|Home | About | Journals | Submit | Contact Us | Français|
In women at increased risk for breast and ovarian cancer, the identification of a BRCA1/2 mutation has important implications for screening and prevention counseling. Uncertainty regarding the role of BRCA1/2 testing in high-risk women from diverse ancestral backgrounds exists due to variability in prevalence estimates of deleterious (disease-associated) mutations in non-White populations. We examined the prevalence of BRCA1/2 mutations in an ethnically diverse group of women referred for genetic testing.
We conducted a cross-sectional analysis to assess the prevalence of BRCA1/2 mutations in a group of non-Ashkenazi Jewish women undergoing genetic testing.
From 1996-2006, 46,276 women meeting study criteria underwent DNA full-sequence analysis of the BRCA1 and BRCA2 genes. Deleterious mutations were identified in 12.5% of subjects, and recurrent deleterious mutations (prevalence > 2%) were identified in all ancestral groups. Women of non-European descent were younger (45.9 yrs, SD11.6) than European (50.0 yrs, SD11.9)(p<0.001). Women of African (15.6%)[OR 1.3(1.1-1.5)] and Latin American (14.8%)[OR 1.2(1.1-1.4)] ancestries had a significantly higher prevalence of deleterious BRCA1/2 mutations compared to women of Western European ancestry (12.1%), primarily due to an increased prevalence of BRCA1 mutations in these two groups. Non-European ethnicity was strongly associated with having a variant of uncertain significance; however, re-classification decreased variant reporting (12.8%→5.9%), with women of African ancestry experiencing the largest decline (58%).
Mutation prevalence is high among women referred for clinical BRCA1/2 testing, and risk is similar across diverse ethnicities. BRCA1/2 testing is integral to cancer risk assessment in all high-risk women.
Each year, over 200,000 American women are diagnosed with breast and ovarian cancer,1 with 5-10% of these attributable to inherited mutations.2-5Knowledge of personal genetic risk is increasingly important as prevention measures have produced significant mortality reduction in high-risk women.4,6-8Deleterious (disease-causing) mutations in the BRCA1 and BRCA2 (BRCA1/2) cancer predisposition genes [Breast Cancer 1 (HGNC1100) and Breast Cancer 2 (HGNC1101)] are the principal cause of hereditary breast-ovarian cancer (HBOC).9,10The lifetime risk of breast cancer in mutation carriers has been reported to be as high as 87%.11,12The risks of ovarian cancer (44% in mutation carriers by age 70) and contra-lateral breast cancer (12-27% in mutation carriers within five years of breast cancer diagnosis) are also elevated.11,13-15In women with a personal or family history suggestive of HBOC, BRCA1/2 full-sequencing and analysis for large genomic rearrangements (LGR) are routinely used to quantify the genetic component of cancer risk.
The association between race/ethnicity (hereafter ethnicity) and BRCA1/2 mutation risk is controversial, potentially hampering prevention efforts in non-white women. Direct comparisons of mutation prevalence by ethnicity are few.16-20Ashkenazi Jews have a markedly elevated risk of HBOC secondary to a high frequency of BRCA1/2 mutations (1 in 40, or 10-fold elevated risk over the general US population), mainly attributable to three well-described founder mutations in BRCA1 (187delAG and 5385insC) and BRCA2 (6174delT).12,21-22For the remainder of high-risk women, ethnicity-specific prevalence of BRCA1/2 are less clearly defined.
