A comparison of BRCA1 and BRCA2 risk assessment models

10034 Background: A number of models of varying complexity are available to calculate the risk of carrying a BRCA1 or BRCA2 mutation causing hereditary breast and ovarian cancer. A >10% risk of carrying a BRCA1 or BRCA2 mutation is commonly used to establish testing eligibility in North America (testing threshold). We have compared these risk assessment models to ascertain their utility in patients from a large breast cancer program in our tertiary referral center. Methods: Risk assessment calculations were performed using the BRCAPRO, Myriad, Manchester, and UPennII models for all new probands referred to our clinic at Mount Sinai Hospital, Toronto (Canada), between May and October 2005. Individuals with a known familial mutation were excluded. Pearson correlation coefficients and pairwise scatterplots were performed for each pair of models. Regression analyses were developed to explore how well the BRCAPRO model can be predicted by other models. Results: 103 probands were included in our analyses (97% female, mean age: 50 yrs, 22% of Ashkenazi Jewish descent, 52.4% affected with breast cancer, 67% with ≥ one 1st degree relative affected with breast or ovarian cancer). The proportion of probands meeting the testing threshold based on each model was: BRCAPRO (34.0%), Myriad (33.0%), Manchester (55.3%), and UPennII (46.6%). Correlations calculated for each pair of models were large (range: 0.48–0.86). In comparing two of the most clinically relevant models (BRCAPRO and UPennII), the proportion of probands meeting the testing threshold was significantly higher using the UPennII model (p=0.0036). Between these two models, a higher correlation was observed for BRCA1 than for BRCA2 (R=0.76, R=0.58, respectively). The UPennII model was the strongest univariate predictor of the BRCAPRO results. Conclusions: All risk assessment models predicted a substantial proportion of patients referred to our clinic to be eligible for testing (≥ 33% of probands). A significantly higher proportion of probands were eligible for testing when using the UPennII model compared to the BRCAPRO model. This difference is in part due to the differential inclusion of prostate/pancreatic cancers and of third degree relatives with cancer in these models. An exploration of the sensitivity and specificity of these models will be presented. No significant financial relationships to disclose.