Malignant gliomas account for approximately 70% of adult malignant primary brain tumors in the United States and are associated with median survival of only 12 to 15 months among patients with glioblastoma, the most common type of glioma [1, 2]. Research on the etiology of glioma over the past several decades has yielded few consistent findings; the only established environmental risk factor is exposure to therapeutic or high-dose ionizing radiation [3–5]. Studies of genetic susceptibility to glioma have been similarly inconsistent, although three recent genome-wide association studies (GWAS) have replicated findings at six susceptibility loci in TERT, CCDC26, CDKN2A-CDKN2B, RTEL1, PHLDB1, and EGFR[6–8]. Despite this progress, the majority of the risk of malignant gliomas remains unexplained.

Given the established link between ionizing radiation and glioma, it has been hypothesized that genetic variants of the DNA repair pathway may affect susceptibility to the disease. Specifically, x-ray cross complementing group 1 (XRCC1) is an important component of the base excision repair system, which is the predominant DNA repair pathway for small base lesions resulting from oxidation and alkylation damage [9]. The 33 kb XRCC1 gene, located on chromosome 19q13.2-13.3, encodes a scaffolding protein which coordinates numerous protein-protein interactions, including with DNA ligase III and DNA polymerase at the site of damage [10, 11].

Although over 300 single nucleotide polymorphisms (SNPs) have been identified in XRCC1 (http://www.ncbi.nlm.nih.gov/projects/SNP), the most extensively studied has been rs25487 in codon 399 of exon 10, a non-synonymous G → A polymorphism changing arginine to glutamine. The frequency of the putative risk allele A has been shown to be 0.366, 0.274, and 0.111 in European American, east Asian, and African HapMap populations, respectively [12]. It has been suggested that this amino acid change results in deficient DNA repair, rendering carriers of the variant allele more susceptible to damage by environmental carcinogens [13, 14]. The 399 G → A polymorphism has been shown to be associated with risk for numerous cancers including lung and prostate cancer among Asians, as well as breast cancer in both Asian and Caucasian populations [15–20]. As studies of the association of 399 G → A with the risk of glioma have been inconsistent, a systematic review and meta-analysis of such studies was conducted following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [21]. The aim was to elucidate the role of the polymorphism in glioma and to explore sources of heterogeneity among the identified studies.

Studies of the XRCC1 protein have demonstrated that it plays an important role in repairing single strand DNA breaks by participating in many processes including detecting such breaks and stabilizing proteins involved in the base excision repair system [35]. Because of the crucial role the protein plays, the effect of DNA variation in the coding sequence has been extensively explored in cancer and many other diseases. Studies have shown that the specific G → A substitution at codon 399 can induce conformational changes including the loss of α-helical structures that may be critical for effective protein-protein interactions [36], and that such variation corresponds with significantly reduced repair rates of irradiation-induced DNA damage [37].

Given the association of 399 G → A polymorphisms with other cancers [29] and the established risk of brain tumor development induced by ionizing radiation [3–5], these variants have been explored in numerous studies for association with glioma yet have yielded conflicting results. Despite the biologic plausibility of an association between XRCC1 399 G → A and risk of glioma, this meta-analysis does not provide evidence of such an association. The initial meta-analysis demonstrated a significant association of the polymorphism with glioma among Caucasians (OR = 1.17, 95% CI 1.05-1.30, P = 0.006) with high corresponding among-study heterogeneity (I ²  = 87%). However, an influence analysis conducted by sequentially removing each study and recalculating the association and heterogeneity demonstrated that the study by Yosunkaya et al. [34], the smallest contributor to the meta-analysis (Figure 2), was disproportionately driving the apparent association and among-study heterogeneity. While the reported methodology of this study did not raise specific concerns, the observed odds ratio employing a dominant model of 7.09 (95% CI 3.83-13.11) is many times higher than has been observed for the association of this polymorphism and any cancer, and is almost certainly a statistical artifact of the small study sample. After exclusion of this study, there was not a statistically significant association demonstrated between the 399 G → A polymorphism and glioma (OR = 1.10, 95% CI 0.98-1.23, P = 0.10) among Caucasians, reduced heterogeneity (I ²  = 7%), and there were no notable associations in stratified analyses.

While it is not possible to rule out the potential for bias in the identified studies, the five studies on which the final meta-analysis was based were deemed to be free of any major sources of bias which would compromise their validity. All studies were sufficiently clear in their description of recruitment methodology and participant characteristics, with the exception of reporting participant age in three studies [22, 27, 30]. All of the studies but one [25] discussed implementation of quality control methods during genotyping, such as use of control samples and replicates to assess concordance. However, only one study [27] reported blinding laboratory personnel to the case/control status of the samples. All studies reported matching controls to cases on age and gender, and all studies except one [25] reported matching on race/ethnicity or geographical location.

Although there was no indication of systematic bias among the final eligible studies as described above, this meta-analysis may have been limited by the small sample sizes of some of the included studies, as well as the lack of data for some planned subgroup analyses. In particular, despite research indicating that the genetic landscape of Grade III versus Grade IV gliomas may differ considerably [38], only two of the included studies in this meta-analysis reported associations for the subtypes separately [22, 30]. This limited our ability to explore associations with the XRCC1 399 G → A polymorphism that may be pertinent to one subtype but not the other. Future investigators should report genotype counts for glioma subtypes separately to permit stratified analyses. Despite these potential limitations, the current literature suggests no statistically significant association between the XRCC1 399 G → A polymorphism and glioma.