A Systematic Review and Meta-Analysis on the Prognostic Value of BRCA Mutations, Homologous Recombination Gene Mutations, and Homologous Recombination Deficiencies in Cancer

Patients with BRCA1/2 mutations (BRCAm), loss-of-function mutations in other homologous recombination repair (HRRm) genes, or tumors that are homologous recombination deficiency positivity (HRD+) demonstrate a robust response to PARPi therapy. We conducted a systematic literature review and meta-analysis to evaluate the prognostic value of BRCAm, HRRm, and HRD+ on overall survival (OS) among those treated by chemotherapy or targeted therapy other than PARPi across tumor types. A total of 135 eligible studies were included. Breast cancer (BC) patients with BRCA1/2m had a similar overall survival (OS) to those with wild-type BRCA1/2 (BRCA1/2 wt) across 18 studies. Ovarian cancer (OC) patients with BRCA1/2m had a significantly longer OS than those with BRCA1/2 wt across 24 studies reporting BRCA1m and BRCA2m, with an HR of 0.7 (0.6–0.8). Less OS data were reported for other tumors: 6 studies for BRCA2m compared with BRCA2 wt in prostate cancer with an HR of 1.9 (1.1–3.2) and 2 studies for BRCA1/2m compared with BRCA1/2 wt in pancreatic cancer with an HR of 1.5 (0.8–3.1). Only 4 studies reported HRD+ by either BRCA m or genomic instability score (GIS) ≥ 42 and OS by HRD status. The HR was 0.67 (0.43–1.02) for OS with HRD+ vs. HRD−. A total of 15 studies reported the association between HRRm and OS of cancers in which one or more HRR genes were examined. The HR was 1.0 (0.7–1.4) comparing patients with HRRm to those with HRR wild-type across tumors. Our findings are useful in improving the precision and efficacy of treatment selection in clinical oncology.


Introduction
Synthetic lethality arises when a combination of mutations in two genes leads to cell death, while mutation of either gene alone has no effect on cell viability [1]. e ability to create synthetic lethal relationships by pairing cancer-associated mutations with pharmacologic agents (at concentrations that would normally be nontoxic to healthy cells) has led to remarkable strides in cancer therapeutics [2]. e discovery that ovarian cancer cells harboring mutations in the homologous recombination repair (HRR) genes BRCA1 and BRCA2 exhibit synthetic lethality when treated with poly adenosine diphosphate (ADP)-ribose polymerase inhibitors (PARPi) has further expanded the application of PARPi in the clinic beyond BRCA1/2 mutant cancers, with efforts to further identify genome-wide synthetic lethal vulnerabilities to this class of drugs [2]. is in turn led to the identification of other HRR gene mutations (HRRm) and characterization of homologous recombination deficient (HRD) cell state that renders cancer cells sensitive to PARPi [3,4]. Current FDA-approved HRD biomarkers predicting response to PARPi include germline BRCA (gBRCA) mutations, platinum sensitivity as a surrogate biomarker for HRD, somatic mutations in HRR genes including BRCA, genomic scar HRD assays, and gene and protein expression profiling [5,6].
e ability to broadly screen across cancer types using different biomarker assays to identify sensitivity to PARPi has produced clinical trials expanding their use as singleagent therapy or in combination with other DNA damage agents, targeted agents, or immunotherapies, across multiple tumor types. PARPi has demonstrated broad application in the treatment of cancer patients with BRCA mutations (BRCAm), HRRm, and HRD positivity. However, little is known regarding whether the presence of these genetic alterations alone affects overall survival (OS) in cancer patients not treated with PARPi or immunotherapy. We performed a systematic review and meta-analysis to examine the prognostic value of these biomarkers across multiple cancer types in predicting OS in cancer patients treated with chemotherapy or targeted therapy other than PARPi.

