A Role for Common Genomic Variants in the Assessment of Familial Breast Cancer

Transcription

1 Published Ahead of Print on October 29, 212 as 1.12/JCO The latest version is at JOURNAL OF CLINICAL ONCOLOGY O R I G I N A L R E P O R T A Role for Common Genomic Variants in the Assessment of Familial Breast Cancer Sarah Sawyer, Gillian Mitchell, Joanne McKinley, Georgia Chenevix-Trench, Jonathan Beesley, Xiao Qing Chen, David Bowtell, Alison H. Trainer, Marion Harris, Geoffrey J. Lindeman, and Paul A. James See accompanying editorial doi: 1.12/JCO Sarah Sawyer, Gillian Mitchell, Joanne McKinley, Alison H. Trainer, and Paul A. James, Peter MacCallum Cancer Centre; Alison H. Trainer and Geoffrey J. Lindeman, Royal Melbourne Hospital; Gillian Mitchell, David Bowtell, Geoffrey J. Lindeman, and Paul A. James, University of Melbourne; Marion Harris, Monash Medical Centre; Geoffrey J. Lindeman, The Walter and Eliza Hall Institute of Medical Research; Paul A. James, Victorian Clinical Genetics Service, Melbourne, Victoria; and Georgia Chenevix-Trench, Jonathan Beesley, and Xiao Qing Chen, Queensland Institute of Medical Research, Brisbane, Queensland, Australia. Submitted January 12, 212; accepted August 2, 212; published online ahead of print at on October 29, 212. Supported by the Victorian Cancer Agency (Grant No. CTTS7). The Australian Ovarian Cancer Study was supported by the U.S. Army Medical Research and Materiel Command under Grant No. DAMD , The Cancer Council Tasmania and The Cancer Foundation of Western Australia, and the National Health and Medical Research Council of Australia. Authors disclosures of potential conflicts of interest and author contributions are found at the end of this article. Corresponding author: Paul A. James, MD, PhD, Familial Cancer Centre, Peter MacCallum Cancer Centre, Locked Bag 1, A Beckett St, Melbourne, VIC 86, Australia; paul.james@ petermac.org. 212 by American Society of Clinical Oncology X/12/399-1/$2. DOI: 1.12/JCO A B S T R A C T Purpose Genome-wide association studies have identified common genomic variants associated with increased susceptibility to breast cancer. In the general population, the risk associated with these known variants seems insufficient to inform clinical management. Their contribution to the development of familial breast cancer is less clear. Patients and Methods We studied 1,143 women with breast cancer who had completed BRCA1 and BRCA2 mutation screening as a result of a high risk for hereditary breast cancer. Genotyping of 22 breast cancer associated genomic variants was performed. A polygenic risk score (PRS), calculated as the sum of the log odds ratios for each allele, was compared with the same metric in 892 controls from the Australian Ovarian Cancer Study. The clinical features associated with the high and low ends of the polygenic risk distribution were compared. Results Women affected by familial breast cancer had a highly significant excess of risk alleles compared with controls (P ). Polygenic risk (measured by the PRS) was greater in women who tested negative for a BRCA1 or BRCA2 mutation compared with mutation carriers (P ). Non-BRCA1/2 women in the top quartile of the polygenic risk distribution were more likely to have had early-onset breast cancer ( 3 of age, odds ratio [OR] 3.37, P.3) and had a higher rate of second breast cancer (OR 1.96, P.2) compared with women with low polygenic risk. Conclusion Genetic testing for common risk variants in women undergoing assessment for familial breast cancer may identify a distinct group of high-risk women in whom the role of risk-reducing interventions should be explored. J Clin Oncol by American Society of Clinical Oncology INTRODUCTION Breast cancer is a common disorder with a significant heritable component. 1,2 Clinical services dedicated to the management of familial breast cancer risk have principally focused on the identification of families segregating rare high-penetrance breast cancer genes, such as BRCA1 and BRCA2, which are associated with the highest lifetime cancer risks. Together, the known high- and moderate-risk genes are thought to account for no more than 25% of the familial aggregation of breast cancer. 