Abstract / Summary
Abstract Background More than 170 breast cancer polygenic risk scores (PRS) are publicly available, yet no systematic comparison guides score selection for clinical use. Guidance is especially needed in hereditary breast and ovarian cancer (HBOC) diagnostics, where many patients meeting clinical criteria carry no pathogenic variant in established risk genes, therefore lacking actionable risk stratification. Methods Scores, including the genome sequencing-optimized BRIDGES PRS, were benchmarked in a monocentric, clinically annotated HBOC cohort of 775 female breast cancer patients (680 non-informative, without a (likely) pathogenic variant in 13 established HBOC genes; 95 informative carriers) and 1,313 cancer-free female controls of European ancestry, all sequenced by clinical-grade GS. PRS were calculated with standardized pipelines and assessed by area under the receiver operating characteristic curve (AUC) and Cohen's d. Benchmarks were stratified by informative status, age at diagnosis, molecular subtype, and breast density. SNP overlap among the top-performing scores was quantified using pairwise Jaccard similarity. Results The best-performing PRS yielded an AUC of 0.656 (95% CI 0.631–0.681; Cohen's d = 0.576; OR 1.77 per standard deviation) in HBOC patients without informative testing. 40.1% of these patients fell in the top PRS quintile versus 20% of controls, and 7.2% in the bottom quintile (top versus bottom OR 5.57, 95% CI 3.93–7.90). The top 20 scores formed a performance plateau (AUC 0.643–0.656) despite ranging from 3,820 to 3.7 million SNPs, and were dominated by near-identical variant composition, though compact scores matched them from far fewer or largely distinct variants. Performance varied by molecular subtype, strongest for hormone receptor-positive tumors (Luminal B-like HER2-positive, AUC 0.689; Luminal A-like, 0.677) and markedly reduced for triple-negative breast cancer (0.568). Conclusions Independently developed breast cancer PRS have converged on comparable performance and a largely shared set of variants. For laboratories already performing GS, computing a PRS adds negligible marginal cost while separating patients without a monogenic finding into groups whose odds of breast cancer differ more than five-fold. PRS thus helps close a major diagnostic gap and supports integration into HBOC counseling, showing that the choice of score is not a barrier to clinical implementation.