Study Design and Search Strategy.
is study was performed as per the the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines. Relevant studies with full-text articles in the last 10 years and conference abstracts in the last 3 years were identified by searching the following databases: Ovid Medical Literature Analysis and Retrieval System Online (MEDLINE), Excerpta Medica database (EMBASE), Cochrane Central Register of Controlled Trials, and Cochrane reviews. Searches were performed on May 21, 2020, using relevant terms in English. Two reviewers independently selected studies according to the inclusion criteria, with a third independent reviewer available to address any discrepancies. Bibliographies from review articles were reviewed thoroughly to identify relevant studies, ensuring that papers and articles not picked up in the original search were also included. Studies involving patients treated with chemotherapy or targeted therapy other than PARPi were included in the analysis.
HRD was defined as having either deleterious or suspected deleterious BRCA1/2m or a genomic instability score ≥42 by the Myriad testing (standard definition), with an alternative definition as only having a genomic instability score ≥42 [5,7]. e genomic instability score is an algorithmic measurement of loss of heterozygosity, telomeric allelic imbalance, and large-scale state transitions using DNA isolated from formalin-fixed paraffin-embedded tumor tissue specimens. e Myriad myChoice ® HRD assay was used in the relevant studies to obtain the genomic instability score. e eligibility criteria are listed in Supplementary Table 1, with a focus on clinical outcomes, defined as OS across all cancers and subtypes harboring BRCA1/2m and HRRm (defined as mutations in one or more of the following genes :  ATM, BARD1, BRCA1, BRCA2, BRIP1, CDK12, CHEK1,  CHEK2, FANCA, FANCL, PALB2, PPP2R2A, MRE11 A,   NBN, RAD50, RAD51, RAD51 B, RAD51 C, RAD51D, and RAD54 L), and HRD status.

Data Analysis.
Hazard ratios (HRs) for OS with corresponding 95% confidence intervals (CIs) were calculated across patients with or without BRCA1/2m, HRRm, and HRD status. Cochrane's Q test and the I 2 statistic were used to assess heterogeneity between studies, with a P value < 0.05 for Cochrane's Q test and I 2 > 50% considered cutoffs for significant heterogeneity [8,9]. Publication bias was assessed by contour-enhanced funnel plots of standard error against the effect estimate. We performed a meta-analysis by tumor type and mutation status using a random-effects model based on the degree of heterogeneity between individual studies and presented data as forest plots. All statistical analyses were performed using STATA (Version 14; Stata Corp., College Station, TX). For studies that presented Kaplan-Meier survival data without reporting HR, we used a previously published methodology for estimating HR from time-to-event analyses [10].

Study Demographics.
Our PRISMA study protocol is shown schematically in Figure 1. We identified 86 outcomes studies on BRCA1/2m, HRRm, and HRD positivity and OS in our systematic review and meta-analysis across types of cancer. Citation lists are presented in Supplementary Tables 2, 3, and 4 for BRCA1/2m, HRRm, and HRD positivity, respectively.

Overall
Survival and BRCA1/2m. We found no association between BRCA1/2m and OS among breast cancer patients (HR � 1.02 (95% CI � 0.80-1.30)) ( Figure 2(a)). Furthermore, stratification revealed no differences in OS in patients with germline BRCA1/2m (Supplementary Figure 1(a)) and in patients with pathogenic variants (HR � 1.30 (95% CI � 0.93-1.81)) (Supplementary Figure 1(b)). Similarly, there was no effect on OS in triplenegative breast cancer patients or subgroup analysis among patients with germline mutations status (HR � 1.10 (95% CI � 0.75-1.60)) (Supplementary Figure 1 Compared to ovarian cancer patients with tumors that were BRCA1/2 wt, ovarian cancer patients with tumors harboring BRCA1/2m had a better OS (HR � 0.67 (95% CI � 0.58-0.77)) ( Figure 2(b)). Similar results were found among patients harboring germline mutations compared with those wild-type patients (HR � 0.69 (95% CI � 0.59-0.81)) (Supplementary Figure 1(e)). Patients with tumor somatic mutations did not have significantly different OS compared with those with wild-type BRCA1/2 (HR � 0.67 (95% CI � 0.23-1.93)). Given only two studies in this subgroup analysis, the results should be interpreted with caution ( Supplementary Figure 1(f )). Similarly, stage III or IV ovarian cancer patients with tumors that were BRCA1/ 2m had significantly better OS than patients with BRCA1/ 2 Journal of Oncology 2 wt tumors of the same stage (HR � 0.64 (95% CI � 0.55-0.75)) (Supplementary Figure 1(g)). ese results should also be interpreted with caution since many studies may not have fully documented the patient treatment history. In addition, the retrospective nature of the studies included in this analysis could be susceptible to selection bias, other potential biases, or confounding. Few studies evaluating OS and BRCA1/2m for other cancer types were found. Our analysis of pancreatic cancer patients did not detect any effects of BRCA1/2m on OS (Figure 2(c)). Taken together, these results display discordant behavior of BRCA1/2m between patients with breast or ovarian cancer, suggesting possible tumor-intrinsic properties of ovarian cancers that combine with the presence of BRCA1/2m to lead to longer survival.

