In this study, we evaluated the performance of the BRCAPRO model in nearly 300 ethnic minority families in the United States and demonstrated that the model had good overall discrimination between BRCA
mutation carriers and noncarriers in minority families. The AUC was 0.75 for all minority groups combined, which is within the range of those reported in non-Hispanic white populations (0.71 to 0.83).9,10,13,14,16,32–34
Of the minority groups examined, BRCAPRO performed the best in Hispanics, as indicated by the highest AUC (0.83) and the smallest Brier score (0.089), in which the AUC was higher than in a previous observation in Hispanics (0.77).18
The model had the worst performance in African Americans (AUC, 0.68; Brier score, 0.208), in which the AUC was lower than our previous observation in African Americans (0.77).17
The performance of BRCARPO was intermediate for the other minority groups, mainly Asian-Americans and Native Americans (AUC, 0.71; Brier score, 0.139), and was similar to that observed in Han Chinese (0.70).35
However, the difference between ethnic groups was not statistically significant. Thus, the data suggest that BRCAPRO performs reasonably well and is applicable to all minorities. However, the sample size of this study may not be sufficient for subgroup comparisons. Real differences possibly exist in BRCAPRO performance across ethnic minority groups, because the same penetrances and non-Ashkenazi white allele frequencies were used in the calculation. These genetic parameters likely vary across ethnicity. Better BRCAPRO performance in Hispanics might occur because Hispanics are genetically closer to other white populations. Thus, mutation prediction could improve if population-specific allele frequencies and penetrance data were available. Our findings underscore the need for population-based studies in these minority populations.36,37
A single instance of breast cancer in families that have limited family structure presents a challenge for mutation prediction with the BRCAPRO model and probably with other models. Weitzel et al15
reported an AUC of 0.67 for the BRCAPRO model in women with breast cancer who did not have any first- or second-degree relatives with breast or ovarian cancers. We observed a similar finding in the current cohort. In terms of accuracy of the model, we showed that the numbers of mutations predicted and observed were virtually the same when the mutation detection method is assumed to have 85% sensitivity. Consistent with previous studies in predominantly white populations,9–11
we found that BRCAPRO overestimated mutation probability for high-risk persons and underestimated mutation probability for low-risk persons. Interestingly, this pattern was also observed for the Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm (BOADICEA) model, which allows for polygenic locus effects,10
and for the Log Odds of the Probability of Carrying an Ancestral Mutation in BRCA1 or BRCA2 for a Defined Personal and Family Cancer History in an Ashkenazi Jewish Woman (LAMBDA) model, which was developed empirically.34
Reasons for this pattern are unclear and may be related to families that have a single breast cancer diagnosis and limited family structure, familial clustering caused by environmental factors, heterogeneous penetrance caused by genetic/environmental modifiers and mutation spectrum, other low penetrance genes, and imperfect sensitivity of molecular methodology. We showed that lack of information in families that have a single breast cancer diagnosis is an important cause of this pattern. Whatever the reasons, genetic counselors should treat the calculated mutation probability from pretest prediction models with caution and should consider the amount of available information when interpreting mutation probability. Given the performance in both discrimination and accuracy, these results indicate that BRCAPRO is a useful risk assessment tool in clinical settings for minority families, but clinicians should exercise judgment in using the numbers to make recommendations.
Incorporation of the proband's test result significantly improved the ability to distinguish likely mutation carriers from noncarriers (ie, AUC increased from 0.76 to 0.90). This is the first study, to our knowledge, to estimate the magnitude of improvement in clinical settings. Thus, it is helpful to recalculate the mutation probability for other family members after the proband has been tested, especially if the proband tests negative. In the two examples shown in , the probands tested negative. For the 36-year-old breast cancer patient in family A, her mutation probability was 0.254 before the proband was tested, and the recalculated probability from BRCAPRO incorporation of the proband's test result was 0.075. Thus, a recommendation not to test for BRCA mutations could be entertained. (Indeed, her BRCA test result was negative.) In contrast, the family member with bilateral breast cancer in family B had a mutation probability of 0.882 before the proband's test result was considered, and her recalculated probability was 0.740 after incorporation of the proband's test result. Thus, a strong recommendation to test for BRCA mutation could be made. (In fact, she had a BRCA1 mutation).
Two family trees with breast and ovarian cancer history.
We did not find that the inclusion of information on prophylactic oophorectomy, a new feature of BRCAPRO, improved mutation prediction. Improvement may be relatively small in the current cohort, because less than 20% of families had at least one member with a history of oophorectomy. Of the 62 individuals who underwent prophylactic oophorectomy, 22 later developed breast cancer after 11 years, on average. As illustrated by Katki,22
distortion caused by ignorance of information on prophylactic oophorectomy is large if the family member has lived many years after the surgery. Therefore, we may not be able to see an influence yet, as prophylactic interventions for familial breast cancer are still relatively recent. Another explanation is that the proband may not know whether her relatives had a prophylactic oophorectomy.
In conclusion, this study demonstrates that the BRCAPRO model performs well in clinic settings and supports its routine use in pretest prediction of BRCA mutations in minority families, especially Hispanic families. Mutation test results of probands, especially if negative, provide additional discriminatory ability for counselees, which may help counselors decide whether to offer other family members testing when one member has already tested negative. The study also highlights the need for population-based studies to estimate penetrance and allele frequencies of BRCA genes in minority populations. Lastly, the inaccuracy in carrier prediction using BRCAPRO for families with a single breast cancer diagnosis is a challenge worthy of additional investigation. Genetic counselors should recognize this limitation when using BRCAPRO or other models to recommend genetic testing in single-diagnosis families.