Prognostic significance of BRCA mutations in ovarian cancer: an updated systematic review with meta-analysis

There is no consensus on the syntheses concerning the impact of BRCA mutation on ovarian cancer survival. A systematic review and meta-analysis of observational studies was conducted that evaluated the impact of BRCA mutations on the survival outcomes of patients with ovarian cancer. The primary outcome measure was overall survival (OS) and secondary outcome was progression-free survival (PFS). We presented data with hazard ratios (HRs) and 95% confidence interval (CI) and pooled them using the random-effects models. From 2,624 unique records, 34 eligible studies including 18,396 patients were identified. BRCA1/2 mutations demonstrated both OS and PFS benefits in patients with ovarian cancer (OS: HR = 0.67, 95% CI, 0.57 to 0.78, I2 = 76.5%, P <0.001; PFS: HR = 0.62, 95% CI, 0.53 to 0.73, I2 = 18.1%, P = 0.261). For BRCA1 mutation carriers, the HRs for OS and PFS benefits were 0.73 (95% CI, 0.63 to 0.86) and 0.68 (95% CI, 0.52 to 0.89), respectively. For BRCA2 mutation carriers, the HRs for OS and PFS benefits were 0.57 (95% CI, 0.45 to 0.73) and 0.48 (95% CI, 0.30 to 0.75), respectively. The results of subgroup analyses for OS stratified by study quality, tumor stage, study design, sample size, number of research center, duration of follow-up, baseline characteristics adjusted and tumor histology were mostly constant across BRCA1/2, BRCA1 and BRCA2 mutation subtypes. In summary, for patients with ovarian cancer, BRCA mutations were associated with improved OS and PFS. Further large-scale prospective cohort studies should be conducted to test its benefits in specific patients.


INTRODUCTION
As two tumor suppressor genes, BRCA1 and BRCA2 mutation are reported to have been associated with increased risk of developing ovarian cancer and breast cancer [1][2][3]. Both of them are involved in DNA damage repair through homologous recombination, contributing to genomic instability and malignant transformation [4][5][6]. Meanwhile, they also involved in cell growth inhibition, gene transcription regulation, apoptosis and other related cellular regulation processes. Previous study reported that patients with BRCA-deficient ovarian cancer had improved survival rates as these patients were reported sensitive to platinum-based chemotherapy [7,8].
Two previous published meta-analyses have reported the prognostic impact of BRCA mutations on ovarian cancer mortality [24,25]. Sun et al. found that patients with ovarian cancer with BRCA dysfunction

Research Paper
Oncotarget 286 www.impactjournals.com/oncotarget status tended to have a better outcome [24]. However, this study investigated the effects of BRCA dysfunction status including mutations, protein expression and its promoter methylation, which did not perform the detailed analyses of BRCA mutations. In the meta-analysis by Zhong et al. they only examined the BRCA1 and BRCA2 mutation separately with limited statistical power without examining BRCA1/2 mutation [25]. Therefore, the purpose of this study was to update the meta-analysis on the impact of BRCA mutation carriers versus noncarriers on mortality in patients with ovarian cancer.

Literature search and study characteristics
From the initial literature search, we yielded 3595 citations. After exclusion of duplicate publications, 2624 citations remained for further review. 45 potentially eligible reports were selected when irrelevant studies were removed. After reading each full manuscript, we finally identified the 34 studies for meta-analysis. As is shown in Figure 1, we follow the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram to conduct this meta-analysis.  Figure 2A).

Characteristics of included studies
Subgroup analyses revealed that studies with adequate adjusted variables, but not with inadequate adjusted variables had statistically significant OS benefit in ovarian cancer patients with BRCA1/2-mutation (adequate adjusted variables, HR = 0.63, 95% CI, 0.53 to 0.75, I 2 = 80.7%, P < 0.001; inadequate adjusted variables, HR = 0.89, 95% CI, 0.72 to 1.10, I 2 = 0, P = 0.992). OS benefits were also indicated in other subgroups and the HRs for all of the different subgroups are summarized in Table 2A.
No evident publication bias was observed by funnel plot asymmetry ( Figure 3A
The results of subgroup analyses for the association between BRCA1-mutation and PFS are presented in Table 2B. We found that ovarian cancer patients with BRCA1-mutation had significantly longer OS than noncarriers, regardless of study quality, sample size, research center or duration of follow-up. Such trend was also noted in studies with cohort study design, adequate baseline characteristics adjusted, all histologic types or conducted in USA or Canada.
The results of subgroup analyses for the association between BRCA2-mutation and OS are presented in Table 2C. We found that ovarian cancer patients with BRCA2-mutation had significantly longer OS than noncarriers, regardless of research center, duration of followup or histologic type. Such trend was also noted in studies with high quality, II-IV disease stage, cohort study design, sample size larger than 200, adequate baseline characteristics adjusted or conducted in USA or Canada.
No evident publication bias was observed by funnel plot asymmetry ( Figure 3C) or through Egger's test (P = 0.54) or Begg's test (P = 0.96) for OS. The trim and fill method applied to further conduct the sensitivity analysis indicated 4 missing studies in the funnel plot for OS ( Figure 3). However, imputing these hypothesized studies did not alter the primary pooled estimates (adjusted HR = 0.38, 95 % CI 0.29 to 0.50). We did not investigate the publication bias for PFS due to the limited number of studies.

