Genetic impact of methylenetetrahydrofolate reductase (MTHFR) polymorphism on the susceptibility to colorectal polyps: a meta-analysis

Two authors (MS and JZ) gathered the relative records through searching the databases, namely, PubMed, WOS (Web of Science), and EMBASE (Excerpta Medica Database), prior to March 2018. The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines were followed [13]. The search terms used with the databases are shown in Additional file 1: Table S1. We independently excluded duplicate and ineligible records based on the following criteria: reviews, mouse data, case reports or trials, meta-analyses, meeting or conference abstracts, other genes, non-SNP or nonpolyp data, or missing genotype data for rs1801133 or rs1801131. Then, the remaining studies were included as eligible case-control studies.

We carefully extracted the data from the above selected studies. The chi-squared test was applied for the calculation of the P value of HWE (Hardy-Weinberg Equilibrium). The included studies should provide the genotype frequency data of the control group, which also must be in line with the requirement of HWE. We summarized the main features of the included studies, such as first author name, publication year, polymorphism genotype frequency, country, ethnicity, genotyping assay, and P value of HWE. We also utilized quality assessment (Newcastle-Ottawa Scale, NOS) to determine the quality score of the enrolled studies. Studies with poor quality (NOS score less than five) were excluded.

We obtained the P_association, risk ratios (RRs) and 95% confidence intervals (CIs) through the association test. The P_{heterogeneity} value of Cochran’s Q statistic > 0.1 or I² value < 50% led us to use a fixed-effects model. Six genetic models were used: allele T vs. allele C for rs1801133, allele C vs. allele A for rs1801131 (allele); TT vs. CC, CC vs. AA (homozygote); CT vs. CC, AC vs. AA (heterozygote); CT + TT vs. CC, AC + CC vs. AA (dominant); TT vs. CC + CT, CC vs. AA+AC (recessive); carrier T vs. carrier C, carrier C vs. carrier A (carrier).

We also carried out a sensitivity analysis and subgroup analyses for all genetic models to evaluate the data stability and source of heterogeneity. Briefly, we omitted each included study in turn to acquire a group of meta-analysis estimations. The omitted study was regarded as the probable heterogeneity source if we detected an obvious alteration of RR and 95% CI value. Subgroup analyses were also carried out, taking the factors of country, ethnicity (Caucasian/Asian) and disease type (hyperplastic polyps/ adenomatous polyps) into consideration.

We initially identified a total of 153 records by searching three databases, namely, PubMed (n = 22), WOS (n = 83), and EMBASE (n = 48). After excluding duplicate records, a total of 115 records were filtered by our criteria. The following 88 records were excluded: reviews (n = 31), mouse data (n = 4), case reports or trials (n = 7), meta-analyses (n = 6), meeting or conference abstracts (n = 8); other genes (n = 9), non-SNP or nonpolyp data (n = 23). Subsequently, twenty-seven full-text articles were evaluated for eligibility. Five articles lacked control or T/T genotype data. Finally, a total of twenty-two articles [8‐10, 14‐32] were selected. We listed the characteristics of eligible studies in the meta-analysis (Table 1). The genotype contributions of all controls in the studies fulfilled the principle of HWE. We found that one article contained two case-control studies, namely, the genotype distribution data in both adenomatous and hyperplastic polyps. In total, twenty-three case-control studies were ultimately included for the overall meta-analysis of MTHFR rs1801133, and ten case-control studies were included for that of MTHFR rs1801131. In addition, one study in which the TT genotype frequency of case and control groups for rs1801133 equaled zero was not included in the meta-analysis under the TT vs. CC (homozygote) and TT vs. CC + CT (recessive) models. The PRISMA-based analysis flowchart is shown in Fig. 1. None of the included studies exhibited poor quality (all NOS scores were larger than five).

First, we carried out a meta-analysis to investigate the genetic relationship between MTHFR rs1801133 and colorectal polyp susceptibility. A total of twenty-three case-control studies with 8321 cases and 17,731 controls were included. As shown in Table 2, compared with the control group, no increased risk of colorectal polyps was detected in the case group under the six genetic models, namely, allele T vs. allele C (P value in test of association =0.156); TT vs. CC (P = 0.454); CT vs. CC (P = 0.077); CT + TT vs. CC (P = 0.079); TT vs. CC + CT (P = 0.847); carrier T vs. carrier C (P = 0.322). We also conducted subgroup analyses by country, ethnicity (Caucasian/Asian) and disease type (hyperplastic polyps/adenomatous polyps). A similar nonsignificant genetic relationship was observed for all the models (all P > 0.05, Table 2). For example, there was no significant difference between the colorectal polyp cases and negative controls in the UK subgroup under the T vs. C allele (Table 2, P = 0.886); TT vs. CC (P = 0.641); CT vs. CC (P = 0.351); CT + TT vs. CC (P = 0.511); TT vs. CC + CT (P = 0.436); or carrier T vs. carrier C (P = 0.831). In the subgroup analysis of “adenomatous polyps”, we also did not observe a statistically significant association under the allele T vs. allele C (Table 2, P = 0.153); TT vs. CC (P = 0.377); CT vs. CC (P = 0.113); CT + TT vs. CC (P = 0.098); TT vs. CC + CT (P = 0.696); and carrier T vs. carrier C (P = 0.331). We show the forest plots of the subgroup analyses based on disease type under the allele T vs. allele C model in Fig. 2. These results revealed that MTHFR rs1801133 does not appear to be significantly linked to susceptibility to colorectal polyps.

Next, ten studies containing 2951 cases and 3527 controls were included in the meta-analysis of MTHFR rs1801131. Pooled analysis in the overall population (Table 3) indicated a null association under all genetic models (all P > 0.05). The results of the subgroup analysis for the UK, containing five studies of 1257 cases/1407 controls, suggested an increased risk in cases of colorectal polyps compared with controls under the genetic models of CC vs. AA (P = 0.032, RR = 1.27, 95% CIs = 1.02, 1.57) and CC vs. AA+AC (P = 0.036, RR = 1.27, 95% CIs = 1.02, 1.60). We showed the related forest plots in Figs. 3 and 4. Nevertheless, no difference between cases and controls was observed in other subgroup meta-analyses (all P > 0.05, Table 3). For example, no increased or decreased risk of adenomatous polyps in cases was detected, compared with controls, under the allele C vs. allele A (Table 3, P = 0.138); CC vs. AA (P = 0.114); AC vs. AA (P = 0.576); AC + CC vs. AA (P = 0.303); CC vs. AA+AC (P = 0.122); or carrier T vs. carrier C (P = 0.376). Thus, the C/C genotype of the MTHFR rs1801131 polymorphism may be related to an enhanced colorectal polyp risk in the UK population.

In addition, we evaluated the between-study heterogeneity and did not detect remarkable heterogeneity in any of the above comparisons (Table 4, all I² < 50.0%, P value of heterogeneity > 0.1). Thus, a fixed-effects model was applied. We also conducted both Begg’s test and Egger’s test to assess the presence of publication bias. As shown in Table 4, the P values of Begg’s test and Egger’s test were larger than 0.05 in all genetic models, indicating the absence of large publication bias. We showed Begg’s funnel plot and the association between the MTHFR rs1801131 polymorphism and colorectal polyp risk under the CC vs. AA model in Fig. 5a. Additionally, similar pooled RRs were detected in our sensitivity analysis under other genetic models (Fig. 5b for CC vs. AA model of MTHFR rs1801131; other data not shown), suggesting the reliability of pooling outcomes.