Individual and combined effects of GSTM1 and GSTT1 polymorphisms on colorectal cancer risk: an updated meta-analysis

Abstract Background. The presence or absence of glutathione S-transferase M1 gene (GSTM1) and glutathione S-transferase T1 gene (GSTT1) polymorphisms, and their combined effects have been suggested as a risk factor for colorectal cancer (CRC). However, the results are inconsistent. Objectives. An updated meta-analysis was performed to solve the controversy. Methods. Meta-analyses of Observational Studies in Epidemiology (MOOSE) guidelines were used. Results. Overall, the GSTM1 null genotype was associated with an increased CRC risk in Caucasians (odds ratio (OR) = 1.14, 95% confidence interval (CI): 1.05–1.23), Asians (OR = 1.19, 95% CI: 1.08–1.32), high-quality studies (OR = 1.12, 95% CI: 1.06–1.18). Moreover, the GSTM1 null genotype was also associated with an increased colon cancer risk (OR = 1.32, 95% CI: 1.16–1.51). The GSTT1 null genotype was also associated with an increased CRC risk in Asians (OR = 1.08, 95% CI: 1.02–1.15) and Caucasians (OR = 1.24, 95% CI: 1.09–1.41). Moreover, The GSTT1 null genotype was associated with an increased rectal cancer risk (OR = 1.13, 95% CI: 1.01–1.27, I2 = 8.3%) in subgroup analysis by tumor location. Last, the GSTM1 null/GSTT1 null genotype was associated with an increased CRC risk in Asians. Conclusion. This meta-analysis indicates that the GSTM1 and GSTT1 null genotypes are associated with increased CRC risk in Asians and Caucasians, and the GSTM1 null/GSTT1 null genotype was associated with increased CRC risk in Asians.


Introduction
Colorectal cancer (CRC) is a common form of cancer, with more than 1.5 million new patients diagnosed every year worldwide [1]. It is a complex chronic disease whose development is affected by genetic and environmental factors [2,3]. CRC incidence rates differ between countries indicating that environmental factors may be associated with an increased cancer risk, although. A twin study indicated that the role of genetic factors is around 35% in CRC [4]. A previous genome-wide association study also indicated that single-nucleotide polymorphisms are important risk factors [5].

Search strategy
Meta-analyses of Observational Studies in Epidemiology (MOOSE) guidelines were used [128]. PubMed, Chinese Biomedical Medical databases (CBM), China National Knowledge Infrastructure (CNKI), and WanFang databases (up to March 15, 2020) were searched to identify eligible studies that analyzed the GSTM1 present/null, GSTT1 present/null, and their combined effects with CRC risk. The following keywords were used: (GSTT1 OR glutathione S-transferase T1 OR GSTM1 OR glutathione S-transferase M1) AND (polymorphism OR variant OR mutation) AND (colorectal OR rectal OR rectum OR colon). The search strategy was designed to be sensitive and broad. We first carefully reviewed the title and abstract of the search results, and then downloaded full articles to identify possible articles. These were evaluated in detail to identify relevant articles. The reference lists of identified articles and reviews was also examined as appropriate. The corresponding author may be contacted by e-mail if only the abstract was available online or the data was incomplete.

Inclusion and exclusion criteria
Inclusion criteria were as follows: (1) articles on the GSTM1 present/null, GSTT1 present/null, and their combined effects with CRC risk; (2) sufficient genotype data to calculate ORs and 95% CIs; and (3) case-control studies. Exclusion criteria were as follows: (1) no raw data; (2) no control; (3) review articles, case reports, editorials, or animal research; (4) duplicate and insufficient data.