Deleterious BRCA1/2 mutations are present in 1/400-1/800 people in the general population.2,11,15In a 2005 clinic-based study examining ethnicity-specific BRCA1/2 mutation rates, Nanda et al. reported a lower frequency of deleterious mutations (27.9% vs 46.2%) in high-risk African American families compared with Whites.23 However, a second clinic-based study of women with early onset (< 50 years) breast cancer found similar rates of deleterious mutations among African American and white women (both 17%).20 More recently, researchers from Northern California have published estimates of BRCA1 mutation prevalence in non-Jewish white (2.2%), Hispanic (3.5%), African American (1.3%), and Asian American (0.5%) women with breast cancer under age 65 years in a population-based study.19
It has been unclear to what degree BRCA1/2 mutation frequency differs among women of diverse ethnicities. More importantly, it is unknown whether ethnicity should be a consideration in BRCA1/2 risk assessment, eligibility for genetic testing, and the adoption of targeted prevention strategies. To begin to address these questions, we examined a comprehensive data repository of BRCA1/2 testing maintained by Myriad Genetic Laboratories, Inc. (Myriad) (Salt Lake City, UT) amassed over 10 years and representing by our estimates >95% of the entire BRCA1/2 testing experience in the US.
The data source for this cross-sectional analysis is a clinical database supported by Myriad Genetic Laboratories, Inc. of individuals tested for mutations in the BRCA1 and BRCA2 genes. Established in 1996, the primary purpose of this database is the organization of personal and family cancer history and mutation data collected on all individuals tested for BRCA1 and BRCA2 mutations through Myriad. The database includes all individuals who have undergone testing, including those receiving: 1)full-sequence DNA analysis of the BRCA1/2 genes; 2)site-specific DNA testing for persons with a known familial mutation; 3)founder panel testing at three sites for two highly prevalent mutations in BRCA1 (187delAG and 5385insC) and one in BRCA2(6174delT) found primarily in the Ashkenazi Jewish population. This database has been used in part to generate BRCA1/2 mutation prevalence estimates accessible by the public for clinical and research purposes (www.myriadtests.com). The structure of this database has been previously described.24
All persons undergoing clinical full-sequence DNA testing for mutations in BRCA1/2 from November 1996 to March 2006 were considered for the study. Individuals electing to undergo BRCA1/2 testing are referred from a wide variety of settings, ranging from private physicians’ offices to high-risk cancer clinics at major academic medical centers. Demographic and personal/family cancer history data are collected directly from a test requisition form (TRF) included in each test kit.24 Following initial receipt of the TRF, demographic and cancer history data on subjects are not updated; however, a particular test result [i.e. mutation classification, such as “deleterious” or variant of uncertain significance “(VUS)”] may be updated over time and, for this study, represents that status as of March 31, 2006. Nearly all subjects are from the US; however, a small number of international subjects(<1%) are included.
Family history data were collected from the TRF. For the purposes of this study, women having 1)breast cancer < 50 years, 2) >2 (including bilateral) breast cancers, or 3) history of ovarian cancer 4)both breast and ovarian cancer were considered to be at elevated risk of a BRCA1/2 mutation. Subjects with a family history of 1)breast cancer in >2 first- and/or second-degree relatives, 2)breast cancer <50 years in a first- or second-degree relative, 3) ovarian cancer in >2 first- and/or second-degree relatives, 4)breast and ovarian cancer in >2 first- and/or second-degree relatives, 5)both breast and ovarian cancer in a single first- or second-degree relative were also considered to have an elevated risk of carrying a BRCA1/2 mutation.
Race/ethnicity was self-reported on the TRF. Traditional racial/ethnic categories (e.g. “white”) were expanded for the TRF to better measure the ethno-geographic variability specific to BRCA1/2 mutations. The TRF allows eight ancestry choices: African, Ashkenazi Jewish, Asian, Central/Eastern European (hereafter Central European), Latin American/Caribbean (hereafter Latin American), Native American, Near/Middle Eastern (hereafter Middle Eastern), and Western/Northern European (hereafter Western European). An “other” option is also provided, allowing subjects to write-in a self-described ancestral designation. As the largest subgroup, Western European ancestry serves as the referent for the majority of analyses performed here. Where indicated, persons of Western and Central European ancestry have been combined in several analyses (referred together as European). For accuracy, racial/ethnic categorizations employed by other authors (e.g. “white”) are preserved in the Discussion; however, while similar, these categories do not directly parallel our own and are introduced for comparative purposes only.