3 Consequently, the majority of diagnostic genetic tests performed in the clinical setting yield uninformative results that provide minimal assistance in the clinical management of the individual and do not contribute to an understanding of the familial breast cancer risk in the family. Major efforts have been made to explain the remaining heritable risk of breast cancer through large genome-wide association studies (GWAS) that seek to identify common variants in the genome associated with increased breast cancer risk. Individually, the single nucleotide polymorphisms (SNPs) identified have only a small effect on breast cancer risk, but collectively may account for a large component of breast cancer heritability. To date, more than 2 risk alleles have been identified in large, high-quality studies that reach the stringent standards of genome-wide significance These studies provide clear evidence that common variants have a role in the 212 by American Society of Clinical Oncology 1 Copyright 212 by American Society of Clinical Oncology

2 Sawyer et al Table 1. Patient Characteristics of 1,212 Index Cases Characteristic Mutation Negative, BRCA1 and BRCA2 Mutation Positive, BRCA1 or BRCA2 Controls Total No. of participants Age at genotyping, Deceased 43 2 Follow-up from diagnosis, Jewish ancestry, n Unclassified variants 25 5 Not tested Unilateral breast cancer Age of diagnosis, Median Range Contralateral breast cancer Synchronous Metachronous Age of first diagnosis, Median 45 4 Range Age of second diagnosis, Median Range Breast/ovary double primaries Age of first diagnosis, Median 51 5 Range Ovarian cancer Age of first diagnosis, Median Range Family history 16 Total first- to third-degree relatives Total No. 24,989 8,34 No. per family Unilateral breast cancer Total No. 1, No. per family Bilateral breast cancer Total No No. per family Male breast cancer Total No No. per family.3.6 Ovarian cancer Total No No. per family.2.57 Isolated cases Breast cancer, all relatives 2, Index cases without a diagnosis of invasive breast cancer were not analyzed further in this study. A total of 16 controls (12%) reported a family member with breast or ovarian cancer. No first- to third-degree relative with breast or ovarian cancer. Includes individuals with more distant family history or young-onset isolated cases. etiology of breast cancer, but the integration of this information into clinical practice is yet to be resolved. Studies have examined the utility of common variants for predicting personal breast cancer risk in the general population, either alone 16 or in combination with established models of personal breast cancer prediction, such as the Breast Cancer Risk Assessment Tool (Gail model) Despite using different sets of SNPs and statistical methodologies, these studies have reported only a modest improvement in the discrimination of breast cancer risk. In the general population, analysis of common variants did not meaningfully contribute to important clinical decision points, such as commencement of early screening with mammography or magnetic resonance imaging, chemoprevention, or surgical prophylaxis. These findings reflect the diversity of etiologic factors that contribute to breast cancer in the general population. In contrast, individuals affected by breast cancer who attend familial cancer clinics represent a distinct group, defined specifically by the apparent significance of hereditary factors in the etiology of their cancers. The risk-management issues are also different in this group, in whom the elevated risk of breast cancer is self-evident, and include questions about the possible utility and interpretation of genetic testing for high-penetrance genes (eg, BRCA1 and BRCA2), the continuing risk of developing a second primary breast cancer, or the potential for developing ovarian or other cancers. A No. of Patients B Probability Risk Alleles Polygenic Risk Score Controls BRCA1/2 mutation No mutation Controls BRCA1/2 mutation No mutation Fig 1. The distribution of polygenic breast cancer risk. The distribution of (A) breast cancer risk alleles and (B) the polygenic risk score in 1,143 women affected by familial breast cancer, with or without a mutation in BRCA1 or BRCA2, compared with population controls (n 893). OR, odds ratio by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

