Prostate Cancer Risks for Male BRCA1 and BRCA2 Mutation Carriers: A Prospective Cohort Study☆

Background BRCA1 and BRCA2 mutations have been associated with prostate cancer (PCa) risk but a wide range of risk estimates have been reported that are based on retrospective studies. Objective To estimate relative and absolute PCa risks associated with BRCA1/2 mutations and to assess risk modification by age, family history, and mutation location. Design, setting, and participants This was a prospective cohort study of male BRCA1 (n = 376) and BRCA2 carriers (n = 447) identified in clinical genetics centres in the UK and Ireland (median follow-up 5.9 and 5.3 yr, respectively). Outcome measurements and statistical analysis Standardised incidence/mortality ratios (SIRs/SMRs) relative to population incidences or mortality rates, absolute risks, and hazard ratios (HRs) were estimated using cohort and survival analysis methods. Results and limitations Sixteen BRCA1 and 26 BRCA2 carriers were diagnosed with PCa during follow-up. BRCA2 carriers had an SIR of 4.45 (95% confidence interval [CI] 2.99–6.61) and absolute PCa risk of 27% (95% CI 17–41%) and 60% (95% CI 43–78%) by ages 75 and 85 yr, respectively. For BRCA1 carriers, the overall SIR was 2.35 (95% CI 1.43–3.88); the corresponding SIR at age <65 yr was 3.57 (95% CI 1.68–7.58). However, the BRCA1 SIR varied between 0.74 and 2.83 in sensitivity analyses to assess potential screening effects. PCa risk for BRCA2 carriers increased with family history (HR per affected relative 1.68, 95% CI 0.99–2.85). BRCA2 mutations in the region bounded by positions c.2831 and c.6401 were associated with an SIR of 2.46 (95% CI 1.07–5.64) compared to population incidences, corresponding to lower PCa risk (HR 0.37, 95% CI 0.14–0.96) than for mutations outside the region. BRCA2 carriers had a stronger association with Gleason score ≥7 (SIR 5.07, 95% CI 3.20–8.02) than Gleason score ≤6 PCa (SIR 3.03, 95% CI 1.24–7.44), and a higher risk of death from PCa (SMR 3.85, 95% CI 1.44–10.3). Limitations include potential screening effects for these known mutation carriers; however, the BRCA2 results were robust to multiple sensitivity analyses. Conclusions The results substantiate PCa risk patterns indicated by retrospective analyses for BRCA2 carriers, including further evidence of association with aggressive PCa, and give some support for a weaker association in BRCA1 carriers. Patient summary In this study we followed unaffected men known to carry mutations in the BRCA1 and BRCA2 genes to investigate whether they are at higher risk of developing prostate cancer compared to the general population. We found that carriers of BRCA2 mutations have a high risk of developing prostate cancer, particularly more aggressive prostate cancer, and that this risk varies by family history of prostate cancer and the location of the mutation within the gene.

There are only a few estimates of absolute risks of PCa for BRCA1/2 mutation carriers and those are based on retrospective studies [ 8 8 _ T D $ D I F F ] [4,6,7,13,15,17,21,22]. Given the rapidly rising incidence of PCa in the prostate-specific antigen (PSA) testing era, retrospective absolute risk estimates may not be representative of the risks for mutation carriers currently seen in genetics clinics. Only two small prospective cohort studies of male BRCA1/2 carriers have been reported [ 8 9 _ T D $ D I F F ] [12,24], the largest of which followed 137 BRCA1 and 71 BRCA2 carriers for an average of 5.1 yr, and did not show an association with PCa [ 9 0 _ T D $ D I F F ] [24].
In the present study, we report age-specific PCa risk estimates for a large prospective cohort of male BRCA1 and BRCA2 carriers. We present relative and absolute risks, investigate variability in these risks by family history and mutation location, and consider the risk of developing highgrade PCa.

2.
Patients and methods Please visit www.eu-acme.org/europeanurology to answer questions on-line. The EU-ACME credits will then be attributed automatically.