Subjects included in the study underwent clinical full-sequence BRCA1/2 analysis (research-related, site-specific, and founder panel testing were excluded), chose only one of the eight provided ethnicity categories, were female, and completed the personal and family cancer history sections of the TRF. Persons with incomplete personal or family cancer history were excluded. Ashkenazi Jews were excluded from this analysis because testing procedures are substantially different in this group—most Ashkenazi women undergo initial founder mutation screening/testing, and, if positive, do not receive full-sequence analysis. Males were excluded due to small numbers after excluding Ashkenazi Jews. In total, from 63,947 persons undergoing full-sequence testing, 10,078 reported >1 ethnicity (or “other”) and 7593 had incomplete cancer histories or were male, leaving 46,276 subjects for this analysis.
Full-sequence DNA analysis of BRCA1 and BRCA2 and, since August 2002, break-point analysis for five large genomic rearrangements in BRCA1(exon13del3835bp, exon13ins6kb, exon14-20del26kb, exon22del510bp and exon8-9del7.1kb) were performed. Technical aspects of these analyses have been previously described in detail.24,25 Mutations identified from sequence analysis were classified into five categories: deleterious and suspected deleterious (herein combined as “deleterious”); VUS; variant-favor polymorphism; and polymorphism.26 The deleterious classification includes all nonsense mutations and all frame-shift mutations that begin at or before the last known nonsense or frame-shift mutation shown to co-segregate with disease. In addition, specific missense mutations and non-coding intervening sequence(IVS) mutations are recognized as deleterious on the basis of data derived from linkage analysis of high risk families, functional assays, biochemical evidence and/or demonstration of abnormal mRNA transcript processing. “Suspected deleterious” are genetic variants for which all of the available evidence indicates a very strong likelihood that the mutation is harmful or deleterious but whose effect on protein function cannot easily be determined. A suspected deleterious result typically is treated clinically as a deleterious (mutation positive) result. Many variants initially classified as VUS have been reclassified based on additional data, using an approach similar to that described by Goldgar et al.,27 segregation analyses, and co-occurrence of known deleterious mutations with VUS in the same individual.28
The age at which subjects underwent testing was treated as a continuous variable. Mean subject age at the time of testing was calculated for each ethnicity and compared by two-sided T-test. Univariate analyses (results adjusted for age) were performed to evaluate the association of ethnicity to the presence of personal and/or familial risk factors predictive of an elevated risk of a BRCA1/2 mutation (Table I), the prevalence of deleterious mutations, and the prevalence of variants of uncertain significance (Table III). Ethnicity was treated as a dummy variable. These results are presented as odds ratios with Western European ethnicity serving as the referent group. All confidence intervals are reported at the 95% significance level. All statistical tests and reported p-values are two-sided (α = 0.05). All analyses were performed using Stata statistical software (Stata Corporation, College Station, TX).
Between November 1996 and March 2006, 46,276 consecutive subjects meeting eligibility criteria underwent BRCA1/2 testing. The racial/ethnic breakdown of the study sample is presented in Table 1. In total, 87.4%(n=40,301) of participants reported European (Western or Central European) ancestry; Middle Eastern represented the smallest ethnic subgroup (n=492, 1.1% subjects).
Ethnicity-specific differences in testing age and strength of cancer history were observed. The mean age at the time of testing among European women was 50.0 years (SD11.9), while that for women of African (45.2 years, SD10.9), Native American (48.5 years, SD11.6), Asian (47.1 years, SD12.5), Latin American (44.7 years, SD11.4), and Middle Eastern (47.4 years, SD12.1) descent (all non-European 45.9 yrs, SD11.6) were all significantly lower (individual p-values all <0.001). To evaluate variability in referral thresholds, we stratified cancer history by ethnicity. The fraction of women at increased risk for a deleterious BRCA1/2 mutation secondary to a personal and/or family history of breast or ovarian cancer was higher for women of African (92.0%) [OR 1.5(1.2-1.8)] and lower for women of Central European (85.2%) [OR 0.8(0.8-0.9)] ancestry compared with Western European (87.2%). Ethnicity-specific frequencies of cancer history characteristics are detailed in Table 2.
