Associations between XRCC3 Thr241Met polymorphisms and breast cancer risk: systematic-review and meta-analysis of 55 case-control studies

Background The X-ray repair cross-complementing group 3 (XRCC3) is an efficient component of homologous recombination and is required for the preservation of chromosomal integrity in mammalian cells. The association between Thr241Met single-nucleotide polymorphism (SNP) in this gene and susceptibility to breast cancer has been assessed in several studies. Yet, reports are controversial. The present meta-analysis has been designed to identify whether this SNP is associated with susceptibility to breast cancer. Methods We performed a systematic review and meta-analysis for retrieving the case-control studies on the associations between T241 M SNP and the risk of breast cancer. Crude odds ratios (ORs) and 95% confidence intervals (CIs) were calculated to verify the association in dominant, recessive, and homozygote inheritance models. Results We included 55 studies containing 30,966 sporadic breast cancer cases, 1174 familial breast cancer cases and 32,890 controls in the meta-analysis. In crude analyses, no association was detected between the mentioned SNP and breast cancer risk in recessive, homozygote or dominant models. However, ethnic based analysis showed that in sporadic breast cancer, the SNP was associated with breast cancer risk in Arab populations in homozygous (OR (95% CI) = 3.649 (2.029–6.563), p = 0.0001) and recessive models (OR (95% CI) = 4.092 (1.806–9.271), p = 0.001). The association was significant in Asian population in dominant model (OR (95% CI) = 1.296, p = 0.029). However, the associations was significant in familial breast cancer in mixed ethnic-based subgroup in homozygote and recessive models (OR (95% CI) = 0.451 (0.309–0.659), p = 0.0001, OR (95% CI) = 0.462 (0.298–0.716), p = 0.001 respectively). Conclusions Taken together, our results in a large sample of both sporadic and familial cases of breast cancer showed insignificant role of Thr241Met in the pathogenesis of this type of malignancy. Such results were more conclusive in sporadic cases. In familial cases, future studies are needed to verify our results. Electronic supplementary material The online version of this article (10.1186/s12881-019-0809-8) contains supplementary material, which is available to authorized users.


Background
Breast cancer ranks first among all women's cancers regarding its incidence and rank second among them regarding its cancer-related mortality rate [1]. Several genetic and environmental factors have been associated with breast cancer risk. Among the most relevant factors is the ability to repair DNA double strand break (DSB). The homologous recombination (HR) and the non-homologous end-joining (NHEJ) pathways have been developed in eukaryotic cells for repair of such defects [2]. Numerous single nucleotide polymorphisms (SNPs) within genes coding the NHEJ pathway have been associated with breast cancer risk [3]. More importantly, the mostly recognized breast cancer susceptibility genes BRCA1 and BRCA2 participate in the process of HR. Deficiencies in HR have been detected both in BRCA1/2 germline mutation-associated and remarkable fraction BRCA1/2 wild-type breast cancer patients [4]. The X-ray repair cross-complementing group 3 (XRCC3) is an efficient component of HR and is required for the preservation of chromosomal integrity in mammalian cells [5]. Consequently, it has been regarded as a supposed candidate gene for breast cancer susceptibility. However, the data regarding its participation in breast cancer risk are inconsistent. Hang et al. conducted a meta-analysis of 48 case-control studies (including 14 studies in breast cancer) and reported that XRCC3 Thr241Met significantly increased risk of breast cancer. However, they suggested that a single larger study should be performed to assess tissue-specific cancer risk in different ethnicities [6]. Garcı'a-Closas et al. meta-analyzed the studies in Caucasian populations (10,979 cases and 10,423 controls) and reported a weak association between homozygous variants for XRCC3 Thr241Met and risk of breast cancer. They concluded that this variant is implausible to have a considerable role in breast cancer risk. However, they suggested studies with larger sample sizes to assess probable underlying gene-gene interactions or associations in ethnic-based subgroups [7]. Lee et al. in their meta-analysis of 12 studies demonstrated that Thr/Met and Met/Met weakly elevated the risk of breast cancer compared to Thr/Thr genotype [8]. Economopoulos et al. conducted a meta-analysis on 20 case-control studies in non-Chinese individuals and three case-control studies on Chinese individuals and reported association between T allele of this polymorphism (corresponding to Met) and breast cancer risk in recessive model. However, the association was only detected in non-Chinese population [9]. He et al. reported the mentioned association in recessive and additive models, but suggested conduction of a study with the larger sample size to assess gene-environment interaction [10]. In another study, He et al. have conducted a meta-analysis of 157 case-control studies including 34 studies in breast cancer (22,917 cases and 24,313 controls) and suggested the XRCC3 Thr241Met as a susceptibility locus for breast cancer, especially in Caucasians [11]. Mao et al. demonstrated a significantly higher risk of breast cancer in heterozygote model but not in other models. Such association was significant in Asians. Based on the reported weak association, they suggested conduction of a study with larger sample size [12]. Finally, using 23 case-control studies, Chai et al. reported association between the mentioned polymorphism and breast cancer risk, especially in Asian populations and in patients without family history of breast cancer [13].
Therefore, according to inconclusive results of the previous meta-analyses and lack of systematic review in this regard, we conducted a systematic review and meta-analysis to assess the association between the Thr241Met SNP (rs861539) within XRCC3 and breast cancer risk in diverse inheritance models.

