BRCA1 Testing—Advances and Retreats

Since the original localization¹ and cloning² of BRCA1, a key question has been: What proportion of the overall burden of breast cancer is attributable to mutations in this susceptibility gene? Previous studies have focused on 4 categories of individuals: members of high-risk families (those with apparent single-gene, autosomal dominant transmission of susceptibility to breast and/or ovarian cancer); genetically distinctive populations such as persons of Ashkenazi extraction; women with early-onset breast cancer; and women seen in high-risk breast cancer clinics (from moderate- and high-risk families). Findings from select published reports in these groups and highlights of findings of 2 articles in this issue of JAMA are summarized in Table 1.^{3 - 18}

The first studies of BRCA1 mutation prevalence suggested that nearly half of the families at high risk for breast cancer carried BRCA1 mutations.^{19 - 20} In the United States, the prevalence was estimated at 39%.¹² However, recent analyses suggest that the actual prevalence of BRCA1 mutations in high-risk families (≥3 cases of breast and/or ovarian cancer) might be as low as 12.8% to 16%.^{11 ,15} The reduced prevalence may be explained by the focus on families with large numbers of breast and ovarian cancer cases in the earlier studies, in comparison with a more representative mix of families with breast cancer alone, and with both breast and ovarian cancer in recent analyses. It now seems clear that the presence of ovarian cancer within a family is a strong predictor of BRCA1 mutations^{7 ,11 ,15} and that families with breast cancer and with no ovarian cancer have a much lower prevalence of BRCA1 mutations.¹⁵ Substantial variation in the prevalence of BRCA1 mutations in high-risk families in other countries has been observed,¹² ranging from less than 20% in Holland,²¹ Germany,^{22 - 23} and Japan²⁴ to 47% in Israel²⁵ and 79% in Russia.²⁶

Women affected with breast cancer at a particularly early age, who are unselected for family history, have also been studied for a possible underlying genetic predisposition. The report by Malone et al¹⁸ in this issue of THE JOURNAL builds on their prior study of BRCA1 mutations in these women.²⁷ Their population-based series of 193 women diagnosed as having breast cancer before age 35 years contained 12 disease-associated BRCA1 mutations (6.2%) (95% confidence interval [CI]=3.2%-10.6%). The authors¹⁸ also studied 208 women with breast cancer whose conditions were diagnosed before age 45 years, all of whom had at least a mother or sister with breast cancer. Hypothetically, this represents a group with a substantial likelihood of inherited risk. However, only 15 germline mutations in BRCA1 (7.2%) (95% CI=4.1%-11.6%) were detected. Mutation screening of 104 young women with no family history of breast cancer identified only 3 mutations (2.9%) (95% CI=0.6%-8.2%). The data indicate that the vast majority of BRCA1 mutations are found in individuals from moderate- to high-risk families. Even when selecting for early age of breast cancer onset, a family history is required for a greater than 3% probability of a BRCA1 mutation. The results of this analysis will prove useful for counseling women who present at high-risk clinics with minimal family histories. However, the low prevalence of BRCA1 mutations in young women with breast cancer who have minimal family history of the disease, argues against genetic screening in this population.

The study by Newman et al¹⁷ in this issue investigates a population-based series of white and black women with breast cancer, unselected for family history, aged 20 to 74 years. Of these 211 women, protein-truncating mutations in BRCA1 were found in 3 women. All 3 of the women were white and had a family history of breast or ovarian cancer. None of the 88 black women within their group had a BRCA1 mutation, although several were found to possess variants of unknown functional importance. The inclusion of this subset of black women is a welcome contribution to the existing literature, which focuses almost exclusively on white women. Clearly, the studies of the white population with breast cancer are not representative of the "melting pot" that is the US population of the 1990s. Further studies of the prevalence and penetrance of both BRCA1 and BRCA2 mutations in minority populations are needed if the entire US population is to be served.

