• Ava Kwong, Connie H N Wong, Dacita T K Suen, Michael Co, Allison W Kurian, Dee W West, and James M Ford.
• Division of Breast Surgery, Department of Surgery, The University of Hong Kong, Hong Kong, China. akwong@asiabreastregistry.com
• World J Surg. 2012 Apr 1; 36 (4): 702-13.

BackgroundBRCA1/2 mutation prediction models (BRCAPRO, Myriad II, Couch, Shattuck-Eidens, BOADICEA) are well established in western cohorts to estimate the probability of BRCA1/2 mutations. Results are conflicting in Asian populations. Most studies did not account for gender-specific prediction. We evaluated the performance of these models in a Chinese cohort, including males, before BRCA1/2 mutation testing.MethodsThe five risk models were used to calculate the probability of BRCA mutations in probands with breast and ovarian cancers; 267 were non-BRCA mutation carriers (247 females and 20 males) and 43 were BRCA mutation carriers (38 females and 5 males).ResultsMean BRCA prediction scores for all models were statistically better for carriers than noncarriers for females but not for males. BRCAPRO overestimated the numbers of female BRCA1/2 mutation carriers at thresholds ≥20% but underestimated if <20%. BRCAPRO and BOADICEA underestimated the number of male BRCA1/2 mutation carriers whilst Myriad II underestimated the number of both male and female carriers. In females, BRCAPRO showed similar discrimination, as measured by the area under the receiver operator characteristic curve (AUC) for BRCA1/2 combined mutation prediction to BOADICEA, but performed better than BOADICEA in BRCA1 mutation prediction (AUC 93% vs. 87%). BOADICEA had the best discrimination for BRCA1/2 combined mutation prediction (AUC 87%) in males.ConclusionsThe variation in model performance underscores the need for research on larger Asian cohorts as prediction models, and the possible need for customizing these models for different ethnic groups and genders.

26 keywords...

hide…