Current Prostate Cancer Genetic Risk Scores Add Limited Predictive Value for Black Men: a prospective cohort study in the UK Biobank and Our Future Health

Abstract Background Genetic risk scores could support risk-stratified prostate cancer screening, but their predictive performance may be limited in Black men, who are at greater risk and are underrepresented in genetic studies. We evaluated whether currently available genetic risk scores predict incident prostate cancer equitably in the UK, on fixed prediction horizons of 2-, 5-, and 10-years. Methods Male participants with imputed genetic data were included from the UK Biobank and Our Future Health cohorts. Logistic regression models combining genetic risk scores with age and/or family history were used to predict diagnosis within 2, 5, and 10 years of either the assessment centre visit or participants’ 40th, 50th, or 60th birthday. Performance was evaluated using the area under the receiver operating characteristic curve, odds ratios per standard deviation increase in genetic risk score, risk stratification by model decile, and net benefit analysis. Primary analyses compared Black and White men based on self-reported ethnicity. Results Adding the 269-variant multiethnic score to age in an integrated risk model improved discrimination in White men across all prediction horizons, increasing AUC by 8.69–11.23%. In Black men, improvement was significant only over 10 years, increasing the AUC by 1.83%. Age alone was a stronger discriminative predictor in Black men, so models combining age and genetic risk had similar AUC between groups. For the best-performing GRS, 10-year odds ratios per standard deviation were 2.18 [2.13–2.23] in White men and 1.62 [1.40–1.88] in Black men. These GRS showed comparable case/control discrimination and association with diagnosis in age-specific screening scenarios in the Our Future Health cohort. Net benefit analysis revealed that the useful range of absolute risk thresholds is wider, and that optimal sensitivity and specificity is typically achieved at higher thresholds, in Black men. Conclusions Models combining age and genetic risk can achieve similar overall discrimination in Black and White men, but current GRS contribute little beyond age alone in Black men. Absolute risk thresholds may need to be higher in Black men to prevent overdiagnosis. Greater representation of underrepresented groups in Genome-Wide Association Studies is needed to improve GRS performance for Black men.