Polygenic score for body mass index in relation to mortality among patients with renal cell cancer

The association between body mass index (BMI) and mortality among individuals with renal cell cancer (RCC) is debated, with some observational studies suggesting a lower mortality associated with higher BMI. However, methodological issues such as confounding and reverse causation may bias these findings. Using BMI-associated genetic variants can avoid these biases and generate more valid estimates. In this prospective cohort study, we included 1264 RCC patients (446 deaths) from the UK Biobank. We created a BMI polygenic score (PGS) based on 336 BMI-associated genetic variants. The association between the PGS and mortality (all-cause and RCC-specific) was evaluated by logistic regression (all RCC cases) and Cox regression (906 incident cases). For comparison, the associations of measured pre-diagnostic BMI and waist-to-hip ratio (WHR) with mortality were quantified by Cox regression among incident cases. We stratified these analyses by time between anthropometric measurement and RCC diagnosis to assess the influence of reverse causation. We did not observe an association between the BMI PGS and all-cause mortality among RCC patients (hazard ratio (HR) per SD increase = 0.98, 95% CI: 0.88,1.10). No association was found for pre-diagnostic BMI (HR per 5 kg/m2 increase = 0.93, 95% CI: 0.83,1.04) or WHR (HR per 0.1 increase = 0.97, 95% CI: 0.83,1.13) with mortality. In patients with anthropometrics measured within 2 years before RCC diagnosis, we observed associations of higher BMI (HR per 5 kg/m2 = 0.76, 95% CI: 0.59,0.98) and WHR (HR = 0.67 per 0.1 increase, 95% CI: 0.45,0.98) with a lower risk of death. Similar patterns were observed for RCC-specific mortality. We found no association between either genetic variants for high BMI or measured pre-diagnostic body adiposity and mortality among RCC patients, and our results suggested a role for reverse causation in the association of obesity with lower mortality. Future studies should be designed carefully to produce unbiased estimates that account for confounding and reverse causation.