DNA Repair Gene XRCC1 Polymorphisms and Head and Neck Cancer Risk: An Updated Meta-Analysis Including 16344 Subjects

Background DNA repair gene X-ray repair cross complementing group 1 (XRCC1) plays an important role in the maintenance of the genomic integrity and protection of cells from DNA damage. Sequence variation in XRCC1 gene may alter head and neck cancer (HNC) susceptibility. However, these results are inconclusive. To derive a more precise estimation of the relationship between XRCC1 polymorphism and HNC risk, we undertook a meta-analysis involving 16,344 subjects. Methods A search of the literature by PubMed, Embase, Web of Science and China National Knowledge Infrastructure was performed to identify studies based on the predetermined inclusion criteria. The odds ratio (OR) with 95% confidence interval (CI) was combined using a random-effects model or a fixed-effects model. Results Twenty-nine studies consisting of 6,719 cases and 9,627 controls were identified and analyzed. Overall, no evidence of significant association was observed between XRCC1 Arg194Trp, XRCC1 Arg280His, XRCC1 Arg399Gln genotypes and the risk of HNC in any genetic models. Subgroup analyses according to ethnicity, tumor site, publication year, genotyping method also detected no significant association in any subgroup, except that oral cancer was associated with Arg194Trp variant in recessive model. Furthermore, no significant effect of these polymorphisms interacted with smoking on HNC risk was detected but Arg194Trp homozygous variant. Conclusion In conclusion, this meta-analysis suggests that the XRCC1 Arg194Trp, Arg280His and Arg399Gln polymorphism may not involve in HNC susceptibility. Further studies about gene-gene and gene-environment interactions in different populations are required.


Introduction
Head and neck cancer (HNC) is now the fifth most common type of cancer in the world [1], with approximately 434,000 new patients diagnosed annually worldwide [2]. Most of the cases involving new patients occur in economically developing countries, such as India, Brazil, and Thailand [3,4]. HNC is generally divided into three groups: oral cavity, pharynx, and larynx. Evolvement of HNC is a multifactorial process associated with various risk factors. Accumulative evidence indicates that tobacco smoking, drinking alcohol, and chewing betel quid are three major risk factors for HNC [5,6]. These environmental carcinogens may induce a defective DNA damage response, which may lead to apoptosis or may result in genomic instability and un-regulated (proliferative) cell growth [7][8][9]. The DNA repair system aims to maintain genomic integrity, and constantly challenge the environmental insults and replication errors. Therefore, the alteration of DNA repair genes could increase the risk of carcinoma in the head and neck [10]. Three important DNA repair pathways, including nucleotide excision repair (NER), base excision repair (BER), and double strand break (DSB), are involved in this process. The x-ray repair cross-complementing group 1 (XRCC1) involved in the BER pathway is thought to play a key role in protecting the genome from a variety of risk factors. Three common single nucleotide polymorphisms in the XRCC1 gene, including Arg194Trp (C to T substitution at exon 6 resulting in an Arg to Trp amino acid change), Arg280His (G to A substitution at exon 9 resulting in an Arg to His amino acid change), and Arg399Gln (G to A substitution at exon 10 resulting in an Arg to Gln amino acid change) are most commonly tested in many studies that have examined different populations.
Multiple studies have evaluated the association of HNC risk with polymorphism in the DNA repair genes XRCC1 Arg194Trp, XRCC1 Arg280His, and XRCC1 Arg399Gln. However, these results are inconsistent. While no association between XRCC1 polymorphisms and HNC risk was demonstrated in some studies [11,12], but Ramachandran et al. [13] and Olshan et al. [14] found a relationship between XRCC1 Arg194Trp and Arg399Gln polymorphisms and the risk of HNC. Olshan et al. [14] performed a stratified analysis to estimate the interaction between XRCC1 polymorphisms and smoking, suggesting that the Arg194Trp and Arg399Gln variants of XRCC1 were associated with the risk of HNC in those cases, but no association was found in Kumar's research [15]. Although Flores-Obando et al. [16] performed a metaanalysis in 2010 on the relationship between XRCC1 polymorphisms and the risk of HNC, subgroup analyses of smoking and genotyping method were not performed. Considering these conflicting results, we conducted an updated meta-analysis to deduce a reasonable conclusion about the relationship between XRCC1 polymorphisms and HNC risk. Subgroup analyses concerning ethnicity, smoking, site of HNC, publication year, and genotyping method were performed. Therefore, the current meta-analysis has a greater ability power to derive a more accurate conclusions than previous meta-analyses.