Overall Survival and BRCA1m
. Similar to what we found in breast cancer patients with BRCA1/2m, we found no significant association in OS in such patients harboring only BRCA1m (HR � 1.12 (95% CI � 0.96-1.13)) ( Figure 3 Finally, we performed a meta-analysis comparing the effects of BRCA2m on OS in prostate cancer patients, which suggested that prostate patients with BRCA2m had worse OS than BRCA2 wt (HR: 1.85 (95% CI � 1.07-3.21)). Similar findings were observed for subgroup analyses by germline mutations or pathogenic mutations (Figure 4

Overall Survival and HRRm.
We next performed a metaanalysis of studies reporting survival outcomes in patients with one, two, and three or more HRRm, as different gene lists and methodology were used across HRRm studies.

Overall Survival and HRD Status.
Analysis of four studies in breast, ovarian, and pancreatic cancer patients suggested a trend of increased OS with HRD positivity (HR � 0.67 (95% CI � 0.43-1.02)) ( Figure 6(a)). However, these results should be interpreted with caution. Similarly, we were only able to identify three studies that reported outcomes in patients with ovarian or gastric cancer using an alternate definition of HRD as a genomic instability score ≥42 regardless of BRCAm status, demonstrating an HR of 0.66 (95% CI � 0.51-0.85) (Figure 6(b)). Given the limited power of this meta-analysis, it is not possible to draw conclusions based on these results.

Publication
Bias. In addition, the funnel plot showed no clear evidence of publication bias in any of these BRCA1/2, HRR, or HRD meta-analyses (data not shown).

