Obesity and incidence of lung cancer: A meta-analysis
Abstract
To date, the relationship between obesity and the incidence of lung cancer remains unclear and inconclusive. Thus, we conducted a meta-analysis of published studies to provide a quantitative evaluation of this association. Relevant studies were identified through PubMed and EMBASE databases from 1966 to December 2011, as well as through the reference lists of retrieved articles. A total of 31 articles were included in this meta-analysis. Overall, excess body weight (body mass index, BMI ≥ 25 kg/m2) was inversely associated with lung cancer incidence (relative risk, RR = 0.79; 95% confidence interval, CI: 0.73–0.85) compared with normal weight (BMI = 18.5-24.9 kg/m2). The association did not change with stratification by sex, study population, study design, and BMI measurement method. However, when stratified by smoking status, the inverse association between excess body weight and lung cancer incidence in current (RR = 0.63, 95% CI: 0.57–0.70) and former (RR = 0.73, 95% CI: 0.58–0.91) smokers was strengthened. In non-smokers, the association was also statistically significant (RR = 0.83, 95% CI: 0.70–0.98), although the link was weakened to some extent. The stratified analyses also showed that excess body weight was inversely associated with squamous cell carcinoma (RR = 0.68, 95% CI: 0.58–0.80) and adenocarcinoma (RR = 0.79, 95% CI: 0.65–0.96). No statistically significant link was found between excess body weight and small cell carcinoma (RR = 0.99, 95% CI: 0.66–1.48). The results of this meta-analysis indicate that overweight and obesity are protective factors against lung cancer, especially in current and former smokers.
Abstract
What's new?
Body mass index (BMI) has been associated with lung cancer risk, though no clear relationship has emerged in the scientific literature. In this meta-analysis, which emphasized control for confounding factors, obesity was associated with a decreased risk for lung cancer, particularly for smokers. Although the biological mechanism underlying this phenomenon is unclear, the results suggest that attention to nutrition and maintenance of body weight could lower lung cancer incidence among smokers.
Lung cancer is one of the most common cancers worldwide in terms of both incidence and mortality.1 Cigarette smoking is the main cause of most lung cancers, but other risk factors such as exposure to asbestos, radon gas, environmental factors, and low dietary consumption of fruits and vegetables have also been identified.2
A number of epidemiological studies also indicated an association between body mass index (BMI) and lung cancer risk. Obesity is a strong risk factor in cancers of the colon, breast (in postmenopausal women), endometrium, kidney (renal cell), stomach, pancreas, gallbladder, liver, and possibly other cancers.3 However, lower lung cancer risk was mostly observed in individuals with higher BMI. Thus, obesity may be a preventive factor for the development of lung cancer.4-13 However, the results are inconclusive and conflicting. Other studies have failed to find such an association,14, 15 or the association disappeared when smoking habits and health status were taken into account.16, 17 Still, other studies reported a positive association between BMI and lung cancer incidence.18, 19 Moreover, the interpretation of the association between low BMI and lung cancer is difficult because smoking is the main cause of lung cancer and is inversely associated with BMI.20 Some studies attributed the inverse association between BMI and lung cancer to the incomplete control of confounding by smoking, by preclinical weight loss among individuals who later developed lung cancer, by competing causes of death, or by a combination of these.6, 14, 16, 19, 21
To clarify the association between BMI and the risk of developing lung cancer, we therefore carried out a meta-analysis of published cohort and case–control studies with an emphasis on the creation of more standardized exposure definitions for better control of the confounding factors.
Material and Methods
Study selection
A literature search was conducted in the PubMed and EMBASE databases for relevant studies published from 1966 to December 2011. No language restriction was applied. The keywords “body mass index,” “BMI,” “overweight,” “obesity,” “leanness,” “body weight”, “body size,” “anthropometric”, and “anthropometry” in combination with “lung cancer,” “lung carcinoma,” or “lung neoplasm” were used. The reference lists of retrieved articles were also manually reviewed to identify additional relevant studies.
