|Home | About | Journals | Submit | Contact Us | Français|
Risk-reducing mastectomy (RRM) and salpingo-oophorectomy (RRSO) are widely used by carriers of BRCA1 or BRCA2 mutations to reduce their risks of breast and ovarian cancer.
To estimate risk and mortality reduction stratified by mutation and prior cancer status.
A prospective multi-center cohort study was used to assess the relationship of RRM and RRSO on cancer outcomes.
Twenty-two clinical and research genetics centers in Europe and North America.
2,482 women identified 1974-2008 and followed until the end of 2009 who tested positive for BRCA1 or BRCA2 mutations.
257 (10%) underwent RRM and 993 (40%) underwent RRSO.
Breast and ovarian cancer risk; cancer-specific and overall mortality.
No breast cancers were diagnosed in women with RRM compared to 7% of women without RRM. In women who underwent RRSO, 1.1% were subsequently diagnosed with ovarian cancer, 11.4% were subsequently diagnosed with breast cancer, and 3% subsequently died from any cause, compared with 5.8% ovarian cancer, 19.2% breast cancer, and 10% overall mortality in women who did not undergo RRSO. RRSO was associated with a lower risk of ovarian cancer in those with (Hazard Ratio (HR) 0.31, 95% CI 0.12-0.82) and without a prior breast cancer (HR 0.15, 0.04-0.63), and a lower risk of first breast cancer in both BRCA1 (HR 0.63, 0.41-0.96) and BRCA2 (HR 0.36, 0.16-0.82) mutation carriers. RRSO was associated with a reduction in all-cause (HR 0.40, 0.26-0.61), breast cancer-specific (HR 0.44, 0.26-0.76), and ovarian cancer-specific (HR 0.25, 0.08-0.75) mortality.
Among a cohort of women with BRCA1 and BRCA2 mutations, the use of RRM was associated with a lower risk of breast cancer, and RRSO was associated with a lower risk of ovarian cancer, first breast cancer, and overall, breast-, and ovarian-cancer specific mortality.
Women who have inherited mutations in the BRCA1 or BRCA2 (BRCA1/2) genes have substantially elevated risks of breast cancer (BC) and ovarian cancer (OC), with a lifetime BC risk of 56%–84% 1-3. The estimated OC risks range from 36%-63% for BRCA1 and 10%-27% for BRCA2 mutation carriers 3-6. Mutation carriers have cancer risk management options that include risk-reducing salpingo-oophorectomy (RRSO), risk-reducing mastectomy (RRM), screening, and chemoprevention. Due to the lack of effective screening for OC, RRSO is strongly recommended once child-bearing is complete.
RRSO has been demonstrated to decrease the risk of both BC and OC in BRCA1/2 mutation carriers 7-16. However, risk and mortality reduction estimates for women with and without a prior history of BC may differ and be of clinical relevance. Likewise, data are emerging that suggest BRCA1 mutation carriers may experience differential benefits from interventions such as RRSO compared to BRCA2 mutation carriers15, 17. We examine here a large cohort of BRCA1/2 mutation carriers followed prospectively. We report cancer risk reduction estimates following RRSO and RRM in a number of scenarios incorporating mutation type (BRCA1 vs. BRCA2), cancer history (prior history of BC vs. none), and use of HT.
Women with inherited, disease-associated BRCA1/2 mutations were identified from 22 centers in the PROSE consortium (see Acknowledgements). Participants were ascertained between 1974 and 2008 (Median: 1999). The PROSE protocol is as previously described12. All participants underwent an informed consent process for participation in research. This protocol was approved by each institution's IRB. Study participants were enrolled as a cohort with time of follow up starting from patient ascertainment into the research program. Genetic testing was performed per institutional guidelines and all patients received post-testing counseling to review medical management options. Women who declined RRSO or PM were offered increased surveillance at all centers according to established guidelines. At US sites, this consisted of annual mammogram and annual MRI for those with breast tissue, and every 6-12 month transvaginal ultrasound and CA125 for those with ovaries in place (www.nccn.com). In the UK, women were offered yearly mammograms, as well as yearly MRI until age 50 (www.nice.org.uk). Ovarian cancer screening consisted of TVUS and CA125 every 4 months (www.instituteforwomenshealth.ucl.ac.uk/academic_research/gynaecologicalcancer/gcrc/ukfocss). Participants were eligible for the study if they had no ovarian cancer diagnosis and no RRSO at the time of ascertainment and had a minimum of 6 months of follow-up; they were excluded if they had a cancer diagnosis within the first six months of follow-up to avoid including cancers that would have been minimally influenced by RRSO or RRM. Participants were followed until the end of 2009 (Median date of follow up: 2005). The median follow up for women was 3.65 years (Range: 0.52-27.4 years) among those who underwent surgery, and 4.29 years (range; 0.5-27.9 years) in controls who did not undergo surgery.
