Breast and Prostate Cancer Risks for Male BRCA1 and BRCA2 Pathogenic Variant Carriers Using Polygenic Risk Scores

Abstract Background Recent population-based female breast cancer and prostate cancer polygenic risk scores (PRS) have been developed. We assessed the associations of these PRS with breast and prostate cancer risks for male BRCA1 and BRCA2 pathogenic variant carriers. Methods 483 BRCA1 and 1318 BRCA2 European ancestry male carriers were available from the Consortium of Investigators of Modifiers of BRCA1/2 (CIMBA). A 147-single nucleotide polymorphism (SNP) prostate cancer PRS (PRSPC) and a 313-SNP breast cancer PRS were evaluated. There were 3 versions of the breast cancer PRS, optimized to predict overall (PRSBC), estrogen receptor (ER)–negative (PRSER-), or ER-positive (PRSER+) breast cancer risk. Results PRSER+ yielded the strongest association with breast cancer risk. The odds ratios (ORs) per PRSER+ standard deviation estimates were 1.40 (95% confidence interval [CI] =1.07 to 1.83) for BRCA1 and 1.33 (95% CI = 1.16 to 1.52) for BRCA2 carriers. PRSPC was associated with prostate cancer risk for BRCA1 (OR = 1.73, 95% CI = 1.28 to 2.33) and BRCA2 (OR = 1.60, 95% CI = 1.34 to 1.91) carriers. The estimated breast cancer odds ratios were larger after adjusting for female relative breast cancer family history. By age 85 years, for BRCA2 carriers, the breast cancer risk varied from 7.7% to 18.4% and prostate cancer risk from 34.1% to 87.6% between the 5th and 95th percentiles of the PRS distributions. Conclusions Population-based prostate and female breast cancer PRS are associated with a wide range of absolute breast and prostate cancer risks for male BRCA1 and BRCA2 carriers. These findings warrant further investigation aimed at providing personalized cancer risks for male carriers and informing clinical management.


Abstract
Background: Recent population-based female breast cancer and prostate cancer polygenic risk scores (PRS) have been developed. We assessed the associations of these PRS with breast and prostate cancer risks for male BRCA1 and BRCA2 pathogenic variant carriers. Methods: 483 BRCA1 and 1318 BRCA2 European ancestry male carriers were available from the Consortium of Investigators of Modifiers of BRCA1/2 (CIMBA). A 147-single nucleotide polymorphism (SNP) prostate cancer PRS (PRS PC ) and a 313-SNP breast cancer PRS were evaluated. There were 3 versions of the breast cancer PRS, optimized to predict overall (PRS BC ), estrogen receptor (ER)-negative (PRS ER-), or ER-positive (PRS ERþ ) breast cancer risk. Results: PRS ERþ yielded the strongest association with breast cancer risk. The odds ratios (ORs) per PRS ERþ standard deviation estimates were 1.40 (95% confidence interval [CI] ¼1.07 to 1.83) for BRCA1 and 1.33 (95% CI ¼ 1.16 to 1.52) for BRCA2 carriers. PRS PC was associated with prostate cancer risk for BRCA1 (OR ¼ 1.73, 95% CI ¼ 1.28 to 2.33) and BRCA2 (OR ¼ 1.60, 95% CI ¼ 1.34 to 1.91) carriers. The estimated breast cancer odds ratios were larger after adjusting for female relative breast cancer family history. By age 85 years, for BRCA2 carriers, the breast cancer risk varied from 7.7% to 18.4% and prostate cancer risk from 34.1% to 87.6% between the 5th and 95th percentiles of the PRS distributions. Conclusions: Population-based prostate and female breast cancer PRS are associated with a wide range of absolute breast and prostate cancer risks for male BRCA1 and BRCA2 carriers. These findings warrant further investigation aimed at providing personalized cancer risks for male carriers and informing clinical management. BRCA1 and BRCA2 pathogenic variants are associated with increased male breast cancer and prostate cancer risks (1)(2)(3)(4). A recent prospective study estimated the lifetime risk of developing prostate cancer to be 29% for BRCA1 and 60% for BRCA2 carriers (5). The risks of developing male breast cancer compared with the general population have been estimated to be 15-to 18-fold higher for BRCA1 and 80-fold higher for BRCA2 carriers (6,7). Up to 1 in 10 male BRCA2 carriers develops breast cancer (8)(9)(10)(11)(12) and displays potentially more aggressive disease relative to sporadic cases (8,12,13).
Polygenic risk scores (PRS) that combine the effects of multiple disease-associated single nucleotide polymorphisms (SNPs) provide marked cancer risk stratification in the general population (14,15) and BRCA1 and BRCA2 carriers (16)(17)(18). Our previous findings suggested the joint effects of PRS and BRCA1 and BRCA2 pathogenic variants may identify men at clinically meaningful breast and prostate cancer risk levels (17). Recent studies have identified additional breast and prostate cancer susceptibility variants (15,19,20) and have refined PRS for these cancers (15,21).
The Breast Cancer Association Consortium recently developed and validated a 313-SNP PRS in European ancestry women, which was further optimized to predict estrogen receptor (ER)specific disease (21). The estimated per standard deviation odds ratio (OR) for the most predictive (ER-positive) PRS was 1.68 (95% confidence interval [CI] ¼ 1.63 to 1.73) (21). A recent evaluation of this PRS in unselected male breast cancer cases showed similar associations with breast cancer risk in men (22). The most recent prostate cancer PRS was developed using 147-SNPs associated with prostate cancer risk in European-ancestry men from the general population (15). The estimated per standard deviation odds ratio for the prostate cancer PRS was 1.86 (95% CI ¼ 1.83 to 1.89) (15).
Male BRCA1 and BRCA2 carriers are likely to benefit from more personalized breast and prostate cancer risk estimates (23). Investigating the extent to which these PRS modify cancer risks may lead to more precise and gender-specific cancer risk assessment and could assist in optimizing cancer screening.
Here, we assessed the associations of the newly developed 313-SNP breast cancer PRS and 147-SNP prostate cancer PRS derived using population-based data, with breast and prostate cancer risks, respectively, for male BRCA1 and BRCA2 carriers. We investigated whether cancer family history influences the associations and if breast cancer associations differed by ER status or tumor grade. Furthermore, we assessed whether associations vary by age or BRCA1 and BRCA2 pathogenic variant characteristics (location; functional effect). We used the results to estimate age-specific absolute risks of developing breast and prostate cancers for male carriers by PRS distribution percentiles.

