Obesity survival paradox in pneumonia: a meta-analysis

1The prevalence of obesity has dramatically increased in the last two decades. The diagnosis of obesity is often based on body mass index (BMI), calculated as weight in kilograms divided by height in meters squared (kg/m2). The ideal BMI is between 18.5 and 24.9. Being overweight is considered as having a BMI between 25 and 29.9, and being classified as obese falls into a BMI of 30.0 or greater. Obesity is associated with an increased risk of cardiovascular disease and type 2 diabetes. However, an inverse relationship between obesity and mortality has been described in patients with heart failure, coronary heart disease, and diabetes. This phenomenon is known as the ‘obesity survival paradox.’

2Pneumonia is one of the most common infectious diseases; however, there is uncertainty about the association between obesity and pneumonia risk or pneumonia mortality. For example, Baik et al. suggested that obesity was directly associated with the development of community-acquired pneumonia (CAP). However, Phung et al. did not find that obesity was significantly associated with pneumonia risk. Takata et al. indicated that mortality risk was not different between obese pneumonia patients and normal weight patients. However, other studies reported that obese subjects with pneumonia had lower mortality compared to normal weight subjects. Thus, whether the ‘obesity survival paradox’ exists in pneumonia is still unclear.

3To date, no meta-analysis has shown whether an ‘obesity survival paradox’ exists for pneumonia. The aim of this meta-analysis was to investigate the relationships between elevated BMI, pneumonia risk, and mortality.

4This meta-analysis was performed according to a predetermined protocol described in the following paragraphs, using standard systematic review techniques, as outlined by the Meta-Analysis of Observational Studies in Epidemiology (MOOSE) criteria.

5A literature search was performed (WN and YZ) using the PubMed search engine, with the database being last accessed on 15 June 2013. The Embase database was also searched for relevant studies published up to June 2013. References from relevant articles were manually checked for further studies. The detailed search strategy is presented in the Additional file 1.

6Two reviewers (WN and YZ) independently screened the abstracts of papers identified by the literature search, retrieved potentially relevant studies and determined study eligibility. Studies were included if: (1) the study design was a prospective or retrospective cohort study; (2) the exposure of interest was BMI; (3) they reported adjusted relative risks (RRs), hazard ratios (HRs) or odds ratios (ORs) with corresponding 95% confidence intervals (CIs), or provided a RR/HR/OR with corresponding 95% CI per unit increment in BMI; and (4) the outcome was pneumonia incidence or mortality. If the same cohort was used in more than one publication, we included the publication that reported the results in greater detail or, if similar, the one with the largest number of cases. Data published only in abstract form were excluded. Case reports, review articles and commentary articles were also excluded. Studies with pediatric participants or pregnant populations were not included.

7From each study, two reviewers (WN and YZ) independently extracted the first author, publication year, study design, location where the study was performed, number of cases and cohort size, gender and age of study participants, follow-up duration, method for assessing height and weight, ascertainment of pneumonia, type of pneumonia, BMI category, adjusted RR/HR/OR and the corresponding 95% CI, and covariates controlled for multivariable analysis. The authors of the relevant studies were contacted by Email if more information was needed.

8Two independent reviewers (WN and YZ) completed the quality assessment. The Newcastle–Ottawa Scale (NOS) was used to evaluate the methodological quality, which scored studies by the selection of the study groups, the comparability of the groups and the ascertainment of the outcome of interest. Discrepancies were resolved by consensus and discussion. The detailed criteria of the methodological quality assessment are in Additional file 2.

9For pneumonia risk and pneumonia mortality risk, we calculated summary RRs and 95% CIs for overweight and obesity versus normal weight. The random effects model was utilized. HRs and ORs were regarded as equivalent to RRs in cohort studies. If a study reported results specifically for men and women, respectively, we combined the sex-specific RR estimates using a fixed-effects model before combining with other studies.

10In dose–response analysis, we calculated the RR per 5-unit increase in the BMI levels for each study. The average of the natural logarithm of the RRs was estimated and the RR from each study was weighted by the inverse of its variance. A two-tailed P < 0.05 was considered statistically significant. We also combined the sex-specific estimates using a fixed-effects model to generate an estimate for both genders combined. The method described by Greenland and Longnecker was used for the dose–response analysis and study-specific slopes (linear trends) and 95% CIs were computed from the natural logs of the RRs and CIs across categories of BMI. This method requires the distribution of case and person-years and the median level of BMI in each category to the corresponding RR for each study (the RRs with estimates for at least three quantitative exposure categories are known). The midpoint between the upper and lower boundary for each BMI category was assigned to the corresponding RR estimate. For studies with an open-ended highest or lowest BMI category, we assumed that the amplitude was the same as the closest adjacent category. Random effects models were used to pool the respective results. The dose–response results in the forest plot were presented for a 5 kg/m2 BMI increment.