Selection
All of the studies met the following inclusion criteria: (1) published in English; (2) examined case-control studies estimating the relationship between XRCC1 polymorphism and the risk of HNC; (3) described genotype frequencies; (4) genotype distribution in controls must be in Hardy-Weinberg equilibrium (HWE); and (5) when duplicated studies were published by the same author obtained from the same patient sample, only the most complete publication study was included in this meta-analysis. Unpublished reports and abstracts were not considered.

Data extraction
The data were collected according to a standard protocol. The following information was extracted from each study: name of the first author, year of publication, country, genotyping methods, ethnicity and source of the cases and controls, characteristics of the sample population, and the genotype numbers from the cases and the controls.

Statistical analysis
We first tested for deviations from the Hardy-Weinberg equilibrium (HWE) in the control groups using the goodness-offit test (Chi-square test or Fisher exact test). The odds ratio (OR) with a corresponding 95% confidence interval (CI) was used to examine the association between XRCC1 polymorphism and HNC risk. The current meta-analysis used the following statistical models, the allelic genetic model, the codominant genetic model (homozygote comparison), and the recessive genetic model. Heterogeneity among the studies was assessed using the chi-square-based Q statistic (P<0.1 for the Q test indicates significant heterogeneity) [17]. We also quantified the effect of heterogeneity using the I 2 statistic [18]. Either the random-effects model (DerSimonian-Laird method [19]) or the fixed-effects model (Mantel-Haenszel method [20]) was used to calculate pooled effect estimates in the presence or absence of heterogeneity, respectively. Finally, potential publication bias was evaluated through Begg's test and Egger's test by visual analysis of the funnel plot [21,22]. P< 0.05 was considered statistically significant publication bias. Genotype frequencies in the control populations according to race were calculated and tests on the equality of proportions was performed for the Asian and Caucasian control populations in order to compare differences in genotype frequencies between the two groups. All statistical analyses were performed with the STATA version 10.0 software (Stata Corporation, College Station, TX).

Studies characteristics
As shown in Figure 1, the computerized search using the search strategy mentioned above delivered 38 publications. Of these, two papers were excluded due to the fact that they did not evaluate the association between HNC risk and XRCC1 polymorphisms [23,24]. Subsequently, five studies were excluded because of the lack of useful genotype data [25][26][27][28][29].
In total, 29 reports, consisting of 6,719 cases and 9,627 controls, matching the inclusion criteria were included in the present meta-analysis. The characteristics are summarized in Table 1. Of those 29 reports, 15 studies were performed on Caucasians, 10 studies were performed on Asians, and four studies were performed on a mixed population. In the 29 studies, 23 focused on the relationship between XRCC1 Arg194Trp polymorphism and HNC risk, 11 focused on Arg280His polymorphism, and 28 investigated the association XRCC1 Polymorphisms and HNC Susceptibility PLOS ONE | www.plosone.org between Arg399Gln polymorphism and HNC risk. In 19 studies, the controls were from a healthy population and in eight studies the controls were from a hospital population.
The distribution of XRCC1 Arg194Trp, XRCC1 Arg280His, and XRCC1 Arg399Gln polymorphism genotype frequencies between the HNC cases and the controls in the 29 studies are shown in Table 2. Noticeably, genotype distribution in the controls of Arg194Trp polymorphism in the study by Demokan et al. [36] and Arg399Gln polymorphism in the study by Dos Reis et al. [46] deviate from HWE, which are excluded in the subgroup analyses.

Meta-analysis results
The overall results of the meta-analysis for XRCC1 polymorphism and the risk of HNC are shown in Table 3. XRCC1 Arg194Trp polymorphism on HNC risk in total population. A total of 22 studies, including 4,487 cases and 6,873 controls, examining the association between XRCC1 Arg194Trp polymorphism and HNC risk were reviewed. There was significant difference in the frequency of the XRCC1 Arg194Trp polymorphism between Caucasians and Asians (34.42% vs. 12.27%, P<0.001). The pooled ORs for total population showed no evidence of a significant association between the variant genotypes of XRCC1 Arg194Trp and the risk of HNC in any genetic model. Significant heterogeneity was found in all genetic models. The forest plot is shown in Figure 2.
XRCC1 Arg194Trp polymorphism on HNC risk in a specific population. Stratified analysis by ethnicity was performed in order to determine the source of heterogeneity among the studies. No significant association of HNC risk with XRCC1 Arg194Trp polymorphism was detected in Asians and Caucasians under any genetic model ( Figures S1-S2). Significant differences between-study heterogeneities were found in the Asians, but they were not found in the Caucasians.
Oral cancer (OC) is the most common form of HNC and it is responsible for more than 90% of head and neck cancers [56].  Table 2. Distribution of XRCC1 genotypes among head and neck cancer cases and controls included in the meta-analysis.  (Table 1). No significant association between the XRCC1 Arg194Trp polymorphism and risk of OC was found in the allelic genetic model and the homozygote comparison, but a significant association was found for the recessive model ( Figure S3). Between-study heterogeneities were detected in the allelic genetic model and the recessive model, but it was not found to be significant in the homozygote comparison.

