Polymorphisms and Pharmacogenomics for the Clinical Efficacy of Methotrexate in Patients with Rheumatoid Arthritis: A Systematic Review and Meta-analysis

Methotrexate (MTX) is widely used and considered a first-line disease modifying anti-rheumatic drug (DMARD) for the treatment of rheumatoid arthritis (RA). Many of the relevant genes have been investigated to estimate the association between gene polymorphisms and MTX effectiveness in RA patients, although inconsistent results have been reported. A systematic review and meta-analysis were performed to identify genetic variants associated with MTX efficacy. A total of 30 publications that included 34 genes and 125 SNPs associated with the transporters, enzymes, and metabolites of MTX or the progression of RA were included in the systematic review (SR), and 21 studies were included in 9 meta-analyses. Associations between MTX response in RA patients in MTHFR 1298A > C (rs1801131), ATIC 347C > G (rs2372536), RFC-1 80G > A (rs1051266), SLC19A1 A > G (rs2838956) and SLC19A1 G > A (rs7499) genetic polymorphisms were found, but not observed between the MTHFR 677C > T (rs1801133), TYMS 28 bp VNTR (rs34743033), MTRR 66A > G (rs1801394), and ABCB1 3435C > T (rs1045642). However, for the polymorphisms not being associated following meta-analysis could still be associated if larger cohorts were used, and studies of other polymorphisms are necessary in large cohorts and a rigorous way, which may provide more accurate results for the effect of the gene polymorphisms on the MTX response.

Scientific RepoRts | 7:44015 | DOI: 10.1038/srep44015 The aforementioned genes are commonly used as important candidate gene polymorphisms in studies of RA response to MTX treatment. All of the genes and pathways included in the present SR are summarized in Fig. 3, where they were highlighted in green; the polymorphisms included in the present meta-analysis were highlighted in red or blue, and the SNPs in red showed associations with the MTX effectiveness in RA patients.

MTHFR 677C > T (rs1801133).
Ten studies were included in the meta-analysis of MTHFR 677C > T (rs1801133), which contained data from a combined total of 579 responders and 677 nonresponders and included six European studies (423 responders and 438 nonresponders) and three South Asian studies (94 responders and 208 nonresponders). The characteristics of these studies are described in Table 2.
When all of the samples were included, the association between the frequency of 3 MTHFR 677C five > T (rs1801133) alleles (CC, CT and TT) and MTX response was not significant in pre allele (OR = 0.969, 95% CI: 0.768-1.222, Z = 0. 26 Moreover, significant between-study heterogeneity was not observed in all of the five models (Table 3).
Stratification by ethnicity did not identify a significant association between the MTHFR 677C > T (rs1801133) 3 allele frequency (CC, CT and TT) and MTX response in the European or South Asian populations in all of the five models (Table 3).
MTHFR 1298A > C (rs1801131). Eight studies were included in the meta-analysis of MTHFR 1298A > C (rs1801131), which contained data from a combined total of 425 responders and 527 nonresponders and included four European studies (270 responders and 288 nonresponders), two East Asian studies (79 responders and 48 nonresponders) and two South Asian studies (76 responders and 191 nonresponders). The characteristics of these studies are described in Table 4.
Muralidharan N et al. 27 327 India (South Asia) Two SNPs in the MTHFR gene, rs17421511 and rs1476413, and one in the DHFR gene, rs1643650, were significantly associated with response to MTX treatment in rheumatoid arthritis, We also found that two SNPs in the ATIC gene, rs16853826 and rs10197559, were associated with toxicity.
Pawlik A et al. 36 Genes from all the three pathways seem to contribute to MTX response in the Indian population.
Ghodke Y et al. 49 34 There were no statistically significant associations of ESR1 and ESR2 gene polymorphisms with response to treatment.
Kurzawski M et al. 54 174 The results of our study suggest that the MTHFR 677T and 1298C alleles may be associated with an increased rate of RA remission in patients treated with MTX receiving high doses of folic acid supplementation.
Wessels JA et al. 19 186  European studies (132 responders and 134 nonresponders), one East Asian study (72 responders and 33 nonresponders) and two South Asian studies (254 responders and 231 nonresponders). The characteristics of these studies are described in Table 5. When all of the samples were included, a significant association between the ATIC 347C > G (rs2372536) 3 allele frequency (CC, CG and GG) and MTX response status was identified in dominant model (OR Moreover, significant between-study heterogeneity was not observed in all of the five models (Table 3).
Stratification by ethnicity identified a significant association between the ATIC 347C > G (rs2372536) 3 allele frequency (CC, CG and GG) and MTX response status in Europeans in pre-allele model(OR  (Table 3).

