Abstract P2-10-01: The Accuracy of BRCA1/2 Mutation Prediction Models in Different Ethnicity and Gender: Experience in a Chinese Cohort

Abstract

Background: Risk models (BRCAPRO, Myriad, Couch and Shattuck-Eidens, BOADICEA) are well established in Caucasian and African American cohorts to estimate the probability of BRCA1/2 mutation. Few studies have suggested its performance limitation in Asian cohorts. Most studies did not account for gender specific prediction. The aim of the study is to evaluate the performance of these models in a Chinese cohort who have breast/ovarian cancer at a pre-genetic test setting. Methods: Four risk assessment models, Boadicea, BRCAPRO, Myriad, Couch and Shattuck-Eidens, were used to perform risk calculations to 217 non-BRCA carriers (198 females and 19 males) and 32 BRCA carriers (28 females and 4 males. Sensitivity, specificity and area under the receiver operator characteristic (ROC) curve were calculated for each model to evaluate for calibration, discrimination and accuracy in BRCA mutation prediction stratified by gender. Results: The mean in prediction score in all models were statistically significantly higher in female BRCA mutation carriers. However, there were no statistically difference in mean prediction score between BRCA carriers and non-carriers in all models for male patients. BRCAPRO slightly over-estimated the total numbers of BRCA1 and BRCA2 female carriers (13 vs. 11 and 20 vs.17), but underestimated the number of BRCA2 male carriers (2.8 vs. 4). While Myriad underestimated both the total numbers of BRCA1/2 male (3.1 vs. 4) and female (25.6 vs. 28) carriers. Boadicea did the closest estimation for both male (3.2 vs. 4) and female (1.5 vs. 11 for BRCA1 and 16.3 vs. 17 in BRCA2). BRCAPRO showed the greatest ROC area for BRCA1 (93%), BRCA2 (73%) and BRCA1/2 (79%) combination mutation prediction and highest sensitivity at conventional thresholds of 10% and 20% in female patients (71.4% vs.60.7%). Boadicea had the greatest ROC area for BRCA2 and BRCA1/2 combination mutation prediction and the same sensitivity at conventional thresholds of 10% and 20% in male patients (92% vs. 93%). Conclusion: All 4 models could perform reasonably well in female patients, but not in male patients. Boadicea has the best performance in male and female Chinese cohort overall. Whereas when comparing females alone, BRCAPRO is most accurate. Citation Information: Cancer Res 2010;70(24 Suppl):Abstract nr P2-10-01.