Gene Polymorphism First author (year) Cases (n) Controls (n) P-value of HWE in controls
Many studies have demonstrated that the interaction between XRCC1 polymorphism and environmental toxins could influence the risk of HNC. Considering that smoking is a major aspect of environmental toxins, we performed a subgroup analysis of six studies to investigate the influence that the interaction of tobacco smoke with XRCC1 polymorphism has on HNC risk. There was a significant association between the joint effect of smoking with XRCC1 Arg194Trp polymorphism and the risk of HNC under homozygote comparison ( Figure 3). No significant association was observed in the allelic genetic model and the recessive model ( Figure 3). Heterogeneity among the studies was not remarkable in any genetic model, except for the recessive model.
Flores-Obando et al. conducted a similar meta-analysis that included studies published before 2010. Considering the inconsistent results between the two studies, we decided to perform a stratified analysis by including studies published after 2010. The result showed no significant association was detected between Arg194Trp polymorphism and HNC risk in any genetic model ( Figure S4). Between-study heterogeneity was not remarkable in this stratified analysis.
The different genotyping methods used in the literature could cause the different genotyping results. Therefore, we performed a subgroup analysis by genotyping methods to investigate the relationship between Arg194Trp and HNC susceptibility. Neither the PCR-RFLP subgroup nor the TaqMan subgroup detected any significant association in the analyses for all genetic models ( Figures S5-S6). Moreover, heterogeneity among the studies was observed in the two stratified analyses under all genetic models, except for the homozygote comparison in the TaqMan subgroup. XRCC1 Arg280His polymorphism on HNC risk in total population. Eleven studies, including 2,972 cases and 3,714 controls, examining the relationship between XRCC1 Arg280His polymorphism and HNC risk were reviewed. There was significant difference in the frequency of the XRCC1 Arg280His polymorphism between Caucasians and Asians (25.75% vs. 9.04%, P<0.001). There was no significant association of HNC risk with variant genotypes of XRCC1 Arg280His in any genetic model. Significant between-study heterogeneity was absence in all genetic models. The forest plot is shown in Figure 4. XRCC1 Arg280His polymorphism on HNC risk in specific a population. We conducted subgroup analyses by ethnicity, tumor site, publication year, and genotyping method to estimate the relationship between XRCC1 Arg280His variant genotypes and the risk of NHC. However, no significant association was observed in any subgroup under different genetic models, and there was no significant heterogeneity among the studies in any stratified analysis (Figures S7-S12).  Only one study evaluated the influence of the interaction between smoking and Arg280His polymorphism on HNC risk; however, we are unable to conduct a further stratified analysis of that study. XRCC1 Arg399Gln polymorphism on HNC risk in total population. There were 27 studies, including 6,466 cases and 9,379 controls, that examined the association between HNC susceptibility and XRCC1 Arg399Gln polymorphism. There was significant difference in the frequency of the XRCC1 Arg399Gln polymorphism between Caucasians and Asians (41.28% vs. 44.65%, P=0.004). Overall, the association between variant genotypes of XRCC1 Arg399Gln polymorphism and HNC susceptibility was not significant under the allelic genetic model, homozygote comparison, and the recessive model. Between-study heterogeneity was detected in all genetic models. The forest plot is shown in Figure 5. XRCC1 Arg399Gln polymorphism on HNC risk in a specific population. In the subgroup analysis by ethnicity, no significant association between XRCC1 Arg399Gln polymorphism and HNC risk was found in Asians ( Figure S13) and Caucasians ( Figure S14). Heterogeneity among the studies was not remarkable in Caucasians; however, significant heterogeneity was detected in Asians under all genetic models.  Four studies, including 663 cases and 1,162 controls, were performed on OC population, and there was no significant association between XRCC1 Arg399Gln polymorphism and HNC susceptibility ( Figure S15). Between-study heterogeneity was found in all genetic models.
In the stratified analysis by smoking in the allelic genetic model and homozygote comparison, four studies were included and no significant association was found ( Figure S16). Seven studies were combined in the recessive model. However, we failed to derive a significant association between HNC risk and Arg399Gln genotype ( Figure S16). Heterogeneity among the studies was observed in all genetic models.
In stratified analysis by publication year of literature published from 2010-2012, significant heterogeneity was detected in all genetic models. Moreover, we found no association between Arg399Gln and HNC risk under any genetic model ( Figure S17).
In subgroup analysis of genotyping method, PCR-RFLP, TaqMan, and sequence analysis were used in the literature for genotyping of XRCC1 Arg399Gln polymorphism. The results showed no significant association between Arg399Gln and