TYMS 28 bp VNTR (rs34743033).
Three studies were included in the meta-analysis of TYMS 28 bp VNTR (rs34743033), which contained data from a combined total of 335 responders and 264 nonresponders. The characteristics of these studies are described in Table 6.
When all of the samples were included, the association between the TYMS 28 bp VNTR (rs34743033) and MTX response status was not significant in pre-allele (OR  Moreover, significant between-study heterogeneity was not observed in all of the five models (Table 3).

MTRR 66A > G (rs1801394).
Two studies were included in the meta-analysis of MTRR 66A > G (rs1801394), which contained data from a combined total of 126 responders and 198 nonresponders. The characteristics of these studies are described in Table 7.
When all of the samples were included, the association between the MTRR 66A > G (rs1801394) allele frequency (AA, AG and GG) and MTX response status was not significant in pre-allele (OR  (Table 3).

RFC-1 80G > A (rs1051266).
Four studies were included in the meta-analysis of RFC-1 80G > A (rs1051266), which contained data from a combined total of 298 responders and 423 nonresponders. The characteristics of these studies are described in Table 8.

SLC19A1 G > A (rs7499).
Two studies were included in the meta-analysis of SLC19A1 G > A (rs7499), which contained data from a combined total of 246 responders and 224 nonresponders. The characteristics of these studies are described in Table 9.
When all of the samples were included, the association between the SLC19A1 G > A (rs7499) 3 allele frequency (GG, GA and AA)and MTX response status was significant in pre-allele (OR = 1.536, 95% CI 1. Moreover, significant between-study heterogeneity was not observed in all of the five models (Table 3).  SLC19A1 A > G (rs2838956). Two studies were included in the meta-analysis of SLC19A1 A > G (rs2838956), which contained data from a combined total of 246 responders and 225 nonresponders. The characteristics of these studies are described in (Table 10). When all of the samples were included, the association between the SLC19A1 A > G (rs2838956) 3 allele frequency (AA, AG and GG) and MTX response status was significant in pre-allele model(OR Moreover, significant between-study heterogeneity was not observed in all of the five models (Table 3).

ABCB1 3435C > T (rs1045642).
Two studies were included in the meta-analysis of ABCB1 3435C > T (rs1045642), which contained data from a combined total of 177 responders and 161 nonresponders. The characteristics of these studies are described in (Table 11).
When all of the samples were included, the association between the ABCB1 3435C > T (rs1045642) allele frequency and MTX response status was not significant in pre-allele (OR