Disease-related mutations were identified in 5780 or 12.5% of subjects (Table 3). Fifteen women were found to carry simultaneous BRCA1 and BRCA2 mutations. Women of African ancestry had the highest prevalence of deleterious mutations (15.6%, vs 12.1% for Western European) [OR 1.3(1.1-1.5)], and had nearly twice as many BRCA1 mutations as BRCA2 (180 versus 100). Women of Middle Eastern descent had the lowest mutation prevalence (9.4%) [OR 0.7(0.5-1.0)]. Overall, BRCA1 mutations were more common than BRCA2 for every ethnicity except Asians, where an equal frequency was seen (6.3% for each gene, 12.7% overall).
Recurrent mutations (mutation prevalence > 2%) were identified in every ancestral subgroup (Tables 4&5). Some, like the BRCA1 Ashkenazi founder mutations 187delAG and 5385insC, were found among nearly all the ancestral groups, while others were unique within each group. The Ashkenazi founder mutation 187delAG was common to every ethnicity except African [range 2.2-15.2%], while 5385insC was seen in Native Americans(3.8%), Western Europeans(5.2%) and Central Europeans(14.9%). Interestingly, while the 187delAG founder mutation is 4-5 times as common in Ashkenazi Jews as 5385insC, the ratio was reversed in women of European decent studied here: the ratio of 5385insC to 187delAG was nearly 2:1 for Western Europeans and 4:1 for Central Europeans.
Recurrent mutations were also identified within each ancestral group. The African (31.5%), Latin American (36.6%), and Middle Eastern subgroups (45.7%) had the largest percentage of total mutations by subgroup represented by recurrent mutations. Aside from the Ashkenazi founder mutations, the BRCA1 C61G mutation was identified the most often (137 subjects total), primarily in Europeans (Western and Central). Other than C61G and the Ashkenazi founder mutations, only the L63X mutation was shared among individuals reporting different ancestral backgrounds [Middle Eastern(7.6%) and Asian(3.3%)].
At the time of analysis, at least one mutation classified as VUS was found in 3,057(6.6%) of the women tested (Table 3). The prevalence of VUS varied considerably by ethnicity, but VUS reporting decreased markedly, particularly for non-European subgroups, over the time of data collection due to reclassification (Figure 1). Overall, women of African ancestry had the highest prevalence of VUS at the time of analysis (16.5%, vs 5.7% for Western European) [OR 3.2(2.8-3.7)]. The prevalence of VUS was inversely correlated with the total number of individuals tested in each subgroup (p <0.001 for trend).
BRCA1/2 mutation prevalence is remarkably similar among women who undergo genetic testing, regardless of ethnicity. Here, we report ethnicity-specific mutation prevalence estimates ranging from 9.4% to 15.6%, with a pooled estimate of 12.6% for women of European (Western and Central) ancestry and 14.1% for all women of non-European ancestry (Latin American, African, Asian, Native American and Middle Eastern). Overall, recurrent mutations (prevalence >2%, including the Ashkenazi founder mutations) represented a proportionally larger fraction of the total mutations identified (as high as 45.7%) among the non-European ethnicities examined compared with women of Western European ancestry. However, despite similar mutation prevalence, testing was less frequently performed in non-European women [n=5,975(12.6%) of all women tested].