3 Common Genomic Variants in Familial Breast Cancer In this study, we examined the role of common genomic variants in familial breast cancer by genotyping an extended panel of previously described SNPs associated with breast cancer risk in a clinically and genetically characterized cohort of women attending familial cancer services. We have found evidence of a strong contribution from common risk variants in this setting and demonstrated an association with specific clinical features. The findings have implications for the genetic counseling and cancer risk management of the significant number of women affected by breast cancer with a family history. PATIENTS AND METHODS Study Population The Victorian Familial Breast Cancer Cohort consists of index cases from families referred to four familial cancer centers (FCCs) throughout Victoria, Australia, between 1997 and 21 who were found to have a high familial risk for breast cancer after clinical assessment. Personal and family histories of cancer were assessed through questionnaire and in person and were verified through local cancer registries. BRCA1 and BRCA2 mutation testing was completed (including screening of all exons and multiplex ligation-dependent probe amplification) for the index case in each family. The criteria defining high familial risk evolved over the period of assessment, in keeping with the published literature, and full details of the ascertainment for cases and controls are provided (Data Supplement). Subjects In this study, 1,143 index cases with at least one diagnosis of invasive breast cancer were analyzed, including 142 with a pathogenic mutation in BRCA1 and 116 in BRCA2. The features of the individuals in each group are shown in Table 1 and the outcome for all families enrolled is shown (Data Supplement). The study was approved by the human research ethics committees at all institutions involved. A population control group of 892 unaffected individuals was recruited by random selection from the Australian electoral roll by the Australian Ovarian Cancer Study as previously described. 21 Table 2. Characteristics of Individuals Affected by Hereditary Breast Cancer According to the Quartiles of the PRS in Women With Familial Breast Cancer (n 1,55 index cases and 863 controls with full genotypes) Characteristic Low, 1st Quartile Intermediate, 2nd-3rd Quartile High, 4th Quartile OR (high v low) 95% CI P No. of individuals Mutation negative BRCA1/2 mutation to.75.1 Controls Percentage of group Mutation negative BRCA1/2 mutation Controls Mutation-negative index cases PRS Mean relative risk Mean age at first diagnosis, NS Age to Proportion No. of patients Age to Proportion No. of patients Contralateral breast cancer to Proportion No. of patients Breast and ovarian cancer.4.11 to 1.71 NS Proportion No. of patients Family history Relatives affected by breast cancer Family history of: Bilateral breast cancer to 1.84 NS Proportion No. of patients Ovarian cancer.8.47 to 1.35 NS Proportion No. of patients Male breast cancer to 5.33 NS Proportion No. of patients Abbreviations: NS, not significant; OR, odds ratio; PRS, polygenic risk score. ORs between the high PRS and low PRS group. Two-tailed t test for means, 2 for proportions. The low- and high-risk groups contain a quarter of the cases each by definition. Mean number of first- to third-degree relatives affected by breast cancer by American Society of Clinical Oncology 3

4 Sawyer et al Genotyping Twenty-two SNPs were selected for genotyping from large, high-quality, two-stage breast cancer GWAS, 4-15,22 where they reached genome-wide significance levels (Data Supplement). Genotyping was performed on the Sequenom iplex platform (Sequenom, San Diego, CA). All variants met the required more than 95% success criteria for genotyping with no significant deviation from the Hardy-Weinberg equilibrium. Of 1,374 individuals eligible for genotyping, 112 were excluded after failing genotyping quality criteria. Analysis The association of individual SNPs with breast cancer risk was estimated, by logistic regression and P values by Cochrane-Armitage trend test. A multiplicative model for the interaction between risk variants was assumed in line with current evidence. 23,24 Individual risk was modeled using the full genotypic information by calculating a polygenic risk score (PRS), a log-additive score in which each risk allele was weighted by the log of the per-allele odds ratio (OR) measured in this cohort, standardized to an OR of 1 in the control population, as previously described. 