Discussion
The pathogenesis of RA is not well understood, and there are considerable challenges in the design of effective medicines to cure RA. MTX is still the gold standard drug for RA and plays antiproliferative and anti-inflammatory roles in RA therapy 11,16 . Although the factors influencing interpatient variability in MTX efficacy remain unclear, genetic factors related to drug metabolism and disease progression may play an important role in this variability. In recent years, extensive pharmacogenomics investigations have been performed to optimize MTX therapy for RA patients through genotyping and/or gene-expression-based tests. These tests were primarily based on mRNA and included transporters, enzymes, metabolites and disease associated genes 11 ; however, the majority of the findings were inconclusive and inconsistent, even for classical candidate gene polymorphisms. Thus, developing effective and practical biomarkers to aid in the prediction of MTX responses in routine clinical practice remains a challenge. The present study performed an SR on the association between polymorphisms and the clinical efficacy of MTX in RA patients using papers published in the PubMed and Embase databases. Furthermore, this review focused on studies that reported the effects of MTX monotherapy and utilized pharmacogenetics, or the analysis of an individual's genetic variation, to predict RA responses to MTX treatment. Methylenetetrahydrofolate reductase (MTHFR) is the best studied gene in the MTX cellular pathway and encodes a protein with several important roles, including the conversion of the prominent circulatory form of folate, 5, 10-methylenetetrahydrofolate required for purine and thymidine synthesis, to 5-methyl tetrahydrofolate, which acts as a carbon donor for the re-methylation of homocysteine to methionine by methionone synthase 3 . MTHFR 677C > T (rs1801133) and 1298A > C (rs1801131) are the most well described two non-synonymous genetic variants, both of which have been reported to be associated with altered phenotypes. Patients with MTHFR 1298AA and MTHFR 677CC were reported to show a greater clinical improvement with MTX 17 . MTHFR 677C > T is a nonsynonymous polymorphism that results in the substitution of alanine with valine at codon 222 of the MTHFR enzyme 3 . It was reported that MTHFR 677TT carriers were statistically significant associated with more than 4-fold increased risk for nonresponse to MTX when compared to MTHFR677C carriers 18 . MTHFR 1298A > C is another nonsynonymous polymorphism that leads to the substitution of glutamine with alanine in the C-terminal regulatory domain of the MTHFR enzyme, which results in decreased enzyme activity 9 . In recent years, extensive investigations have been performed to identify the association between these two SNPs and MTX efficacy; however, the results were inconsistent. Ghodke-Puranik Y et al. 15 reported that MTHFR 1298A allele (AA-AC) were more likely to have better MTX efficacy relative to those with MTHFR 1298 CC in Indian (South Asian) patients. However many other investigations did not shown an association between MTHFR 1298A > C allele and the MTX response in RA patients. The variability in individual study findings may  GG (dominant model)). % weight: the percentage weight attributed to each study in the meta-analysis; OR: odds ratio. Point estimates of the ORs for each study (black squares) and the corresponding 95% confidence intervals (CI) (horizontal lines) are shown, with the size of the black square representing the relative weight of study. The diamond represents the overall pooled estimate. arise due to the fact that each includes a small sample size thereby reducing the power to accurately estimate effect sizes. In the last decade, three meta-analyses were performed in relatively large samples, and the results suggested that both of the two SNPs were not associated with the efficacy of MTX in RA 3,12,13 . The present study updated the meta-analysis, and a significant association was not observed between either the 677C > T ATIC is an important gene in the adenosine pathway, which is involved in the de novo synthesis of purine and converts aminoimidazole carboxamide adenosine ribonucleotide (AICAR) into formyl-AICAR and has been mapped to chromosome 2q35. MTX is polyglutamylated to form MTX polyglutamates after entering cells  and directly inhibits ATIC, which leads to an intracellular accumulation of AICAR, and causing the release of adenosine into the extracellular space. The adenosine released diminishes the adherence of neutrophils and inhibits the function of natural killer cells, monocytes/macrophages and T-lymphocytes, thus producing potent anti-inflammatory effects 14 . ATIC 347C/G (rs2372536) polymorphism on exon 5 is the most commonly studied ATIC polymorphism in RA, and resulting in threonine to serine substitution at position 116 of the gene. Wessels JA et al. 19 reported that individuals carrying the AMPD1 T allele and the ITPA 94CC and ATIC 347 CC genotypes were two to three times more likely to have a good clinical response to MTX. However, a lack of association has been reported between the ATIC 347C > G gene polymorphism and the MTX treatment response 15,20-23 .   -allele model)). % weight: the percentage weight attributed to each study in the meta-analysis; OR: odds ratio. Point estimates of the ORs for each study (black squares) and the corresponding 95% confidence intervals (CI) (horizontal lines) are shown, with the size of the black square representing the relative weight of study. The diamond represents the overall pooled estimate.