Study selection criteria
A published article was included according to the following criteria: (1) cohort or case–control study in which lung cancer incidence was an outcome; (2) clear description of overweight or obesity as defined by a BMI in kg/m2; and (3) estimates of relative risk (hazard ratio, rate ratio, odds ratio, or standardized incidence ratio) with their corresponding 95% confidence intervals (CI) were reported.
Data extraction
The following information was extracted from each study: first author's name, publication date, location where the study was conducted, study design, duration of follow-up (cohort studies) or data collection (case–control studies), sample size, sex and age range of participants, BMI categories(lowest–highest), method of assessment of weight and height, lung cancer diagnosis method, and point estimates [relative risk (RR), odds ratio (OR), or standardized incidence ratio (SIR)] and corresponding 95% CI. For each study, we extracted the risk estimates that were adjusted for the greatest number of potential confounders.
Statistical analysis
Summary relative risk estimates were calculated for two categories of BMI (BMI = weight in kilograms/height in meters2) as defined by the World Health Organization (WHO) for adults: overweight (BMI = 25–29.9 kg/m2) and obesity (BMI ≥ 30 kg/m2) compared with normal weight (BMI = 18.5–24.9 kg/m2) as the reference category. We also created an “excess body weight” category that included both overweight and obese individuals (BMI ≥ 25 kg/m2). When non-standard BMI categories were used, we selected the category that was most similar to those defined by the WHO. The DerSimonian and Laird random effects model was employed to combine the study-specific relative risks.22
We examined heterogeneity among studies using the Q and I2 statistics.23 For the Q statistic, statistical significance was set at P < 0.1. When substantial heterogeneity was detected, the random effects model was used.22 Otherwise, the pooled estimate was based on the fixed effects model. We also conducted a sensitivity analysis in which one study was removed and the rest were analyzed to assess whether the results were affected statistically significantly. Publication bias was evaluated using funnel plots and the Egger test. P < 0.1 was considered to indicate statistically significant publication bias.24 In the presence of publication bias, we used the “trim and fill” method to correct such bias.25 Subgroup analyses were carried out based on the study design (cohort and case–control studies), gender (male and female), smoking status (current smokers, former smokers, and non-smokers), study population (non-Asians and Asians), body size assessment (measured and self-reported), and lung cancer histology (small cell carcinoma, squamous cell carcinoma, and adenocarcinoma). All statistical analyses were performed using STATA12.0 (StataCorp, College Station, TX, USA).
Abbreviations
BMI: body mass index; CI: confidence interval; F: fixed model; OR: odds ratio; R: random model; RR: relative risk; SIR: standardized incidence ratio
Results
Study characteristics
A total of 3445 papers relevant to the search words were identified. After screening titles and abstracts and reviewing the full-text articles, 20 cohort studies and 11 case–control studies (7 using population-based controls and 4 using hospital-based controls) were obtained (Fig. 1). Of these studies, 11 were conducted in Europe, 9 in the United States, 8 in Asia, 2 in Canada, and 1 in Israel. The main characteristics of the studies included in the meta-analysis are summarized in Table 1.
Flowchart of study selection.
 |
Overall analyses
As shown in Figure 2, the overall analysis of all studies revealed a statistically significant inverse association between BMI and lung cancer risk (overweight: RR = 0.74, 95% CI: 0.68–0.80; obesity: RR = 0.71, 95% CI: 0.62–0.80; excess weight: RR = 0.79, 95% CI: 0.73–0.85) compared to normal weight (BMI = 18.5–24.9 kg/m2). After stratifying by study design, a statistically significant inverse link between BMI and lung cancer risk was observed for the cohort studies (overweight: RR = 0.78, 95% CI: 0.72–0.84; obesity: RR = 0.80, 95% CI: 0.73–0.88; excess weight: RR = 0.78, 95% CI: 0.72–0.84). For the case–control studies, a statistically significant inverse link was observed (overweight: RR = 0.68, 95% CI: 0.57–0.82; obesity: RR = 0.56, 95% CI: 0.40–0.79; excess weight: RR = 0.65, 95% CI: 0.52–0.81). In population-based case–control studies, the RRs of lung cancer for the “overweight,” “obesity,” and “excess weight” groups were 0.75 (95% CI: 0.62–0.90), 0.70 (95% CI: 0.46–1.06), and 0.72 (95% CI: 0.57–0.91), respectively. In hospital-based case–control studies, the RRs of lung cancer for “overweight,” “obesity,” and “excess weight” groups were 0.46 (95% CI: 0.35–0.62), 0.41 (95% CI: 0.22–0.75), and 0.52 (95% CI: 0.32–0.85), respectively. No sex-based differences were observed in the association between BMI and lung cancer risk. When stratified by race, the association between BMI and the risk of lung cancer was similar for both non-Asians and Asians (Table 2).