16,18 Scores for individuals with complete genotypic information were analyzed further. The distributions of the risk alleles and the PRSs in all groups were consistent with the expected normal distribution on examination of Q-Q plots (Data Supplement). The corresponding distributions were fitted from the measured parameters, and a cross-validation analysis was used to examine the model for over-fitting. The clinical characteristics associated with intervals of the distributions were assessed and compared. Relative and absolute risks of breast cancer were calculated based on annual and lifetime breast cancer risks specific for the Australian population. 24 The rate of second primary breast cancer was assessed by Kaplan-Meier estimate curves and a Cox regression model that included the effect of risk-reducing salpingo-oophorectomy, age of first diagnosis, and birth year. Additional details of the analysis and crossvalidation are described (Data Supplement). RESULTS A total of 1,143 index cases from high-risk breast cancer families and 892 population controls were genotyped, with an overall success rate of greater than 98%. The genomic variants examined showed a strong association with breast cancer risk, both individually and collectively. The cohort was not powered to assess the effect of individual SNPs; however, the association with breast cancer reached an unadjusted significance of P.5 for 12 variants, and the remaining SNPs all showed the expected trend toward increased risk (Data Supplement). The sum of the risk alleles and the PRS, calculated from the measured log ORs for the 22 SNPs, best fitted a normal distribution in cases and population controls. Breast cancer risk, estimated by either parameter, was substantially greater in the group of women with familial breast cancer who did not harbor a BRCA1 or BRCA2 gene mutation compared with the control group, and the difference was highly significant: mean PRS.3 (95% CI,.26 to.33) versus (95% CI,.3 to.3; P ; Fig 1). More than 7% of the mutation-negative group had a PRS greater than the average for the control population. The mean PRS in the mutation-negative women was equivalent to an average (expectation) OR of The sum of the risk alleles and fitted distributions are shown in Figure 1. In comparison with the mutation-negative group, women with a mutation in the high-penetrance genes BRCA1 or BRCA2 had a lower level of polygenic risk by either measure: mean PRS.12 (95% CI,.5 to.19) versus.3 (95% CI,.26 to.33; P.1). For mutation-negative women, differences in the mean PRS were associated with diagnosis of breast cancer before age 35 (mean PRS.39 v.28; P.4) and the occurrence of contralateral breast cancer (mean PRS.4 v.28; P.2). To determine whether this corresponded to clinically important differences for a significant number of women, the cohort of mutation-negative women was divided into high- and low-risk groups consisting of the highest and lowest quartiles of the PRS (the middle two quartiles forming an intermediate-risk group). When the features of the three groups were compared, significant differences were confirmed in the frequency of early-onset and second primary breast cancers (Table 2). Compared with the low-risk group, the high polygenic risk quartile had a trend toward an increased risk of a breast cancer diagnosis before age 35 (OR 1.88; 95% CI,.99 to 3.25; P.5) that reached significance for diagnoses made before age 3 (OR 3.37; 95% CI, 1.3 to 1.26; P.3). In addition, an almost two-fold increase in the frequency of second primary breast cancers was demonstrated in the group with the highest polygenic risk score compared with the lowrisk group (OR 1.96; 95% CI, 1.9 to 3.52). In contrast, there was no difference in the occurrence of breast and ovarian double primary cancers between the groups. The occurrence of contralateral breast cancer is an important clinical and psychological consideration for women with a familial breast cancer. The rate of contralateral cancer was analyzed with the first diagnosis as a starting point for follow-up to the time of last contact, prophylactic mastectomy, death, or age 75. The rate in the highest risk women was significantly increased compared with both the low- and intermediate-risk group (log-rank P.2 and.33 respectively, Fig 2). The difference in the rate of contralateral cancers remained significant when the age of first diagnosis, birth year, and the effect