Methods
Statistical analyses were performed using R-3.6.3 (R Foundation for Statistical Computing, Vienna, Austria) (commands can be found in the Supplementary Methods, available online).

Study Participants and Genotyping
Male BRCA1 and BRCA2 pathogenic variant carriers were recruited through 40 studies from 19 countries participating in the Consortium of Investigators of Modifiers of BRCA1 and BRCA2 (CIMBA) (24). The majority of male carriers were ascertained through families attending cancer genetic clinics (96.9%; Supplementary Tables 1 and 2, available online). In this setting, individuals are referred to clinical genetics because of strong family or personal cancer history. The first individual in a family, screened for mutations, tends to be an affected individual diagnosed at a young age, most often a female relative with a young age at breast cancer diagnosis (24). When a pathogenic variant is identified, then other family members are tested for the same variant. All participants were aged 18 years or older and provided written informed consent. All studies were approved by local ethical review committees. A total of 1989 male BRCA1 and BRCA2 carriers of European ancestry were included in the present study, by selecting all available men with a breast or prostate cancer diagnosis and matched controls. Details of matching, genotyping, and quality control processes have been described previously (17) and in Supplementary Table 2 (available online).
Data collected included breast or prostate cancer diagnoses; age at diagnosis or interview; prostate cancer Gleason score; breast cancer ER status and grade; and family history of prostate, male breast, and female breast cancers among first-and second-degree relatives. BRCA1 and BRCA2 pathogenic variants (detailed pathogenicity description: http://cimba.ccge.medschl. cam.ac.uk/files/CIMBA_Mutation_Classification_guidelines_ May16.pdf) were categorized according to their known or predicted effect on protein function: class I included loss-offunction variants expected to yield unstable or no protein; class II included variants likely to produce stable mutant proteins (25). Pathology data were obtained from pathology reviews; medical, pathology or tumor registry records; or immunohistochemical staining of tissue microarrays (26).

Polygenic Risk Scores
PRS were constructed as the weighted sums of alleles (Supplementary Methods, available online) for 313-SNPs for breast cancer (21) and 147-SNPs for prostate cancer (15) (Supplementary Tables 3 and 4, available online). Three breast cancer PRS were evaluated, optimized to predict overall (PRS BC ), ER-negative (PRS ER-), and ER-positive (PRS ERþ ) breast cancer (21). These PRS were scaled to the female population-based control PRS standard deviation (21). The prostate cancer PRS (PRS PC ) was scaled to the standard deviation calculated from population-based controls (15).

