Breast Cancer Risk After Bilateral Prophylactic Oophorectomy in BRCA1 Mutation Carriers

Abstract

Women who carry germline BRCA1 mutations have a greatly increased risk of breast and ovarian cancers when compared with the general population. The clinical management of women with BRCA1 mutations may include bilateral prophylactic oophorectomy (i.e., the surgical removal of both ovaries). The rationale for this surgery is that removing ovarian epithelium reduces ovarian cancer risk. In premenopausal women, an additional benefit from this surgery is a decrease in ovarian hormone exposure, which could in turn reduce breast cancer risk. However, limited data are available to guide specific recommendations regarding the use of this surgery to reduce cancer risk in women with a germline BRCA1 mutation ( 1 ). Some evidence suggests that ovarian cancer risk may be reduced in high-risk women who have undergone this surgery ( 4 - 7 ) have shown a decreased breast cancer risk among oophorectomized women, and oophorectomy has been used to treat breast cancer. To evaluate whether bilateral prophylactic oophorectomy alters the risk of developing breast cancer in women who have BRCA1 mutations, we compared the incidence of breast cancer in BRCA1 mutation carriers who had and had not undergone this surgery.

M ethods

Study Participants

All women with germline, disease-causing BRCA1 mutations who reported having undergone an oophorectomy were identified as potential study subjects from the registry databases of five participating institutions: Creighton University (Omaha, NE), the Dana-Farber Cancer Institute (Boston, MA), the Fox Chase Cancer Center (Philadelphia, PA), the University of Pennsylvania (Philadelphia), and the University of Utah (Salt Lake City). While these represent geographically distinct groups, all of these centers ascertained study participants through similar clinical and research programs involving genetic screening for women at increased risk of breast and/or ovarian cancers. The population of inference, therefore, reflects a relatively homogeneous set of high-risk women. Women were included in the study sample if they had undergone bilateral oophorectomy prior to or at the time of enrollment in these registries or if they reported having had this procedure during periodic follow-up by the collaborating institutions. Women were excluded from the study sample if they had only unilateral oophorectomies, if they had undergone mastectomy prior to their oophorectomy, or if they had a personal history of breast or ovarian cancer at or before the time of their oophorectomy. Surgical subjects were, therefore, included only if their surgery was not performed to treat ovarian or related peritoneal cancers.

After a set of eligible surgical subjects was identified, a matched set of control subjects was selected from women in the registries at each of the five collaborating centers. Potential control subjects were eligible if they had inherited a confirmed disease-causing BRCA1 mutation, were alive and had both ovaries (i.e., no history of oophorectomy), had no history of breast or ovarian cancer, and had no history of prophylactic mastectomy at or before the time of the surgical subject's surgery. Control subjects were matched to surgical subjects on year of birth (±5 years) and on the collaborative institution from which they were ascertained. Women who had inherited germline BRCA2 mutation were excluded as control subjects. All eligible control subjects that could be matched to a surgical subject were selected for analysis. While at least one matched control subject was selected for each surgical subject, we selected more than one matched control subject per surgical subject whenever possible. Criteria for entry into registries, data collection, and follow-up were undertaken at each collaborating center without regard to surgical status.

By use of these criteria, we identified 43 surgical subjects and 79 control subjects. Of these subjects, 44 were ascertained at Creighton University, 26 at the Dana-Farber Cancer Institute, 18 at the Fox Chase Cancer Center, 16 at the University of Pennsylvania, and 18 at the University of Utah. Fifty (41%) study subjects were unrelated to one another. The remainder consisted of individuals who were related to at least one other person in the sample. Related subjects consisted of two (n = 24 individuals in 12 families, 20%), three (n = 9 individuals in three families, 7%), or four or more (n = 39 individuals in five families, 32%) women from the same family.