11Nonlinear dose–response curves were plotted using restricted cubic splines for each study, using knots fixed at percentiles 10%, 50% and 90% through the distribution; then these were combined using multivariate meta-analysis.

12Statistical heterogeneity among studies was evaluated using the Q and I2 statistics. For the I2 metric, we considered low, moderate and high I2 values to be 25%, 50% and 75%, respectively. We examined the role of several potential sources of heterogeneity by subgroup analyses according to study design, gender, ascertainment of case, pneumonia type, assessment of anthropometry, and duration of follow-up. Meta-regression was also performed to find the sources of heterogeneity. Sensitivity analysis was conducted by excluding one study at a time to explore whether the results were driven by one large study or by a study with an extreme result. Potential small study effects, such as publication bias, were investigated with funnel plots.

13All statistical analyses were performed with the Stata software (version 12.0, Stata Corporation, College Station, Texas). A threshold of P < 0.1 was used to decide whether heterogeneity was present. In other cases, P values were two sided with a significance level of 0.05.

14The process of identifying relevant studies is shown in Figure 1. The initial search produced 1,035 studies from the PubMed and Embase databases. After exclusion of duplicates and irrelevant studies, 115 potentially eligible studies were selected. After detailed evaluations, 21 studies were selected for final meta-analysis. A manual search of reference lists from these studies did not yield any new eligible study. Several studies investigated the association between BMI and mortality. We contacted these authors to get additional data on pneumonia mortality. Dr. Sun Ha Jee and colleague shared their data. Finally, 22 studies were included in this meta-analysis.

15Twelve cohort studies (n = 1,536,623) investigated the association between BMI and pneumonia risk, and ten studies (n = 1,375,482) assessed the association between BMI and pneumonia mortality. There were seven retrospective cohort studies and fifteen prospective cohort studies. The durations of follow-up varied from 1 year to 15.8 years. Sixteen studies collected measured BMI; three studies collected self-reported data. The characteristics of each study are presented in Table 1. The methodological quality assessment is provided in Additional file 2. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist for meta-analysis is provided in Additional file 3.

16Compared with normal weight individuals, overweight and obese individuals were associated with a significantly increased risk of pneumonia (RR = 1.33, 95% CI 1.04 to 1.71, P = 0.02, I2 = 87%). A small-study effect was demonstrated using a funnel plot [see Additional file 4]. Ten studies reported RRs for categorized BMI levels. Thus, we included these studies for the dose–response analysis. The summary RR was 1.04 (95% CI 1.01 to 1.07, P = 0.01, I2 = 84%; Figure 2). A potentially nonlinear dose–response relationship was not detected (P > 0.05; Figure 3). We found evidence of a small-study effect as assessed by funnel plot [see Additional file 5].

17We conducted a sensitivity analysis by omitting one study at a time and calculating the pooled RRs for the remainder of the studies. This sensitivity analysis showed that the results were not changed (data not shown). The potential sources of heterogeneity were explored by stratifying analyses. Studies comparing overweight and obese subjects with normal weight subjects, and studies assessing dose–response associations were explored, respectively. The positive relationship between BMI and risk of pneumonia was significant in subgroups by ascertainment of case and assessment of weight and height [see Additional file 6]. However, in subgroups of retrospective cohort studies, male population, female population, hospital-acquired pneumonia, and longer follow-up duration, the positive relationships were not statistically significant [see Additional file 6]. A meta-regression found that assessment of weight and height might be the source of the high heterogeneity.

18Overweight and obese individuals were associated with decreased mortality risk (RR = 0.83, 95% CI 0.77 to 0.91, P < 0.01, I2 = 34%). A small-study effect was revealed by the funnel plot [see Additional file 7].

19Six cohort studies were identified in the dose–response analysis. The summary RR was 0.95 (95% CI 0.93 to 0.98, P < 0.01, I2 = 77%; Figure 4). There was no evidence of a nonlinear relationship with BMI (P = 0.44; Figure 5). The shape of the funnel plot was asymmetrical, suggesting that there was a small-study effect [see Additional file 8].

20In sensitivity analyses, no result was significantly altered when excluding studies one by one (data not shown). Stratified analyses were defined by study design, gender, assessment of weight and height, and follow-up duration. Studies comparing overweight and obese subjects with normal weight subjects, and studies assessing dose–response associations were explored, respectively. The inverse relationship between BMI and risk of mortality was significant in subgroups that were defined by ascertainment of case and study design. However, in the female subgroup or studies with shorter follow-up duration, the inverse relationship was not statistically significant [see Additional file 9]. Study design was found to be the major source of heterogeneity by meta-regression.

21In this meta-analysis, we showed that an ‘obesity survival paradox’ might exist for pneumonia. On the one hand, there was a positive association of obesity with pneumonia risk. On the other hand, pneumonia mortality was lower for patients with high BMI compared to normal BMI.

