A comparison of analytic approaches for investigating the obesity paradox in kidney cancer

Body mass index (BMI) and kidney cancer mortality are inconsistently associated in the scientific literature. To understand how study design affects results, we contrasted associations between pre-diagnosis BMI and mortality under different analytic scenarios in a large, population-based prospective cohort study.

Using data from the NIH-AARP Diet and Health Study (1995-2011), we constructed two cohorts: a "full at-risk" cohort with no kidney cancer history at baseline (n = 252,845) and an "incident cancer" subset who developed kidney cancer during follow-up (n = 1,652). Cox Proportional Hazards models estimated hazard ratios (HR) and 95% confidence intervals (CI) between pre-diagnosis BMI and mortality for different outcomes (all-cause and cancer-specific mortality), in the different cohorts (full at-risk vs. incident cancer cohort), and with different covariates (minimally vs. fully adjusted). For the incident cancer cohort, we also examined time to mortality using different timescales: from enrollment or diagnosis.

In the full at-risk study population, higher pre-diagnosis BMI was associated with greater cancer-specific mortality in fully adjusted multivariable models, particularly for obese participants [HR, (95% CI): 1.76, (1.38-2.25)]. This association was less pronounced in the incident cancer cohort [1.50, (1.09-2.07)]. BMI was not strongly associated with all-cause mortality in either cohort in fully adjusted models [full cohort: 1.03, (1.01, 1.06); incident cancer cohort: 1.20, (0.97, 1.48)].

Populations characterized by high adult BMI will likely experience greater population burdens of mortality from kidney cancer, partially because of higher rates of kidney cancer diagnosis. Questions regarding overall mortality burden and post-diagnosis cancer survivorship are distinct and require different study designs.