All BRCA1 mutations were disease causing. Mutation testing was undertaken by use of a variety of methods at each of the participating institutions, but the majority of mutations were determined for the purpose of clinical testing and therefore reflect the relatively consistent and high-quality standards used in a clinical setting. The BRCA1 mutation status of all of the subjects was confirmed by direct mutation testing with full written informed consent under research protocols approved by the human subjects review boards at each participating institution. Carriers of missense variants of unknown functional significance were excluded from the study sample. The identified mutations spanned the majority of the gene's coding region, ranging from Met1Ile (methionine to isoleucine at amino acid position 1) to 5438insC (an insertion of cytosine at nucleotide position 5438). Mutations included deletions (including large genomic deletions), nonsense mutations, insertions, and disease-associated missense mutations. Among the commonly identified mutations in this sample were 185delAG (n = 19; 16%) and 5382insC (an insertion of cytosine at position 5382; n = 6; 5%). Women with BRCA2 mutations were not included in this study because of relatively small numbers available in our study population and because their risk of breast and ovarian cancers (and possibly patterns of surgery use) may differ from BRCA1 mutation carriers. Women were included who had surgery and were later identified as having BRCA1 mutations and who were first identified as having BRCA1 mutations and later underwent this surgery.

Data Collection and Statistical Analysis

Vital status and cancer occurrence information were obtained by use of the ongoing follow-up records for each study subject from existing clinical research programs and from follow-up telephone interviews and/or self-administered questionnaires. For women who were deceased based on records maintained for each family, we reviewed medical records and family history reports to establish date of death and whether any malignancy had been diagnosed in that subject. Living women were interviewed by telephone to assess current vital status and occurrences of cancer. We obtained a self-reported reproductive history and history of hormone replacement therapy (HRT) use by interview. Occurrences of postsurgery cancer were verified by review of medical records, operative notes, and/or pathology reports.

Cox proportional hazards models were used to evaluate breast cancer incidence by surgical status using SAS (v.6.11; SAS Institute, Inc., Cary NC). Only confounders that were statistically significant in any analysis were used to adjust the effect of surgery. Age at menarche was the only such variable identified in any analysis. Therefore, the only confounder variable considered was age at menarche. To correct for nonindependence of observations among subjects from the same family, we used the robust variance-covariance estimation method of Lin and Wei ( 8 ), as implemented in the software STATA (release 5) (Stata Corp., College Station, TX). The widths of the 95% confidence intervals (CIs) from the robust models were not uniformly changed compared with those of the standard models: some 95% CIs were narrower and some were wider than the standard models. Furthermore, the inferences from both the robust and nonrobust analyses were identical. Therefore, only the standard model results are presented. Both surgical subjects and control subjects were followed retrospectively from birth until the occurrence of the first event of interest. First, the first diagnosis of a primary invasive breast cancer was considered to be the primary event of interest. Second, the subjects were censored if they developed ovarian cancer or peritoneal carcinomatosis, had a prophylactic mastectomy, died, or were lost to follow-up; they were censored at the date of each subject's last contact for follow-up if none of these events occurred. Our follow-up strategy was chosen to provide an estimate of lifetime breast cancer risk reduction, as measured by an adjusted hazard ratio (HR). Because our study design did not allow inclusion of any women who developed cancer prior to the time of surgery, the risk reduction estimates presented here are conditional on surviving cancer free until the time of surgery. Analysis of breast cancer risk over various periods of follow-up after surgery was undertaken by use of a “landmark analysis” in which women were analyzed after 0-5 years of follow-up (with cancer-free women censored at 5 years of follow-up), between 5 and 10 years of follow-up (with cancer-free women censored at 10 years of follow-up), and more than 10 years of follow-up. All reported statistical inferences were based on two-sided hypothesis tests.

R esults

Surgical subjects who underwent bilateral prophylactic oophorectomy were followed for an average of 9.6 years after surgery (range, <1-36 years) and control subjects were followed for an average of 8.1 years (range, <1-43 years) after the time of the matched subject's surgery. Forty-nine percent of all subjects were followed for at least 5 years after the surgical subject's surgery. The mean length of follow-up did not differ statistically significantly between surgical subjects and control subjects (F _ANOVA = .766; P = .384). However, the statistically nonsignificant difference in mean follow-up between the two groups supported our choice to use survival analysis models. No statistically significant differences overall were noted in the distribution of parity, age at first live birth, age at menarche, mean year of birth, age at time of the surgical subject's surgery, or ascertainment location between those who did and did not have bilateral prophylactic oophorectomy (Table 1 ).