Data extraction and quality score assessment
Two investigators independently extracted data using Excel. Any disagreement was solved by iteration, discussion, and consensus. The following data were extracted from eligible studies: (1) first author's name, (2) publication year, (3) country, (4) source of controls (hospital-based and population-based case-control studies), (5) sample size, (6) genotyping method, and (6) genotype distribution of the GSTM1, GSTT1, and their combined effects in cases and controls. Different ethnicities included "Caucasians", "Asians", "Indians", and "Africans". If ethnicity was not stated or if the sample size could not be separated, the term "Mixed populations" was used. Two investigators independently assessed the quality of each individual study. The quality assessment criteria (Table 1) were obtained from two previous meta-analyses [129,130]. The highest value is obtained from the quality assessment was nine; studies of quality scoring ≥ 6 were considered as high quality.

Statistical analysis
We used crude odds ratios (ORs) and 95% confidence intervals (CIs) to estimate the association on the above issues. The genetic model of the individual GSTM1 and GSTT1 polymorphisms was null vs. present. Their combined effects used the following five genetic models: − − vs. + +, − − vs. + −, − − vs.− +, − − vs. (+ −) + (− +), and − − vs. (+ −) + (− +) + (+ +). − − referred to the GSTM1 null/GSTT1 null genotype, + − referred to the GSTM1 present/GSTT1 null genotype, − + referred to the GSTM1 null/GSTT1 present genotype, and + + referred to the GSTM1 present/GSTT1 present genotype. Heterogeneity among studies was tested using the I 2 value [131]. A fixed-effects model (Mantel-Haenszel method) was used when I 2 ≤ 50% [132]; otherwise, a random-effects model (DerSimonian and Laird method) was considered [133] if I 2 > 50%. However, these studies cannot be pooled into together when I 2 value > 75%. Subgroup analyses were performed by ethnicity, source of controls, tumor location, smoking history, gender, quality score, and tumor site. Then, a sensitivity analysis was carried out to assess the stability, a single study was excluded one at a time. Publication bias was tested by using Begg's funnel and Egger's test (significant publication bias was considered if P < 0.05). A nonparametric "trim and fill" method was applied to accredit missing studies if publication bias was detected. Finally, a meta-regression analysis was applied to assess the heterogeneity source. All results were calculated using Stata version 9.0 (Stata Corporation, College Station, TX, U.S.A.).

Controls Cases
Controls Cases

Heterogeneity and sensitivity analyses
Significant heterogeneity was detected in the meta-analysis, as shown in Tables 4-6. A meta-regression analysis revealed that sample size (P=0.002) was the source of heterogeneity for the GSTM1 present/null polymorphism. Concerning the GSTT1 present/null polymorphism and the combined effects of GSTM1 and GSTT1, meta-regression analysis did not reveal a source of heterogeneity under any genetic model. Additionally, I 2 > 75% as shown in Tables  4-6. When the study of Laso et al. [82] was excluded, the values of heterogeneity dropped and the GSTT1 null genotype was associated with an increased CRC risk in the following subgroups: Caucasians (OR = 1.24, 95% CI: 1.09-1.41, I 2 = 70.8%) and hospital-based studies (OR = 1.19, 95% CI: 1.06-1.35, I 2 = 54.5%). When the study of Martínez et al. [38] was excluded, the I 2 value dropped and no significant association was found between the combined effects of When the study of Gorukmez [49] was deleted, the I 2 value dropped and no significant association was observed between the combined effects of GSTM1 and GSTT1 polymorphisms and CRC risk in low-quality studies (− − vs. + −: OR = 1.23, 95% CI: 0.58-2.59, I 2 = 68.3%). When the study of Cong et al. [33] was excluded, the I 2 value dropped and no significant association was observed between the combined effects of GSTM1 present/null and GSTT1 present/null polymorphisms and CRC risk in population-based studies (− − vs.