Associations Between PRS and Cancer Risks
PRS associations with breast and prostate cancer risks were assessed simultaneously using multinomial logistic regression to estimate per standard deviation odds ratios. Men without breast or prostate cancer diagnoses were considered controls. Breast and prostate cancer cases were defined by considering the first occurring cancer. Instances in which breast and prostate cancers were diagnosed simultaneously were considered as breast cancer cases. Statistical models were adjusted for 3 ancestry informative principal components (proxy adjustment for study and/or country , as a direct adjustment would result in too few controls and cases within each study and/or country; Supplementary Table 1, available online) and age. Models using the combined sample of carriers were adjusted for BRCA1 and BRCA2 status. To account for relatedness, we estimated robust variances by clustering on family membership (27,28). The primary analyses assumed a continuous PRS. Categorical PRS associations were evaluated using the quartiles of the PRS distributions in the combined BRCA1 and BRCA2 carrier controls.
Because the distribution of tumor ER status in male carriers may differ from the distributions in the general population (26), we assessed the associations between all 3 versions of the breast cancer PRS with overall breast cancer risk and ER-specific disease. Associations with ER-positive and ER-negative breast cancer were assessed simultaneously by considering ER negative, ER positive, or unknown as distinct multinomial outcomes. We also assessed the associations with breast cancer gradespecific risk by considering grade 1, grade 2, grade 3, or unknown grade as separate multinomial outcomes. A case-only logistic regression also was undertaken that considered grades 1 and 2 as controls and grade 3 as cases.
To assess the PRS PC association with disease aggressiveness, we partitioned prostate cancers into those with Gleason scores less than 7, 7 or greater, or unknown, and these were used as distinct multinomial outcomes. A case-only logistic regression assessed differences in the associations with Gleason scores less than 7 (controls) and Gleason scores of 7 or greater (cases).
Discriminatory ability of each PRS was assessed by calculating the area under the receiver operator characteristic curve (AUC). Under the sampling design, the majority of male carriers were identified through clinical genetics. Therefore, the majority of both affected and unaffected carriers are expected to have family history of cancer. To determine whether this introduces any biases in the PRS associations, we fitted models that were adjusted for family history in first-and second-degree relatives.
To determine whether PRS associations varied by age (continuous), pathogenic variant location, or pathogenic variant effects on protein function (class I or class II variants), we estimated interaction terms between these factors with the PRS, and statistical significance was assessed using likelihood ratio tests (LRT). Pathogenic variants were categorized based on previously reported nucleotide position differences in breast and ovarian, or prostate cancer risks (29-31).
We undertook a sensitivity analysis to test for PRS heterogeneity across study countries (Supplementary Methods, available online).
All statistical tests were 2-sided, and a P value of less than .05 was considered statistically significant.

Predicted Age-Specific Absolute and 10-Year Cancer Risks by PRS
We predicted absolute risks up to age 85 years and 10-year risks of developing breast and prostate cancers by PRS distribution percentiles, assuming the estimated PRS odds ratio follows a log-linear model across the entire PRS range (Supplementary Methods, available online) (32).

Study Participants and Genotyping
After quality control, the analyses included 483 BRCA1 (

Associations With Breast Cancer Risk
The associations between the breast cancer PRS and male breast cancer risk for carriers are shown in Table 1 and Supplementary Tables 5 and 6 (available online). The PRS ERþ yielded the strongest associations with overall breast cancer risk for BRCA1 (OR ¼ 1.40, 95% CI ¼ 1.07 to 1.83) and BRCA2 (OR ¼ 1.33, 95% CI ¼ 1.16 to 1.52) carriers. The PRS BC resulted in nearly identical associations as the PRS ERþ . There was no statistically significant evidence that the PRS ERþ associations differed by country (P heterogeneity .48; Supplementary Figure 3, available online). In the joint analysis of BRCA1 and BRCA2 carriers, men in the uppermost PRS ERþ quartile had approximately twofold increased breast cancer risk (OR ¼ 2.10, 95% CI ¼ 1.43 to 3.08) compared with men in the lowest quartile (Supplementary Table 6, available online).
Most breast cancers among the male carriers were ER positive (95.7%). The odds ratio for the association between the PRS ERþ and ER-positive breast cancer risk for BRCA1 carriers (OR ¼ 1.79, 95% CI ¼ 1.30 to 2.48; Table 1) was somewhat higher compared with the odds ratio for overall breast cancer. The number of ER-negative cancers was too small to assess associations with ER-negative disease. There was no statistically significant evidence for differences in the associations of any of the PRS by grade (Table 1; Supplementary Table 6, available online).
The ability of PRS ERþ to discriminate between controls and breast cancer cases was estimated as an AUC of 0.60 (95% CI ¼ 0.51 to 0.69) for BRCA1 and 0.59 (95% CI ¼ 0.55 to 0.63) for BRCA2 carriers.