Publication bias
Both Begg's test and Egger's test were performed to assess the publication bias of the literature. Visual analysis of the funnel plots did not present any evidence of obvious asymmetry for any genetic model in the overall meta-analyses of XRCC1 Arg194Trp, Arg280His, and Arg399Gln ( Figures  6-8). However, obvious evidence of publication bias was revealed in the XRCC1 Arg194Trp Asian group under all genetic models. In XRCC1 Arg399Gln smoking stratified analysis, potential publication bias was not revealed in Begg's test under any genetic model, but it was presented in the  Egger's test. Neither the Begg's test nor the Egger's test detected any obvious evidence of publication bias in other stratified analyses for all genetic models (Table 3).

Discussion
DNA repair mechanisms play a critical role in the protection of cells from DNA damage and in the maintenance of genomic integrity. The protein encoded by the XRCC1 gene is a scaffolding protein that associates with DNA ligase I, DNA ligase III, polynucleotide kinase (PNK), DNA polymerase β, and poly polymerase, which are parts of the DNA repair system. The interaction of XRCC1 with DNA ligase III could increase the endocellular stability of ligase. The joint effects of XRCC1 and PNK stimulate the 5'-kinase and 3'-phospatase activities. All of these conditions promote the repair of DNA. Therefore, sequence variation in the XRCC1 gene is suggested to alter cancer's susceptibility. The most common variant genotypes of XRCC1, including the Arg194Trp, Arg399Gln, and Arg280His genes, are described and a number of studies have investigated the genetic effect of the XRCC1 Arg194Trp, Arg280His, and Arg399Gln polymorphisms on HNC susceptibility with inconsistent results. This diversity motivates the current updated meta-analysis that may help us to explore a more robust estimate of the effect of XRCC1 polymorphism on the risk of HNC. In the present meta-analysis of 6,719 cases and 9,627 controls, no evidence of a significant association between HNC susceptibility and any type of XRCC1 variant genotype was detected.
A previous meta-analysis, conducted in 2010 by Flores-Obando et al., evaluated the relationship between XRCC1 polymorphisms and the risk of HNC based on 15 publications including 2,330 cases and 3,834 controls for Arg194Trp, four publications including 879 cases and 926 controls for Arg280His, and 15 studies including 3,582 cases and 5,347 controls for Arg399Gln polymorphism. We updated this metaanalysis by adding the sample sizes. A total of 22 studies, including 4,487 cases and 6,873 controls, evaluated the association between the XRCC1 Arg194Trp gene and HNC risk; 11 studies, including 2,972 cases and 3,714 controls, evaluated the association between Arg280His and HNC risk; and 27 studies, including 6,466 cases and 9,379 controls, XRCC1 Polymorphisms and HNC Susceptibility PLOS ONE | www.plosone.org evaluated the association between Arg399Gln polymorphism and HNC risk. There are some discrepancies between the Flores-Obando et al. meta-analysis and ours. A marginal association between XRCC1 Arg399Gln polymorphism and HNC risk was detected under the recessive genetic model on Caucasians in the meta-analysis conducted by Flores-Obando et al., but it was not found in ours. These diverse results may, generally, be due to the differences in the studies included in the meta-analysis. Nasopharyngeal carcinoma (NPC) is one type of HNC, which has a striking geographic and ethnic distribution, with particularly high rates observed among Asians. The literature on NPC was not included in the Flores-Obando et al. study, but it was included in ours. In the abovementioned stratified analysis result, seven articles were shared the Flores-Obando et al. study and our study, and our study included an additional seven articles, including newly published literature and NPC literature. The results of these seven studies account for 54.94% weight (Figure 9), which caused the discrepancy between these two meta-analyses. In the assessment of the effect of XRCC1 Arg194Trp variant genotypes on HNC susceptibility, the meta-OR conducted by Flores-Obando et al. found a significant association between the Arg194Trp variant and HNC risk for homozygote comparison in the overall population and in the Asian group, which was not detected in our meta-analysis. Similarly, in our study, an additional nine studies and four studies, comprised of recently published research and NPC studies, included in the overall population group and the Asian group, respectively. The results of these additional studies account for 53.46% and 53.79% weight (Figures 10-11  not performed in the Flores-Obando's study, but it was included in this meta-analysis. We also conducted stratified analyses by genotyping methods and publication year, and all the results revealed no association between XRCC1 polymorphism and cancer risk. Therefore, our meta-analysis has stronger evidence to clarify the associations.
