Obesity Paradox and Lung Cancer Mortality: The Contributing Roles of Airflow Limitation and Pre-Chronic Obstructive Pulmonary Disease.

[RATIONALE] Increased body mass index (BMI; measured in kilograms per square meter) has been consistently associated with reduced mortality from lung cancer (LC), relative to low BMI, and termed the "obesity paradox." Although the basis of the obesity paradox remains unknown, mediating effects from sex, smoking status, diabetes mellitus (DM), and methodological issues (including bias) have been suggested causes.

[OBJECTIVE] Our aim was to examine whether respiratory comorbidity may contribute to this paradox. Methods: In this secondary analysis of 18,463 high-risk participants in the National Lung Screening Trial, we examined factors contributing to LC mortality (primary endpoint) using stratification analyses and regression models according to baseline demographics and comorbidity; specifically, respiratory-related comorbidity based on lung function and/or clinical history.

[FINDINGS] With increasing BMI, both respiratory and LC mortality decreased (P < 0.001), consistent with the obesity paradox. However, increasing BMI was associated with a linear decrease in the prevalence of airflow limitation (halving) and linear increases in both pre-COPD (twofold) and DM (eightfold) across BMI septiles (all Ps < 0.0001). In a sequentially constructed competing risk model for LC death, and after adjustment for smoking, age, sex, BMI, and other comorbidities, we found that airflow limitation, pre-COPD, and DM remained significant predictors of increased LC death (P < 0.01), albeit from opposite ends of the BMI continuum. When subjects with airflow limitation, pre-COPD, and DM were sequentially removed, the obesity paradox for LC mortality was substantially attenuated and almost abolished.

[INTERPRETATION] We propose that the obesity paradox in high-risk ever-smokers who develop lung cancer results, in large part, from the stronger deleterious effect of airflow limitation on LC mortality, with a lesser effect associated with DM-pre-COPD, where each predominate at opposite ends of the BMI continuum.