4255 known BRCA1/2 mutation carriers in the PROSE study were considered for inclusion: 12 were excluded because they had both BRCA1 & BRCA2 mutations, 525 because they underwent RRSO before ascertainment date, 363 because they were diagnosed with ovarian cancer before ascertainment date, 738 because we had less than 6 months of follow-up, and 135 because they were incident cases. Participants who had undergone RRM prior to ascertainment were excluded from all breast cancer incidence analyses. BC included invasive cancers and ductal carcinoma in-situ.
Unaffected women (without RRM) were followed prospectively from the time of ascertainment. Exposed women with RRM after ascertainment were followed from the age of their RRM. Our primary outcome was BC. If no BC occurred, women were censored at the date of OC, death, or date of last contact.
Women who underwent RRSO after ascertainment were considered “exposed”. If they did not undergo RRSO, they were considered “unexposed”. Exposed women were followed from age of RRSO and unexposed women were followed from age of ascertainment. The outcomes of interest were: 1) OC, 2) BC, 3) second primary BC in those with a prior BC, and 4) mortality.
For OC endpoints, women were followed until OC or censoring at death or date of last contact. Women were excluded if they were diagnosed with an occult OC at RRSO. When missing data were encountered, the individual was dropped from the analysis that involved the missing data point, but the individual was included in other analyses where complete data were available; in fact, because many of the data items were required for enrollment, missing data was only applicable to ovarian cancer endpoints, with missing OCP data. For BC endpoints, women were excluded if they underwent RRM prior to ascertainment. Women who had RRM after ascertainment but before RRSO were considered unexposed and were censored at RRM. Women were followed until BC or were censored at OC, RRM, death, or last contact. For analysis of second primary BC, the aforementioned censoring criteria were applied, and women were also censored at the time of a contralateral mastectomy (CM). A second BC was defined as any contralateral BC or an ipsilateral BC more than five years after the first.
For the mortality analysis, the inclusion and follow-up methods were similar to the RRSO analyses described above. However, women were censored at the date of last contact, with the primary outcome being death.
The effect of RRSO/RRM on cancer incidence and mortality was analyzed by Cox proportional hazards models. A robust variance-covariance estimation method was used to correct for non-independence of observations among participants from the same family or within centers 18. Adjustment for year of birth was undertaken in all analyses using Cox regression. Oral contraceptive use was adjusted for when OC was the outcome. Adjustment for center of ascertainment was undertaken by stratifying analyses by center to avoid imposing linear constraints in the model. Surgical participants were followed from age of RRSO or RRM. Non-surgical participants were followed from age of ascertainment. All analyses were undertaken using STATA8 (College Station, TX). All statistical tests were based on two-sided hypotheses, and inferences of statistical significance were made at the p=0.05 level.
RRM was associated with a decreased risk of breast cancer in BRCA1/2 mutation carriers: No BC events were seen in women who underwent RRM during 3 years of prospective follow-up. In contrast, 7% of women without RRM over a similar follow up period were diagnosed with BC. BRCA1BRCA2
RRSO was associated with a decreased risk of OC. Among those with no prior breast cancer, the risk reduction estimate in all BRCA1 mutation carriers was Hazard ratio (HR)=0.31 (95% CI: 0.12-0.82). No OC events were seen in BRCA2 mutation carriers without prior breast cancer who underwent RRSO during 6 years of prospective follow-up. In contrast, 3% of women without RRSO over a similar follow up period were diagnosed with OCBRCA2 Among women with a prior diagnosis of BC, the risk reduction in BRCA1 mutation carriers was HR=0.15 (95%CI: 0.07-0.33), while no post -RRSO OC were diagnosed in BRCA2 mutation carriers.