The relationship between HNC susceptibility and variant genotypes of XRCC1 might be affected by the tumor sites. Accordingly, we also performed stratified analysis in the oral cancer group. The result of this subgroup analysis showed no evidence for a significant association between XRCC1 polymorphism and the risk of OC in any genetic model, except for the Arg194Trp variant in the recessive model. These results generally agreed with the meta-analysis conducted by Zhou et al. [57], but there were still several differences. Two studies by Sturgis et al. [45] and Matullo et al. [49], which included oral cavity, pharynx, and larynx cancer cases, were not excluded from the Zhou et al. study. The two studies [31,48] were conducted by the same first author and the patient population was obtained from the same hospital; as a result, there is a suspicion that the findings are a duplication of a previous publication. Therefore, only one study met the inclusion criteria to participate in our research.
Recent studies have reported on the associated risk of XRCC1 polymorphism cross lifestyle factors in the progression of head and neck cancer. Cigarette smoking is a major subject of investigation in various cancers. In 1997, the World Health Organization reported that there were 1.1 billion smokers worldwide and smoking-related cancers accounted for 22% of all cancers. Hence, subgroup analysis to estimate the interaction between the genotypes of XRCC1 Arg194Trp and Arg399Gln and smoking on HNC risk was performed. The results showed that the interaction of smoking and Arg399Gln variant genotypes displayed no statistical significance in all three genetic models. A significant association between the joint effect of smoking and XRCC1 Arg194Trp polymorphism and HNC susceptibility was detected under a homozygote comparison (OR= 2.53, 95% CI= 1. 16-5.53, P= 0.020), but no statistical significance was revealed in the other genetic models. In the forest plot of HNC risk associated with the interaction between smoking and Arg194Trp, the result of a study by Cao et al. accounted for 71.15% weight (Figure 12), which may mean that XRCC1 Arg194Trp variants are nominally associated with HNC susceptibility in smokers even at a lenient threshold for statistical significance (P= 0.05). Hence, careful consideration is needed for the lack of signals with strong credibility that emerged from this subgroup analysis. The results suggest that XRCC1 Arg194Trp polymorphism may have a small involvement in the pathogenesis of HNC in smokers.
Between-study heterogeneity is a well-known problem that is unavoidable. In our meta-analysis, heterogeneity was detected in the Arg194Trp and Arg399Gln polymorphism total population groups ( Table 3). The source of heterogeneity may arise from many aspects, such as the region of study, the sample size of the case and the control group, and the genotyping method. In order to explain the main reasons for the heterogeneity across studies, stratified analyses by ethnicity and genotyping method were performed. The results showed that in both the Arg194Trp group and the Arg399Gln group, significant heterogeneity was observed in the Asian population subgroup and in the PCR-RFLP analysis subgroup under all different genetic models. This signifies that the source of total population group heterogeneity may come from different races and different genotyping methods.
Publication bias is a well-known problem that was not found by funnel plot for the overall meta-analyses of the XRCC1 polymorphisms (Figures 6-8). However, we found a potential publication bias in the Arg194Trp Asians stratified analysis and the Arg399Gln smoking stratified analysis. The reasons for this could arise from many aspects. For instance, our meta-analysis took into consideration only fully published studies. Positive results tend to be accepted by journals. In addition, language bias may also have existed.
Some limitations should be considered in our meta-analysis. First, some of the included studies in our meta-analysis contained a smaller sample size, which might result in a lack of ability to detect the possible risk for XRCC1 polymorphism. Second, due to the limited number of studies, subgroup analysis was not performed in Africans [12]. Third, this study was based on unadjusted estimates, while a more precise analysis could be performed if individual data were available.
Despite of the limitations mentioned above, the results of the current meta-analysis suggest that XRCC1 Arg194Trp, Arg280His, and Arg399Gln polymorphism is not involved in