Date of Award

Spring 2025

Document Type

Open Access Dissertation

Department

Epidemiology and Biostatistics

First Advisor

Anthony J Alberg

Abstract

Background: Lung cancer is the leading cause of cancer-related deaths in the U.S. Simultaneously, the U.S. is grappling with a growing obesity epidemic, a major risk factor for several cancers and metabolic disorders, including diabetes mellitus, which is associated with an increased risk of cancer in various organs. However, research on the relationship between body mass index (BMI) and lung cancer has generated conflicting results. Studies using measured BMI (either objectively measured or self-reported) suggest that lower BMI is inversely associated with lung cancer risk. In contrast some of the more recent Mendelian randomization (MR) studies, which use genetically proxied BMI, suggest increased BMI may be associated with increased lung cancer risk. However, MR studies face uncertainties due to potential pleiotropic effects, and lack of strong correlation with measured BMI. No previous systematic reviews and meta-analyses on this topic have included MR studies. Further, due to lingering concerns that uncontrolled confounding may explain the inverse association observed between measured BMI and lung cancer makes it important to investigate the association between BMI and lung cancer using marginal structural models (MSMs) and causal mediation analysis, because these methods specifically address this issue. Aims: The aims of this study were: (1) to conduct a systematic review and meta-analysis on the association between BMI and lung cancer risk; (2) to determine the association between BMI and lung cancer risk analyzing data from a prospective cohort study using causal inference methods (MSMs) and compare with the results generated analyzing the data using the more routinely used Cox-proportional Hazards (Cox-PH) model; and (3) in the same cohort study to examine the direct and indirect effects of smoking on lung cancer, with BMI as a mediator, and the direct and indirect effects of BMI on lung cancer, with diabetes mellitus as a mediator. Methods: For Aim 1, a systematic review and meta-analysis synthesized the available evidence on the association between BMI and lung cancer risk, including studies that used measured/self-reported BMI and studies that used genetically proxied BMI (MR studies). We searched PUBMED, CINAHL, Embase, and Web of Science for cohort and nested case-control studies published until August 31, 2024. Two independent reviewers screened articles for eligibility, conducted data extraction and quality assessments for included papers, and resolved discrepancies through discussion. Meta-analyses using random-effects models estimated pooled relative risks (RRs) and odds ratios (ORs) with 95% confidence intervals (CIs). Dose-response analyses evaluated the relationship between increasing BMI and lung cancer risk, and E-values were calculated to assess the potential influence of unmeasured confounding. Aim 2 was a prospective cohort study using data from the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial. The study population was comprised of 120,863 participants aged 55-74 years, who were recruited and followed-up from1993-2009. The data were analyzed to examine the association between BMI and lung cancer risk using causal inference methods (MSMs) and compared with results from Cox-PH models. BMI was measured at baseline and earlier life stages (ages 20 and 50), and BMI trajectories were also evaluated. Participants' self-reported smoking history, demographic factors, family history of lung cancer, and lifestyle behaviors were collected. Both MSMs and Cox-PH models with inverse probability of treatment and censoring weights accounted for time-varying confounders. Covariates included sex, education, race, occupation, comorbidities, cigarette smoking, dietary intake, and physical activity. Sensitivity analyses excluded participants with short follow-ups and included diabetes as an additional covariate. Aim 3 was also a prospective cohort study using data from the PLCO Cancer Screening Trial. We performed causal mediation analysis to investigate the direct and indirect effects of smoking on lung cancer with BMI as a mediator, and the effects of BMI on lung cancer mediated by diabetes mellitus. Analyses was focused on baseline smoking status, with BMI and diabetes mellitus as mediators. Lung cancer incidence was the primary outcome. Cox-PH models were used to calculate hazard ratios (HRs) with 95% CIs for direct and indirect effects. The CAUSALMED procedure in SAS (version 9.4) estimated natural direct and indirect effects, adjusting for potential confounders such as sex, education, occupation, family history of