of risk-reducing salpingo-oophorectomy were included as covariants in a Cox regression analysis (Data Supplement) with a hazard ratio of 2.8 (95% CI, 1.17 to 3.7) between the highest and lowest quartile. In the familial breast cancer cohort as a whole, the level of polygenic risk was inversely correlated with the chance that a mutation was present in either BRCA1 or BRCA2. Mutations were CBC (cumulative incidence) High PRS BCRA1/2 mutation Intermediate PRS Low PRS Log-rank P = Time () Fig 2. Cumulative incidence of contralateral breast cancer (CBC) from time of first diagnosis. Kaplan-Meier estimate curve for follow-up from the first diagnosis of breast cancer to age 75 or censor points ( ) of last follow-up, prophylactic mastectomy, death, or diagnosis of ovarian cancer. The rate of CBC in the combined group of BRCA1 and BRCA2 mutation carriers (gold) is shown for comparison. Analysis demonstrates significant differences between the three groups of mutation-negative women (excluding mutation carriers): log-rank P.27. (*) For the comparison between high and low polygenic risk score groups, P.2. * by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

5 Common Genomic Variants in Familial Breast Cancer Breast Cancers (proportion) At-Risk Population (proportion) Fig 3. Receiver operating characteristic curve for the predictive quality of the polygenic risk score based on 22 common genomic variants in women with familial breast cancer (negative for BRCA1 or BRCA2 mutation). Area under the curve.654 (95% CI,.628 to.68). The measured discrimination is greater than similar measures in population-based studies (.55 to.6 17,18,2 ) reflecting both the incorporation of additional variants and the selection of familial cases. found only half as frequently in the women assessed as having a high level of polygenic risk compared with the low-risk group (OR.46; 95% CI,.29 to.75). However, the probability of a BRCA1 or BRCA2 mutation, estimated on the basis of personal and family history for each individual (by the BRCAPro model 25,26 ), actually showed a small increase corresponding to the increasing level of polygenic risk, from an average probability of 17% in the low PRS group to 23% in the quartile with the highest polygenic risk (P.1). As a result, the proportion of BRCA1/2 mutations predicted in each stratum was poorly calibrated to the actual number of mutations detected, with a significant underestimation of mutations in the women with a low polygenic risk score and a corresponding overestimation of the number of mutations in women with a high level of polygenic risk (Data Supplement). The ability of the polygenic model, incorporating the measured effect of the 22 genomic variants to predict the incidence of breast cancer, was examined in the setting of a high-risk family history, in which no BRCA1 or BRCA2 mutation is present. On the basis of a population risk of breast cancer to 75 of 9%, 24 the difference in the calculated relative risk of breast cancer from the lowest to the highest quartile was a greater than 4.5-fold increase, equivalent to an average absolute life time risk of 6% in the lowest quartile to 27% in the highest quartile. The overall discrimination of the PRS for breast cancer, calculated by the area under the receiver operating characteristic, was.65 (95% CI,.63 to.68; Fig 3). DISCUSSION GWASs have been successful in identifying common variants associated with an increased risk of breast cancer, predominantly from population-based cohorts. Currently little is known about the functional basis of these associations, but it is hoped that these new insights will lead to a deeper understanding of breast cancer etiology and biology, although clinical applications arising from this direction are likely to remain some way off. Neither, to date, have the data from GWASs been found to add significantly to the currently available tools for breast cancer risk assessment in the general population. 19 Consequently, despite the extensive research effort that GWASs involve, there is currently no clear direction for the clinical translation of the outcomes of these studies. Here, we demonstrate for the first time the potential for direct clinical application of the results of the breast cancer GWAS by applying a new approach, focusing on families clustering the cancer of interest. We have shown that polygenic risk, defined by the common genomic variants identified in GWAS, is a major contributor to both incidence and clinical characteristics in a substantial subgroup of the breast cancers that occur in this context. The magnitude of the effect of these genomic variants observed in this setting was greater than described in previous, population-based analyses. 