Associations With Prostate Cancer Risk
The estimated associations between the PRS PC and prostate cancer risk for male carriers are reported in Table 2 and Supplementary Tables 5 and 7 (available online). The odds ratios per PRS PC standard deviation were estimated to be 1.73 (95% CI ¼ 1.28 to 2.33) for BRCA1 and 1.60 (95% CI ¼ 1.34 to 1.91) for BRCA2 carriers. There was no statistically significant evidence that the PRS PC associations differed by country (P heterogeneity .14; Supplementary Figure 4, available online). In the joint analysis of BRCA1 and BRCA2 carriers, men in the top PRS PC quartile had a prostate cancer odds ratio of 3.35 (95% CI ¼ 2.06 to 5.42) compared with men in the lowest quartile (Supplementary Table 7, available online).

Adjusting for Cancer Family History
Adjusting for family history of male breast cancer did not influence the PRS ERþ associations with breast cancer risk (Table 1;  Supplementary Table 8, available online). However, the odds ratio estimates were somewhat larger when adjusting for female breast cancer family history (Table 1; Supplementary Table 9, available online).
The associations of PRS PC with prostate cancer risk remained similar after adjusting for prostate cancer family history ( Table 2; Supplementary Table 10, available online).

PRS Interactions With Age and Gene Pathogenic Variants Characteristics
There was little evidence for odds ratio estimate variability with age, for both the breast and prostate cancer PRS (P LRT .43; Table 3).
The PRS ERþ and PRS PC odds ratios with breast or prostate cancer risks appeared to be larger for class II variant (pathogenic variants likely to yield stable mutant proteins) carriers compared with class I BRCA1 and BRCA2 variant carriers (Table 3). However, these differences were not statistically significant (P LRT .26).
There was no statistically significant evidence that the PRS ERþ (P LRT .61) or PRS PC (P LRT ¼ .52) associations differed by the pathogenic variant location in the gene (Table 3).

Absolute Risks of Developing Breast and Prostate Cancer
The absolute risks of developing breast cancer by age 85 years for BRCA2 carriers was predicted to be 7.7% at the 5th and 18.4% at the 95th PRS ERþ distribution percentiles ( Figure 1). The 10year risks of developing breast cancer at 50 years were 0.8% at the 5th and 2.0% at the 95th PRS ERþ distribution percentiles for BRCA2 carriers (Figure 2). The corresponding risks at age 75 years were 3.7% and 9.3%, respectively.
The predicted absolute risks of developing prostate cancer by age 85 years were 13.1% at the 5th and 50.4% at the 95th PRS PC distribution percentiles for BRCA1 carriers (Figure 1). The corresponding risks for BRCA2 carriers were 34.1% and 87.6%. BRCA2 carriers had 10-year risks of 2.1% and 10.1% at the 5th and 95th PRS PC percentiles at age 50 years, respectively. The corresponding risks at age 75 years were 25.5% and 77.0% ( Figure 2).