Abstract
Background: BRCA1 and BRCA2 mutations have been associated with prostate cancer (PCa) risk but a wide range of risk estimates have been reported that are based on retrospective studies. Objective: To estimate relative and absolute PCa risks associated with BRCA1/2 mutations and to assess risk modification by age, family history, and mutation location. Design, setting, and participants: This was a prospective cohort study of male BRCA1 (n = 376) and BRCA2 carriers (n = 447) identified in clinical genetics centres in the UK and Ireland (median follow-up 5.9 and 5.3 yr, respectively). Outcome measurements and statistical analysis: Standardised incidence/mortality ratios (SIRs/SMRs) relative to population incidences or mortality rates, absolute risks, and hazard ratios (HRs) were estimated using cohort and survival analysis methods.

Statistical analysis
We prospectively followed the participants from completion of their baseline questionnaire until their age at diagnosis of PCa, age of death, age at the end-of-follow-up, or age 85 yr, whichever occurred first. A diagnosis of another cancer or of prostatic intraepithelial neoplasia was not considered as a censoring event. Analogously, we followed the participants for deaths due to PCa. We compared the observed PCa incidence and PCa mortality to those expected from population incidences and PCa-specific mortality rates (Office for National Statistics, https://www.ons.gov.uk/) using standardised incidence ratios (SIRs) or standardised mortality ratios (SMRs) computed via Poisson regression. We used the Kaplan-Meier estimator to estimate absolute risks, and Cox regression to test for differences in risk between subgroups.
We classified men who had at least one first-or seconddegree relative diagnosed with PCa as having a positive PCa family history, and assessed trends in risks according to the number of affected relatives. We investigated differences in risk by mutation position using prespecified definitions of regions for which different associations with PCa risk have been identified in published studies [ 9 2 _ T D $ D I F F ] [6,10,13,22,23]. To assess the association of BRCA1/2 mutations with clinical PCa subtypes according to biopsy Gleason score (GS), we compared the observed number of PCa diagnoses by GS subtype to those expected given population GS-specific incidences. We used competing risk estimators to estimate the absolute risk for these clinical subtypes. Because data on GS were not available for all PCas, we used multiple imputation to avoid omission of PCa events.
For the main analysis, we included men with previous non-prostate cancers, did not censor for non-prostate cancers during follow-up, and considered follow-up up to the last questionnaire if available after the last record linkage. We assessed the impact of these assumptions in sensitivity analyses. We also repeated the analysis after omitting pathogenic missense mutations to assess the impact of such less clearly deleterious mutations.
Mutation carriers may be offered a different screening and diagnosis regimen than men in the general population [ 9 3 _ T D $ D I F F ] [25]. We performed further analyses to assess the potential impact of such differential screening. First, we performed landmark analyses where follow-up was initiated at 6 or 12 mo after baseline. Second, on the basis of previous findings that observed PCa incidences are 1.4-1.9 times higher for men undergoing PSA screening at regular intervals in comparison to unscreened men

Prostate cancer
In total, 16 of 376 BRCA1 and 26 of 447 BRCA2 mutation carriers were diagnosed with PCa during median follow-up of 5.9 and 5.3 yr, respectively ( online) had a significantly lower risk of PCa than men with mutations outside this region (HR 0.37, 95% CI 0.14-0.96). However, mutations both within (SIR 2.46, 95% CI 1.07-5.64) and outside (SIR 5.88, 95% CI 3.75-9.22) the OCCR were associated with higher [ 9 4 _ T D $ D I F F ] than population PCa risk. [ ( F i g . _ 1 ) T D $ F I G ]  [2,22] with follow-up initiated at 6 mo after study entry. Family history was defined as having at least one first-or second-degree relative with a prostate cancer diagnosis at the time of study entry.    Table 3). The proportional hazards assumption was violated for this model (Schoenfeld residuals test, p = 0.005); the corresponding Kaplan-Meier curves revealed that the risks were similar between OCCR and non-OCCR mutation carriers at younger ages but deviated at older ages ( Fig. 1E and 1[ 9 5 _ T D $ D I F F ] F). The difference in risk between OCCR and non-OCCR mutation carriers (wide definition) was not statistically significant but was of similar magnitude after adjusting for family history (adjusted HR 0.40, 95% CI 0.15-1.07) and after omitting Ashkenazi mutation carriers (HR 0.43, 95% CI 0.15-1.24; Table 3).