In evaluating the impact of genetic testing as a cancer prevention modality, our prevalence estimates are unique in their practical relevance. Unlike a high-risk clinic-based23 or a population-based sample19, ours represents a referral population reflecting a diverse range of personal and familial risk factors, yet one that is also guided by clinically-relevant forces (belief systems, provider biases, and healthcare disparities). Comparatively, the strength of this study is its inclusion of all comers referred for testing, regardless of personal or family history (Nanda et al.23 studied only “high-risk” families; John et al.19 studied only women with incident breast cancer). In their study, John et al. report a positive association of BRCA1 prevalence with Hispanic ethnicity (OR1.3, 1.0-1.7) but an inverse association with Asian ethnicity (OR0.2, 0.1-0.3). We also found a high prevalence of BRCA1 mutations in Latin American women, much of it attributable to a previously described high frequency of the 187delAG mutation (Tables 3 and and44)29, but found a prevalence of BRCA1/2 mutations in Asians (12.7%) comparable to that of Europeans (12.6%). Nonetheless, it is difficult to directly compare these important findings to our own because incident ovarian cancers were not included in John et al, and only a portion (51% of high-risk, 57% of average risk) of the population was tested (and only for BRCA1).
Striking racial/ethnic variability in the number of women undergoing genetic testing in the United States exists. In our sample, 87.4% of tested subjects reported European ancestry, 4.2% Latin American ancestry, 3.8% African ancestry, and 2.6% Asian ancestry (see Table I). Data from the US 2000 Census estimate that 74% of the US population is white (this figure includes most Hispanics, who represent 14.8% of the population), 13.4% African American, and 4.4% Asian.30 While the standard US Census racial designations are clearly different than the ancestral categories offered on the TRF, analysis of the data suggests that genetic testing for hereditary breast cancer has been performed disproportionately more often in white women.
Differences in breast and ovarian cancer epidemiology, disease biology, and healthcare access/utilization may contribute to ethnicity-specific variability in the number of women receiving genetic testing. In the United States, the incidence of both breast and ovarian cancer is higher in white women than other racial/ethnic groups.31-32 Nonetheless, breast cancer mortality for African American women exceeds that of white women, and is nearly double that of Asian American, Hispanic, and Native American counterparts.31,33 Ethnicity-specific variation in risk factor exposure,34 including dietary,35-36 body size/weight,37-38 behavioral,39 and reproductive/hormonal risks,40-41 may also contribute to observed differences in cancer incidence and genetic testing uptake.
There may also be inherent biologic differences in tumors among ethnic groups. African American women have a high proportion of early-onset (pre-menopausal) breast cancers, particularly those displaying basal-like and/or triple-negative histopathology.42-44 Because early-onset breast cancer is an established hereditary risk criterion, particularly for BRCA1-related breast cancers, disease biology may in part explain the lower age observed among women of African origin in our cohort, as well as the higher proportion of BRCA1 mutations (64.3%) seen in this group.
Healthcare access and/or utilization barriers to testing must also be considered. To investigate provider-based differences in testing referral, we examined whether women of European ancestry were referred at lower thresholds by calculating the fraction of women meeting criteria for elevated risk. As seen in Table 1, absolute risk differences between ethnicities were overall minimal, on the order of only a few percentage points. Thus, in this study, those women who ultimately received BRCA1/2 testing were of comparable pre-test risk, and so potential disparities in referral thresholds or practices45 would alone be unlikely to explain the population disproportions observed. Other sources of differences in genetic testing uptake related to healthcare access/utilization have been documented and should also be considered, including ethnicity-specific socio-economic barriers (lack of health insurance46 or access to primary care47) cultural differences (perception of risk45,48-49or transmission of family cancer history49), fears of genetic discrimination,50 differences in medical knowledge-base,45,47-48and responses to genetic counseling.51
Recurrent mutations were common, with at least one of the three Ashkenazi founder mutations being identified at elevated frequency in nearly every ethnicity. While unknown or undisclosed Ashkenazi Jewish ancestry in these populations may in part explain this finding, a de novo mutation occurring at the same site as the Ashkenazi founder 187delAG must also be considered.29,52Among women of African ancestry, five recurrent mutations with prevalence >2% were observed. A previously described 943ins10 founder mutations in BRCA1 was the most commonly seen in our sample.53 For the Middle Eastern, Latin American, and African subgroups, recurrent mutations represented a sizable proportion of all mutations detected (45.7%, 36.6%, and 31.5%, respectively), although these numbers are small compared to the >90% of Ashkenazi Jewish individuals who have an Ashkenazi founder mutation. Extensive haplotype analysis has not been conducted to determine whether all of these represent true founder mutations (mutations sharing a common haplotype), but the low rate of de novo mutations in the BRCA1/2 genes would suggest that most, if not all, may be traced to common ancestral origins. Though unlikely, the possibility exists that the prevalence of recurrent mutations in our dataset could be slightly inflated if two relatives independently underwent full-sequence testing. Nonetheless, it is standard practice that once a mutation has been identified, carrier status is confirmed in family members by single-site DNA analysis.