Forest plot for the association between excess weight and lung cancer risk in the general population. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
 |
Statistically significant heterogeneity was observed in all the study results (I2 > 50%, P < 0.001) except for some subgroup analyses (Table 2). No indication of publication bias was observed in the literature on BMI and lung cancer risk in “overweight” and “obesity” groups either based on the Egger's test (P = 0.20, 0.35) or Begg's test (P = 0.26, 0.69) results (Fig. 3). For BMI and lung cancer risk in the “excess weight” group, the funnel plot showed some asymmetry, indicating some evidence of bias. However, when the “trim and fill” approach was performed, data were unchanged, suggesting that the effect of publication bias could be negligible.
Funnel plot for all studies included in the meta-analysis of obesity and lung cancer risk. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
Smoking status
Smoking status is potentially the most likely confounder of the inverse relationship between BMI and lung cancer risk. After stratifying by smoking status, we found that the inverse association was strengthened in both current and former smokers compared with the overall meta-analysis. The pooled RRs of lung cancer for “overweight,” “obesity,” and “excess weight” groups were, respectively, 0.69 (95% CI: 0.62–0.77), 0.68 (95% CI: 0.58–0.79), and 0.63 (95% CI: 0.57–0.70) for current smokers and 0.72 (95% CI: 0.55–0.95), 0.71 (95% CI: 0.54–0.92), and 0.65 (95% CI: 0.43–0.97) for former smokers. When we restricted the meta-analysis to the studies that focused on non-smokers, the pooled RRs of lung cancer for “overweight,” “obesity,” and “excess weight” groups were 0.89 (95% CI: 0.70–1.12), 0.81 (95% CI: 0.64–1.01), and 0.83 (95% CI: 0.70–0.98), respectively. Such inverse association also existed in non-smokers, but the statistical significance disappeared in the “overweight” and “obesity” groups. When we considered all the studies, the inverse association became statistically significant again. Further stratification by sex revealed that the inverse association existed in female non-smokers (Fig. 4), whereas it disappeared in male non-smokers (Table 2).
Forest plot for the association between excess weight and lung cancer risk among female non-smokers. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
Histological subtypes
When stratified by histological type, the pooled RRs of squamous cell carcinoma for “excess weight,” “overweight,” and “obesity” groups were 0.68 (95% CI: 0.58–0.80), 0.63 (95% CI: 0.42–0.94), 0.34 (95% CI: 0.12–0.91), respectively. We also found an inverse association with BMI for adenocarcinoma, but without statistical significance in the “obesity” group (Table 2). For small cell carcinoma, the pooled RRs for “excess weight,” “overweight,” and “obesity” were, respectively, 0.99 (95% CI: 0.66–1.48), 1.06 (95% CI: 0.78–1.43), 0.95 (95% CI: 0.39–2.31), showing no association with BMI. Given the small number of studies obtained, we could not evaluate the publication bias in these subgroup analyses.
Sensitivity analyses
In the sensitivity analyses in which one study was removed and the rest were analyzed, the pooled RRs were similar with the overall pooled RRs (data not shown), supporting the robustness of our results.
Discussion
To our knowledge, this study is the first meta-analysis on the association between BMI and the incidence of lung cancer. An inverse association between BMI and lung cancer incidence was established in the general population, and the association was strengthened among current and former smokers. Thus, smokers with low BMI are more susceptible to lung cancer. Inconsistent results were obtained in studies that analyzed the association of BMI and lung cancer incidence in non-smokers. In our research, the results of the meta-analysis supported the inverse association between BMI and lung cancer incidence.
