Effects of BRCA1- and BRCA2-Related Mutations on Ovarian and Breast Cancer Survival: A Meta-analysis

BRCA1 and BRCA2 are 2 distinct tumor suppressor genes that play an integral role in response to cellular stress via the activation of DNA repair processes (1). Individuals with mutations in these 2 genes are at an increased risk of developing breast, ovarian, and other cancers. The lifetime risk of breast cancer among BRCA mutation carriers is 45% to 80% and for ovarian cancer 45% to 60% (2). Several studies have investigated prognoses among BRCA mutation carriers and noncarriers. Patients with ovarian cancer with BRCA mutations have been reported to have better outcomes than noncarriers (3–6). While other studies (7–10) demonstrated that outcomes were superior only among women with BRCA2 mutations, a recent study involving 873 patients did not find survival differences between BRCA1 or BRCA2 mutation carriers and noncarriers (11).

Research on breast cancer prognoses and BRCA mutations has also yielded inconsistent results. Multiple studies have demonstrated that BRCA1-related breast cancers were more likely to be triple-negative, which was correlated with poor prognosis (12–14). A meta-analysis published in 2010 also linked BRCA1 mutation to decreased overall (OS) and progression-free survival (PFS) among patients with breast cancer, although BRCA2 mutation was not associated with differential survival outcomes (15). Additional studies, published since 2010, have generated varying results on the role of BRCA mutations in breast cancer prognoses. While a multi-country study failed to show differences in breast cancer survival rates associated with BRCA mutations (16), Cortesi and colleagues (17) reported favorable 10-year breast cancer outcomes among BRCA1 carriers. Meanwhile, a large-scale study involving 2,967 patients demonstrated adverse effect of BRCA2 mutations on breast cancer–specific survival (18). These inconsistencies in the research may be due to variations in sample size and the relative rarity of mutation carriers.

Despite our increased understanding of the pathophysiology of BRCA-mutated cancers, specific treatments for these diseases have not yet been established. A novel class of anticancer agents, PARP inhibitors, has shown strong activity in BRCA-related cancers by using the inherent homologous recombination defectiveness of these tumors (19, 20). Enhanced understanding of the role of BRCA mutation status in ovarian and breast cancer survival will provide more accurate prognostic information and can improve clinical decision-making with respect to trial design and cancer treatment. The objective of this study was to perform a meta-analysis to evaluate the effects of BRCA1 and BRCA2 mutations on OS and PFS of ovarian and breast cancer.

We followed the PRISMA Statement guidelines to design, analyze, and report our meta-analytic findings (21). Only the study-level summary data were used for the analyses.

We reviewed PubMed and EMBASE databases for articles published before February 10, 2014. We identified studies by using Medical Subject Headings (MeSH) including mutation, BRCA1, Genes, BRCA2, Survival, Prognosis, Ovarian Neoplasms, Hereditary Breast and Ovarian Cancer Syndrome, Ovarian epithelial cancer, Breast Neoplasms. These terms were also combined with keyword and manual searches of references in all selected studies.

The inclusion criteria were as follows: (i) comparisons were made between patients with breast or ovarian cancer with BRCA1 or BRCA2 mutations and noncarriers confirmed by mutation analyses; (ii) outcomes were survival-related, such as OS or PFS; (iii) outcomes involved human subjects; and (iv) articles were published in English. In addition, articles were excluded if (i) control patients not confirmed of noncarriers; (ii) they were review articles; (iii) subjects had metastatic or recurrent cancer; (iv) comparisons were made between carriers or did not evaluate BRCA1 and 2 mutations separately; and (v) studies were for cancer specific–survival only.

