Abstract 3506: African-specific improvement of a polygenic hazard score to predict age of onset of prostate cancer

Abstract

Abstract Introduction: Polygenic hazard models can be used in predicting age of onset of prostate cancer. A recently published model, PHS46, showed reduced performance in men with African genetic ancestry - the hazard ratio for aggressive prostate cancer in individuals with high values of PHS46 was approximately half as large for those with African genetic ancestry (2.4) as it was for those with European (5.6) or Asian (5.2) ancestry. We used a cross-validated search to identify SNPs that would add statistical value to a base model of PHS46 among African men and help bridge the performance gap in this population. Material and Methods: Anonymized genotypic data were obtained from 6,253 men with African genetic ancestry from the PRACTICAL consortium. Ten iterations of a ten-fold cross-validation search were conducted, where SNPs were added greedily and stepwise to the PHS46 Cox proportional hazards model using age of onset of any prostate cancer as the dependent variable. Individuals without prostate cancer were censored at age of last follow-up. SNPs were selected for the final, enhanced-PHS46 model if they were selected in more than 50% of the folds. The predictive performances of PHS46 and enhanced-PHS46 were compared using the same cross-validated approach. Results: Three SNPs (rs76229939, rs74421890, and rs5013678) were selected for inclusion in enhanced-PHS46 model. All three SNPs are located on chromosome 8q.24. Enhanced-PHS46 model showed substantial improvements in all performance metrics measured, including a 20% reduction in the relative hazard of those in the bottom 20% compared to the middle 40% (0.65 to 0.53), a 75% increase in the relative hazard of those in the upper 20% compared to the bottom 20% (2.47 to 4.34) and a 95% increase in the relative hazard of those in the upper 2% compared to the middle 40% (2.1 to 4.1). Similar improvements were demonstrated using the age of onset of aggressive prostate cancer. Conclusions: We identified three SNPs (rs76229939, rs74421890, and rs5013678) that substantially improved the performance of PHS46 in a dataset of men with African genetic ancestry. A strategy of building on established models might be applied to other groups generally under-represented in genome-wide association studies. Citation Format: Roshan Karunamuni, Minh-Phuong C. Huynh-Le, Chun C. Fan, Ian G. Mills, Ole A. Andreassen, Anders M. Dale, Tyler M. Seibert, The PRACTICAL Consortium. African-specific improvement of a polygenic hazard score to predict age of onset of prostate cancer [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 3506.