Approximately one third of the women developed breast cancer during the postsurgery follow-up period. Fig. 1 shows the cumulative incidence of breast cancer in women with and without surgery. These incidences do not represent lifetime breast cancer risks in BRCA1 mutation carriers because most subjects were followed only until the time of censoring or the diagnosis of breast cancer and, therefore, had not passed through their entire period of breast cancer risk.

Bilateral prophylactic oophorectomy was associated with a statistically significantly reduced risk of developing breast cancer in the total sample (HR = 0.53; 95% CI = 0.33-0.84; Table 2 ). While complete information about menopausal status on all participants was not available, six surgical subjects had surgery after 50 years of age and were likely to be perimenopausal or postmenopausal (mean age at time of surgery, 56.7 years; range, 52-63 years). When these six women and their 10 matched control subjects were removed from the sample, we estimated the HR to be 0.57 (95% CI = 0.36-0.92), a value similar to that in the total sample. While the sample of surgical subjects older than age 50 years and their matched control subjects was very small (n = 16), the HR estimate in this group (0.93; 95% CI = 0.22-3.92) suggested that surgery after age 50 years was not associated with a substantial breast cancer risk reduction. Because many women elect to undergo surgery after childbearing, we also evaluated the effect of surgery among parous women (n = 104). In this group, reduction in breast cancer risk by surgery was of similar magnitude to that estimated in the whole sample (HR = 0.49; 95% CI = 0.30-0.79).

Reduction in breast cancer risk after surgery may also depend on duration of postsurgery follow-up. When we limited our analyses to women who had been followed for less than 5, 5-10, and 10 or more years after the surgical subject's surgery (Table 2 ), risk reduction was estimated to be HR = 0.55 (95% CI = 0.36-0.85), HR = 0.28 (95% CI = 0.08-0.94), and HR = 0.33 (95% CI = 0.12-0.91), respectively. Subjects who were followed at least 10 years after surgery and were parous (HR = 0.35; 95% CI = 0.13-0.95) or had surgery before age 50 years (HR = 0.34; 95% CI = 0.12-0.96) also experienced a substantial reduction in breast cancer risk.

HRT has been used after bilateral prophylactic oophorectomy to ameliorate the symptoms of surgically induced menopause. Because HRT may also increase breast cancer risk ( 9 ), we evaluated whether HRT use affected postsurgery breast cancer risk. Sufficient information about dose, preparation, timing, or duration of HRT use was not available on the majority of study subjects. Self-reported ever/never use of HRT was available on 91 of 122 subjects. We also assumed that eight women younger than age 50 years without surgery did not receive HRT, although data were not available regarding HRT use in these women. Therefore, we had or inferred HRT use information on 32 (74%) of 43 surgical subjects and on 67 (85%) of 79 control subjects. Of these women, 22 (69%) of 32 who had undergone surgery had any HRT exposure, while only four (6%) of 67 control subjects had any HRT exposure.

HRT use was not a statistically significant independent predictor of breast cancer outcome in a multivariate Cox model that included surgery (χ ² = 1.40; P = .237). After we excluded women who had used HRT or who had no HRT data available, we found that the effect of surgery on subsequent breast cancer risk was lower (HR = 0.42; 95% CI = 0.22-0.81) than estimated in the total sample (HR = 0.53). The HRs among parous women who had no HRT exposure (HR = 0.35; 95% CI = 0.17-0.71) or who underwent surgery before age 50 years (HR = 0.46; 95% CI = 0.23-0.90) were lower than those estimated in the total sample.

D iscussion

We report that women who carry a germline BRCA1 mutation who have had bilateral prophylactic oophorectomy may experience a substantial reduction in breast cancer risk. Our observations suggest that decreased exposure to ovarian hormones after surgery may alter breast cancer risk in BRCA1 mutation carriers. Our results have clinical relevance for women who have inherited a germline mutation in BRCA1 and want to decrease their breast cancer risk.