Initial screening of potentially eligible records was performed by 2 investigators (Q. Zhong, L. Zhang). Subsequent full-text record screening was performed independently by 2 investigators (Q. Zhong, W.-T. Hwang). Disagreements were resolved by consensus. Baseline characteristics and outcomes were extracted from the selected articles. We chose OS and PFS as our endpoints for meta-analysis and treated disease-free survival (DFS) as PFS. Various endpoints for PFS were reported in the selected breast cancer studies, including local recurrence-free survival (LRFS; refs. 22, 23), distant DFS (DDFS; refs. 16, 23, 24), DFS (17, 25, 26), metastasis-free survival (MFS; refs. 22, 27), and contralateral RFS (CRFS; refs. 23, 27). Because DDFS and MFS shared the same definition in these articles, we operationally defined PFS to include also MFS or DDFS for studies that did not provide DFS.

The quality of each study was independently rated by 2 investigators (Q. Zhong, W.-T. Hwang), using the Newcastle–Ottawa Scale (28) and Altman framework (29).

HR was used as a measure of the prognostic value. HR > 1 indicated poor survival for the group with BRCA mutations. HRs and 95% confidence intervals (CI) were extracted from articles. For those studies in which HRs and CIs were not available, we used the method proposed by Parmar and colleagues to derive estimates from survival curves (30). To improve accuracy, we excluded HRs derived from studies if the number of events in one group was less than 5 (31, 32). Heterogeneity was assessed by the χ² test and expressed by I² index (33), which describes the percentage of total variation across studies that is due to heterogeneity rather than chance (25% low heterogeneity, 50% medium, 75% high). Random-effects models were initially used to obtain the summary HRs and 95% CIs, and if there was no heterogeneity among studies, fixed-effects models were used to estimate pooled HRs (34). If results of both univariate and multivariate Cox regression analyses were reported, we chose multivariate models for a more accurate estimate of the effect of BRCA mutation. Subgroup analyses were performed on the basis of variables including stage, grade, histologic subtype, optimal debulking rate, tumor size, nodal status, estrogen receptor (ER) status, chemotherapy, and hormone therapy rates and whether a multivariate or univariate Cox regression was used. Publication bias was assessed by inspecting the symmetry of the funnel plot and tested with Begg and Egger's tests (35, 36). STATA 12.0 was used to test for publication bias. All other analyses were conducted using Review Manager 5.2.

Figure 1 summarizes the identification of relevant studies. We screened 1,201 articles for eligibility and identified 27 studies [including 2 related publications from the same study (refs. 13, 26)] for meta-analysis. Among them, 20 were retrospective cohort studies and 7 were prospective cohort studies. The research quality among the selected studies was high, with a median quality score of 0.86 (range, 0.67–0.92).

Fourteen studies were included in the meta-analysis of ovarian cancer. Characteristics of ovarian cancer patient cohorts are presented in Table 1. The studies were conducted in 8 countries (Poland, Italy, Israel, United Kingdom, Canada, United States, Hong Kong, and Australia) and published between 1999 and 2014. The median number of women evaluated per study was 251 (range, 48–3,879), with a total of 9,588 patients, including 1,722 BRCA1 mutation carriers, 659 BRCA2 mutation carriers, and 7,207 noncarriers. The reported mean or median age for studies was similar, ranging from 54.4 to 65.4 years. The percentage of women with serous tumors varied from 31% to 100%, with 2 studies focused exclusively on serous cancers (8, 9). Between 62% and 100% of patients had stage III–IV disease, whereas 12% to 100% of patients had grade 3 tumors. Optimal debulking rate varied from 51% to 89%, but data were missing in 8 studies (3, 4, 7, 10, 37–40). Eight studies (3–5, 7, 10, 11, 37, 41) reported median or mean follow-up time, ranging from 1.5 to 6.9 years. Ten studies (71%; refs. 3–11, 37) conducted a multivariate analysis to adjust for confounding variables such as age, stage, debulking status, cancer or family history, menopausal status, grade, histology, platinum sensitivity, year of diagnosis, or morphology. The most commonly used BRCA mutation detecting methods were PCR and sequencing in selected studies. BRCA-associated ovarian cancers were more likely to be characterized by high-grade serous histology and advanced stage. On the basis of 7 of 14 ovarian cancer studies that had mean age data available, we found that the mean ages for the mutation groups appeared to be younger, particularly for BRCA1 mutation carriers: 43 and 47 years for BRCA1 and BRCA2 carriers, respectively, compared with 49 years for noncarriers (Supplementary Table S1).