Discussion
e prognostic value of pathogenic mutations in BRCA1/2 and other HRR genes and HRD positivity is not fully understood in cancer patients not treated with PARPi. In this systematic review and meta-analysis, the presence of BRCA1/2m was significantly associated with better OS in ovarian cancer patients, but not in other cancer types. e results should be interpreted with caution due to the lack of full-treatment history and potential bias and confounding of included retrospective studies. For HRRm, no associations between HRRm and OS were observed across studies. In subgroup analyses, we observed a positive association between ATMm and urothelial cancer, but inverse associations between HRRm and pancreatic and ovarian cancers. ese findings should be interpreted with caution due to a limited number of studies and studies with variable design and quality.
e UK prospective cohort (POSH) study assessed OS in 2,733 women below 40 years of age at first diagnosis with invasive breast cancer harboring BRCA1/2m. ose researchers found insignificant differences in survival, but triple-negative breast cancer patients harboring BRCA1/2m had a slight survival advantage in the first several years following their primary diagnosis [13]. A second study comparing 3,345 women with stages I-III breast cancer, 233 of who harbored a BRCA1m, also found a survival rate similar to that in women without that mutation, with improved survival following oophorectomy [14]. A third study in early onset triple-negative breast cancer patients assessing the presence of BRCAm and OS showed better outcomes, likely due to the increased response to anthracyclines and taxane-based chemotherapies [15]. Finally, a fourth study evaluating the outcomes of germline BRCA1/2m in patients with advanced high-grade serous ovarian cancer revealed longer progression-free survival compared to patients without germline mutations [16].
ese studies echo our findings in both breast (Figure 2(a)) and ovarian (Figure 2(b)  germline BRCA1/2m, as well as in our analysis of breast and ovarian cancer patients with BRCA1m or BRCA2m alone, helping to validate our meta-analysis results across cancer types. A recent study linking HRD scores in samples from patients with high-grade serious ovarian cancer (HGSOC) with clinical prognosis found that BRCA1/2m was more common in those patients' samples with HRD scores ≥63 and had a better prognosis compared to patients with HDR scores ≤62. HRD caused by gene alterations was associated with a better prognosis than HRD caused by epigenetic changes or unknown variant changes [17]. Similarly,     Journal of Oncology another study used an HDR score cutoff of ≥33 (≥42 is currently used in the clinic) and found that this cutoff was associated with better OS in patients with epithelial ovarian cancer [18]. Analysis of HRRm other than BRCA1/2m in the TCGA project revealed that only patients with homozygous deletions in CHEK1 and PTEN showed high HRD scores, but not patients with mutations in ATM, ATR, FANCA, FANCD2, FANCM, or PALB2 [17]. Long-term survival in these patients depended on aggressive debulking of their primary disease, suggesting that patients with pathogenic HRRm should undergo surgical resection. Similarly, our meta-analysis also showed that patients with mutations in HRR genes did not have worse OS ( Figure 5(b)), and HRD positivity was associated with improved outcomes (Figures 6(a) and 6(b)). As HRD positivity is increasingly being used as a means to guide the use of neoadjuvant chemotherapy, PARPi, or other targeted therapy across multiple cancer types, our metaanalysis suggests prognostic differences in these patients with HRD positivity, which may be targetable to improve outcomes [2,19]. While the present study has limitations of small numbers of publications in certain subgroups, a relatively large number of publications with BRCA1/2m in breast and ovarian cancers were identified but not in other cancer types. Additionally, there were insufficient data points for HRR genes to perform a meta-analysis for each gene and group genes together in the summary. Furthermore, the methods used to identify HRRm and HRD in tumors vary by study and continue to evolve, and assays utilizing RNA or methylation techniques were excluded. Due to the limited number of published studies and inconsistent methodology and definitions of HRRm and HRD positivity, these results should be interpreted with caution.
It is puzzling that the results are not in consistent directions. is is the value of this comprehensive review that we brought to the literature. It is the reason that the treatments have to be tested in each of the cancers, their subtypes, and their mutation profiles. We have added this in the text. In addition, different studies could use different lists to determine pathogenic or VUS for BRCA mutations. ere are some common rules: truncation mutations (stop gained, frameshift, and splice site), homozygous copy number deletions, and functional rearrangements. However, for those missense mutations, due to database curation and updating, the list to determine pathogenic or VUS could be different. Studies typically do not report such details in their publications. Given this is a literature review, we rely on the original papers regarding the definition of pathogenic and VUS mutations, as one of the limitations of the literature review.
In conclusion, this systematic review and meta-analysis evaluated the prognostic value of BRCA1/2 and HRR pathway gene mutations and HRD positivity in multiple cancers. ese findings should prove useful in improving the precision and efficacy of treatment selection in clinical oncology. Given the significantly improved outcomes following treatment with PARPi and augmented synthetic lethality to platinum agents in patients with tumors that have markers of HRD, the development and standardization of biomarker assays could have important clinical implications in discovering novel synthetic lethal combination therapies to improve outcomes for cancer patients.

Data Availability
e data used to support the findings of this study are publicly available and listed in the supplementary material of this article.  Supplementary Figure 1(a). BRCA1 and BRCA2: a metaanalysis of OS among breast cancer patients with germline tumor testing only. Supplementary Figure 1(b