Skip to Main Content
Principal Investigator
Rajat Das Gupta
M.B.B.S., M.P.H.
University of South Carolina
Position Title
Graduate Research Assistant
About this CDAS Project
PLCO (Learn more about this study)
Project ID
Initial CDAS Request Approval
Aug 2, 2022
Clarifying the relationship between body mass index and lung cancer using causal inference methods
Overweight and obesity are well established risk factors for several cancers (i.e.: cancers of the uterus, gallbladder, kidney, cervix, thyroid, and leukemia). However, the potential association between body mass index (BMI) and lung cancer has yielded divergent and complex findings. The results of studies using traditional epidemiological methods have consistently yielded results showing an inverse association between BMI and the risk of lung cancer, a finding consistently observed for the major histologic subtypes of lung cancer. In contrast, the evidence from studies that have applied Mendelian randomization methods have instead tended to show that the risk of lung cancer increases as BMI increases. To advance understanding of the association between BMI and lung cancer, we propose to use inverse probability treatment weight (IPTW) of marginal structural models (MSMs) using the PLCO data and compare these findings with the results of a traditional multivariable adjusted Cox proportional hazards model. The primary independent variable is baseline BMI. At study baseline, height and weight of the participants were collected. BMI will be calculated as the weight measured in kilogram (kg) divided by the square of the height in meter squared (m2). BMI at 20 years and 50 years of age will be included in ancillary analyses. The following variables will be considered as covariates: age, gender, race, smoking history, family history of cancer, alcohol use, dietary intake/fruits and vegetables intake, occupation, exercise, work activities, and randomization arm (intervention/control). Hazards ratios from Cox proportional hazards models will be used to estimate the association between BMI and lung cancer. The time varying variables will be adjusted during the analyses. In a separate aim, we propose a causal mediation approach to estimate the total, direct and indirect effects of smoking on lung cancer mediated through BMI using the counterfactual framework. The findings of this study hold promise to add insight into the role of BMI on the risk of lung cancer.

The overall goal of this study is to advance understanding of the role of BMI on the risk of developing lung cancer as well as to understand the role of BMI as a mediator in the causal pathway of smoking and lung cancer. We propose a cohort study using PLCO data to include all the eligible participants who enrolled in the study at baseline (between November 1993 and July 2001), except for those with a prior history of lung cancer and were diagnosed of other cancers. The specific aims are to: 1. Determine the association between BMI and lung cancer risk using causal inference methods (marginal structural model) among the PLCO cohort and compare it with findings using traditional epidemiological methods (Cox proportional hazards model). 2. Determine the direct and indirect effect of smoking on lung cancer considering BMI as a mediator.


Dr. Anthony Alberg, Department of Epidemiology and Biostatistics, Arnold School of Public Health University of South Carolina