Figure 8. Forest plot showing the association between the RFC-1 80G > A (rs1051266) single-nucleotide polymorphism and the efficacy of methotrexate (G vs. A (Pre
Scientific RepoRts | 7:44015 | DOI: 10.1038/srep44015 One meta-analysis found that the ATIC 347C > G polymorphism may be associated with non-responsiveness to MTX in Caucasian patients but not in Asian RA patients 14 . In the present meta-analysis, when all of the samples were included, a significant association between the ATIC 347C > G (rs2372536) 3 allele frequency (CC, CG and GG) and MTX response status was identified in dominant and codominant model but not in pre-allele (CC vs. CG + GG), recessive model (GG vs. CG + CC) and homozygotic (CC vs. GG) model. Further more, stratification by ethnicity identified the significant associations between the ATIC 347C > G (rs2372536) 3 allele frequency (CC, CG and GG) and MTX response status in Europeans in pre-allele, dominant and codominant model (CG vs. CC + GG). The results were consistent with the results of a previous study despite differences in the ethnicity classification method.  TYMS is a key enzyme in de novo thymidylate synthesis, and it is directly inhibited by MTX-PG. The TYMS gene has a tandem repeat polymorphism (two or three repeats of a 28 bp unit) in the enhancer region in the 5′ -UTR. Lima A. et al. 24 reported that the TYMS 28 bp VNTR (rs34743033) 3R3R polymorphism was associated with non-response to MTX. However, Wessels JA. et al. 19 and Jekic B et al. 25 did not find an association between this gene and the response to MTX. Three studies with a total of 603 European RA patients were included in the   present meta-analysis 19,24,25 , and significant associations were not observed. Until now, Asian RA patients were not included in the research into this SNP. MTR and MTRR participate in folate metabolism and are also involved in the metabolism of adenosine. MTRR is an auxiliary factor of MTR and catalyzes the regeneration of the methylcoamine, maintains sufficient activation of MTR, and is indirectly involved in the process of in vivo methylation. It was reported that the MTRR 66A > G gene polymorphism might affect the activity of the enzyme and the pharmacological effects of MTX, and MTR AG and MTRR G allele seems association with the poor response of MTX in RA patients 26 . Two studies were included in the present meta-analysis, including one for East Asian RA patients (n = 107) and another for South Asian RA patients (n = 217), and no significant association was observed between the MTRR 66A > G (rs1801394) genotype and MTX effectiveness.
Solute carriers (SLCs; especially SLC19A1/RFC-1) and ABCs (ABCC1-4, ABCB1 and ABCG2) are two groups of MTX transporters that influence cellular MTX uptake and efflux. The RFC-1 80G > A (rs1051266), SLC19A1 G > A (rs7499), SLC19A1 A > G (rs2838956), and ABCB1 3435C > T (rs1045642) polymorphisms were included in the present meta-analysis.  Table 11. Summary of the analyzed studies and the distribution of methylenetetrahydrofolate reductase ABCB1 3435C > T (rs1045642) genotypes.  For RFC-1 80G > A (rs1051266), RFC-1 is a constitutively expressed folate transport protein that has high affinity for MTX and is involved in transport of folate and MTX into the cell; the 80G > A variant maps within exon 2 of the RFC1 gene on chromosome 21 and encodes a substitution of histidine for the arginine at amino acid position 27. Ghodke-Puranik Y et al. 15 reported that those with an RFC1 80A allele (AA-GA) had better response to MTX than those with the RFC1 80 GG genotype. Drozdzik M et al. 27 found that the patients with RFC-1 AA genotype responded to the therapy more effectively than carriers of AG and GG genotypes. Five studies with a total of 403 responders and 551 nonresponders were included in the present meta-analysis, and significant associations was observed between the allele frequency (GG, GA and AA) and MTX response status in pre-allele, dominant, recessive and homozygotic model, but not in codominant model when all of the samples were included. Moreover, stratification by ethnicity identified a significant association between the RFC-1 80G > A (rs1051266) allele frequency (GG, GA and AA) and MTX response status in Europeans in pre-allele, recessive and homozygotic model, and South Asian populations in pre-allele, dominant model and homozygotic model. This result was consistent with two previous meta-analyses, which found that the RFC-1 80G > A polymorphism is associated with responsiveness to MTX therapy 1,2 , even though the inclusion and exclusion criteria are different. In the present study, we only focused on the association between gene polymorphisms and the response to MTX monotherapy in RA patients and did not investigate toxicity 28 and gene-gene interactions 29 . In addition, combined MTX and biologic disease-modifying anti-rheumatic drug (bDMARD) treatment 30-32 studies and reviews 1,33 were excluded from the meta-analysis of the RFC1 80G > A (rs1051266) polymorphism. Remarkably, the same SNP (rs1051266) was identified by a different name (SLC19A1 G > A) in the research from Lima A. et al. 34 , but was excluded in the present meta-analysis because it did not conforming to Hardy-Weinberg equilibrium.
The present meta-analysis of SLC19A1 A > G (rs2838956) found significant associations between the SLC19A1 A > G (rs2838956) 3 allele frequency (AA, AG and GG) and MTX response status in pre-allele recessive (A vs. G) and homozygotic model (AA vs. GG). For the SLC19A1 G > A (rs7499) gene polymorphism, this meta-analysis showed a significant association between the frequency of 3 alleles (GG, GA and AA) and MTX response status in in pre-allele, dominant, recessive and homozygotic model when all 480 patients were included in the present study. However, because of the small sample size, the association between SLC19A1 (rs2838956 and rs7499) and the response to MTX in RA patients require further verification.  Furthermore, the ABCB1 3435C > T (rs1045642) polymorphism, Takatori R et al. 35 found that patients with ABCB1 3435CC and 3435CT showed higher therapeutic effects of MTX, which is inconsistent with the results of Lima A et al. 34 . When all of the samples were included in the present study, the association between the ABCB1 3435C > T (rs1045642) 3 allele frequency and MTX response status was not significant. This result is consistent with a previous meta-analysis that showed a negative association between the ABCB1 C3435T polymorphism and RA susceptibility or responsiveness to MTX 5 .
In addition to the above MTX transporter genes, an increased likelihood of non-response has been reported to be associated with SLC22A11 rs11231809 T carriers; ABCC1 rs246240 G carriers; ABCC1 rs3784864 G carriers; the CGG haplotype for ABCC1 rs35592, rs2074087 and rs3784864; and the CGG haplotype for ABCC1 rs35592, rs246240 and rs3784864 34 .
Many RA progression genes have been included in research investigating the association between gene polymorphisms and MTX response. SNPs in AIF-1 36 , ESR a (ESR1) and ESR b (ESR2) 37 , PTPN22 38 , HLA-DRB1 and HLA-DQB1 39 , TGFB1 40 , TLR4 40 , CXCL9 and CXCL10 41 have been evaluated, although most of these studies showed a negative association between these polymorphisms and MTX effectiveness in RA patients with the exception of the AIF1 CC (rs2259571) genotype, which showed a poorer response to therapy with MTX 36 , and HLA-DRB1*03, which Ali AA et al. 39 found to be significantly associated with nonresponders to MTX treatment, and suggested that Pakistani patients with this genotype are less likely to benefit from MTX.
The lack of a significant association in this meta-analysis may represent a true result, but the possibility of a false-negative finding requires consideration. Certain limitations of our meta-analysis warrant consideration. First, the possibility of publication bias is always a concern. Although our analysis did not observe clear evidence of such a bias, it should be recognized that publication bias is difficult to exclude with certainty, especially when the number of incorporated studies is small. Second, publication bias could have distorted our meta-analysis because of the small number of included studies. We included 10, 8, 5, 3, and 4 studies in the meta-analysis of the MTHFR (677C > T (rs1801133) and 1298A > C (rs1801131)), ATIC 347C > G (rs2372536), TYMS 28 bp VNTR (rs34743033), and RFC-1 80G > A (rs1051266) polymorphisms, respectively, and 2 studies in the meta-analysis of the MTRR 66A > G (rs1801394), SLC19A1 (G > A (rs7499) and A > G (rs2838956)), and ABCB1 3435C > T (rs1045642) polymorphisms. Third, heterogeneity and confounding factors may have affected the meta-analysis. Variables such as sex, rheumatoid factor status, disease duration, and even the baseline DAS28 all have the potential to influence this analysis.
Even though the genetic researches showed inconsisit results in the previous researches and meta-analysis, the genetics still seem to be a powerful supplemental method of the experssion and the biomarker studies in the future research into MTX response and the combination of the above research techniques should be helpful to understanding the MTX efficacy. Given the small effect size still a choke point of the polymorphisms associated research, genotyped these and other polymorphisms within the candidate genes in large sample size study are required. Furthermore, the ethnic group, sex, rheumatoid factor status, disease duration, MTX dose, treatment duration, MTX with or without the combination of the folic acid and even the baseline disease activity of the cohorts might greatly influence the correlation of genetic polymorphisms and the MTX efficacy, so the standardized research and treatment protocal is needed to improve the quality of the genetics researches.
Taken together, this SR and meta-analysis demonstrated associations between MTX response in RA patients in MTHFR 1298A > C (rs1801131), ATIC 347C > G (rs2372536), RFC-1 80G > A (rs1051266), SLC19A1 A > G (rs2838956) and SLC19A1 G > A (rs7499) genetic polymorphisms, but not in the MTHFR 677C > T (rs1801133), TYMS 28 bp VNTR (rs34743033), MTRR 66A > G (rs1801394), and ABCB1 3435C > T (rs1045642) genetic polymorphisms. However, for the polymorphisms not being associated following meta-analysis (e.g. those in MTHFR 677C > T (rs1801133) could still be associated if larger cohorts were used, and studies of other polymorphisms are necessary in large cohorts and a rigorous way, which may provide more accurate results for the effect of the gene polymorphisms on the MTX treatment response.

