Association of ERCC1 C8092A and ERCC2 Lys751Gln Polymorphisms with the Risk of Glioma: A Meta-Analysis

Objectives To comprehensively evaluate the association of ERCC1 C8092A and ERCC2 Lys751Gln polymorphisms with the risk of glioma. Methods Potential studies were searched and selected through the Pubmed/MEDLINE, EMBASE, the China National Knowledge Infrastructure (CNKI) platforms, WanFang and VIP database up to June 2013. Two investigators independently reviewed full text and included studies met inclusion criteria. Combined odds ratios (ORs) and 95% confidence intervals (95% CIs) were calculated in a fixed-effects model or a random-effects model according to results of heterogeneity test. All analyses were performed by Revman 5.2 and Stata 10.0 software. Results A total of 10 studies were included in our meta-analysis, including 3,580 glioma patients and 4,728 controls. Overall, ERCC1 C8092A polymorphism was associated with the risk of glioma (AA vs. CC: OR = 1.29, 95%CI: 1.07–1.55, P = 0.01; recessive model: OR = 1.29; 95% CI: 1.07–1.55, P = 0.01). When stratified by ethnicity, significant association was only observed in the Chinese population (AA vs. CC: OR = 1.37, 95%CI: 1.03–1.81, P = 0.03; recessive model: OR = 1.34; 95% CI: 1.02–1.75, P = 0.04). For ERCC2 Lys751Gln polymorphism, no significant association was found between ERCC2 Lys751Gln polymorphism and the risk of glioma in different genetic models. A significant association of ERCC2 Lys751Gln polymorphism with the risk of glioma was identified in the Caucasian population under recessive model (OR = 0.87; 95% CI: 0.78–0.98, P = 0.02), but not in the Chinese population. Conclusion This meta-analysis suggested that the AA genotype of ERCC1 C8092A polymorphism might increase the susceptibility of glioma in the Chinese population. And the TT genotype of ERCC2 Lys751Gln polymorphism may decrease the risk of glioma in the Caucasian population. But the small number of studies and moderate methodological quality require cautious interpretation of the study results.


Introduction
Glioma is one of the most common brain tumors, accounting for approximately 80% of all brain tumors [1]. In a multi-center cross-sectional study in China, it is estimated that the agestandardized prevalence of primary brain tumor is 22.52 per 100,000 (95%CI: 13.22-31.82) for the entire population, and the overall prevalence in females (27.94 per 100,000) was higher than that in males (17.64 per 100,000). Of 272 newly diagnosed brain tumors, glioma accounted for 29.78% [2]. Although the prevalence of glioma is relatively lower than other cancers, survival rates of glioma patients is still poor, especially the 5 year survival rate of glioblastoma case, less than 3% [3,4,5].
Like other cancers, the development and progression of glioma are also determined by genetic and environmental factors. Currently, there are several confirmed environmental risk factors, including ionizing radiation, ultraviolet (UV) rays, diet and smoking. It is evidenced that the low exposure of ionizing radiation and ultraviolet (UV) rays would reduce the susceptibility of glioma [6,7]. Ionizing radiation and UV rays can result in several kinds of DNA damage, such as oxidative DNA damage, and single-and double-strand breaks in DNA chains [8]. Consequently, DNA damages would result in glioma development and progression. However, major DNA repair pathways exist in the human body for preventing DNA damage and mutagenesis including base-excision repair (BER), nucleotide excision repair (NER), and homologous recombination repair (HRR) [9]. Due to the strong contributions to genomic integrity and fidelity that DNA repair genes encompass, many studies have focused on polymorphisms within them in relation to glioma susceptibility. Among these genes, excision repair cross-complementing group 1 (ERCC1) and ERCC2 gene are considered to be related with the susceptibility of glioma due to their roles as rate limiting enzymes in association to NER. The NER pathway shows effective and prominent repair on bulky DNA lesions and UV damage [10,11].
Several single nucleotide polymorphisms (SNPs) have been identified in ERCC1 and ERCC2 genes. But ERCC1 C8092A and ERCC2 Lys751Gln polymorphisms have significant influences on gene expression and DNA repair functions. ERCC1 C8092A polymorphism, located in the 39-untranslated region of the ERCC1 gene, has been shown to influence the level of ERCC1 mRNA expression and may affect mRNA stability for ERCC1 [12,13]. And individuals with ERCC2 Lys/ Lys 751 genotype had poorer repair proficiency based on the number of chromatid aberrations (OR = 7.5, 95%CI = 1.01-87.7), suggesting this polymorphism is associated with an increased risk of suboptimal DNA repair [14,15]. These two polymorphisms have been widely investigated in different ethnic populations. In a study on adult glioma cases and controls in a Caucasian population, it was presented that patients with oligoastrocytoma were significantly more likely to be homozygous in ERCC1 C8092A polymorphism than the controls (CC versus CA/AA: OR = 4.6, 95% CI: 1.6-13.2), however, no significant results were found in other histologic types of glioma [16]. Another study in the Caucasian population observed that ERCC1 C8092A polymorphism was associated with risk of glioma in recessive model (AA versus CC/CA: OR = 1.86, 95%CI: 1.01-3.46) [17], however, no significant findings were reported in other studies in the Caucasian or other ethnic populations [18,19,20]. In a Chinese population the variants of ERCC2 Lys751Gln polymorphism was significantly different between glioma cases and controls (x 2 = 6.015, P = 0.014) [21]. In addition, among Caucasians, glioma cases were significantly more likely than controls to be homozygous for ERCC2 Lys751Gln polymorphism (OR = 3.2, 95% CI: 1.1-9.3) [22].
However, other association studies did not show any positive results [17,18,23].
Evidently, numerous studies have been done to identify the association of ERCC1 C8092A and ERCC2 Lys751Gln polymorphisms with the risk of glioma in different ethnic populations. But the findings were not consistent. In this study, we performed a meta-analysis of the available studies in different ethnic populations to evaluate the effects of ERCC1 C8092A and ERCC2 Lys751Gln polymorphisms on the susceptibility to glioma.