Discussion
We evaluated the associations of the most recently developed breast and prostate cancer PRS with site-specific cancer risks in the largest case-control study of male BRCA1 and BRCA2 carriers available to date. Our findings showed that these PRS, developed using population-based data, are associated with breast and prostate cancer risks for male BRCA1 and BRCA2 carriers. Despite the modest estimated AUCs, our results demonstrate that because male carriers are already at elevated risks of developing breast and prostate cancers, these PRS can lead to large differences in the absolute cancer risks for carriers across PRS percentiles. The breast cancer grade specific odds ratios were estimated by partitioning breast cancer status into multinomial outcomes for grade 1, grade 2, grade 3, or grade unknown. e The case-only breast cancer grade analysis was a logistic regression considering grade 1 and grade 2 breast cancers combined as controls and grade 3 breast cancers as cases.
f The ER-specific breast cancer odds ratios were estimated by partitioning breast cancer status into distinct multinomial outcomes for ER negative, ER positive, or ER status unknown. g Grade 1 and grade 2 combined for BRCA1 carriers (to ensure adequate sample size to estimate associations). PRS BC and PRS ERþ were associated with larger odds ratio estimates than PRS ER-in predicting breast cancer risk, consistent with the fact that most breast cancers in men are ER positive, including those harboring BRCA1 and BRCA2 pathogenic variants (26). Similarly, when assessing associations with ERpositive breast cancer risk, PRS BC and PRS ERþ showed the strongest associations for BRCA1 and BRCA2 carriers. There were no differences in PRS associations by breast cancer grade.
The 147-SNP PRS PC (15) yielded larger per standard deviation odds ratio estimates than a previously evaluated 103-SNP prostate cancer PRS (17). There was some evidence that PRS PC may be associated with a higher odds ratio for more aggressive disease (Gleason score 7) for BRCA1 carriers. This pattern was not observed for BRCA2 carriers, who tend to develop more aggressive disease (5). If this finding is replicated by larger studies, the PRS may prove to be useful in cancer prevention and surveillance by identifying BRCA1 carriers at greater risk of developing aggressive prostate cancers. PRS associations with breast or prostate cancer risk, adjusted for family history of male breast or prostate cancer, were similar to unadjusted estimates, suggesting that cancer family history in male relatives does not alter PRS associations. Adjusting for family history of female breast cancer resulted in somewhat larger odds ratio estimates for the breast cancer PRS compared with unadjusted estimates. This observation is consistent with male carriers being identified and recruited into our studies mostly based on their female relatives' breast cancers.
There was little evidence supporting variability in PRS associations by age or pathogenic variant characteristics. However, larger sample sizes are required to reliably assess such differences, and the current analyses were likely underpowered.
Previous studies (18,33) suggest the magnitude of the breast cancer PRS associations is attenuated in female BRCA1 and BRCA2 carriers compared with associations seen in the general population (21). As seen for female carriers, the estimated breast cancer odds ratios for male carriers were attenuated compared with estimates for women in the general population (21). Similarly, the estimated prostate cancer odds ratio estimate for male carriers was attenuated compared with population-based data (15). Taken together, these observations suggest there is a deviation from the multiplicative model for the joint effects of BRCA1 and BRCA2 pathogenic variants and the PRS for male and female carriers. These observed attenuations for BRCA1 and BRCA2 carriers are unlikely to be an overestimation of the effects in the general population ["winner's curse" (34)], as they have been validated in independent prospective cohorts (21). The lower odds ratios for the breast and prostate cancer PRS in male BRCA1 and BRCA2 carriers, compared with the general population, may reflect a general attenuation of the effect sizes of common variants on genetic risk in the presence of a pathogenic variant in a high-risk gene (35,36). This supposition may also explain the larger PRS odds ratios for BRCA1 carriers, who are at lower risk compared with BRCA2 carriers (37). However, given the current study design, we cannot rule out that the observed attenuations in effect size are related to ascertainment biases. Although adjusting for family history did not change the odds ratio estimates substantially, residual confounding may still remain. Large-scale population studies will be required to address this. If the attenuations in the PRS effect size are real, they would result in a smaller range of cancer risks for BRCA1 and BRCA2 carriers compared with using the PRS effect sizes estimated from general population data.
Although breast cancer risk stratification might not currently be feasible for men in the general population, male BRCA1 and BRCA2 carriers may represent a group likely to benefit from a more refined stratification of their individual breast and prostate cancer risks, to better inform their clinical management. At present, limited recommendations based on lowlevel evidence or expert opinion are available for male carriers. Current guidelines recommend clinical breast examinations beginning at ages 30-35 years and suggest mammographic screening on an individual basis, whereas clinical prostate cancer screening, particularly for BRCA2 carriers, is recommended from ages 40 to 45 years (38)(39)(40).