Thirteen studies were included in the meta-analysis of breast cancer. Characteristics of breast cancer patient cohorts are presented in Table 2. The studies were conducted in 12 countries (Poland, Italy, Israel, Norway, United Kingdom, Netherlands, Canada, France, Finland, United States, Australia, and Germany) and published between 2000 and 2013. A total of 10,016 patients (ranging from 85 to 3,345 per study) were included, among them 890 BRCA1 mutation carriers, 342 BRCA2 mutation carriers, and 8,784 noncarriers. The reported mean or median age ranged from 42.6 to 62.1 years across eligible studies. The percentage of women with tumor > 2 cm varied from 9% to 66%, with 0% to 49% nude positive, 23% to 49% grade 3 tumors, 37% to 67% ER-positive, 25% to 72% receiving chemotherapy, and 19% to 56% receiving hormone therapy. Among the 9 studies (12, 13, 16, 17, 22–25, 27) that reported median or mean follow-up time, the median duration was 6.3, ranging from 4.5 to 7.9 years. Nine studies (69%; refs. 12, 13, 16, 17, 22, 24, 42–44) conducted a multivariate analysis of HRs to adjust for age, tumor size, nodal or ER status, oophorectomy, chemotherapy, hormone therapy, year of diagnosis, or nuclear grade. PCR and sequencing were the most commonly used BRCA mutation detecting methods across selected studies. After analyzing data from 6 studies (5 for BRCA2) of 13 breast cancer studies that had mean age data available, we found that BRCA mutation carriers were comparatively younger in age, particularly for BRCA1 mutation carriers. The mean age was 52 years for BRCA1 mutation carriers and 54 years for BRCA2 mutation carriers, compared with 58 years for noncarriers. Furthermore, BRCA1 mutation carriers were more likely to be ER- and progesterone receptor (PR)–negative, have lower histologic grade, and receive chemotherapy (Supplementary Table S2).

Fourteen studies were included in the analysis of the effect of BRCA1 mutation on ovarian cancer OS. The results of our meta-analysis showed that BRCA1 mutation carriers were associated with better OS than noncarriers, with a pooled HR of 0.76 (95% CI, 0.70–0.83). No significant heterogeneity was found across the studies (I² = 17%, P = 0.27). A stronger effect was found using 12 studies assessing OS between BRCA2 carriers and noncarriers. BRCA2 mutation carriers were associated with even better OS than noncarriers, with a pooled HR of 0.58 (95% CI, 0.50–0.66). There was an indication of slight heterogeneity across the studies, but it did not reach statistical significance (I² = 40%, P = 0.08). Associations between BRCA1 or BRCA2 mutations and ovarian cancer PFS were also evaluated, respectively. Meta-analyses results (3 studies for BRCA1 mutation, 2 studies for BRCA2 mutation) demonstrated that both BRCA1 and BRCA2 mutation carriers were associated with better PFS than noncarriers, with pooled HR of 0.65 (95% CI, 0.52–0.81) for BRCA1 mutation carriers and 0.61 (95% CI, 0.47–0.80) for BRCA2 mutation carriers. No significant heterogeneity was found across the studies (I² = 0% for BRCA1 mutation carriers and I² = 59% for BRCA2 mutation carriers, all P > 0.10; Fig. 2).