RRSO was associated with a decreased risk of BC in both BRCA1 (HR=0.63, 95%CI: 0.41-0.96) and BRCA2 (HR: 0.36, 95%CI: 0.16-0.82) mutation carriers with no prior diagnosis of BC. There was evidence for an age effect with a reduction in BC risk among BRCA1 mutation carriers who had their RRSO before age 50 (HR=0.51 95%CI: 0.32-0.82), but no suggestion of BC risk reduction in women who underwent RRSO after age 50 (HR=1.36; 95%CI: 0.26-7.05); a test of interaction was significant with p=0.027 (HR: 0.62, 95%CI 0.41-.95). In BRCA1 and BRCA2 mutation carriers with a prior diagnosis of BC, there was no evidence for reduction in risk of a second primary BC diagnosisBRCA1BRCA2. BRCA1BRCA2BRCA1BRCA2
RRSO was associated with significantly lower all-cause mortality in those with no prior cancer (HR=0.45, 95%CI: 0.21-0.95) and those with prior BC (HR=0.30, 95%CI: 0.17-0.53; Table 4a).BRCA1BRCA2When analyzed by mutation status, RRSO was associated with a significantly lower all-cause mortality in BRCA1 mutation carriers overall [0.38 (0.24-0.62)]. With fewer participants and fewer events, all-cause mortality in BRCA2 mutation carriers was not statistically significant [0.52 (0.22-1.23)].
RRSO was associated with a lower BC-specific mortality (HR=0.44, 95% CI: 0.26-0.76) (Table 4b) and OC-specific mortality (HR=0.24, 95% CI: 0.08-0.73) (Table 4c). In BRCA1 mutation carriers, RRSO was associated with improved BC- (HR=0.38, 95%CI: 0.2-0.72; Table 4b) and OC-specific mortality (HR=0.25, 95%CI: 0.08-0.75; Table 4c). There were no OC deaths following RRSO in BRCA2 mutation carriers, nor were there any BC deaths in BRCA2 mutation carriers who underwent RRSO prior to cancer diagnosis. In an exploratory analysis, overall survival was associated with RRSO in those <50 (HR 0.41, 0.25-0.67) and in those ≥50 (HR 0.37, 0.15-0.94); however, a test of interaction was not significant suggesting no difference in overall mortality benefit between these two groups. There have been eight deaths after RRSO in those without cancer prior to RRSO: two deaths from BC, three from post-RRSO primary peritoneal cancer, one from leukemia, one from stomach cancer, and one non-cancer related (Supplementary Table 2). Although 151 (34%) of those with no prior cancer also underwent RRM, neither of the two who died from BC had undergone RRM. In those patients with BC prior to RRSO, 23 died including 19 BC deaths and one OC death.
The clinical management of cancer risk in BRCA1 and BRCA2 mutation carriers is complex and is best informed by accurate knowledge of the outcomes of interventions. Most prior studies that have investigated the effect of RRSO or RRM on BC risk either did not examine effects by mutation status 7, 9, 12 or by prior cancer diagnosis 14, 19. There may be little added benefit of RRSO on BC risk if women have chemotherapy-induced menopause, or if they are already receiving hormonal therapy. We hypothesize that both of these factors are important in determining precise estimates of risk reduction. Our results confirm that RRM is associated with a significant reduction in BC risk. In addition, RRSO is associated with a significant decrease in OC risk in both BRCA1 and BRCA2 mutation carriers, and in those with and without a prior BC. There is a significant reduction in BC risk following RRSO in both BRCA1 and BRCA2 mutation carriers with no prior cancer. Overall mortality was improved in women undergoing RRSO.
RRM is a highly effective strategy for BC risk reduction 20, 21. In our prospective analysis, no BC were observed after RRM. The observation of no prospectively identified BC cases may be due to biases in prior retrospective studies, or improved surgical techniques in recent prospective analyses.
RRSO is highly effective in reducing ovarian and fallopian tube cancers in both BRCA1 and BRCA2 mutation carriers and in those with and without a prior BC. Precise estimates of risk reduction following RRSO are needed to balance the increasingly recognized health risks caused by premature menopause22-26. We observed no primary peritoneal cancers following RRSO in BRCA2 mutation carriers, although such cancers have been reported 10. In BRCA1 mutation carriers, RRSO was associated with a 70% reduction in risk of OC in those without a prior BC and 85% in those with a prior BC. Whether these numbers are truly different, reflect censoring from death due to BC, or a protective effect following BC treatment is unknown.
