Kaul, M.

Abstract Background Artificial intelligence (AI)–based, mammography breast cancer risk prediction models show improved discriminatory accuracy relative to clinical risk models. However, data on their calibration are limited. This study compared model performance of 3 clinical breast cancer risk models: Gail, Tyrer–Cuzick v8, and Breast Cancer Surveillance Consortium v3 with the MIRAI AI-risk model. Methods Digital mammograms were ascertained from a screening mammography cohort of 12 308 women within the Mayo Clinic Biobank with 250 incident breast cancers (176 invasive) within 5 years. We predicted 5-year breast cancer risk, estimated discriminatory accuracy (C-index), and calibration (observed-to-expected ratio) of both overall and invasive breast cancer and compared estimates using bootstrapping approaches. Results MIRAI demonstrated similar or improved discriminatory accuracy of overall breast cancer (C-index = 0.71, 95% confidence interval [CI] = 0.68 to 0.74) and invasive breast cancer (C-index = 0.71, 95% CI = 0.67 to 0.75) compared with clinical models (overall breast cancer: C-index = 0.59-0.68; invasive breast cancer: C-index = 0.60-0.68). MIRAI’s calibration for risk of overall breast cancer (observed-to-expected ratio = 0.96, 95% CI = 0.85 to 1.08) was improved compared with Gail (observed-to-expected ratio = 1.22, 95% CI = 1.07 to 1.38) and Breast Cancer Surveillance Consortium (observed-to-expected ratio = 1.38, 95% CI = 1.22 to 1.56) but similar to TC with volumetric percent density and polygenic risk score (observed-to-expected ratio = 0.99, 95% CI = 0.87 to 1.13). However, for low-risk women (approximately 50%), MIRAI overestimated risk of overall breast cancer. MIRAI also overestimated risk of invasive breast cancer across the risk spectrum (observed-to-expected ratio = 0.68, 95% CI = 0.58 to 0.78), whereas clinical models had good calibration (observed-to-expected ratio = 0.86–0.99). Conclusion MIRAI demonstrated stronger discriminatory accuracy than clinical models for 5-year overall and invasive breast cancer risk prediction but overestimated risk for both breast cancer endpoints. AI-based risk models should consider discriminatory accuracy and calibration for invasive cancer before implementation.

Abstract Background: Pathogenic variants (PVs) in PALB2 are known to confer high-to-moderate risk for breast cancer. However, breast cancer risk estimates for PALB2 PVs for different populations are not well defined and associations with other cancers are still unclear. Using data from a large clinical genetic testing cohort we have established cancer-specific risks associated with PALB2 PVs. Methods: We analyzed the associations of PALB2 PVs with breast, ovarian, prostate, and pancreatic cancers in individuals tested at Ambry Genetics. Cancer-specific odds ratios (ORs) and 95% confidence intervals (CI) were estimated by comparing PV carrier frequencies in cancer cases to population controls from AllofUs matched by sex and adjusting for age and genetic ancestry and gnomAD v4.1 matched by sex and adjusting for ancestry. Two-sided p-values are reported. Results: Of 1,119,667 individuals tested, 357,109 had breast cancer, 37,468 ovarian cancer, 22,112 pancreatic cancer and 32,738 prostate cancer. PALB2 PVs were observed in 0.79% of breast cancer cases, 0.30% of ovarian cancer cases, 0.78% of pancreatic cancer cases, and 0.47% of prostate cancer cases, with mean age at diagnosis of 54.2 years for breast, 60.1 years for ovarian, 62.8 years for pancreatic and 66.7 years for prostate. Whereas only 2.4% of breast cancer cases with PVs had a personal history of ovarian or pancreatic, 13.4% of ovarian cases and 33.3% of pancreatic cases had a history of breast cancer. Analysis of PV types found that frameshift PVs accounted for 50% of PVs, nonsense for 30%, canonical splice sites for 5% and copy number variants for 8-9% of cases in all cancer types. Among 10,077 individuals with multiple primary cancers, PALB2 PVs were detected in 0.65% of cases, with the highest frequencies observed in combinations of breast and pancreatic cancer (1.87%) and pancreatic and prostate cancer (0.79%). To refine estimates of risks for breast cancer, association studies were performed separately with GnomAD v4.1 and AllofUs controls. PALB2 PVs were consistently associated with high risks of breast cancer with lower 95%CI above 4 (GnomADv.4.1 OR=6.85, 95%CI=5.94-7.96, p=3.05x10-145; AllofUs OR=5.65, 95%CI=4.83-6.65, p=2.54x10-155). This was significantly different and increased over breast cancer risk estimates from the CARRIERS and BRIDGES population-based studies (OR=4.67, 95%CI=3.73-5.85, p=3.95x10-27) and the UK Biobank population-based study (OR=3.87, 95%CI=3.09-4.83, p=4.62x10-27). Similarly, high risks of pancreatic cancer were confirmed (GnomADv.4.1 OR=5.14, 95%CI=4.15-6.32, p=3.38x10-52; AllofUs OR=5.59, 95%CI=3.92-7.85, p=2.78x10-18). Importantly, associations between PALB2 PVs and moderate risks of both ovarian cancer (GnomADv.4.1 OR=2.68, 95%CI=2.11-3.40, p=5.94x10-16; AllofUs OR=2.15, 95%CI=1.62-2.85, p=2.97x10-7) and prostate cancer (enriched for aggressive disease) (GnomADv.4.1 OR=2.45, 95%CI=2.01-2.97, p=1.52x10-19) were observed. Conclusions: This large-scale study provides robust estimates of PALB2 PV prevalence and associated cancer risks in a U.S. clinical genetic testing population. The findings provide refined risk estimates for breast and pancreatic cancers but also establish that PALB2 PVs confer increased risk for ovarian and prostate cancers. These results help clarify ongoing uncertainty, particularly for ovarian and prostate cancer, and highlight the importance of PALB2 in broader hereditary cancer risk assessment and multigene panel testing. Citation Format: Y. Y. Tan, C. Hu, H. Huang, N. J. Boddicker, W. Chen, M. Kaul, T. J. Rao, A. N. Monteiro, T. Pesaran, R. Karam, S. M. Domchek, S. Yadav, M. E. Richardson, F. J. Couch. Cancer-specific risks associated with germline PALB2 pathogenic variants from a large clinical genetic testing cohort [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr RF5-01.