DISCUSSION
The aim of this meta-analysis was to examine the association between BRCA mutation status and ovarian cancer survival (OS and PFS). By pooling the outcomes of 18,396 ovarian cancer patients from 34 individual studies, we found that BRCA mutation (BRCA1/2, BRCA1 and BRCA2) carriers had significantly improved OS and PFS benefits in ovarian cancer patients. Subgroup analysis revealed that this survival benefits remained constant This meta-analysis showed that patients who were BRCA mutation carriers had a 33%, 27% and 43% reduction in all-cause mortality for BRCA1/2, BRCA1 and BRCA2 mutants respectively, while patients had a 38%, 32% and 52% reduction in progression-free mortality for BRCA1/2, BRCA1 and BRCA2 mutants, respectively. Individually, however, some of the studies had reported contradictory findings [7,[19][20][21][22]: some studies have indicated significantly reduced all-cause mortality or progression-free mortality among BRCA mutation carriers [9][10][11][12][13][14], whereas Kotsopoulos et al. reported that the mortality risk in ovarian cancer patients was significantly poorer for BRCA mutation carriers than for non-carriers (HR = 1.67; 95% CI 1.34 to 2.08) [19]. The present metaanalysis with the largest number of patients investigated both survival (OS) and progression outcomes (PFS) for ovarian cancer, incorporating not only the general BRCA mutation status but also two subtypes, including BRCA1 and BRCA2 mutation status.
It has been reported that germline BRCA1/2 mutations occur in approximately 10 to 20% of patients with invasive epithelial ovarian cancers [7,[9][10][11][12][13][14][19][20][21][22], and more than 20% of patients with high-grade serous ovarian cancer [12]. BRCA1/2 tumor suppressor genes are reported to be involved in DNA repair through homologous recombination, through which pathway genes are unable to repair DNA double-strand, resulting   Oncotarget 296 www.impactjournals.com/oncotarget in genomic instability and having a tendency to malignant transformation [3]. On the other hand, the impairment of this pathway can also influence DNA cross-links by tumor cells, which can be induced by cisplatin, a chemotherapy agent for ovarian cancer. It has been indicated that BRCAdeficient patients can have better survival outcomes through the increase in the response rate to platinum-based chemotherapy [7,8].
The findings of this updated meta-analysis are generally consistent with and further extend the other two published systematic reviews and meta-analyses in several important ways. First, our study had added greater statistical power to the associations between BRCA mutations and ovarian cancer survival with more detailed subgroup analyses. For example, the present meta-analysis involved approximately 2.3 times as  [24,25]. As a matter of fact, 11 recent published cohort studies were involved in the analyses. Second, three mutation subtypes (BRCA1/2, BRCA1 and BRCA2) were thoroughly investigated in contrast to the earlier two meta-analyses (including only BRCA1 and BRCA2 subtypes). Thirdly, we did detailed subgroup analyses under a broader range of study level circumstances to examine the potential sources of heterogeneity. However, our findings concur with the previous meta-analyses. Inter-study heterogeneity was found very high for a number of analyses, which was probably due to the very variation in population characteristics and BRCA mutation detection methods.
Thus, caution is required when interpreting these findings. Moreover, one important advantage of this meta-analyses lies in that we have thoroughly tested the influence of publication bias through Begg's test, Egger's test and sensitivity analysis and confirmed the robustness of the findings.
Several limitations of this meta-analysis are required to be addressed. We acknowledge that the results of this meta-analysis were derived from published data rather than from studies of individual patient data. Thus, we could not obtained the detailed characteristics of each individual from the involved studies including patient age, tumor stage, sample size, and follow-up period, which to some extent were contributory factors to the heterogeneity, but an attempt was made to account for this variation by conducting subgroup analyses. Another potential limitation of the study is that we also include some conference abstracts in the analysis. It is likely that the results may differ to certain extent between the conference abstracts and future updated full publication. However, we proposed that such differences are very likely to be relatively mild. Moreover, the method for the detection of BRCA mutation varied among studies, which may also a source of substantial heterogeneity. Some of the included studies did not report complete data for analysis, and could have potentially affected the results of multivariate analysis. Most of the included studies adequately adjusted for some known confounders, in particular patients age, tumor stage and grade or chemotherapy. However, some studies did not assess these factors and we acknowledge this limitation. As these were all studies with small sample size, it is unlikely to have affected the results of the analysis substantially. Although no obvious evidence of publication bias was noted in each subset of metaanalysis, it was still a major concern. Due to the time taken to conduct this meta-analysis, further relevant studies concerning this topic may have been published. However, given the relative paucity of suitable studies identified through the last 20 years from 1996 to 2016, we proposed that there were probably very few studies in number and they would not substantially affect the general conclusions of this study.
Despite all of these limitations, however, our meta-analysis with a large sample size of over 18,396 participants, and used the appropriate analyses to investigate the heterogeneity and publication bias among the different studies, showed that in patients with ovarian cancer, BRCA mutation (irrespective of its subtypes) carriers had better OS and PFS than non-carriers. Whether the results may have therapeutic implications remains to be elucidated with further larger, well-designed studies in specific ovarian cancer patients.