Methods
The methodology for this study was based on the Preferred Reporting Items for SRs and Meta-Analyses (PRISMA) statement 42 . Ethical approval was not necessary for this meta-analysis because the results included pooled data from individual studies that received ethics approval.
Published study identification and selection for meta-analysis. All studies investigating the relationship between a genetic variant and MTX treatment response in RA published before February 2016 were identified using computer-based searches of the PubMed database and Embase database (OvidSP) using the following combination of keywords: 'methotrexate[Title/Abstract] AND (polymorphism[Title/Abstract] OR polymorphisms[Title/Abstract] OR genetic[Title/Abstract])) AND ("arthritis, rheumatoid"[MeSH Terms] OR ("arthritis"[All Fields] AND "rheumatoid"[All Fields]) OR "rheumatoid arthritis"[All Fields] OR ("rheumatoid"[All Fields] AND "arthritis"[All Fields]))' . Details of the search flow are provided in Fig. 1. The titles alone were initially reviewed for suitability, and then the abstracts of these titles were obtained and reviewed to determine the full-text retrieval suitability. Data were then extracted as described in the following section from suitable full-text reports.
Inclusion and exclusion criteria. The following inclusion criteria have been used: (1) evaluation of the associations between the gene polymorphism (or nucleotide tandem repeat)and the efficacy of MTX treatment in adult patients with RA; (2) detailed genotype data could be acquired to calculate the odds ratios (ORs) and 95% confidence intervals (CIs); (3) per-reviewed publications in English or Chinese. Exclusion criteria include (1) duplication of previous publications; (2) comment, review, editorial and conference abstract; (3) inability to ascertain the number of null and wild genotypes or alleles; (4) studies not conforming to Hardy-Weinberg Scientific RepoRts | 7:44015 | DOI: 10.1038/srep44015 equilibrium; (5)studies with no SNP site or no gene sequence; (6)non-English or Chinese publications. Each study was screened in duplicate by two independent reviewers (QQ and HJ) per the guidelines of the Human Genomic Epidemiology (HuGE) Review Handbook. Of note, for studies of MTX efficacy, all measures of disease activity were accepted, which mainly included the 44-joint count Disease Activity Score (DAS44) or the 28-joint count DAS (DAS28) or Physician's global assessment of disease (VAS score) and the ACR 20% or ACR 50% improvement response criteria (ACR20 or ACR50) 1 .