Registration
We conducted the present systematic review protocol according to the preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) [14]. We also registered the study protocol on the international prospective register of systematic review (PROSPERO) network. The registration number was CRD42018104217.

Information source and searching strategy
We searched PubMed, Scopus, EMBASE, Web of Science and ProQuest databases, the key journals (Breast Cancer Research and Treatment, Cancer Research), conferences/ congress research papers (as Grey literature) and the reference list of the included primary studies until March 2018 T(1990/01/01:2018/03/31) using the following syntaxes: "x-ray repair cross-complementing group 3" or "XRCC3"and"polymorphisms" or "single nucleotide polymorphism" and "breast tumor" or "breast cancer" and "rs861539" or "c.722C > T" or "p.Thr241Met" or "T241 M" (see Additional file 1). The complete search syntaxes were developed based on MeSH database and Emtree. The syntaxes for each database are shown in supplementary file. We did not implement any language restriction.

Eligibility criteria and selection process
We included: i) all observational studies such as crosssectional, case-control and cohort studies ii) studies that assessed associations between Thr241Met within XRCC3 and breast cancer risk. iii) Studies with available genotype frequencies in both case and control groups. We excluded books, reviews, editorial, letters and articles       Quality assessment and data extraction Methodological quality assessment (risk of bias assessment) was based on the Newcastle-Ottawa Scale (NOS). Checklist of each study was filled with two reviewers independently. Any disagreements (between two reviewers) were resolved by the discussion or consensus otherwise opinion of third expert reviewer. For assessing total quality status in primary study we used sum score of quality items. According to this score, we classified the papers in three groups (Good, Fair, Poor) [6]. Data was extracted by two reviewers as described above. Dataincluded general information of studies, study eligibility, method, risk of bias assessment and results including odds ratio.

Data synthesis (meta-analysis)
All of data analyses were performed in two distinct groups of familial breast cancer and sporadic breast cancer. Data were analyzed using STATA 13 software. Association between the mentioned SNP and breast cancer risk were analyzed by pooling odds ratio (ORs) with 95% confidence interval (CIs) in three models including dominant (TM + MM vs.TT), recessive (MM vs. TM + TT), and homozygote (MM vs.TT) models using STATA metan module. Z test was applied to assess the significance of the ORs, The heterogeneity between included publications was evaluated using I 2 parameter as described previously [14] where the higher values indicate higher level of heterogeneity. Furthermore, we checked heterogeneity by the chisquare-based Q-test (Heterogeneity was considered statistically significant if p < 0.05) (Egger et al., 1997). We combined genotype frequencies to calculate univariable (crude) OR. In addition, combination of adjusted OR values was based on the similarity of adjusted OR values restricted in two models including age-adjusted (association between rs861539 and breast cancer after controlling age of patients) and age and other factors. The random-effects model was used to combine parameters acquired from discrete studies due to methodological variation. Sensitivity analyses were performed using leave-one-out sensitivity analysis to indicate the effect of the quality score on the results. Subgroup analyses were done for evaluating potential sources of heterogeneity based on ethnicity, case selection methods case group (hospital vs. population), methodological quality status (Good, Fair, Poor) and-case enrollment strategies (incident vs. prevalent).

Meta-analysis results
Initially, we conducted the analysis in the familial and sporadic studies after using the random-effects model. Random model was used for analysis of associations in three inheritance models based on its more conservative nature. Final results for familial and sporadic studies are shown in Tables 3 and 4. The forest plots for each model are depicted in Figs. 2 and 3.
No significant associations were detected between the mentioned SNP and breast cancer risk in any inheritance model either in familial or in sporadic breast cancer cases.
Subsequently, we appraised the associations based on the study-base for selecting case/control (society) subgroup (hospital-based vs. population-based). In sporadic cases, the associations were significant in populationbased studies in homozygote and recessive models (OR (95% CI) = 0.869 (0.796-0.950), p = 0.002 and OR (95% CI) = 0.868 (0.806-0.934), p = 0.0001 respectively). The Interaction p-value was calculated as 0.655 which shows inappropriateness of such subgroup analysis strategy. No significant associations were found in society-based analysis in familial cases (Additional file 2: Figure S1 and Additional file 3: Figure S2). We also assessed the associations in methodological quality subgroups (Based on NOS scores) and found no significant association in sporadic (Interaction p-value = 0.891) but in familial cases we found the association in studies with fair quality in homozygote and recessive models (OR (95% CI) = 0.504 (0.304-0.834), p = 0.008, OR (95% CI) = 0.515 (0.297-0.894), p = 0.018 respectively) (Additional file 4: Figure S3 and Additional file 5: Figure S4).
Finally, we evaluated associations based on the case enrollment strategy (Incident vs. Prevalent). No significant associations were detected either in sporadic or familial cases (Interaction p-value = 0.22) (Additional file 6: Figure  S5 and Additional file 7: Figure S6).