To our knowledge, our study is the first to show an association between surgery and a statistically significant reduction in breast cancer risk among BRCA1 mutation carriers. Brinton et al. ( 4 ) reported that women who underwent this surgery before the age of 40 years had a 45% reduction in breast cancer risk compared with women who underwent natural menopause. Meijer and van Lindert ( 5 ) reported that surgery performed before the age of natural menopause statistically significantly reduced breast cancer risk, even with HRT use. Parazzini et al. ( 6 ) reported a 20% reduction in breast cancer risk after bilateral prophylactic oophorectomy done at the time of hysterectomy in Italian premenopausal women and that this protection increased from the date of surgery. Schairer et al. ( 7 ) reported that Swedish women less than 50 years of age experienced a 50% reduction in breast cancer risk within 10 years of bilateral oophorectomy. Surgery after age 50 years conferred no risk reduction. Finally, Struewing et al. ( 3 ) suggested that bilateral prophylactic oophorectomy may reduce breast cancer risk in genetically high-risk women, but their sample size was not large enough to achieve statistical significance. On the basis of our results, women who carry germline BRCA1 mutations and who undergo bilateral prophylactic oophorectomy may experience a reduction in breast cancer risk that is as large as or larger than that reported in women who were not characterized with respect to BRCA1 mutation status ( 4 - 7 ).

Limitations of this study include a relatively small sample size and a lack of data on some confounder variables. With the use of 122 study participants, we had sufficient statistical power to identify a statistically significant association between bilateral prophylactic oophorectomy and breast cancer risk reduction. However, our CIs remain relatively wide, and some may wish to interpret the upper bound of the CIs as the most conservative estimate of risk reduction. Similarly, we did not have sufficient statistical power to evaluate the effect of some potentially important factors including reproductive history. As would be the case in a randomized clinical trial or a case-control study, competing mortality that may have excluded some study subjects from analysis could not be assessed. Survival bias could produce an apparent decrease in the protective effect of surgery (i.e., bias toward the null hypothesis because fewer breast cancer cases among control subjects might be recorded), while a similar bias in surgical subjects could result in an apparent increase in the protective effect of surgery (i.e., bias away from the null hypothesis). The matching criteria used here may have minimized these effects to some degree. However, survivor bias is more likely to have underestimated rather than overestimated the true breast cancer risk reduction associated with surgery. Larger prospective studies, now being undertaken by our group, are required to address potential survival biases, to address the effects of surgery in specific substrata, and to refine CIs.

Confounding by indication may also have influenced our results if breast cancer risk was lower in women from families with a history of ovarian cancer, and this pattern of breast and ovarian cancer risk affected who chose to undergo surgery. Mutations occurring in the ovarian cancer cluster region (OCCR) of BRCA2 confer higher ovarian versus breast cancer risk than mutations in other parts of BRCA2 ( 10 ). Thus, women carrying OCCR mutations may have a greater family history of ovarian cancer, may have preferentially sought out surgery, and may be at decreased breast cancer risk. However, only BRCA1 mutation carriers were studied here, and no OCCR region has been identified in BRCA1. Mutations in the 3` region of BRCA1 confer a deficit of ovarian cancer risk ( 11 ). Women carrying these mutations may have less ovarian cancer in their family, may have been less likely to seek out surgery, and may be at increased breast cancer risk. Thus, our analyses may have underestimated risk reduction by surgery. This study provides limited opportunities to directly address the potential for confounding by indication. Analyses stratified by ovarian cancer family history are infeasible because more than 80% of our families have at least one ovarian cancer. An analysis of breast cancer-only families would consist of fewer than 10 eligible surgical subjects and 15 control subjects. These are insufficient data to evaluate bias in the point estimates of risk. Further stratification by HRT use or other factors will result in even smaller sample sizes. Finally, it is unlikely that a randomized clinical trial of this surgery could be undertaken to address this issue, since randomization to surgery or no surgery is unlikely to be accepted by women who carry BRCA1 mutations.

The use of HRT may increase breast cancer risk even in the absence of endogenous ovarian hormones. However, risk of breast cancer after HRT use has not been evaluated in BRCA1 mutation carriers ( 9 ). In this study, complete information about postsurgery HRT was not available. We were able to obtain or infer HRT use data from 81% of our sample subjects and inferred that any increase in breast cancer risk in women with HRT exposure is moderate at best: HR estimates in the total sample were only marginally higher than those in women without HRT exposure (Table 2 ). We conclude that HRT use did not negate the finding that bilateral prophylactic oophorectomy is associated with a reduction in breast cancer risk. However, the data available in this study were limited, and additional analyses with more complete HRT data will be required to confirm and extend these observations.