Publication bias
Begg's funnel plot and Egger's test were used to assess publication bias in the meta-analysis. The Begg's funnel plot shape and Egger's test (P<0.001) revealed obvious publication bias between the GSTM1 present/null polymorphism  Figure 13 shows the Begg's funnel plots by the trim and fill method; 24 missing studies should be added to this. Notably, log OR and 95% CI did not alter significantly when the trim and fill method was used. No significant publication bias was observed for the GSTT1 present/null polymorphism (P=0.195). Concerning their combined effects, no publication bias was detected under any genetic model (

Discussion
Strange et al. [106] in 1991 first reported an association between the GSTM1 null genotype and colon adenocarcinoma risk. Chenevix-Trench et al. [21] first analyzed the association between the GSTT1 null genotype and CRC risk in 1996. Deakin et al. [52] first examined their combined effects with CRC risk in 1996. Since then, many case-control studies have investigated the associations but the results are still inconsistent. Hence, an updated meta-analysis was performed to explore the GSTM1 null genotype, GSTT1 null genotype, and their combined effects with CRC risk.
Overall, this meta-analysis indicates that the GSTM1 and GSTT1 null genotypes are associated with increased CRC risk in Asians and Caucasians, and the GSTM1 null/GSTT1 null genotype was associated with increased CRC risk in Asians, but not in Africans and Indians. In addition, the GSTM1 null genotype was associated with colon cancer risk but not rectal cancer, while conversely that the GSTT1 null genotype was associated with rectal caner but not colon cancer.
Actually, it may not be uncommon that the same polymorphism played different roles in cancer risk among different ethnic population, because cancer is a complicated multi-genetic disease, and different genetic backgrounds  [134]. In addition, the differences might arise by chance because studies in Indians and Africans with small sample size may have insufficient statistical power to generate an authoritative risk estimate [135]. Therefore, a large population-based case-control study is required to confirm the GSTM1, GSTT1 and their combined effects with CRC risk in Indians and Africans. Nine [32,33,46,59,90,93,99,101,105] and seven [38,45,48,75,81,83,85] studies indicated that the GSTM1 null genotype was associated with an increased CRC risk in Asians and Caucasians, respectively. Five [33,47,78,93,102] and eight [26,38,49,52,62,80,82,89] studies indicated that the GSTT1 null genotype had a significantly increased CRC risk in Asians and Caucasians, respectively. Moreover, five studies [33,41,57,90,99] reported a significant association between their combined effects and CRC risk in Asians. The results of present study strongly supported these findings.
Subgroup analysis by source of control found a significant association in hospital-based studies, but not in population-based studies in the present meta-analysis. However, hospital-based controls are not likely to replace the general population because they may have more bias than population-based studies [136]. Therefore, the results of hospital-based controls should be carefully explained. Heterogeneity is a common problem in meta-analyses. The present study observed several high levels of heterogeneity (I 2 > 75%), and the results of meta-regression analysis indicated that sample size was the source of heterogeneity between the GSTM1 null genotype and CRC risk. Small sample size studies may be important confounding bias in molecular epidemiological studies, because random error and bias were common in the studies with small sample sizes, and the results were unreliable [137]. Furthermore, small sample studies were easier to accept if there was a positive report as they tend to yield false-positive results because they may be not rigorous and are often of low-quality. In addition, several value of I 2 > 75% dropped when a single study was excluded, the results indicate that source of heterogeneity also may be from one or multiple small sample or low quality studies. Figure 13 indicates that the asymmetry of the funnel plot was caused by studies with low-quality small samples.