Meta-analysis of 11 studies showed a worse breast cancer OS among BRCA1 mutation carriers than noncarriers. Because a significant heterogeneity was found across the studies (I² = 59%, P = 0.007), the pooled HR of 1.5 (95% CI, 1.11–2.04) was calculated on the basis of a random-effects model. The result of meta-analysis of 5 studies comparing breast cancer OS between BRCA2 carriers and noncarriers indicated no association between mutation and OS, with a pooled HR of 0.97 (95% CI, 0.78–1.22), and no significant heterogeneity was found across the studies (I² = 0%). Eight and 2 studies were included to assess the association between BRCA1 or BRCA2 mutations and breast cancer PFS, respectively. Meta-analyses results did not reveal any association between these 2 mutations and PFS, with a pooled HR of 1.35 (95% CI, 0.95–1.90) for BRCA1 mutation carriers and 0.95 (95% CI, 0.69–1.30) for BRCA2 mutation carriers. A random-effects model was used for the BRCA1 mutation meta-analysis due to a significant heterogeneity among studies (I² = 55%, P = 0.03). No significant heterogeneity among studies was found in the meta-analysis of BRCA2 mutation (I² = 0%; Fig. 3).

Subgroup analysis was conducted to investigate potential sources of heterogeneity among studies and to assess the consistency of conclusions among different subpopulations of patients.

For ovarian cancer, BRCA1 and BRCA2 mutation carriers have significantly longer OS than noncarriers, regardless of tumor stage, grade, or histologic subtype.

The results of subgroup analyses for BRCA1 mutation in breast cancer outcomes were less consistent due to the limited number of studies available. Interestingly, a subgroup of studies using multivariate analysis showed that BRCA1 mutation carriers had borderline poorer breast cancer OS (HR, 1.40; 95% CI, 1.00–1.98; P = 0.05) than noncarriers, whereas studies involving univariate analysis yielded no association between BRCA1 mutation and breast cancer OS (Table 3).

Formal investigation using Begg and Egger's tests indicated no publication bias in the meta-analyses for associations of BRCA1 or BRCA2 mutations with ovarian cancer OS and BRCA1 mutation with breast cancer OS and PFS. We did not test publication bias for the meta-analyses of ovarian cancer PFS and breast cancer OS and PFS for BRCA2 mutation because too few studies were available to make a valid statistical test (Supplementary Fig. S1).

This is the first meta-analysis that investigated both ovarian and breast cancer survival in patients with BRCA1 or BRCA2 mutations compared with noncarriers in one report. Our results demonstrated that BRCA mutations had a differential impact on ovarian and breast cancer survival. Both BRCA1 and BRCA2 mutations were positively associated with ovarian cancer OS and PFS and BRCA2 mutation appeared to have a greater effect upon OS, regardless of tumor stage, grade, or histologic subtype. In breast cancer outcomes, the impact of BRCA mutations was just the opposite. BRCA1 mutation carriers had worse prognoses than noncarriers for OS but were not different from noncarriers in PFS. BRCA2 mutation was not associated with breast cancer prognosis.

Our results, with respect to ovarian cancer OS, were consistent with the previous meta-analysis by Sun and colleagues (45). For ovarian cancer PFS, although Sun and colleagues did not separate BRCA1 and BRCA2 mutations in their analysis, they still found a favorable PFS in BRCA1/2 mutations carriers, which was consistent with our finding. After we incorporated 4 additional studies (12, 16, 17, 27) into our meta-analysis, our breast cancer OS results were also consistent with the previous meta-analysis published by Lee and colleagues (15), but we did not replicate their findings on the effect of BRCA1 mutation on breast cancer PFS. We believed this may be due to the fact that Lee and colleagues analyzed short-term (i.e., <5 years) and long-term (i.e., >5 years) PFS separately and fewer studies were analyzed.

At this time, the mechanism driving the association between BRCA1 or BRCA2 mutation and survival is not clear. It may be rooted in different functions of the BRCA genes in the dsDNA repair pathway. As we know, postoperative platinum-based chemotherapy is the main treatment for ovarian cancer. Most studies suggested that the ovarian cancer survival advantage observed among BRCA carriers could be mediated through improved response to platinum-based agents, due to an impaired ability to repair dsDNA breaks through homologous recombination (46). The fact that BRCA2 mutation improved ovarian cancer survival may be due to the gene's more direct involvement in homologous recombination which can attenuate or abolish the interaction with RAD51, resulting in failure to load RAD51 to DNA damage sites. The BRCA1 mutation may have less impact on RAD51-mediated homologous recombination, which could account for the disparity in outcomes between BRCA1 and BRCA2 carriers (47, 48). For breast cancer, we often rely upon different types of treatments, including hormone therapy, radiotherapy, and chemotherapy. Research on breast cancer also indicated that BRCA1 mutation carriers had more p53 mutations, resulting in worse prognosis (49). The different breast cancer outcomes observed among BRCA1 and BRCA2 carriers could also be due to their distinct clinical characteristics. As mentioned previously, BRCA1-related breast cancers were more likely to be triple-negative, which made treatment more difficult.