By combining these studies with others, a number of family history characteristics that predict the presence of a BRCA1 mutation can be identified. These include early average age of onset in the family, increased number of relatives with ovarian cancer or both breast and ovarian cancer, Ashkenazi ancestry, and a total of 4 or more relatives with breast cancer.^{11 ,15 ,18 ,28} No correlation between the presence of bilateral breast cancer in a family and increased probability of being a BRCA1 mutation carrier has been observed.^{15 ,18} Some early attempts at generating BRCA1 mutation risk models based on these characteristics have been published.^{11 ,15 ,28} An important, and often overlooked, aspect of these risk models is family size, or the number of women at risk for cancer in a given family. For example, if one ignores family size and concentrates only on number of affected individuals, then 3 affected women in a family of 8 women at risk might be considered moderate risk, while 4 or 5 affected women from a family of 20 women at risk is labeled high risk. Some attempt to adjust for family size should help to ensure that small families of high risk are appropriately identified. The models contain wide statistical variation in many of the risk estimates but have proved to be of clinical importance since they provide a useful framework for identifying families in which genetic testing may be informative. Using the current literature, it is now possible to provide almost every white woman with a family history of breast cancer an estimate of her risk of carrying an underlying BRCA1 mutation. Of course, this does not identify an individual's risk of actually developing cancer. Further studies that evaluate large numbers of mutation carriers are needed to generate risk profiles and penetrance estimates for women who carry either BRCA1 or BRCA2 mutations.

The technical aspects of BRCA1 mutation screening must also be carefully considered. Current mutation screening methods, including direct DNA sequencing, are relatively insensitive techniques in that they fail to detect many BRCA1 inactivating mutations. Failure to detect BRCA1 mutations in families with breast cancer is potentially a substantial source of error in BRCA1 mutation prevalence estimates. In fact, a recent analysis of Dutch families²⁹ suggests that as many as 30% to 40% of disease-associated BRCA1 mutations are not being identified. The undetected mutations include large genomic deletions, which have been identified in several families, and are known to be major founder mutations in Dutch patients with breast cancer,^{29 - 30} and mutations that interfere with expression or stability of the BRCA1 transcript.² Another source of error in the prevalence estimates is the inability to discriminate between disease-associated missense mutations and "polymorphic" variants. As many as 30% of all sequence variants identified in mutation studies of BRCA1 are missense mutations,^{15 ,18} a large number of which remain unclassified at a functional level. Although functional assays have been successfully used to identify several disease-associated BRCA1 missense mutations,^{31 - 32} more extensive functional assays are needed. Improving knowledge of the role of BRCA1 as a transcription factor³³ and in DNA repair³⁴ may provide the basis for these assays. If 30% to 40% of disease-associated mutations are not being identified, what are the implications for patient counseling? Can patients with breast cancer and their families be appropriately counseled about their risk of carrying a BRCA1 mutation in this setting? A 30% to 40% error in the prevalence estimates currently used for patient counseling is very large, even for estimates that are meant to act as rough guidelines at best. To improve the quality of the information presented to patients and families who are considering mutation testing, some attempt is needed to correct the currently published estimates of BRCA1 mutation frequency. This may take the form of identification of loss of expression and genomic deletion mutations in a number of populations. Unfortunately, classification of missense mutations must wait for the development of improved functional BRCA1 assays. The BRCA1 mutation screening for clinical and research purposes may continue, but it would seem appropriate that supplemental methods of analysis for detection of these mutations be included in all testing protocols.

Many questions remain regarding the proper application of BRCA1 (and BRCA2) testing. The low prevalence of BRCA1 mutations in the population-based series of Newman et al¹⁷ and Malone et al¹⁸ strongly suggests that widespread clinical testing of individuals and families of low risk is premature. However, testing may be appropriate for motivated members of low- to moderate-risk families, given that some women are carriers of BRCA1 mutations in the absence of striking family history. Clinical application of BRCA1 testing is most appropriate when trained health care professionals use the available clinical and statistical tools to identify high-risk families in which testing may be clinically beneficial. This testing procedure always should include careful counseling by appropriately trained individuals. While the medical advantages of testing for high-risk families are evident, patients and families should be encouraged to join clinical research protocols so that further studies on familial breast cancer and ovarian cancer can be carried out. Finally, it is important to note that a sizable proportion of BRCA1 mutations may have been overlooked using current screening methods, and that current risk estimates for the presence of BRCA1 mutations may contain significant error. However, the currently published risk estimates probably will continue in use until improved studies are completed.