The potential biological mechanism under this phenomenon is that leanness may be involved in the carcinogenic progress of smoking. Some studies have shown that BMI was inversely correlated with the level of urinary 8-hydroxydeoxyguanosine, which serves as an indicator of oxidative DNA damage in smokers, suggesting that BMI may serve as an independent factor for host susceptibility to smoking-related cancers.43, 44 If BMI is truly associated with lung cancer risk, the next step would be to explore the biological basis of the association. In fact, one recent study reported that one allele of the fat mass and obesity-associated (FTO) gene, which has been linked with increased BMI, was associated with a decreased risk of lung cancer. This finding seems to support the hypothesis that BMI is inversely associated with lung cancer risk and independent of smoking and weight loss because of preclinical disease.42 However, more studies related to BMI and lung cancer risk should be conducted with a primary focus on non-smokers, especially in males, to clarify such issue.
In a subgroup analysis stratified based on lung cancer histology, we found that excess body weight was inversely associated with the risk of squamous cell carcinoma and adenocarcinoma, whereas no association was observed between BMI and small cell carcinoma. Compared with adenocarcinoma, small cell carcinoma is more strongly associated with smoking.45 Thus, this question cannot be explained by smoking. Unlike non-small cell lung cancer cells, small cell lung cancer cells originate from neuroendocrine cells in the bronchus and can secrete ectopic hormones and peptides such as adrenocorticotropic hormone and gastrin-releasing peptide.46 These characteristics may determine that small cell lung cancer has no association with BMI.
Our study presents a number of advantages. First, it included a large sample size (26,066 lung cancer cases and 79,915,395 participants) and a broad time span (from 1963 to 2006). Second, more comparable BMI categories were created for each study, and subgroup analyses stratified by smoking status, sex, study population, study design, and BMI measurement method were conducted. Thus, the effect of potential confounders was minimized.
We also acknowledged several potential limitations in this meta-analysis. First, all the studies included in the meta-analysis have been adjusted for smoking and some other confounders. Since the variation in duration of smoking and degrees of inhalation always makes statistical adjustment inadequate, and smoking status is strongly related to both BMI and lung cancer risk,47 inadequate adjustment for smoking may result in a spurious association between lung cancer risk and BMI.6, 14 Unmeasured confounders such as physical activity, dietary constitution, and medical history may also exert some effects on the results. Second, the association between BMI and lung cancer risk may be caused by preclinical weight loss or competing causes of death.14 However, in some studies,6, 7, 16 the exclusion of cases presenting poor health in the first several years of follow-up and the selection of BMI measured when participants were healthy had a slight effect on the inverse association. Third, several sources of bias should be considered in the design and conduct of the individual studies, as well as in the synthesis of the studies, for instance, the inherited limitation of case–control studies. Actually, a lower pooled relative risk was observed more frequently in the case–control studies than in the cohort studies (Table 2). Some evidence of publication bias was found in this analysis because related studies were identified from limited databases and studies with null results tend to go unpublished. The presence of publication bias could lead to an inaccurate estimation of the true association between obesity and lung cancer risk. However, the results remained unchanged after using the “trim and fill” method, suggesting that the publication bias in this meta-analysis could be negligible. In addition, some studies in the meta-analysis adopted self-reported body size to evaluate BMI. The self-reported method has always been viewed to cause small systematic errors from the overestimation of height and underestimation of weight.48 When we stratified by the body size assessment method, some differences between the two methods were observed. However, the inverse association between BMI and lung cancer risk was not changed.
Conclusion
In conclusion, obesity is significantly inversely associated with lung cancer incidence in the general population of both sexes. Moreover, leanness is more strongly associated with lung cancer risk among current and former smokers. An inverse relationship may also exist between BMI and lung cancer incidence in non-smokers, but this finding will need further investigation. Our findings on the inverse association between BMI and the risk of developing lung cancer in smokers suggest that smokers should improve their nutritional status and maintain a suitable body weight.
Acknowledgements
This work was supported by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD).
