10545 Background: Women with a family history of breast cancer (BC) have an increased risk of developing the disease. Several BC risk prediction models are used in clinical practice to estimate future BC risk, however, their comparative performance in women with a family history of BC is uncertain. The aim of this study was to systematically identify, describe, and critically appraise BC risk prediction models developed or validated in women with a family history of BC, and to synthesize their performance in predicting BC occurrence in this population. Methods: We searched MEDLINE, Embase, CINAHL, and ISI Web of Science up to February 2022, with a targeted MEDLINE update to 19 December 2024. We included studies that developed or externally validated BC risk prediction models in women with a family history of BC, provided family history was included as a predictor. Data were extracted using the CHARMS checklist, and risk of bias was assessed using PROBAST. Where data from more than three validation studies were available, random-effects meta-analyses of calibration (observed/expected O/E ratios) and discrimination (c-statistic) were conducted. Results: Forty-five studies were included. Twelve models were externally validated in the target population; four were validated more than three times and included in meta-analyses. Reporting quality was variable, and most studies had high or unclear risk of bias. The Gail (BCRAT) model was well calibrated (pooled O/E 1.06, 95% CI 0.91–1.25) but showed modest discrimination (c-statistic 0.61, 95% CI 0.57–0.66). BOADICEA was also well calibrated (O/E 0.98, 95% CI 0.90–1.17) with modest discrimination (c-statistic 0.65, 95% CI 0.58–0.71). Tyrer-Cuzick (IBIS) overpredicted risk (O/E 0.86, 95% CI 0.74–0.98), while BRCAPRO underpredicted risk (O/E 1.44, 95% CI 1.25–1.62). Discriminatory accuracy was similar across Tyrer-Cuzick v8, BOADICEA, and BRCAPRO, with pooled c-statistics of 0.64 (0.58 to 0.71), 0.65 (0.58 to 0.71),0.64 (0.54 to 0.73) respectively and slightly higher than Gail (pooled c-statistic 0.61 (0.57 to 0.66)) . Conclusions: In women with a family history of BC, Gail and BOADICEA are well calibrated, while Tyrer-Cuzick overpredicts and BRCAPRO underpredicts risk. No model demonstrated clearly superior discrimination. Considering both calibration and discrimination, BOADICEA may be useful for clinical risk assessment in this population, although conclusions are limited by study heterogeneity, small numbers of validation studies in the target population, and high or unclear risk of bias. Improved model performance and reporting are needed. This abstract is based on a post-peer review version of a Cochrane Review. Upon acceptance, the final version is expected to be published in the Cochrane Database of Systematic Reviews (www.thecochranelibrary.com).
McGarrigle et al. (Wed,) studied this question.