Abstract PS16-01: Longitudinal validation in the UK Biobank of a breast cancer risk assessment tool that combines a polygenic score for all ancestries with traditional risk factors

Le résumé fourni par la source

Abstract Background: Polygenic risk scores (PRS) have been shown to improve predictive accuracy when incorporated into traditional breast cancer (BC) risk assessment tools. However, most PRS have demonstrated suboptimal performance among women of non-European ancestry. We improved a previously reported multiple-ancestry PRS (MA-PRS) to create a second-generation MA-PRS consisting of 56 ancestry-informative and 329 BC-associated single-nucleotide polymorphisms (SNPs). MA-PRS achieved accuracy for diverse populations by characterizing genetic ancestry at each BC SNP and applying ancestry-specific SNP risks and frequencies. Here, we present longitudinal validation of a combined risk score (CRS) that integrates the second-generation MA-PRS with Version 8 of the Tyrer-Cuzick (TC) model using data from the UK Biobank (UKBB). Methods: The study cohort comprised 198,872 female participants from the UKBB with no history of cancer at the time of study enrollment. Data was dispensed on September 20, 2023. The primary outcome was the time from enrollment to diagnosis of invasive BC or censoring. CRS calibration was assessed by comparing observed (O) to expected (E) incident BCs. The ability of CRS and TC to discriminate between women diagnosed with BC versus women who were unaffected throughout study follow-up was measured by Cox proportional hazards models adjusted for age and ten genetic principal components. We used Kaplan-Meier analysis to examine BC incidence for patients classified as high-risk or low/moderate-risk according to a 5% 10-year risk threshold. Analyses were conducted in both the full cohort and in a subset of 11,563 female participants of self-reported non-White ancestry. Results: After a median follow-up of 11.8 years, 7,127 (3.6%) participants from the overall cohort and 315 (4.4%) non-White participants were diagnosed with invasive BC. The CRS was well calibrated in the overall cohort (O/E 1.00; 95% CI 0.97-1.02) and in the non-White sub-cohort (O/E 0.91; 95% CI 0.81-1.02). The incorporation of MA-PRS led to significantly improved discriminatory accuracy of CRS compared to TC in both the overall cohort (p=8.2´10^-320) and in the non-White sub-cohort (p=3.1´10^-04). More participants were identified as high risk by CRS (13.0%) than by TC alone (8.4%). Among those classified as high risk by TC, 39.3% were low/moderate risk by CRS, and of those classified as low/moderate risk by TC, 8.6% were high risk by CRS. In cases where CRS and TC classifications disagreed, CRS was more accurate in predicting incident BC. Conclusion: The CRS, incorporating a second-generation MA-PRS, was well-calibrated in predicting BC and significantly improved upon a traditional risk factor model in UKBB participants unaffected with BC at the time of enrollment. Clinical use of CRS has the potential to improve BC survival through more accurate identification of individuals at high risk. Citation Format: Timothy Simmons, Elisha Hughes, Srikanth Jammulapati, Dmitry Pruss, Eudora Hu, Ryan Bernhisel, Allison Kurian, Holly J. Pederson, Pat Whitworth, Sarah Ratzel, Jeff Jasper, Katie Johansen Taber, Alexander Gutin. Longitudinal validation in the UK Biobank of a breast cancer risk assessment tool that combines a polygenic score for all ancestries with traditional risk factors [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr PS16-01.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.