16,2 An enhanced effect of the risk variants in the familial setting has been suggested in A Genetic testing: BRCA1/BRCA2 2%* 8% BRCA1 or 2 mutation Familial breast cancer Uninformative result B 2%* BRCA1 or 2 mutation Familial breast cancer Genetic testing: BRCA1/BRCA2 + common variants 2%* 2%* No BRCA1/BRCA2 mutation High PRS Intermediate PRS Risk management issues: CBC - - Ov Ca AD risk Low PRS Fig 4. Proposed model for the integration of common variants into genetic testing for hereditary breast cancer. (A) In the model of current practice, genetic testing identifies a high penetrance mutation in approximately one in five women. (B) The addition of information on common genomic variants has the potential to subdivide the group of mutation-negative women, particularly with respect to the ongoing risk of contralateral breast cancer (CBC) and the residual chance of a dominantly acting highpenetrance gene mutation. AD risk, the residual likelihood of a dominantly segregating high penetrance mutation; Ov Ca, risk of epithelial ovarian cancer; PRS, polygenic risk score. (1) Increased risk; ( ), risk unknown or no evidence of an increased risk; ( ) certain; (2) decreased risk; (*) approximate values; ( ) in the absence of a family history to suggest an independent risk of ovarian cancer by American Society of Clinical Oncology 5

6 Sawyer et al other studies based on familial cohorts 8,27 and reflects the fact that the cases involved are highly selected to ensure the importance of hereditary factors in the etiology of their cancers. As a result, the 22 variants genotyped would together account for 18.5% of the variance of breast cancer risk in this cohort compared with 1.2% by the same estimate based on the published ORs, most of which are derived from population-based cohorts. 16 Our findings suggest that the assessment of polygenic risk can help to distinguish a genetically defined group of women with high familial risk who have a characteristic set of clinical features increased frequency of early-onset cancers, approximately two-fold increase in the rate of contralateral breast cancer diagnosis, less than half the risk of a high-penetrance mutation in BRCA1 or BRCA2, and no excess of ovarian cancer. Together these features suggest a model for the interpretation of genomic variants in the familial cancer setting that can be used as a framework for further evaluation through clinical research (Fig 4). Critically, the information from common variants showed the capacity to identify individuals among the women without a BRCA1 or BRCA2 mutation who continue to have a high risk of breast cancer. No mutation is found for the majority of individuals with a family history who undergo genetic testing, and advice about risk management in this group is currently based on limited information and tends to be nonspecific in nature. Follow-up studies in this setting report a low uptake of both medical and surgical interventions. 28,29 Determining the future cancer risks more accurately for even a proportion of women who do not have a BRCA1 or BRCA2 mutation has the potential to improve significantly the risk management decisions of a considerable number of women. In this study, the same information could not be derived from features in the clinical or family history. The excess of risk alleles was not seen to the same degree in individuals in whom a BRCA1 or BRCA2 mutation was identified and in whom the high-risk mutation would be expected to be the major causative factor underlying their diagnosis. Despite this finding, the overlap in the PRS between the groups with apparently separate genetic etiologies remains considerable, and the mean PRS in BRCA mutation carriers affected by breast cancer was also significantly higher than that seen in the unaffected controls (P.17), which is consistent with the evidence that many of the same variants act as modifiers of the risk associated with mutations at the BRCA1 or BRCA2 loci. 3,31 In this study, the difference between the high polygenic risk and