Despite the potential benefits of bilateral prophylactic oophorectomy in breast cancer risk reduction in BRCA1 mutation carriers, the costs and benefits of this surgery must be weighed. For example, the surgery itself may cause some risk of morbidity and mortality ( 12 ). The appropriate choice of surgical technique (e.g., laparotomy versus laparoscopy) is not clear ( 13 ). The primary negative side effect of surgery is the induction of premature menopause, which is associated with increased risks of osteoporosis and cardiovascular disease ( 14 , 15 ). Hot flashes, vaginal dryness, sexual dysfunction, sleep disturbances, and cognitive changes associated with menopause may also substantially affect quality of life. Although HRT moderates the risk of developing osteoporosis or cardiovascular disease, there is some concern that this therapy may be contraindicated after surgery if HRT increases the risk of breast cancer. Finally, other nonsurgical options for breast or ovarian cancer prevention should be considered. These options include the use of compounds that ablate the production of ovarian hormones and may provide a nonsurgical alternative to oophorectomy for breast cancer risk reduction. However, most women who carry germline BRCA1 mutations undergo surgery to reduce ovarian cancer risk. It is not clear whether nonsurgical ovarian hormone ablation will reduce ovarian cancer risk to the same degree as surgery.

Aside from the medical consequences of bilateral prophylactic oophorectomy, very little is known about the psychosocial effect prophylactic surgery has on women. Lerman et al. ( 16 ) reported that, among women who have undergone appropriate genetic counseling and are members of families with a BRCA1 or BRCA2 mutation, 48% were considering surgery 1 month after counseling; however, only 2% had actually undergone this surgery 6 months later. Clearly, the removal of ovaries resulting in premature menopause could have a profound effect on a woman's body image and lifestyle. A potentially negative consequence of surgery is that women may have a false sense of security after surgery and may not continue with appropriate cancer screening. In our sample, 10 women (23%) who underwent surgery subsequently developed breast cancer. Thus, while surgery may reduce cancer risk, it does not completely eliminate the occurrence of breast cancer. Breast cancer screening and prevention options should, therefore, continue after surgery.

Fig. 1.

Cumulative incidence of breast cancer in bilateral prophylactic surgery subjects and nonsurgical control subjects carrying BRCA1 mutations. Because most women were followed only until the time of censoring or until the diagnosis of breast cancer, the incidences reported here do not represent lifetime breast cancer risks in BRCA1 mutation carriers. The cumulative incidences of breast cancer at ages 45, 60, and 75 years are 15.2%, 25.3%, and 31.6%, respectively, for control subjects and 11.6%, 14.0%, and 18.6%, respectively, for surgical subjects.

Open in new tabDownload slide

Supported by Public Health Service grants P30CA16520 (supplement to T. R. Rebbeck and B. L. Weber), CA57601 (B. L. Weber), CA74415 (S. L. Neuhausen), and N01CN6700 (to S. L. Neuhausen) from the National Cancer Institute, National Institutes of Health, Department of Health and Human Services; by the University of Pennsylvania Cancer Center (to T. R. Rebbeck and B. L. Weber); by The Breast Cancer Research Foundation (to B. L. Weber); by the Dana-Farber Women's Cancers Program (to J. E. Garber); by Department of Defense grants DAMD-17-96-I-6088 (A. K. Godwin), DAMD-17-94-J-4260 (S. L. Neuhausen), DAMD-17-94-J-4340 (H. T. Lynch), and DAMD-17-97-I-7112 (H. T. Lynch); by The Utah Cancer registry and the Utah State Department of Health (to S. L. Neuhausen); and by the Nebraska State Cancer and Smoking-related Diseases Research Program (LB595) (to H. T. Lynch).

A. Eisen is a member of the Speakers' Bureau of Zeneca Pharmaceuticals (Salt Lake City, UT). L. Cannon-Albright holds stock in, conducts research sponsored by, and serves on the scientific advisory board of Myriad Genetics, Inc., Salt Lake City, UT. B. L. Weber holds stock options in and is a member of the clinical advisory board of Myriad Genetics Inc.

We thank Drs. Steven Gruber and Patricia Peyser for their helpful discussions of this research.

References