RRSO was associated with a significantly decreased risk of BC in those with no prior BC with both BRCA1 (37% reduction) and BRCA2 mutations (64% reduction). In mutation carriers with a prior BC, RRSO had no effect on second primary BC risk. In women with sporadic BC, the benefit of RRSO when added to standard adjuvant treatment is uncertain and is the subject of multiple ongoing clinical trials. Chemotherapy often leads to cessation of menses, so any effect of ovarian ablation from RRSO may be achieved in some women by chemotherapy. In premenopausal women with estrogen receptor (ER) positive tumors, hormonal therapy in addition to chemotherapy significantly improves disease-free survival27, 28. Our data are in contrast to prior reports demonstrating benefit of oophorectomy in preventing contralateral BC29, 30. Differences in adjuvant therapy use could explain these discrepancies; however, a limitation of our study is the absence of detailed treatment information. It is important to note that OC risk is independent of menopause; menopause either naturally occurring or chemotherapy-induced is not known to decrease the risk of OC. Regardless of the effect of oophorectomy on second BC, oophorectomy is essential to reduce the risk of OC, which can be a significant cause of morbidity and mortality in women with early stage breast cancer 29.
Our data suggest that RRSO may be associated with a lower BC risk in BRCA2 mutation carriers than in BRCA1 mutation carriers (64% vs. 37%). Kauff et al 15 observed a statistically significant BC risk reduction following RRSO in BRCA2 but not BRCA1 mutation carriers. The potentially larger risk reduction associated with RRSO in BRCA2 compared with BRCA1 mutation carriers is of interest given the high proportion of ER positive breast tumors in BRCA2 mutation carriers compared with BRCA1 mutation carriers31. Additional research is required to address this issue.
We are still unable to provide definitive data with respect to the timing of RRSO on the efficacy of BC risk reduction as the numbers in each subgroup remain small with a limited number of events. Eisen et al. 14 reported that the BC risk reduction with RRSO was greater in BRCA1/2 mutation carriers who underwent surgery before age 50 than in women who underwent surgery after age 50. The results of the present analyses are consistent with this finding for unaffected women who undergo RRSO before or after age 50.
The importance of understanding the optimal age at RRSO is underscored by several reports32 in the general population that suggest that oophorectomy in women under age 45 is associated with increased mortality, particularly if HT is not used 26 Although these data are not directly applicable to BRCA1/2 mutations who have markedly increased risks of breast and ovarian cancer and therefore a different risk/benefit profile, issues of timing and the safety of HRT are important. The Women's Health Initiative in postmenopausal women did not demonstrate a cardiovascular benefit overall from HT 33, but younger women going through natural menopause may derive such benefit 33, 34. It is possible that BRCA1/2 mutation carriers undergoing abrupt surgical menopause to reduce OC risk and receive HT may derive health benefits. Two prior studies have examined HT in BRCA1/2 mutation carriers. Rebbeck et al.35 examined 462 patients and reported no increased BC risk with post-RRSO HT use. Eisen et al.14 examined HT in women with and without RRSO and observed no increased risk associated with HT. Further work is needed regarding this important issue.BRCA1BRCA1BRCA2BRCA1/2
We observed an association of RRSO with a significant reduction in all-cause, BC-specific, and OC-specific mortality. We previously reported that RRSO was associated with a 90% reduction in BC–specific mortality, a 95% reduction in gynecologic cancer–specific mortality, and a 76% reduction in overall mortality12. Our present estimates are consistent with those reports. The apparent lesser effect on mortality in BRCA2 vs. BRCA1 mutation carriers may be due to the lower risk of OC in BRCA2 mutation carriers as well as the more aggressive biological features of BRCA1-associated BC, however, more data is needed to adequately address this important question.
We recognize a number of limitations of the study. We designed our study to maximize follow up time and statistical power. However, this choice could have compromised our results in a number of ways. Our study design was powered to detect effects of RRSO and RRM stratified on BRCA1/2 as well as the other groups defined in the tables. The observation of statistically significant p-values in many strata provides evidence that we had sufficient power for our pre-planned hypotheses. A few strata-specific analyses did not achieve statistical significance, suggesting that we did not have sufficient power to detect some smaller than anticipated effects. Therefore, analyses in some substrata may require additional study after accrual of much larger sample sizes.
