Validation and context-dependent effects of a prostate cancer polygenic risk score in the All of Us Research Program

Abstract

Polygenic risk scores (PRSs) have demonstrated strong potential for improving prostate cancer risk stratification. However, it is unknown whether the clinical utility of prostate cancer PRS varies by demographic, lifestyle, and socioeconomic factors. We validated a previously developed multi-ancestry PRS of 451 prostate cancer risk variants and evaluated context-dependent effects using genetic and clinical data from the diverse All of Us Research Program, including 7,577 indivisuals with prostate cancer and 90,608 control individuals across six genetic ancestry groups. In ancestry-stratified testing, the PRS showed strong associations with prostate cancer risk, with odds ratios (ORs) per standard deviation (SD) increase ranging from 1.61 (95% confidence interval [CI] = 1.02-2.64, p = 0.05) in Middle Eastern to 2.19 (95% CI = 1.98-2.42, p = 2.2 × 10^-51) in American populations. Age-stratified analyses showed reduced PRS effects with increasing age. Across modifiable lifestyle and healthcare access factors, PRS effects were larger in those with a higher body mass index (OR = 2.15 vs. 1.96 in individuals with obesity and normal weight, respectively, p = 0.03), in never or former smokers vs. current smokers (OR = 2.06, 2.37, and 1.93, respectively, p = 0.06), and in those recently accessing healthcare (OR = 2.21 vs. 1.88, p = 0.05), highlighting important context-specific modifiers. We did not observe context-dependent effects of other socioeconomic factors, such as income, education, and insurance. In a phenome-wide association study (PheWAS), the PRS was associated with 14 clinical outcomes, including known prostate cancer-related conditions. These findings confirm the predictive strength of the multi-ancestry prostate cancer PRS across diverse populations and underscore the importance of accounting for demographic-, lifestyle-, and healthcare-related contexts when applying PRSs in clinical and public health settings.

Keywords: All of Us; absolute risk; context-dependent effects; genetics; large-scale biobanks; multi-ancestry; phenome-wide association study; polygenic risk score; prostate cancer; risk modeling.