Variants of uncertain significance (VUS) were identified in 6.2% (n=2874) of subjects, excluding individuals with simultaneous deleterious mutations (n=183). Genetically distinct and/or under-tested populations will contain VUS not seen in the White/European reference population. For example persons of African ancestry have a disproportionately large number of ancient BRCA1 haplotypes,54 and have been tested at only a fraction of the frequency (see Table I), so their DNA sequences predictably vary from the European-influenced North American haplotypes with greater frequency. VUS reporting has declined with increased volume of testing, most notably in women of African ancestry [37% (2002)→17% (2006)]. Re-classification of the VUS reported in the current study should, however, have minor impact on reported prevalence estimates of deleterious mutations.
This work has several important limitations. As a non-probability (opportunity) sample, our study population is subject to multiple selection biases that may influence receipt of testing and therefore bias prevalence estimates. Self-reported race/ethnicity or personal/family cancer histories collected from the TRF (particularly among high-risk persons) are subject to mis-classification, imprecision (e.g. changing self-reported ethnic identity over time), and cancer history recall bias. Finally, because family structure and size are not specifically queried on the TRF, individual risk estimation using popular clinical risk assessment tools is not possible. Nonetheless, because these data reflect empirical mutation prevalence in a clinically relevant population, they remain valuable for risk assessment. Several comparative analyses of the Myriad prevalence tables to other risk assessment tools have been published in recent years.55
The exclusion of Ashkenazi Jews, though limiting our ability to directly compare findings to the other ethnicities studied here, is nonetheless essential because mutation prevalence and testing procedures (i.e. use of the founder panel) are so different for this group that a reasonable comparison cannot be made. Moreover, provider referral thresholds for testing are also very different, with several authorities recently advocating population screening of Ashkenazi women. We excluded persons reporting no ethnicity because of the inability to draw conclusions regarding shared ethnicity and its association to mutation frequency, and those reporting multiple ethnicities in an attempt to enhance the genetic homogeneity of our subgroups, thereby increasing the clinical relevance of conclusions to a subject's predominant ancestral background.
Finally, the technological challenges of diagnostic BRCA1/2 DNA testing are many, including the dynamic process of VUS detection and reclassification. More recently, multi-level testing for LGR, estimated to occur in 7-10% of sequence negative individuals with an a priori risk of 30% or higher,56 has been implemented, improving mutation detection. However, while important at the individual level, the impact of rearrangement testing on overall mutation prevalence estimates will likely be minimal.
The identification and cloning of the BRCA1 and BRCA2 cancer predisposition genes opened the doorway to a new era of genetics-based cancer prevention and risk assessment.9,10 BRCA1/2 mutation prevalence is high and nearly identical among women of diverse ethnicities undergoing clinical genetic testing. Nonetheless, testing volumes are disproportionately low among women from non-European ancestries, and likely reflect the complex social, economic, and cultural factors governing healthcare access and utilization. Clinical genetic testing is an integral component of hereditary breast-ovarian cancer risk assessment and should be considered in all high-risk women regardless of race and/or ethnic background.
Dr. Hall is a recipient of a Mentored Research Scholar Grant from the American Cancer Society (MRSG-07-232-01-CPHPS to MJH).