rare, high-penetrance mutation groups may be underestimated, as undetected mutations in rarer genes associated with breast cancer such as TP53, PTEN, PALB2, ATM, or STK11 are likely to be present in some individuals, potentially with a corresponding distribution of common risk variants similar to the BRCA1 and BRCA2 carriers. Further investigation is needed to explore whether the distinction between polygenic and monogenic etiologies can be exploited to help identify the individuals and families most likely to harbor a mutation in a new, high-penetrance gene. Individuals at the low end of the scale of polygenic risk would then be candidates for more extensive genetic screening through technologies such as tailored gene arrays or whole-exome sequencing. A limitation of this study is that only the genotype of the index case from each family was analyzed. As a result, features such as contralateral breast cancers, early onset cases, and the extent of family history are all affected by the ascertainment; however, this effect applies equally to all groups within the cohort. Analyzing a single case from each family avoids the complex problem of the interdependence of polygenic risk between related individuals, but the extent to which the observed excess of risk alleles contributes to the history in family members cannot be assessed. In our cohort, polygenic risk in the index case was not a surrogate measure of the extent of the family history, with no significant correlation found between the level of polygenic risk and family history parameters such as the number of breast cancers or the age of onset in family members. Despite extensive evidence of the importance of common genomic variants to individual breast cancer risk, application of these insights to improve screening and management remains a challenge. Analysis of risk alleles in the general population suggests that the contribution of common variants to the assessment of risk is likely to be important for only a small proportion of women at the extreme ends of the distribution We describe an alternative approach in which we have selected a group of individuals with strong indicators of a hereditary element to their diagnosis of breast cancer and have found that in a group of patients who make up a large component of the referrals to familial cancer clinics and cancer genetics services, polygenic risk is a major predisposing factor in their diagnosis. Although debate continues around the utility of this information in the general population, our study suggests that common risk variants have potential for direct application in the setting of a positive family history. AUTHORS DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST The author(s) indicated no potential conflicts of interest. AUTHOR CONTRIBUTIONS Conception and design: Gillian Mitchell, Geoffrey J. Lindeman, Paul A. James Provision of study materials or patients: Gillian Mitchell, Georgia Chenevix-Trench, David Bowtell, Marion Harris, Geoffrey J. Lindeman, Paul A. James Collection and assembly of data: Sarah Sawyer, Gillian Mitchell, Joanne McKinley, Georgia Chenevix-Trench, Jonathan Beesley, Xiao Qing Chen, David Bowtell, Marion Harris, Geoffrey J. Lindeman, Paul A. James Data analysis and interpretation: Sarah Sawyer, Gillian Mitchell, Georgia Chenevix-Trench, Jonathan Beesley, Alison H. Trainer, Marion Harris, Geoffrey J. Lindeman, Paul A. James Manuscript writing: All authors Final approval of manuscript: All authors REFERENCES 1. Collaborative Group on Hormonal Factors in Breast Cancer: Familial breast cancer: Collaborative reanalysis of individual data from 52 epidemiological studies including 58,29 women with breast cancer and 11,986 women without the disease. Lancet 358: , Lichtenstein P, Holm NV, Verkasalo PK, et al: Environmental and heritable factors in the causation of cancer: Analyses of cohorts of twins from Sweden, Denmark, and Finland. N Engl J Med 343:78-85, 2 3. Thompson D, Easton D: The genetic epidemiology of breast cancer genes. J Mammary Gland Biol Neoplasia 9: , Ahmed S, Thomas G, Ghoussaini M, et al: Newly discovered breast cancer susceptibility loci on 3p24 and 17q23.2. Nat Genet 41:585-59, by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

