A multi-institution study of the accuracy of BRCAPRO in predicting BRCA1/BRCA2 mutations in women with ovarian cancer.

Abstract

1520 Background: 10-15% of ovarian cancer patients have a germline BRCA1 or BRCA2 mutation, with significant management implications for both patients and relatives. Genetic testing decisions are guided in part by the estimated likelihood of identifying a mutation. The BRCAPRO model uses personal and family history of breast and ovarian cancer to calculate the likelihood of a BRCA1/2 mutation. This study’s purpose was to assess the ability of BRCAPRO to accurately determine this likelihood. Methods: BRCAPRO scores were calculated using CancerGene v5.1 for 589 ovarian cancer patients referred for genetic counseling at three institutions. The study population was divided into quintiles by BRCAPRO score, with cutpoints chosen such that each quintile represented 20% of the sample. Chi-square goodness-of-fit test was used to compare observed BRCA1/2 mutations to the number predicted. ANOVA models were used to assess factors impacting BRCAPRO accuracy. Results: 180/589 (31%) ovarian cancer patients tested positive for a BRCA1/2 mutation. At BRCAPRO scores under 40%, more mutations were observed than expected (93 observed vs. 34.1 expected, p<0.001). If patients with BRCAPRO scores <10% had not been offered genetic testing, almost one-third of mutations (51/180, 28%) would have been missed. Multivariate analysis demonstrated that BRCAPRO underestimated risk for high grade serous ovarian cancers but overestimated risk for other histologies (p<0.0001), underestimation increased as age at diagnosis decreased (p=0.02), and model performance varied by institution (p=0.02). Conclusions: Ovarian cancer patients classified as low risk by BRCAPRO are more likely to test positive than predicted, therefore the BRCAPRO prediction could falsely reassure patients considering genetic testing. BRCAPRO performance could be improved by incorporating factors such as ovarian cancer histology. Alternatively, given the high prevalence of BRCA1/2 mutations in high grade serous ovarian cancer and the apparent limitations of using family history to predict mutation probability, BRCA1/2 genetic testing could be offered to high grade serous ovarian cancer patients regardless of family history.