In this article, we provided an analysis of the largest US collection of both AJ and non-AJ families who have been ascertained for mutation analysis of BRCA1 and BRCA2. Because of our large sample size, the estimates that we provide are more accurate than those provided by previous studies.
Several findings have direct impact on counseling and screening. First, our study reported the future risks of developing cancer for cancer-free mutation carriers. Risks given in this form can be immediately used by a genotyped individual to make preventative decisions for herself and her relatives. They can also be used in establishing cancer screening recommendations at various ages for mutation carriers.
Second, our study found the RRs and BFs of breast cancer to be high in early ages, and then they rapidly decrease. This decreasing trend is the result of both a decreasing risk for carriers and an increasing risk for noncarriers. It is widely believed and used informally in genetic counseling practice that early-onset breast cancers are a strong indicator for germline mutations. Our study provided concrete quantitative evidence for this hypothesis. The high BFs that we obtained for the age groups from 20 to 49 years clearly indicate the importance of considering mutation analysis in counselees presenting with breast cancer in this age range. In the age interval of 60 to 69 years, the BF for breast cancer cases decreased to nearly 1, suggesting only weak evidence in favor of a mutation for women diagnosed with cancer in this age range.
Last, we found consistently high RRs and high BFs for ovarian cancer for all ages. This is in accordance with the literature and empirical observation that the presence of any ovarian cancer diagnosis is strong evidence towards an inherited mutation, regardless of the age of diagnosis.11
Prophylactic oophorectomy is widely accepted as a risk reduction procedure in women who are found to be BRCA1/2
mutation carriers. Bilateral prophylactic oophorectomy (BPO) is reported to cut the risk of ovarian cancer by 96% (95% CI, 84% to 99%) and the risk to breast cancer by 53% (95% CI, 23% to 71%) among carriers.37
In the CGN centers, 27% to more than 60% of unaffected women who received a positive mutation analysis result underwent prophylactic oophorectomy within a year after the mutation testing38,39
; this includes centers where all carrier women received a BPO unless they were less than 30 years old and/or had a desire to have children. By combining the results of our study and the risk reduction estimates just given, we derived the future risks for an asymptomatic woman who underwent BPO, which are reported in .
Predicted Cancer Risk for a Female Mutation Carrier Who Has Undergone Bilateral Prophylactic Oophorectomy
Ignoring the status of BPO leads to the underestimation of penetrance in prospective studies.40
However, this is unlikely to affect our analysis because, at the time the family history was collected, the probands and their relatives had not yet been genotyped. The rate of oophorectomy is only high for those women who are known carriers; the rate is low among women who do not yet know their genotype.41
Although our study includes mostly families having a strong history of breast or ovarian cancer, by using a retrospective likelihood, we can appropriately account for this sampling design, and in fact, we obtain penetrance estimates that are similar to those obtained from population-based designs. In addition, this approach allows more powerful estimates for probabilities of events that are rare in population-based samples, such as early-onset breast cancers. With the use of a Mendelian likelihood, we take advantage of all biologic information in the family history and are able to obtain penetrance estimates with tight CIs. A limitation of the likelihood formulation that we used is the lack of a parameter to model residual genetic or familial factors beyond the risks conferred by BRCA1
. If these factors have a strong impact on familial risk of cancer, it is possible that our model could overestimate absolute risks for cancer as a result of BRCA1
. However, King et al18
suggested that heterogeneity in the penetrance among BRCA1/2
carriers is unlikely to be significant.
We conducted a sensitivity analysis examining how our estimates change if we use different assumptions on the combined allele frequency of BRCA1 and BRCA2 deleterious variants. The combined frequency is strongly negatively correlated with breast cancer risk for BRCA1 mutation carriers.
Our statistical methodology has taken extensive care in accounting for family ascertainment issues, site heterogeneity, and sensitivities of mutation analysis techniques (see Appendix
). Yet, because of the scope and computational complexity of the project, some simplifying assumptions had to be made. An important one is that testing results are conditionally independent of the family history and other parameters given the genotype. This could be violated in practice if there are additional sources of familial aggregation.42,43
Second, the chosen form of the penetrance function is based on an RR between carriers and noncarriers that is constant within 10-year age intervals. We have also explored more parsimonious parameterization, which gave similar results. We present the nonparametric results to avoid artifacts that may be introduced by specific parameterization.
A limitation of our study is the lack of further ethnicity indicators beyond Ashkenazim. Minority groups, such as the African Americans, are often under-represented in breast cancer studies.44,45
Families of African ancestry have recently been shown to have a different mutation spectrum than that of well-studied populations.46
However, the performance of the carrier probability prediction software BRCAPRO is similar among families of different ethnic origins, suggesting similarity in genetic characteristics despite the difference in mutation spectra.
In summary, we found that the penetrance of BRCA
mutations in the United States is largely consistent with previous studies on Western populations given the large CIs on existing estimates. However, the absolute cumulative risks are on the lower end of the spectrum.13–19,35
Women newly diagnosed with breast or ovarian cancer can directly use our results to understand the implications for their risk of carrying a BRCA1/2
mutation. Asymptomatic women in whom mutations have been identified can use our results to understand their risk of developing breast and ovarian cancer, both before and after prophylactic oophorectomy.