GS-specific PCa
For BRCA1 carriers, the SIR was higher for GS 6 (SIR 3.

Sensitivity analyses
The estimated SIRs remained similar under alternative inclusion or censoring assumptions (  Table 5). When follow-up was restricted to the period before initiation of the IMPACT screening trial [ 7 6 _ T D $ D I F F ] [27], in addition to the entire follow-up for participants from non-IMPACTrecruiting centres, there was no association with PCa risk for BRCA1 carriers (SIR 0.74, 95% CI 0.18-3.04). However, this was based on a small sample size and the 95% CI overlapped with that for the estimate for BRCA1 carriers from IMPACTrecruiting centres with follow-up after October 2005 (SIR 2.83, 95% CI 1.67-4.81). The point estimates were similar for BRCA2 carriers followed without potential overlap with the IMPACT trial period and recruiting centres (SIR 3.57, 95% CI 1.29-9.85) and those whose follow-up potentially overlapped with IMPACT (SIR 4.54, 95% CI 2.96-6.99). The SIR for ages <65 yr for BRCA2 carriers with no potential overlap with IMPACT was 6.75 (95% CI 1.98-23.0; Table 5).

Discussion
We estimated the risk of PCa for male BRCA1 and BRCA2 mutation carriers using data from a large prospective cohort. The results substantiate previous reports from retrospective studies of a strong association between BRCA2 mutations and PCa risk, and give some support for a similar but weaker association for mutations in the BRCA1 gene, particularly at younger ages. Depending on the assumptions, we found that BRCA2 carriers are at a two to five times higher risk of PCa compared to men in the general population, which is consistent with previous RR estimates in the range 2-6 [ 8 0 _ T D $ D I F F ] [3][4][5][6][7][8][9][10][11][12][13]. Our BRCA2 RR estimates did not vary substantially with age, in contrast to previous studies suggesting higher RRs at younger ages [ 8 1 _ T D $ D I F F ] [4,6,[13][14][15]. However, the higher RR estimate at ages <65 yr for the subset of BRCA2 carriers with no potential overlap with the IMPACT screening trial suggests that the similarities in associations by age might be due to potential screening effects. However, owing to the small number of events at younger ages, the precision of the estimates was low. In line with previous studies [ 9 6 _ T D $ D I F F ] [3,5,[7][8][9][10]12,13,[16][17][18][19][20][21], our findings indicate that BRCA1 mutations are at most associated with a moderate PCa risk at younger ages, with RR estimates in the range 2-4 for ages <65 yr. The evidence for an association is weak at older ages, with our RR estimates varying between 1 and 2. Much larger studies are required to clarify the association between BRCA1 mutations and PCa risk.
The estimated cumulative risk of developing PCa by age 85 yr was 29% (95% CI 17-45%) for BRCA1 and 60% (95% CI 43-78%) for BRCA2 carriers. However, absolute PCa risks depend on the screening regimen used, and the PCa risks were lower in analyses that assessed the impact of potentially prevalent cancers and the excess PCa risk among PSA-screened individuals. Although our RR estimates are similar to previous estimates, the absolute risk estimates from the present study are higher than estimates from retrospective studies. Previous absolute PCa risk estimates by ages 65-80 yr range from 3% to 9% for BRCA1 carriers [ 9 7 _ T D $ D I F F ] [7,17,21] and from 15% to 34% for BRCA2 carriers [ 9 8 _ T D $ D I F F ] [4,6,7,13,15,22] (Supplementary Table 2[ 9 9 _ T D $ D I F F ] ). It is plausible that absolute risk estimates based on historical data are not representative of the absolute PCa risks for BRCA1/2 carriers in the PSA testing era. Prospective risk estimates may be more informative for counselling current mutation carriers.
Only two previous prospective studies on PCa risk for male BRCA1/2 carriers have been reported and were limited by small sample sizes and wide CIs for their RR estimates, and neither presented absolute risk estimates. In a prospective cohort of 62 carriers from the USA, BRCA2 mutations were associated with higher PCa risk (SIR 4.89, 95% CI 1.96-10.08) but there was no significant association for BRCA1 carriers (SIR 3.81, 95% CI 0.77-11.13) [ 1 0 0 _ T D $ D I F F ] [12]. An Israeli study observed only three prospective PCas in 210 unaffected BRCA1/2 carriers (median follow-up 5.1 yr) and chose not to report a prospective RR estimate [ 1 0 1 _ T D $ D I F F ] [24].