A total of 13 meta-analyses [115][116][117][118][119][120][121][122][123][124][125]126,127] were conducted between 2010 and 2019 reported on the associations between the GSTM1 present/null and/or GSTT1 present/null polymorphisms with CRC risk. Cai et al. [115] examined 17 studies that included 5907 CRC cases and 9726 controls to explore the association between the GSTM1 null genotype and CRC risk in Asians, reporting that the GSTM1 null genotype was associated with an increased CRC risk. Liao et al. [116] examined 23 studies including 5058 cases and 5999 controls to show that the GSTT1 null genotype was associated with an increased CRC risk in Caucasians and Asians. Wan et al. [117] identified 30 studies of 7635 cases and 12,911 controls in all races, and demonstrated that the GSTT1 null genotype was associated with an increased CRC risk in Caucasians. Teng et al. [118] examined 13 studies (including 2225 cases and 3990 controls) to assess the GSTM1 null genotype with CRC risk and they found that the GSTM1 null genotype was associated with an increased CRC risk in Chinese. Gao et al. [119] assessed the association of the GSTM1 null genotype with CRC risk in all races (including 10,009 cases and 15,070 controls from 36 studies) and indicated that the GSTM1 null genotype was associated with an increased risk of CRC, especially in Caucasians. Qin et al. [120] selected 46 studies including 15,373 cases and 21,238 controls to show that the GSTT1 null genotype may contribute to an increased CRC risk in Asians and Caucasians. Wang et al. [121] (19 studies including 3130 cases and 6423 controls) found that the null genotypes of GSTM1 and GSTT1 and the dual null genotype of GSTM1/GSTT1 were not associated with CRC risk in Chinese population. The examination of 44 studies of GSTM1 (11,998 CRC cases and 17,552 controls) and 34 studies of GSTT1 (8596 CRC cases and 13,589 controls) by Economopoulos and Sergentanis [122] indicated that the GSTM1 and GSTT1 null genotypes were associated with an increased CRC risk in Caucasians. Li et al. [123] analyzed 33 studies (including 8502 CRC Asian cases and 13,699 controls) and indicated that the GSTM1 null genotype conferred susceptibility to CRC, especially in Chinese population. Xu et al. [124] examined 13 publications of 4832 cases and 7045 controls, demonstrating that the GSTT1 null genotype was associated with an increased CRC risk in Asians. Zhong et al. [125] conducted an association of 12 studies involving 4517 cases and 6607 controls, and suggested that the GSTT1 null genotype contributed to an increased CRC risk in Asians. Du et al. [126] examined 12 studies of GSTM1 and 8 studies of GSTT1, and found no association on the GSTM1 or GSTT1 null genotype with CRC risk. Huang et al. [127] selected 55 studies including 17,498 cases and 26,441 controls to show that the GSTM1 null genotype was a risk factor for CRC.
The current meta-analysis has several advantages over previous meta-analyses [115][116][117][118][119][120][121][122][123][124][125]126,127]. First, the sample size was much larger, with 86 case-control studies including 24,931 CRC cases and 36,537 controls evaluated for the GSTM1 present/null polymorphism, 64 case-control studies including 19,725 CRC cases and 28,725 controls for the GSTT1 present/null polymorphism, and 33 case-control studies including 8306 CRC cases and 14,369 controls for their combined effects in all races. Second, this is the first meta-analysis to explore their combined effects in overall population. Third, we used a meta-regression analysis method to explore the source of heterogeneity. Finally, the current meta-analysis included the most recent relevant publications to produce more accurate results. Similar to previous meta-analyses, our study also has several limitations. First, only published articles were selected. Hence, publication bias may be found as shown in Figure 13. Moreover, positive results are known to be published more readily than negative ones. If negative results were included, an underestimation of the effect may be observed. Second, some case-control studies were based on hospital-based controls. These controls with non-cancerous disease may influence the pooled results in this study. Therefore, the use of population-based control studies may be more appropriate than hospital-based control studies. Third, only one study on Africans and Indians were included in the present study. Further new original studies were need on these issues in Africans and Indians.
In summary, the present study indicates that the GSTM1 null genotype is associated with increased CRC risk in Asians and Caucasians, the GSTT1 null genotype is associated with increased CRC risk in Asians, and the GSTM1 null/GSTT1 null genotype was associated with increased CRC risk in Asians. Further investigations involving large population-based studies should be conducted to explore the associations on the GSTM1 null genotype, GSTT1 null genotype and their combined effects with CRC risk.

Data Availability
All relevant data are within the paper.