GSTM1 polymorphism contribute to colorectal cancer in Asian populations: a prospective meta-analysis

Glutathione S-transferases (GSTs) are enzymes which expressed in many tissues and play important roles in neutralization of toxic compounds, and protecting hosts against cancer. Among several GSTs, Glutathione S-transferases mu (GSTM) has been drawn attention upon the association with the genetic risk for many types of cancers. But whether the GSTM1 polymorphisms confer the susceptibility to colorectal cancer in Asians has not been well established. We searched the PubMed database with GSTM1, polymorphism and colorectal cancer, attempting to identify the eligible studies. In total, 33 case-control studies in Asian populations with 8502 colorectal cancer patients and 13699 controls were included in the current meta-analysis. The association between the polymorphism and susceptibility to colorectal cancer was evaluated by the odds ratio (OR) and 95% confidence intervals (CI). The pooled meta-analysis suggested that GSTM1 null variant was correlated to the colorectal cancer risk in Asians. There was a marginal heterogeneity among these eligible studies. Nevertheless, cumulative meta-analysis observed a trend of an obvious association between the GSTM1 null genotype and colorectal cancer risk in Asians. In summary, the meta-analysis suggested that GSTM1 null polymorphism confer the susceptibility to colorectal cancer in Asians, especially in Chinese populations.

GSTs 13 . The null variant (GSTM1*0 allele) is the most common variant of the GSTM1, leading to the loss of enzyme activity, and the variant-carriers were proven to be associated with increased risk to cancers 9 . GSTM1 has been identified to be involved in the pathogenesis and development of certain cancers, inclusive of colorectal cancer 14 .
Many association studies and evidence were conducted and highlighted the GSTM1 null variant was correlated with the risk of cancers 15,16 . Moreover, the GSTM1 null genotype has been demonstrated to be linked to early onset for colorectal cancer 17 . However, the previous studies which aimed to investigate the association of GSTM1 polymorphism and the colorectal cancer susceptibility in Chinese populations were controversial, and so were the genetic results in Asians. Inclusive of 33 case-control studies, the present meta-analysis was performed to further explore the relationship between GSTM1 null variant and the susceptibility to colorectal cancer more comprehensively in Asians.

Methods
Literature search and inclusion criteria. We systematically searched the PubMed database (http:// www.ncbi.nlm.nih.gov/pubmed/) to identify the eligible case-control studies using the following keywords: ("glutathione S-transferase" or "GST" or "glutathione S-transferase M1" or "GSTM1"), ("polymorphism" or "Single Nucleotide Polymorphism") and ("colorectal cancer" or "CRC" or "colorectal carcinoma" or "colorectal adenoma"). There was no language limitation in the literature search. Titles and abstracts of these searching studies were primarily screened and full papers were further retrieved to confirm eligibility, the reference lists were also examined to find other relevant studies. These studies were reported from 1996 to 2014. These given studies included into our meta-analysis had to meet the following inclusion criteria: (1) case-control validation study, (2) the studies should estimate the association of the GSTM1 null variant with risk to colorectal cancer, (3) the studies should provide odds ratio (OR) with 95% CI or available data, (4) the studies should be conducted in Asian populations. Reviews and duplicate studies were excluded from the analysis. If the same case-control study were overlapped in multiple publications, only the most complete or most recent literature was included in the present study.
Data extraction. The available data originated in the eligible studies was independently extracted by two co-authors. The following information was collected: first author's name, year of publication, country of the study conducted, ethnicity of participants, numbers of cases and controls and the genotype distributions of GSTM1 null variant. All subjects followed the principles of the Declaration of Helsinki.
Statistical analysis. The association between GSTM1 null genotype with colorectal cancer risk in Asians was calculated by pooled OR with 95% CI. The ORs were evaluated according to the extracted data. A 95% CI was used for standard for statistical significance and 95% CI without 1 for OR indicated that the genotype may increase or decrease the cancer genetic risk significantly. Firstly, the I 2 statistic was estimated to quantify the heterogeneity among all eligible studies 18 , and an I 2 < 50% suggested low heterogeneity. In general, the meta-analysis was performed using fixed-effect 19 or random-effect 20 models according to the effect estimates in the presence (I 2 ≤ 50%) or absence (I 2 > 50%) of significant heterogeneity. So, the random-effect model should be chosen once the obvious heterogeneity was observed. Interesting, the heterogeneity was also considered to be significant when P < 0. 10 18,21,22 . But it was widely accepted that the cutoff for significance of heterogeneity is P < 0.05 in the meta-analysis 23 . Furthermore, Sensitivity analysis was conducted by omitting those studies in turns to estimate the overall pooled ORs in the present study. Additionally, the Begg's funnel plot and the Egger's regression plot were considered to be preferred method to assess the publication bias 24,25 . An asymmetric funnel plot suggested a relationship between effect and study size, indicating the possibility of either publication bias or a systematic difference between smaller and larger studies 26,27 . We further performed a cumulative meta-analysis to investigate a framework for updating a case-control-effect from all eligible studies and to assess how much the genetic effect changes as statistical power accumulates, and to find the trend in risk effect 28 . In the cumulative meta-analysis, studies were ordered by publication year, and the pooled ORs were calculated at the end of each study. All statistical analyses were assessed by using STATA software, version 12.0 (StataCorp LP, College Station, TX, USA). All P values were two-sided.