7 Common Genomic Variants in Familial Breast Cancer 5. Cox A, Dunning AM, Garcia-Closas M, et al: A common coding variant in CASP8 is associated with breast cancer risk. Nat Genet 39: , Dunning AM, Healey CS, Baynes C, et al: Association of ESR1 gene tagging SNPs with breast cancer risk. Hum Mol Genet 18: , Easton DF, Pooley KA, Dunning AM, et al: Genome-wide association study identifies novel breast cancer susceptibility loci. Nature 447: , Turnbull C, Ahmed S, Morrison J, et al: Genome-wide association study identifies five new breast cancer susceptibility loci. Nat Genet 42:54-57, Antoniou AC, Wang X, Fredericksen ZS, et al: A locus on 19p13 modifies risk of breast cancer in BRCA1 mutation carriers and is associated with hormone receptor-negative breast cancer in the general population. Nat Genet 42: , Fletcher O, Johnson N, Orr N, et al: Novel breast cancer susceptibility locus at 9q31.2: Results of a genome-wide association study. J Natl Cancer Inst 13: , Sehrawat B, Sridharan M, Ghosh S, et al: Potential novel candidate polymorphisms identified in genome-wide association study for breast cancer susceptibility. Hum Genet 13: , Stacey SN, Manolescu A, Sulem P, et al: Common variants on chromosomes 2q35 and 16q12 confer susceptibility to estrogen receptorpositive breast cancer. Nat Genet 39: , Stacey SN, Manolescu A, Sulem P, et al: Common variants on chromosome 5p12 confer susceptibility to estrogen receptor-positive breast cancer. Nat Genet 4:73-76, Thomas G, Jacobs KB, Kraft P, et al: A multistage genome-wide association study in breast cancer identifies two new risk alleles at 1p11.2 and 14q24.1 (RAD51L1). Nat Genet 41: , Ghoussaini M, Fletcher O, Michailidou K, et al: Genome-wide association analysis identifies three new breast cancer susceptibility loci. Nat Genet 44: , Pharoah PD, Antoniou AC, Easton DF, et al: Polygenes, risk prediction, and targeted prevention of breast cancer. N Engl J Med 358: , Comen E, Balistreri L, Gönen M, et al: Discriminatory accuracy and potential clinical utility of genomic profiling for breast cancer risk in BRCAnegative women. Breast Cancer Res Treat 127: , Gail MH: Discriminatory accuracy from singlenucleotide polymorphisms in models to predict breast cancer risk. J Natl Cancer Inst 1: , Gail MH: Value of adding single-nucleotide polymorphism genotypes to a breast cancer risk model. J Natl Cancer Inst 11: , Wacholder S, Hartge P, Prentice R, et al: Performance of common genetic variants in breastcancer risk models. N Engl J Med 362: , Merritt MA, Green AC, Nagle CM, et al: Talcum powder, chronic pelvic inflammation and NSAIDs in relation to risk of epithelial ovarian cancer. Int J Cancer 122:17-176, Milne RL, Benítez J, Nevanlinna H, et al: Risk of estrogen receptor-positive and -negative breast cancer and single-nucleotide polymorphism 2q35- rs J Natl Cancer Inst 11: , Antoniou AC, Cunningham AP, Peto J, et al: The BOADICEA model of genetic susceptibility to breast and ovarian cancers: Updates and extensions. Br J Cancer 98: , Australian Institute of Health and Welfare: Australian Cancer Incidence and Mortality Books, ed 21. Canberra, Australian Capital Territory Australia, AIHW, Berry DA, Iversen ES Jr, Gudbjartsson DF, et al: BRCAPRO validation, sensitivity of genetic testing of BRCA1/BRCA2, and prevalence of other breast cancer susceptibility genes. J Clin Oncol 2: , Chen S, Wang W, Broman KW, et al: Bayes Mendel: An R environment for Mendelian risk prediction. Stat Appl Genet Mol Biol 3:Article21, Latif A, Hadfield KD, Roberts SA, et al: Breast cancer susceptibility variants alter risks in familial disease. J Med Genet 47: , Schwartz MD, Isaacs C, Graves KD, et al: Long-term outcomes of BRCA1/BRCA2 testing: Risk reduction and surveillance. Cancer 118:51-517, Morgan D, Sylvester H, Lucas FL, et al: Cancer prevention and screening practices among women at risk for hereditary breast and ovarian cancer after genetic counseling in the community setting. Fam Cancer 8: , Antoniou AC, Beesley J, McGuffog L, et al: Common breast cancer susceptibility alleles and the risk of breast cancer for BRCA1 and BRCA2 mutation carriers: Implications for risk prediction. Cancer Res 7: , Wang X, Pankratz VS, Fredericksen Z, et al: Common variants associated with breast cancer in genome-wide association studies are modifiers of breast cancer risk in BRCA1 and BRCA2 mutation carriers. Hum Mol Genet 19: , 21 Acknowledgment We thank Rebecca Driessen and Antonis Antoniou for their contributions by American Society of Clinical Oncology 7