22In a recent meta-analysis, Phung and coworkers found a J-shaped relationship between BMI and risk of CAP and a U-shaped relationship between BMI and risk of influenza-related pneumonia. Our results were partly similar to this previous report. In the current meta-analysis, we only investigated the association between higher BMI and pneumonia risk, but did not assess the association between subnormal BMI and pneumonia risk. Thus, the shape of our dose–response relationship deviated from J-shaped or U-shaped.

23There were several potential explanations for why obese individuals may have higher risk of pneumonia. First, obesity is often accompanied by co-morbid conditions, such as gastroesophageal reflux disease. When reflux of gastric fluid occurs, the fluid can be aspirated into the respiratory tract resulting in pneumonia. Second, obesity has been shown to be an independent predictor of diabetes and asthma. These two diseases are also important risk factors for pneumonia. Third, a recent study reported that a higher BMI led to lower 25(OH)D3 levels. More recently, Aregbesola and colleagues found that subjects with a lower serum 25(OH)D3 concentration had a higher risk of pneumonia. Fourth, leptin (ob/ob) and leptin receptor (db/db) deficient mice showed severe immune abnormalities and greater susceptibility to viral and bacterial infection. Individuals with this leptin defect also exhibited greater susceptibility to respiratory infections. Therefore, leptin plays an important role in the human immune response to infectious disease. Although there were elevated leptin levels in obese subjects, leptin resistance often coexisted with these persons. Taken together, these results suggest that individuals with high BMI might have an increased risk of pneumonia compared to subjects with normal BMI.

24This present meta-analysis suggested a survival advantage for obese patients with pneumonia. A study by LaCroix et al. showed that the risk of pneumonia mortality was 2.6 times higher in men of the lowest BMI quartile compared to the highest quartile. Another study by Salive et al. indicated that the highest two quartiles of BMI had a significantly reduced risk of mortality compared with the lowest quartile. We propose three explanations for the inverse relationship between obesity and the risk of pneumonia mortality. First, obese individuals have a higher risk of developing coronary heart disease, type II diabetes, and heart failure. Thus, obese patients with pneumonia may receive optimal medical treatment or aggressive treatment. This may lead to a reduction in mortality. Second, tumor necrosis factor-alpha (TNF-α) is a potential proinflammatory cytokine that plays a critical role in inflammatory and immune responses. Puren et al. indicated that the plasma level of TNF-α is a marker of pneumonia severity. Adipose tissue is known to produce soluble TNF-α receptors. Additionally, recent studies showed that obese patients with pneumonia had lower pneumonia severity index scores and plasma levels of C-reactive protein. Third, as with other diseases, patients with pneumonia who are at a normal weight may not have enough metabolic reserve to counteract the increased catabolic stress. These patients may be particularly vulnerable to the adverse pathophysiologic consequences of a limited metabolic reserve.

25There are several limitations of the current study. First, a meta-analysis of observational studies inherits the limitation of the original studies. Although most studies adjusted for potential confounders, such as age, gender, smoking and underlying diseases, the possibility of residual confounding cannot be ruled out. Because this meta-analysis investigated only BMI, we cannot exclude the possibility that the observed associations may be confounded by other lifestyle factors, such as lower physical activity or dietary factors. Second, the number of available studies that were included in this meta-analysis was moderate. Therefore, the results could be influenced by some factors,such as random error. Third, most of the studies used International Classification of Diseases (ICD)-9 or ICD-10 codes to classify pneumonia. van de Garde and coworkers suggested that ICD-9 codes showed modest sensitivity for detecting CAP, leaving at least one quarter of pneumonia cases undetected. Fourth, statistical heterogeneity was detected when quantitative pooling was performed. In addition, small-study effects were detected in this study. Thus, caution with interpretation of the results is necessary, and these results should be confirmed by future studies.

26This meta-analysis suggests that obese individuals may be at higher risk for pneumonia, but they might have a lower mortality risk. Additional prospective studies with adjustment for more confounding factors are warranted before a conclusion can be drawn.

27BMI: body mass index; CAP: community-acquired pneumonia; CI: confidence interval; HAP: hospital-acquired pneumonia; ICD: International Classification of Diseases; NOS: Newcastle–Ottawa Scale; RR: risk ratio; TNF-α: tumor necrosis factor-alpha.

28The authors declare that they have no competing interests.

29WN searched the papers, created tables, and wrote the manuscript. YZ searched the papers, created tables, and wrote the manuscript. SHJ analyzed and provided data. KJJ analyzed and provided data. BL reviewed the study proposal and contributed to the manuscript writing. QX reviewed the study proposal and contributed to the manuscript writing. All authors read and approved the final manuscript.

30The pre-publication history for this paper can be accessed here:

31http://www.biomedcentral.com/1741-7015/12/61/prepub