Abstract for: Risk Factors and Predictive Modeling of Total and Aggressive Prostate Cancer in African Men

Prostate cancer (PCa) is a major health burden among men in Sub-Saharan Africa, where mortality rates are among the highest globally and patients often present with advanced disease. PSA screening remains limited, and information on key risk factors, including age, family history of PCa, and polygenic risk scores, may help identify high-risk men. Using harmonized MADCaP consortium data, this study examined demographic and clinical risk factors for overall and high-grade PCa. Epidemiologic, clinical, and family history data from 2,505 prostate cancer (PCa) cases and 2,220 age-matched controls enrolled at seven centers across West and South Africa were analyzed. Polygenic risk scores were derived from germline SNP data. Logistic and polytomous logistic regression models evaluated risk factors for overall and high-grade PCa. Machine learning models were developed to predict high-grade disease and compared their performance with traditional regression approaches. A first-degree family history of PCa (OR=3.24; 95% CI: 2.30–4.55) and higher polygenic risk scores (OR=1.90; 95% CI: 1.71–2.12) were independently associated with increased PCa risk. However, neither factor was differentially associated with high-grade versus low-grade PCa. Machine learning models identified younger age at diagnosis (<60 years), serum PSA >20 ng/mL, and severe symptom burden as predictors of aggressive disease, whereas higher education and comorbidity burden were associated with low-grade PCa. Findings demonstrate that age, family history, and polygenic risk scores are strong predictors of PCa risk among African men and may help guide risk-stratified screening strategies. In contrast, elevated PSA levels and severe symptom burden were associated with aggressive disease. Leveraging the largest multicenter MADCaP cohort in Sub-Saharan Africa, this study provides important evidence to inform early detection efforts, although findings should be interpreted cautiously given the potential for observational bias.