Abstract PD14-06: PD14-06 Polygenic risk score added to risk calculator improves prediction of breast cancer in U.S. Black women

Abstract

Abstract We recently derived an absolute breast cancer risk prediction model, the Black Women’s Health Study (BWHS) model, for breast cancer in U.S. Black women using data from three large case-control studies and validated it in independent prospective data from the Black Women’s Health Study (Palmer 2021). The BWHS model includes epidemiologic risk factors as well as family history of breast cancer and family history of prostate cancer. It does not include genetic variants because at the time of model development breast cancer polygenic risk scores performed poorly in women of predominantly African ancestry, primarily due to differences in allele frequency and linkage disequilibrium. More recently, Gao et al. (2022) developed and tested a polygenic risk score (PRS) using 56,943 SNPs for breast cancer in women of African ancestry (AA) based on 9,235 breast cancer cases and 10,184 controls from a large pooled analysis of studies from African American and African women; the c-statistic from cross-validation was 0.581, considerably better than in previous efforts. We evaluated whether adding this AA-PRS to the BWHS risk prediction model would improve risk stratification. We conducted a nested case-control study of 901 breast cancer cases and 1,576 controls matched on age and most recent questionnaire completed from among BWHS participants for whom genome-wide association data were available and who had not been included in the collaboration from which the PRS was derived and tested. We examined discriminatory accuracy, estimated by the area under the receiver operating characteristic curve (AUC), for the risk prediction model alone, the PRS alone, and the combination of risk prediction model and PRS, controlling for the matching factor “questionnaire cycle”. We conducted the analyses within strata of 5-year age and then combined results using inverse-variance weighting. In preliminary analyses, the AUC was 0.579 for the risk prediction model alone and 0.600 for the AA-PRS alone. When the AA-PRS and the BWHS risk prediction model were both used as predictors in a logistic regression model, the AUC increased from 0.579 to 0.622. This improvement in risk stratification is similar to what Kachuri et al. (2020) obtained in an analysis of U.K. Biobank data, where adding a PRS to epidemiologic and personal risk factors showed an improvement from 0.572 to 0.635 in women of European ancestry. The present study provides external validation of a recently derived AA PRS and demonstrates the potential for improving risk stratification for U.S. Black women by adding a PRS to a breast cancer risk prediction model that already includes family history of breast cancer. Citation Format: Gary R. Zirpoli, Kathryn L. Lunetta, Ruth M. Pfeiffer, Julie R. Palmer. PD14-06 Polygenic risk score added to risk calculator improves prediction of breast cancer in U.S. Black women [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr PD14-06.