent and severe RTIs [4,5]. Frequent contact to small children [21,22], vitamin D deficiency [23,24], and lack of physical activity [25,26] constitute other exposures associated with heightened RTI risks. Moreover, higher levels of education were associated with a lower risk of CAP [27]. Based on those previous findings we investigated their role as possible confounders. The association between obesity and RTIs remained nearly unchanged after adjustment for age, gender, educational status, contact to children, smoking status, sports activity and nutrition scores, suggesting that the association is not markedly confounded by the effects of these factors on both BMI and the risk of infections. Also additional adjustment by measured serum vitamin D in a subgroup for which measurements were available did not change the risk estimates considerably. This supports arguments that the observed associations between obesity and RTI burden are due to physiological differences in the immune responsiveness between obese and non-obese individuals rather than lifestyle differences. In addition, some chronic diseases, foremost asthma and COPD, are associated with both an increased risk of RTIs and obesity [28][29][30][31][32]. Considering these associations we investigated the effect of asthma, COPD and a comorbidity scoresummarizing the other chronic conditionson the relationship between obesity and individual RTIs and the RTI diary score. Adjusting for these conditions individually and even more so in a combined fashion resulted in a considerable attenuation of the association between obesity and considered RTI outcomes. Hence part of the association between infections and obesity might be explainable by associations of co-morbidities with both. We see a gender difference in the observed associations with more noticeable findings for women. A significantly    increased risk for combined RTIs was also restricted to women in a Danish blood donor cohort [12]. Several lines of research support this notion: Szabova et al. and Ilavska et al. reported gender-dependent effects of obesity on the immune system [33,34]. The effect of BMI on a variety of immune parameters including those with relevance for immune defence was much more apparent in women than in men [34]. NK cells (CD3-/CD16+/CD56+), represent first-line cells for the clearing of virus-infected cells. Reduced levels of these cells reported for obese women, but not for respective men, might underlie the gender effect seen in our study. We also investigated a potential effect modification by sports activity and nutrition. Interestingly, an association between obesity and RTIs was evident only for those obese individuals who reported a higher level of sports activity. Thus, only the group of obese people who engaged in more intensive sports activity reported RTIs more frequently whereas obese people with low sports activity and non-obese with low or high sports activity showed comparable lower prevalences for most outcomes. We hypothesize that oxidative stress induced by vigorous aerobic as well as anaerobic sports activity is exacerbated in people with obesity, but not in normal weight individuals. Evidence supporting this has been previously published [35]. An imbalanced oxidative stress status may have negative consequences on mounting an appropriate immune response towards respiratory pathogens. Excessive reactive oxygen species (ROS) was shown to hinder T cell responses to viral infection [36] and ROS accumulation was detected in autophagy-deficient effector T cells rendering them incapable of controlling viral infections [37].
A similar surprising result was found when studying the effect modification by dietary patterns. Here we queried the participants' dietary habits and classified them as adhering to a more favourable or more unfavourable dietary pattern according to Winkler et al. [38]. Aware of the limitations of a one-time assessment of a habitual diet, we found a more pronounced relationship between obesity and infections among obese people who reported an apparent healthier diet. Thus, again only the group of obese individuals who presumably eat a healthier diet showed an increased risk of RTIs. The question arises as to whether misreporting of dietary habits among these individuals with and without RTIs may explain the puzzle. One can imagine that obese individuals may have an increased perception of RTI related symptoms experiencing the contradiction between living a healthy lifestyle and being affected by excess weight and frequent infections. On the other hand the inconspicuous results from the non-obese population with respect to favourable and unfavourable diet pattern would somewhat argue against this explanation. Alternatively, among the group of people with obesity a genetically defined subgroup may exist predisposing to both, excess body weight and proneness to infections.

Strengths and limitations
As strengths of our study we count 1) its sample size, allowing for the analysis of effect modification, 2) its prospective design involving 18 months infection diaries for the exploration of the relationship between BMI and subsequent RTI frequency and severity, 3) the comprehensive information on lifestyle and co-morbidities allowing to study the interplay of such factors on their effect on infections, and 4) the wide range of outcome indicators considered. The uniformity of the results with respect to these outcomes also suggests that in the field of airway infection morbidity, studies may be comparable despite the fact that they often concentrate on different RTI outcomes. In line with the majority of epidemiological studies in this area of research, our study suffers from some limitations, including the reliance on self-reported outcomes and exposure data with the risk of misclassification. However, we found -for instance -a good agreement between BMI derived from self-reported weight and height data and BMI calculated from measured values available for a sub-cohort (n = 508). Moreover, differential misclassification which would substantially bias the relationship between obesity and RTIs is rather unexpected in this setting. The disproportional selection of women into the study may negatively impact the generalizability of some of our results.