The results indicate that PCa risks for mutation carriers increase with the number of affected relatives, consistent with findings in the general population [ 1 0 2 _ T D $ D I F F ] [28]. This is also consistent with the hypothesis that other familial factors modify PCa risks for mutation carriers, and with recent observations that common PCa susceptibility genetic variants [ 1 0 3 _ T D $ D I F F ] [29] modify PCa risk for BRCA1/2 carriers [ 1 0 4 _ T D $ D I F F ] [30]. This emphasises the importance of considering family history and other risk-modifying genetic factors when counselling male BRCA1/2 carriers. However, it is possible that mutation carriers with a family history of PCa are more likely to be screened or biopsied than mutation carriers without a PCa family history; this may also partly explain the higher risk observed.
We found that BRCA1 carriers were at higher risk of GS 6 disease, but after omitting diagnoses in the initial 6 mo after study recruitment, the associations with highand low-grade disease were similar. BRCA1 carriers were not at a significantly higher risk of PCa mortality, although the CI for the SMR estimate was wide. A lack of association between BRCA1 mutations and PCa grade is in line with published data [ 8 2 _ T D $ D I F F ] [8,9], and the higher SIR for GS 6 disease might reflect a higher propensity for diagnosing indolent low-grade tumours that would not have been detected in the absence of the discovery of a deleterious mutation. Conversely, our results suggest that BRCA2 mutations are associated with a more aggressive PCa phenotype; the association was stronger for GS 7 than for GS 6 tumours. Furthermore, we observed a significant association between BRCA2 mutations and PCa mortality. Associations with high-grade disease and PCa mortality are consistent with previous reports for BRCA2 carriers [ 8 2 _ T D $ D I F F ] [8,9], and suggest that the BRCA2 findings are less affected by screening effects. BRCA2 mutations both within and outside the OCCR were associated with elevated PCa risk. However, our results suggest that carriers of mutations within the OCCR are at comparatively lower risk than carriers of mutations outside the OCCR, consistent with previous findings [ 1 0 5 _ T D $ D I F F ] [6,22]. They are also consistent with reports of lower PCa risk for carriers of the BRCA2 c.5946delT Ashkenazi Jewish founder mutation, which is located in the OCCR [ 1 0 6 _ T D $ D I F F ] [23]. Conversely, the results are in contrast to a UK study that reported a HR of 2.92 (95% CI 1.54-5.54) for OCCR compared to non-OCCR mutations [ 1 0 7 _ T D $ D I F F ] [10]. However, this study was based on a retrospective cohort of BRCA2 carriers and their relatives and the analyses were not adjusted for the ascertainment process.
Strengths of our study include the nationwide recruitment of mutation carriers, which supports the generalisability of our findings. Furthermore, this is the largest prospective cohort of men with deleterious BRCA1/2 mutations to date, and the prospective study design allows for direct estimation of both relative and absolute risks. We have provided risk estimates by family history and mutation location.
Although this is the largest prospective study reported to date, the precision of our estimates is still limited by a moderate sample size and the number of incident PCas and PCa deaths. The results by GS are limited by potential inaccuracies in tumour grading based on biopsies; however, since mutation carriers were recruited through a UK-wide study and SIRs were computed relative to national GS-specific incidences (which will have similar inaccuracies), variability in pathological grading is unlikely to have resulted in a systematic bias. Other limitations include possible oversampling of men with a family history of PCa as a result of the recruitment