Results
Characteristics of the case-control studies. The flow chart of eligible studies was shown in Fig. 1.
Firstly, 89 potential studies were screened after literature search followed with the search strategy, but only 33 eligible studies remained according to the inclusion criteria. All these association studies were published from 1996 to 2014, and 19 studies were in English, and the other 14 studies were published in Chinese, which shown in the Table 1. In total, there are 22201 subjects, comprising of 8502 colorectal cancer patients and 13699 matched controls were included in the current study from these 33 studies. The characteristics of those eligible studies in Asians were shown in Table 1.
The cumulative meta-analysis further showed a trend of an obvious association between the GSTM1 null genotype and colorectal cancer risk in Asians as information accumulated by year (Fig. 3a), moreover, the cumulative analysis accumulated by the sample size also supported the result (Fig. 3b).
Publication bias. The funnel plot, Begg's adjust rank correlation test, Egger's regression test, trim and fill method are four corresponding methods applied to assess the publication bias in the current meta-analysis. Funnel plot, Begg's funnel plot, Egger's regression, plot trim and fill funnel plot were produced by these four methods respectively (Fig. 4a-d). There was no obvious asymmetry the shape of the funnel plot and the Begg's funnel plot (Fig. 4a). However, publication bias was marginal when the Begg's rank correlation method (z = 2.34, P = 0.019 < 0.05) was used, the publication bias was more significant in the Egger's regression test (t = 3.51, P = 0.001 < 0.05) (supplementary Figure S3). Nine missing studies should be filled in the trim and fill method, furthermore, LogRR and its 95%CI altered significantly after the application of trim and fill method (supplementary Figure S4). The combined analyses indicated that there is publication bias in the present study.
Sub-group analysis. There were 21 studies of Chinese, 6 in Japanese, 2 of Taiwan subjects, each single study in Korean, Singapore, India and Iran in the meta-analysis. Sub-analysis based on the different countries was conducted to verify the heterogeneity, it seemed that there may exist a marginal heterogeneity in Chinese populations (P = 0. 023, I 2 = 42%)

Discussion
The etiology of colorectal cancer has not been well established, and previous evidence suggested that the pathogenesis of colorectal cancer was intricate and influenced by complicated interactions between genetic factors and environmental risk. Previous studies reported that the GSTM1 null genotype was in high prevalence in human population, about 40-60% in Europeans 29 and about 50% in Asians 30 . It has been widely accepted that some susceptible genes correlated to colorectal cancer risk, besides smoking, diet and other environmental factors. Up to date, xenobiotic-metabolizing enzymes were primarily concerned on the pathogenesis of cancers 31 .
GSTM1 is one of the most main subtypes GSTs, which are more effective on protecting host from cancer than others 13 . GSTM1 null genotype is the most common polymorphism and has been proven to be associated with risk of several cancers 15,16 , including colorectal cancer 17 . There were many association studies on relationship between GSTM1 null polymorphism and colorectal cancer in various ethnicities 32 . Previous validation evidence and meta-analysis studies supported that the GSTM1 null genotype was significantly related to increased risk to colorectal cancer in Caucasian populations 32 , but the association in Asians has not been revealed yet. Large amount of validation studies have been conducted to detect the association in Asians, but produced inconsistent results. Therefore, we preformed the current meta-analysis to further explore whether the GSTM1 null polymorphism associated with the susceptibility to colorectal cancer.
In our meta-analysis, 33 eligible studies with 8502 colorectal cancer patients and 13699 healthy controls were pooled to calculate the association between GSTM1 polymorphism and colorectal cancer. Significant association between GSTM1 null variant and colorectal cancer in Asian populations was observed in the pooled meta-analysis under both fixed-effect model and random-effect model (Fig. 2a,b). Even though publication bias analyses suggested that obvious bias may exist in the included studies ( Fig. 4a-d). The sensitivity analysis indicated the results of meta-analysis were credible and stable (supplementary Figure S1). Furthermore, the cumulative meta-analysis accumulated by publication year or the sample size further confirmed the obvious association of the GSTM1 null variant colorectal cancer in Asians (Fig. 3a,b). In summary, the meta-analysis verified that the GSTM1 null variant was linked to the genetic risk of colorectal cancer in Asians, which in accordance with the previous meta-analysis study 26 .
Compared to the previous meta-analyses, more validation studies were pooled in the present meta-analysis. The pooled meta-analysis, the cumulative meta-analyses by publication year and the sample size, and the sensitivity analysis correspondingly supported the GSTM1 null polymorphism contribute the genetic risk to colorectal cancer in Asians, especially in Chinese. Nevertheless, it should be noted that there are several limitations in the meta-analysis. First, the association between GSTM1 null variant and colorectal cancer was the only aspect we focused on in this meta-analysis, but the other potential susceptible factors, such as age, sex and smoking status were not considered in the current study, because great majority of these eligible studies did not provide the available information or data. Second, there   exist a marginal heterogeneity in Asians (P = 0. 01, I 2 = 40.4%). Third, evident bias also exists accordingly to the publication bias analyses. More case-control studies should be added to conduct the meta-analysis, then to further detect the potential gene-gene and gene-environment association.
In summary, the meta-analyses implied that the GSTM1 null variant was significantly associated with the susceptibility to colorectal cancer in Asians, which supporting the genetic factors play vital roles in the pathogenesis of colorectal cancer. Further validation studies should be included to solidify the current conclusions.