Literature search and study selection
PubMed and EMBASE were searched for studies published up to February 2016 for the following searching terms: (ovary/ovarian/oophor* and cancer/neoplas*/ tumor*/tumour*/ cancer*/carcinoma*/malignan*/ neoplasms) and (BRCA1/2 and mutation*/mutated) and mortality/survival/prognosis. Mesh (Pubmed) and Emtree (Embase) terms combined with free text words were used for searching. Detailed search terms and strategies for the two databases are presented in Supplementary Appendix 1. In addition, we also conducted the manual searches of references in all eligible studies to identify potential missing publications that were not identified during the preliminary literature searches. We did not place any restrictions on the searches.
Studies were considered eligible if they met the following inclusion criteria: observational studies (cohort or case-control studies) that investigated patients with ovarian cancer assessed for BRCA mutation status (BRCA1 or BRCA2 mutation status). The outcome measures included OS and PFS, measured as the relative risk (RR), the odds ratio (OR), or the hazard ratio (HR) along with the 95% confidence interval (CI) (or sufficient data for calculating them). We did not include studies with unpublished data. If multiple reports contained the duplicated datasets, the report with the largest or the most recent data was included for analysis. Two investigators independently conducted the literature review (KX and SHY) and any discrepancies were resolved by discussion or by a senior investigator (YCZ).

Data extraction and quality assessment
Two investigators independently extracted data from each included study using a predefined standardized data extraction form including the pertinent issues that concerned the characteristics and survival outcomes of the ovarian cancer patients. For each article, the following information was extracted: authors and published years, study design, sample size, inclusion period, research country, disease stage, tumor histology, BRCA mutation detection methods, research center involved, duration of follow-up and adjusted variables. The extracted data were crosschecked and any disagreements were resolved by discussion.

Outcome measures
The primary outcome measure was OS defined as the time from initial ovarian cancer diagnosis to death due to any causes. Secondary outcome was PFS defined as the time from diagnosis to the first confirmed sign of cancer recurrence, or progression (disease relapse or metastasis) or death from any cause.

Quality assessment
The nine-star Newcastle-ottawa Scale (NOS) [44] was used to assess the study quality for each study. Three domains associated with the selection of study population, data comparability and exposure (case-control studies) or outcome (cohort studies) assessment were evaluated. The NOS score ranged from 0 to 9 with a score > 7 indicating high quality. Two investigators scored each study, and any discrepancies were resolved by a third investigator.

Statistical analysis
All statistical analyses were performed using Stata statistical software (version 12.0; Stata Corporation, College Station, TX, USA). Pooled HRs for OS and PFS with 95% CIs were calculated using random-effects model due to the potential substantial heterogeneity between studies [45]. Heterogeneity across studies was examined by I 2 statistic with an I 2 ≥ 50% indicating the presence of significant heterogeneity [46]. We further investigated potential heterogeneity by subgroup analyses stratified by study quality, tumor stage, study design, sample size, number of research center, duration of follow-up, baseline characteristics adjusted, mutation ratio and tumor histology for OS and PFS across BRCA1/2, BRCA1 and BRCA2 mutation subgroups. Publication bias was evaluated by observing the asymmetry of funnel plots and using the Begg-Mazumdar rank correlation test and Egger's test [47,48]. The Duval and Tweedie trimand-fill method was also applied to conduct sensitivity analysis [49]. A two-sided P ≤ 0.05 was considered statistically significant.