Publication searching
We searched for relevant studies up to June 2013 in both English and Chinese through the Pubmed/MEDLINE, EMBASE, the China National Knowledge Infrastructure (CNKI) platforms, WanFang and VIP database with the following terms and their combinations: ''ERCC1'', ''ERCC2 or XPD'' or ''ERCC'', ''glioma'' and ''polymorphism or variant''. We tried to identify potential relevant studies from the whole reference lists by orderly reviewing title, abstract and full text.

Selection criteria
The inclusion criteria were as follows: a). Research focused on the association of ERCC1 C8092A (rs3212986) or ERCC2 Lys751Gln (rs13181) polymorphisms with the risk of glioma; b). Case-control studies; c). Genotype and allele data available. Studies were excluded for following reasons: a). unpublished papers, dissertations, conference articles, reviews and duplication of publications (select the study in a larger sample size); b). data unavailable for calculating genotype or allele frequencies; c).

Data extraction
All the following information was separately extracted by two investigators, including: the first author, year of publication, country (ethnicity), allele frequencies and genotype distributions in glioma cases and controls, number of glioma cases and controls. Extracted data were compared and cross referenced. Inconsistencies were discussed and corrected together.

Quality assessment
We evaluated the methodological quality of included studies using the Strengthening the Reporting of Genetic Association studies (STREGA) recommendations, extended the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist [24]. The resulting STREGA checklist provides additions to 12 of the 22 items on the STROBE checklist. Two investigators independently assessed the quality of all included studies to determine whether they meet each item or not.

Data analysis
All meta-analysis was performed by Review manager 5.2 software (The Cochrane Information Management System). The pooled odds ratios (OR) and 95% confidence interval (CI) were calculated for measuring the genetic association between ERCC1 or ERCC2 polymorphisms and the risk of glioma. The betweenstudy heterogeneity and the I 2 statistic for estimation of inconsistency was analyzed using the heterogeneity Q statistic test. The following cut-off points indicated different degrees of heterogeneity: I 2 = 0-25%, no heterogeneity; I 2 = 25-50%, moderate heterogeneity; I 2 = 50-75%, large heterogeneity; I 2 = 75-100%, extreme heterogeneity [25]. If there was no significant between-study heterogeneity (P $ 0.10), the method of Mantel-Haenszel was used to calculate the pooled OR (95%CI) in a fixed effects model. Otherwise, the DerSimonian-Laird method was performed for evaluating the pooled OR (95%CI) in a random effects model. Begg's funnel plot and Egger's test were performed to assess publication bias among the literatures by Stata 10 software. Funnel plots were used to evaluate publication bias. P, 0.05 was considered statistically significant.