Certain limitations must be considered when interpreting study findings. First, our meta-analysis only encompassed a total of 27 studies. Studies specific to BRCA2 mutations for breast cancer were less common in the literature and are therefore underrepresented in this analysis. The limited availability of BRCA2 mutation studies complicated our ability to test publication bias. This is partially due to the fact that we excluded a number of studies that did not evaluate BRCA1 and BRCA2 separately or compared mutation carriers to sporadic patients with cancer without testing mutation status on control patients. Second, because of the variety of endpoints reported in breast cancer studies, we operationally defined breast cancer PFS to include MFS or DDFS for studies that did not provide DFS and had to exclude LRFS and CRFS results (which were considered to be parts of PFS for some studies) from a few studies. Third, most studies were focused on the Western population; only one study included Asian women (3). Additional studies are needed to better understand the effects of BRCA1 and BRCA2 mutations on survival among carriers in Asia and other ethnic or geographic regions. Finally, nearly half of the studies selected for our breast cancer meta-analysis failed to provide sufficient data on patient characteristics, which may have impacted the validity of subgroup analysis.

Despite these limitations, our meta-analysis rigorously evaluated the effects of BRCA1 and BRCA2 mutations on both ovarian and breast cancer survival using all eligible studies in the same setting that included intensive subgroup analyses. Our results may have important implications for the clinical management of both ovarian and breast cancer. The results can be used immediately by health care professionals for patient counseling regarding expected survival. Also, there is a widespread agreement that PARP inhibitors (PARPi) can affect BRCA-related cancers by using the inherent homologous recombination defectiveness of the tumors in a synthetically lethal manner. Many phase I and II trials have reported the successful applications of PARPi in patients with ovarian and breast cancer with BRCA mutations and phase III trials are under way (19, 20, 50). Given the prognostic information provided by different BRCA status and the possibility for personalized treatments in carriers and noncarriers, the next step would be to see how this knowledge can be used to optimize treatments and testing targeted therapy in the future. The design of upcoming clinical trials could be stratified by BRCA status to avoid potential bias introduced by unequal numbers of carriers in treatment groups or between study cohorts. Routine testing of BRCA1 and BRCA2 mutation status of ovarian cancer and BRCA1 mutation status of breast cancer may now be warranted.

No potential conflicts of interest were disclosed.

Conception and design: X. Zhao, L. Zhang, W.-T. Hwang

Development of methodology: Q. Zhong, H.-L. Peng, W.-T. Hwang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Q. Zhong, W.-T. Hwang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Q. Zhong, H.-L. Peng, X. Zhao, W.-T. Hwang

Writing, review, and/or revision of the manuscript: Q. Zhong, H.-L. Peng, L. Zhang, W.-T. Hwang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W.-T. Hwang

Study supervision: L. Zhang, W.-T. Hwang

This work was supported, in whole or in part, by NIH (P50-CA83638 to W.-T. Hwang and L. Zhang; R01-CA142776 to L. Zhang); Department of Defense (W81XWH-10-1-0082 to L. Zhang); Ovarian Cancer Research Fund (to L. Zhang); the Basser Research Center for BRCA grant (to L. Zhang); and Marsha Rivkin Center for Ovarian Cancer Research grant (to L. Zhang). Q. Zhong was supported by a scholarship from Sichuan University and the National Natural Science Foundation of China (0040215401068).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.