Publication bias
We conducted both Begg's funnel plot and Egger's test for appraisal of the publication bias in sporadic and familial studies separately. The calculated parameters are shown in Tables 3 and 4. Moreover, the outlines of the funnel plots were rather symmetric implying absence of any significant publication bias (Figs. 6 and 7).

Adjusted OR
As we did not detected any association between the mentioned SNP and breast cancer risk in crude analysis, we subsequently assessed associations considering the effects of confounder variables using adjusted ORs. We retrieved adjusted ORs and confounder variables from the publications. Subsequently, we categorized confounder variables to two groups: 1. Age 2. Other variables including body mass index, smoking, hazardous life style and contraceptive use. Analyses were performed in sporadic subgroup based on the three inheritance models (Fig. 8). There was no significant association between this SNP and risk of sporadic breast cancer in any inheritance model considering adjusted ORs.

Sensitivity analysis and cumulative meta-analysis
To assess the strength of the association results, we conducted a leave-one-out sensitivity analysis by repeatedly removing one study at a time and re-measuring the summary OR. The summary ORs did not change, showing that our results were not originated from any certain study (Table 1).

Discussion
In the present meta-analysis, we assessed the associations between Thr241Met SNP and familial/ sporadic breast cancer based on the results of 55 studies containing 30,966 sporadic breast cancer cases, 1174 familial breast cancer cases and 32,890 controls. Crude analyses revealed no associations. In spite of assessing potential confounder variables and adjusting odds ratio of the primary studies, we did not find any association.
In sporadic cases, the narrow confidence intervals indicate the high power of the meta-analysis, so the results are conclusive. However, in familial cases, the wide confidence intervals imply that further studies are needed to reach conclusive results. Based on such findings, we predict that inclusion of further studies would not change the results of the meta-analysis. Sensitivity analyses by repeatedly removing one study at a time showed that the results of crude analysis were consistent result, therefore signifying the robustness of the study according to sensitivity analysis results, no relation between quality of studies with results and non-considerable publication bias.
Another strong point of our study was that we considered adjusted ORs to control the effects of confounding variables. Such approach further verified our results.
Through calculation of Interaction p values we determined subgroup analysis based on ethnicity as being the most strategy in this regard. Ethnic based analysis showed that in sporadic breast cancer, the SNP was associated with breast cancer risk in Arab and Mixed populations in homozygous and recessive models. The association was significant in Asian population in dominant model. However, no associations were detected in familial breast cancer in any ethnic-based subgroup and any inheritance model. The detected associations between this SNP and risk of sporadic breast cancer in certain populations had wide confidence intervals which necessitate extra studies. The same situation has been seen in familial breast cancer cases in ethnic-based subgroup analyses.
Chai et al. have performed a meta-analysis of 23 casecontrols studies on association between Thr241Met SNP and breast cancer. Their meta-analysis of the pooled data of 13,513 cases and 14,100 controls association between the mentioned SNP and breast cancer risk in recessive and homozygote models in total populations as well as within Asian populations [14]. Our study had the advantage of including higher numbers of cases and controls and assessment of adjusted ORs and sensitivity analysis. The results of our ethnic-based analysis were consistent with their results regarding the observed association in Asian population but not regarding the associated model. Although they found association between this SNP and risk of sporadic breast cancer, we disapprove such association based on the obtained conclusive results.
In brief, we have implemented the high quality systematic review and meta-analysis including comprehensiveness (inclusion of 5 databases), inclusion of grey literature (theses) and duplicate implementation of all steps of systematic review and meta-analysis (independent implementation of search, screening, selection, quality assessment and data extraction by two authors). In addition, priori principle (establishment and registration of protocol) was applied.
Our study had some limitations. Based on the unavailability of sufficient data from the primary studies, we could not assess the association between the mentioned SNP and breast cancer risk in pre−/post-menopause subgroups. In addition, the adjusted OR values of the primary studies were based on different parameters which might influence the validity of this kind of statistical analysis. Finally, there were some limitations in the primary studies and we did not find any genotyping data according to breast cancer subtypes except for 3 studies in triple negative breast cancer. Due to the low number of primary studies, the result of meta-analysis based on breast cancer subtypes was not reliable. So, we did not performed this type of analysis.

Conclusion
Taken together, our results in a large sample of both sporadic and familial cases of breast cancer showed insignificant role of Thr241Met in the pathogenesis of this type of malignancy. Such results were more conclusive in sporadic cases. In familial cases, future studies are needed to verify our results.