Ideally, the evaluation of risk reducing surgery on cancer risk and mortality reduction would involve a randomized trial design. However, it is accepted in the field that a randomized approach would neither be acceptable nor ethical. As a result, this field of research is limited to undertaking observational studies that have methodological limitations. An observational design requires that statistical methods be used to correct for factors that may influence relative risk estimates. We have attempted to correct for a number of limitations of the observational study design by using the recommended analysis approaches of Klaren 36 and Hartmann 37. However, additional analytical corrections could be considered in future analyses, including the effect of competing risks, or the consideration of time-dependency of RRSO/RRM and other covariates. As a result of these limitations, our results cannot be definitively inferred to be causal in nature.
An additional limitation of an observational study design is the need to determine appropriate follow-up periods for participants. Once choice is whether to follow individuals from the time of ascertainment vs. the time from genetic testing. All individuals in our cohort underwent genetic testing, however some individuals were ascertained prior to their genetic testing; the median ascertainment year was 1999 and median genetic test disclosure date 2001. To address the implications of this choice, we performed an exploratory analysis examining overall mortality using the time of genetic testing (rather than ascertainment) as the starting point, with all other criteria met as in the methods section. With this approach, we see a significant mortality benefit: RRSO was associated with a decreased risk of overall mortality in the entire group (HR=0.36; 0.20-0.62), in BRCA1 mutation carriers (HR 0.42; 0.22-0.81), and in BRCA2 mutation carriers (HR 0.10, 0.01-0.77). However, due to smaller number we lose power to address other questions and therefore have continued to use our initial, preplanned analysis as the primary analysis.
Although all women who chose to forego RRSO were counseled to undergo intensive screening, we do not have detailed information on compliance of these recommendations at all centers. However, there are no data that ovarian cancer screening is effective in reducing the risk of developing ovarian cancer, or in dying from ovarian cancer38. Therefore, we feel that it is unlikely that compliance would significantly alter our results related to ovarian cancer endpoints of incidence and mortality. Intensive breast cancer surveillance does not reduce the risk of developing breast cancer, but aims to improve early detection. Due to our lack of detailed information on breast MRI compliance, we cannot conclude that RRSO improved breast cancer specific mortality compared to optimal screening; however, we do see that women who choose RRSO are associated with better outcomes in terms of breast cancer risk, ovarian cancer risk, and ovarian cancer specific- survival, none of which would be anticipated to be affected by compliance to intensive breast screening. We also see an association between RRSO and breast-cancer specific and overall survival, when compared to women who have not chosen RRSO. BRCA1BRCA2BRCA1BRCA2
This paper is dedicated to the late Andrew Shenton, who contributed significantly to this research.
This study was supported by grants from the Public Health Service (R01-CA83855 and R01-CA102776 to TRR), the University of Pennsylvania Cancer Center (to TRR), the Cancer Genetics Network (HHSN21620074400C to SMD and CI), the Marjorie Cohen Research Fund (to SMD) the Dana-Farber/Harvard Cancer Center SPORE in BC P50 CA-089393 (to JEG), the Department of Defense (DAMD-17-96-I-6088 to AKG; DAMD-17-94-J-4340 and DAMD-17-97-I-7112 to HTL; DAMD-17-03-1-0619 to SMD), P30-CA51008-15 (to Georgetown University), The Utah Cancer registry (funded by Public Health Service Grant NO1-CN-6700) and the Utah State Department of Health, the Nebraska State Cancer and Smoking-Related Diseases Research Program (LB595 to HTL), P30-CA-16042 (to PAG), Cancer Research UK Grant Number C5047/A7357 (to RE), and NCI P30 CA51008-12 (to CI). OIO is Doris Duke Distinguished Clinical Scientist. RE acknowledges The Support of the NIHR to The Biomedical Research Centre at The Institute of Cancer Research and The Royal Marsden NHS Foundation Trust. None of these funding agencies had involvement in the study design; collection, analysis, or interpretation of data; in the writing of the report; nor in the decision to submit the paper for publication.