Characteristics of eligible publications
The initial search identified a total of 40 studies for ERCC1 C8092A and ERCC2 Lys751Gln, of which 10 studies met the selection criteria [16,17,18,19,20,21,22,23,26,27]. Among the 30 excluded articles, two were dissertations, two were conference articles, three were reviews, 16 were not gene polymorphism studies, three were not association studies on the risk of glioma, two were not the polymorphism we studied, one was not a study on glioma and one was a replication (Figure 1).
Likewise, seven studies reported ERCC2 Lys751Gln, including 2,758 glioma cases and 3,847 controls (Table 1)  Main results of meta-analysis ERCC1 C8092A. The between-study heterogeneity was not significant in all the comparisons (P . 0.1). In addition, no between-study heterogeneity was identified using I 2 statistic (Table  2). Therefore, we used the fixed effects model to calculate the pooled OR (95%CI). A significant association was identified comparing AA with CC (OR = 1.29, 95%CI: 1.07-1.55, P = 0.01). Significant results were also observed in recessive model (OR = 1.29; 95% CI: 1.07-1.55, P = 0.01). There was no significant association in other genetic models. Subgroup analysis showed that ERCC1 8092AA was significantly associated with the risk of glioma compared with ERCC1 8092CC in Chinese population (Z = 2.19, P = 0.03), but was not found in Caucasian population (Z = 1.55, P = 0.12) (Figure 2). Similar results were obtained in recessive model (Figure 3).
ERCC2 Lys751Gln. For the recessive model, the betweenstudy heterogeneity was significant (P = 0.03), and large betweenstudy heterogeneity was identified by I 2 statistic (Table 2). Therefore, the random effects model was used to calculate the pooled OR (95%CI) for the recessive model. For other genetic models, the fixed effects model was performed for calculating pooled OR (95%CI). No significant association was found between ERCC2 Lys751Gln polymorphism and the risk of glioma in different genetic models. However, subgroup analysis showed that a significant association was identified comparing TT with Gcarriers in the Caucasian population (Z = 2.42, P = 0.02), but this was not found in the Chinese population (Z = 0.19, P = 0.85) ( Figure 4).
Next, stratified analyses for different types of glioma including glioblastoma multiforme and nonglioblastoma multiforme were performed in ERCC2 Lys751Gln polymorphism. There was no significant association in both glioblastoma multiforme and nonglioblastoma multiforme groups (Table 3).

Sensitivity analysis
To further confirm the combined results, a sensitivity analysis was conducted by changing the fixed or random effects model. It was shown that the combined results between fixed effects model and random effects model for ERCC1 C8092A polymorphism were no different. However, a significant change was identified in recessive model for ERCC2 Lys751Gln polymorphism. The combined OR (95%CI) was 0.88 (0.79-0.98) estimated by fixed effects model (Z = 2.38, P = 0.02). We then estimated the influence of individual studies on the combined results by omitting one study at a time. No significant differences were found in different genetic models for ERCC1 C8092A polymorphism. When deleting a study in the Chinese population [23], the ERCC2 Lys751Gln polymorphism showed significant effects on the risk of glioma, but the significance was much less (Z = 2.10, P = 0.04, OR = 0.89, 95%CI: 0.79-0.99). Therefore, our results were generally and statistically reliable.

Publication bias
In this study, Begg's test and Egger's test did not show any evidence of publication bias for both ERCC1 C8092A (Begg's test: z = 0.00, P = 1.00; Egger's test: t = 1.55, P = 0.18) and ERCC2 Lys751Gln polymorphisms (Begg's test: z = 0.30, P = 0.76; Egger's test: t = 1.03, P = 0.35). In the funnel plot ( Figure 5 and 6), studies of ERCC1 C8092A and ERCC2 Lys751Gln polymorphisms showed symmetric distribution. All studies were nearly located in the region of 95% CI, suggesting that publication bias was not significant in this meta-analysis.