through clinical genetics centres. While this allowed us to obtain estimates applicable to mutation carriers both with and without a family history, the overall risk might be somewhat overestimated compared to average BRCA1/2 carriers in the population. In addition, known mutation carriers who undergo genetic counselling may receive enhanced screening compared to men from the general population. More specifically, during the study period the IMPACT screening trial [ 7 6 _ T D $ D I F F ] [27] also recruited male BRCA1/2 carriers, and therefore some overlap between IMPACT and EM-BRACE is likely. Given the background prevalence of indolent PCas that are undetectable in the absence of screening [ 1 0 8 _ T D $ D I F F ] [31] and our observed clustering of PCa diagnoses shortly after study entry, it is plausible that some of these PCas would not have been discovered in the absence of diagnostic measures taken as a result of the discovery of a mutation. When we initiated follow-up at 6 or 12 mo after study entry, the estimated RRs were attenuated for both BRCA1 and BRCA2 carriers, but remained statistically significant. Furthermore, known mutation carriers may undergo a different screening regimen over an extended period of time in comparison to men in the general population [ 9 3 _ T D $ D I F F ] [25]. To assess this we compared the PCa incidence observed to that expected from population incidences adjusted by screening effect sizes estimated in the ERSPC trial [ 7 5 _ T D $ D I F F ] [26]. The SIRs for BRCA2 carriers remained significant, but the excess risk for BRCA1 carriers was not consistently significant, and was significant only for ages <65 yr. This adjustment is limited by the assumption of a constant average screening effect on the population PCa incidences [ 1 0 9 _ T D $ D I F F ] based on published ERSPC estimates [ 7 5 _ T D $ D I F F ] [26]. The ERSPC data also suggest that the effect of screening may be time-dependent, with a probable decrease in screening effect sizes with time since initiation of screening [ 7 5 _ T D $ D I F F ] [26]. This time dependence was not considered in our analysis and can result in potential overestimation of SIRs if the true effect of screening on population incidences is higher than the assumed average during the follow-up period. However, our adjustment used the highest published average PSA screening effect size from ERSPC, and assumes that no screening occurs in the general population, which is unlikely given the rates of opportunistic screening [ 1 1 0 _ T D $ D I F F ] [32] and may result in an attenuation of the SIR estimates. After using both a 6-month landmark to control for the detection of prevalent PCas[ 1 1 1 _ T D $ D I F F ] , and higher population incidences, the SIRs remained significant only for BRCA2 carriers. However, these may represent extreme overadjustments. Finally, when we restricted the follow-up to [ 1 1 2 _ T D $ D I F F ] centres and/or time periods not overlapping with the IMPACT recruitment, we found no association between BRCA1 mutations and PCa risk. This might suggest that the association observed for BRCA1 carriers is driven by screening-induced diagnoses of indolent tumours, but caution is needed in the interpretation as the sample size used for this subgroup analysis was small. By contrast, the strength of the association was similar for BRCA2 carriers regardless of potential overlap with IMPACT. Assuming that clinically significant tumours are likely to be diagnosed regardless of screening regimen, this observation is consistent with the hypothesis that BRCA2 mutations are associated with a risk of more aggressive disease. It provides further evidence that the association between BRCA2 mutations and PCa risk is unlikely to be explained by screening effects.

Conclusions
This prospective analysis substantiates previous reports on the RR of PCa for BRCA1 and BRCA2 mutation carriers from retrospective studies, and provides direct estimates of absolute PCa risk by family history and mutation characteristics. The results will be informative in the counselling of men who carry BRCA1 or BRCA2 mutations.