The PRS percentile-specific absolute risks varied substantially over the PRS distribution, consistent with previous studies in male (17) and female (16,18) BRCA1 and BRCA2 carriers. At least twofold increased risk is often considered a clinically The Gleason score prostate cancer odds ratios were estimated by partitioning prostate cancer status into distinct multinomial outcomes for GS < 7, GS 7, or GS unknown. e The case-only prostate cancer analysis was a logistic regression considering GS < 7 prostate cancers as "controls" and GS 7 prostate cancers as "cases".
actionable level for breast and prostate cancers (41). Our findings may inform the development of age-specific clinical recommendations and provide guidance on when to start riskadapted screening, based on their PRS percentile-specific 10- year risks. Overall, refined risk estimates may be useful to distinguish male carriers at higher risk, who may benefit from enhanced and/or earlier screening, and identify carriers at lower risk, who may opt for more limited or postponed surveillance. The associations with breast cancer risk are reported for the ER-positive breast cancer PRS (PRS ERþ ). CI ¼ confidence interval; OCCR ¼ ovarian cancer cluster region; OR ¼ odds ratio per PRS standard deviation, estimated from a multinomial logistic regression; PCCR ¼ prostate cancer cluster region; PRS ¼ polygenic risk score; NA ¼ not applicable. b P value was calculated using a 2-sided Wald test, unless otherwise indicated. c The PRS term is applicable at age 0 years and the PRS x age interaction term is a per-year effect. Age in years. d P values were calculated using a 2-sided likelihood ratio test. The likelihood ratio test compared the model that estimated the interaction term with a nested model that omitted the interaction term. e Class I pathogenic variant ¼ loss-of-function pathogenic variants expected to result in unstable or no protein; class II pathogenic variant ¼ pathogenic variants likely to yield stable mutant proteins.   Identification of men at lower risk of prostate cancer by PRS stratification has been shown to be useful in reducing overdiagnosis in the general population, resulting in a reduction in the harms associated with prostate-specific antigen testing (42). Similar arguments may apply to male carriers in whom PRS prediction may further improve screening efficacy. Strengths of this study include the fact that this is an independent validation of the most recently derived breast (21) and prostate (15) cancer PRS derived from population-based data. We benefited from the availability of Gleason scores and breast cancer ER status and grade; hence, we could assess subtypespecific associations. Finally, we assumed recent prospectively estimated prostate cancer incidence rates (5) to predict absolute prostate cancer risks, which may be more representative of risks for carriers currently seen in clinical genetics centers.
Study limitations include the limited sample size to assess PRS associations with cancer risks for subgroups of male carriers. However, these data remain the largest male BRCA1 and BRCA2 carrier case-control study with available genotype data. The breast (21) and prostate (15) cancer PRS do not include male breast cancer-specific risk-associated SNPs or SNPs that may specifically be associated with prostate cancer risk for carriers. If such SNPs exist, further improvement may be gained in risk prediction by including them in PRS. The absolute risk calculations assumed that the PRS odds ratio behaves log linearly over the PRS range. It was difficult to evaluate this assumption in the present analyses because of the limited sample size of male carriers. However, empirical evidence based on larger sample sizes of female carriers (18) or in the general population (15,21) suggests that this assumption is plausible. Additionally, the absolute breast and prostate cancer risk predictions by PRS will require validation in large prospective studies of male carriers with long-term follow-up, although such studies remain a challenge. Finally, the PRS that we investigated were derived using European ancestry data; hence, our estimated associations and predicted risks may not be applicable to non-European ancestry carriers.
PRS are now used in cancer risk-stratified screening trials and implementation studies in the general population (43)(44)(45)(46)(47). They are commercially available and are used in multifactorial cancer-risk prediction models for women (48,49). We found that PRS derived from population-based data are associated with breast and prostate cancer risks and lead to meaningful risk stratification for male carriers. These findings may potentially be used to provide more personalized cancer risk predictions and therefore assist clinical management decisions. Future implementation studies should determine if optimal strategies exist for incorporating these PRS into genetic counseling and risk assessment to clarify whether they can influence the clinical management decisions of male BRCA1 or BRCA2 carriers.

Funding
The CIMBA data management and data analysis were supported by Cancer Research UK grants C12292/A20861 and PPRPGM-Nov20\100002. The research leading to these results has received funding from the Italian Association for     Disclaimers: The content of this manuscript does not necessarily reflect the views or policies of the National Cancer Institute or any of the collaborating centers in the Breast Cancer Family Registry (BCFR), nor does mention of trade names, commercial products, or organizations imply endorsement by the US government or the BCFR. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Data Availability
The complete dataset is not publicly available because of restraints imposed by the ethical committees of individual studies. Requests to access the complete dataset, which is subject to General Data Protection Regulation (GDPR) rules, can be made to the Data Access Coordinating Committee (DACC) of CIMBA, following the process described on the CIMBA website (http://cimba.ccge.medschl.cam.ac.uk/projects/data-accessrequests/). Submitted applications are reviewed by the CIMBA DACC every 3 months.