lung cancer, second-hand smoke exposure, and physical activity. Results: Aim 1: A total of 45 studies (40 studies that measured BMI or used self-reported BMI and 5 MR studies) met the inclusion criteria. Meta-analysis of studies with measured/self-reported BMI (n=40) indicated that each 1 kg/m² increase in BMI was associated with a 4% reduction in lung cancer risk (RR: 0.96; 95% CI: 0.95–0.97). Higher BMI (overweight and obesity) was inversely associated with lung cancer risk (RR: 0.79; 95% CI: 0.74–0.83), while underweight individuals had a significantly higher risk (RR: 1.43; 95% CI: 1.25–1.64). Overall dose-response analyses revealed a non-linear inverse association between BMI and lung cancer. E-values suggested that strong unmeasured confounding would be required to nullify the observed associations between measured/self-reported BMI and lung cancer. MR studies (n=5) showed a non-significant increase in lung cancer risk with each 1 kg/m² increase in genetically predicted BMI (OR: 1.05; 95% CI: 0.99–1.12). Aim 2: Baseline BMI was inversely associated with lung cancer risk across both MSMs and Cox-PH models. Among the never smokers, overweight and obese individuals had lower lung cancer risk compared to normal-weight never smokers. Among the current and former smokers, within all the smoking strata lung cancer risk decreased with increasing BMI, with the MSM model showing consistently stronger inverse HRs compared to the Cox-PH model. While the Cox-PH model suggested uncertainty with wide confidence intervals, the MSM highlighted a clearer trend with greater statistical precision, particularly among current smokers, with lung cancer risk decreasing as BMI increased. These findings remained consistent when considering BMI at ages 20 and 50. Sensitivity analyses excluding the first year and the first five years of follow-up, as well as including diabetes as a covariate, did not significantly change the results. Aim 3: Among 142,245 participants, current smokers with more than 45 years of smoking history had a 32-fold higher lung cancer risk compared to never smokers (HR: 32.08; 95% CI: 27.42–36.74). The controlled direct effect was slightly lower (HR: 30.30; 95% CI: 25.84–34.75). The natural indirect effect of BMI was positive among current smokers, particularly those with longer smoking histories, indicating a modest increase in lung cancer risk mediated by BMI. In contrast, former smokers had an inverse or null indirect effect. For BMI and lung cancer, an inverse association was observed, with each 1 kg/m² increase in BMI linked to a 2% decrease in lung cancer risk (HR: 0.98; 95% CI: 0.96–0.99). The mediation effect of diabetes mellitus on this association was not statistically significant (HR: 1.000; 95% CI: 0.999–1.003). Conclusion: This research program offers insights into the controversial relationship between BMI and lung cancer risk. Aim 1, the systematic review and meta-analysis, provides strong, consistent evidence from studies using measured or self-reported BMI that higher BMI is inversely associated with lung cancer risk. In contrast, MR studies did not show a significant association between BMI and lung cancer, though the summary OR was in the direction of increased risk. Aim 2, an analysis within the PLCO cohort using MSMs, further confirmed an inverse association between BMI and lung cancer risk. In Aim 3, causal mediation analysis demonstrated that BMI positively mediates the relationship between smoking and lung cancer risk in current smokers, while it shows an inverse or null mediation effect in former smokers. Diabetes mellitus does not significantly mediate the BMI-lung cancer association. The findings of this dissertation project indicate a potential causal inverse relationship between BMI and lung cancer, supported by temporality, strength of association (including dose-response relationship), consistency, and biological plausibility. This relationship is evident across subgroups (i.e.: gender, histological subtypes and BMI measurement technique) and smoking statuses, with biological mechanisms like reduced carcinogen-DNA adducts and oxidative DNA damage providing further support, though additional research is needed. Future studies should incorporate other indices of body fat, such as central obesity measures (e.g., waist-to-hip ratio, waist circumference), to better understand risk factors and biological mechanisms. Further research employing sophisticated body composition tools, such as imaging technologies, alongside molecular techniques like metabolomics, could provide deeper insights into the biological mechanisms at play.

Rights

© 2025, Rajat Das Gupta

Download

Available for download on Monday, May 31, 2027

Included in

Epidemiology Commons

Share

COinS