Discussion
ERCC1 and ERCC2 protein, key elements in transcriptioncoupled NER pathway, are absolutely critical for efficient DNA repair capacity. These two proteins are essential for the maintenance of genetic stability, thus, deficits in ERCC1 and ERCC2 proteins are associated with cancer susceptibility [28]. Studies have found that ERCC1 and ERCC2 polymorphisms are associated with high levels of DNA adducts or lower DNA repair capacity [14,29,30,31]. Therefore, an increasing amount of studies attempt to identify the association of ERCC1 and ERCC2 polymorphisms on the susceptibility of glioma.
To comprehensively estimate the associations of ERCC1 C8092A and ERCC2 Lys751Gln polymorphisms with the risk of glioma in different ethnic populations, meta-analyses in 10 studies including 3,580 glioma patients and 4,728 controls was carried out. To our knowledge, this is to date the largest meta-analysis conducted for ERCC1 C8092A and ERCC2 Lys751Gln polymorphisms in glioma. The data from meta-analysis showed a significant increase in frequency of ERCC1 8092AA genotype in glioma patients than in controls, indicating that ERCC1 8092AA genotype contribute to increases the risk of glioma with combined OR of 1.29 (1.07-1.55). When the analysis was stratified by ethnicity, a significant association was identified in the Chinese population with combined OR of 1.37 (1.03-1.81). However, the significant association disappeared for Caucasians. Similar results were identified in the recessive model of ERCC1 C8092A polymorphism. Remarkably, none of the original researches in Chinese populations reported the significant association between ERCC1 C8092A polymorphism and the risk of glioma. Therefore, the positive results of this meta-analysis might have been caused by small sample size. However, ERCC1 C8092A polymorphism influence the expression of ERCC1 and the DNA repair function of NER pathway may be weakened, a possible explanation of our findings from the point of the pathogenesis [13].
For ERCC2 Lys751Gln polymorphism, no significant association between ERCC2 Lys751Gln polymorphism and the risk of glioma was observed. However, TT genotype in ERCC2 Lys751Gln polymorphism may significantly decrease the risk of glioma in the Caucasian population with combined OR of 0.87 (0.78-0.97), but not in the Chinese population. Although 2230 glioma patients and 3393 controls were included in the subgroup analysis in the Caucasian population, only one study with larger sample size presented significant results (OR = 0.84, 95%CI: 0.72-0.98), which may influence the combined results [18]. When this study was deleted from the subgroup analysis, significant results in the Caucasian population disappeared. Therefore, the finding between ERCC2 Lys751Gln polymorphism and the risk of glioma in Caucasian population should be addressed with caution. Furthermore, a stratified analysis for different types of glioma in ERCC2 Lys751Gln polymorphism was performed, and no significant association was identified in either groups. However, only two studies provided the available data for glioma classification. In addition, glioma patients were only separated into two groups, including glioblastoma multiforme and nonglioblastoma multiforme. Therefore, we cannot explicitly estimate the effects of ERCC2 Lys751Gln polymorphism on the risk of different kinds of glioma in our study. Table 3. Stratified analysis of the association of ERCC2 Lys751Gln polymorphism with the risk of different type of glioma. In this study, the heterogeneity Q statistic test and the I 2 statistic were used to test the between-study heterogeneity. For all studies, the significant heterogeneity existed only under recessive model in ERCC2 Lys751Gln polymorphism. When the studies were stratified according to ethnicity, significant heterogeneity still existed in the Chinese population, but no statistically significant heterogeneity was observed in the Caucasian population. This suggests that significant heterogeneity might be attributed to differing ERCC2 Lys751Gln polymorphism by ethnicity.
Sensitivity analysis was performed by changing the fixed or random effects model and omitting one study each time to observe the change of effects. For ERCC1 C8092A polymorphism, no significant change was observed. However, the effect of ERCC2 Lys751Gln polymorphism on the risk of glioma changed significantly, though the combined OR was very close to 1. We then looked for publication bias using Begg's test, Egger's test and funnel plot. No significant publication bias existed in this metaanalysis. This indicated that the results in this meta-analysis were generally reliable.
Although publication bias was not identified in this metaanalysis, there are some potential limitations. First, only ten eligible studies were included in this meta-analysis. Therefore, in the subgroup analyses by ethnicity, the number of cases and controls was relatively small, which may lead to low statistical power in identifying the association. Furthermore, different pathological types of glioma were not considered, which may be the source of between-study heterogeneity. Unfortunately, only two studies in Caucasian populations provide genetic data of ERCC2 Lys751Gln polymorphism for glioblastoma multiforme and nonglioblastoma multiforme, respectively. Lastly, our meta-  analysis was largely performed by unadjusted estimates, because of the limitations in selected studies that presented adjusted estimates. Although adjusted estimates were shown, the estimates were not adjusted by the same confounders. It was also difficult to present the combined estimates by adjusted potential confounders.
In conclusion, our meta-analysis suggested that ERCC1 8092AA genotype was associated with the higher susceptibility of glioma in the Chinese population. However, TT genotype of ERCC2 Lys751Gln polymorphism might decrease the risk of glioma in Caucasian population. Due to the small number of studies and moderate methodological quality, the results require cautious interpretation. Studies with larger sample size and more specified information in pathological types of glioma are needed to confirm our results for both Chinese and Caucasian populations.