Data extraction.
References were screened and data were extracted independently by 2 authors (QQ and HJ) using a predetermined data collection template. To resolve discrepancies on the inclusion of studies and interpretation of data, a third investigator (XC) was consulted, and consensus was reached by discussion. The following data were recorded: first author's last name, year of publication, location of study, inclusion and exclusion criteria, sample size, MTX dose, SNP analysis results, treatment duration, demographic details of patients, follow-up period, and outcomes.
Statistical analyses. Hardy-Weinberg equilibrium (HWE) was accessed for each study by Chi-square test in response groups, and P < 0.05 was considered a significant departure from HWE. Studies that did not use a categorical outcome for response or did not publish necessary genotype counts per response category were excluded, if these data could not be obtained directly from the authors. The gene SNPs detected in more than two studies were included in the meta-analysis. Genotype frequencies for the MTHFR (677C > T (rs1801133) and 1298A > C (rs1801131)), ATIC 347C > G (rs2372536), TYMS 28 bp VNTR (rs34743033), MTRR 66A > G (rs1801394), RFC-1 80G > A (rs1051266), SLC19A1 (G > A (rs7499) and A > G (rs2838956)), and ABCB1 3435C > T (rs1045642) polymorphisms were determined. In this process pre-allele, dominant, recessive, codominant, and homozygotic model were performed and allowed for the inclusion of a maximum number of studies 12,43,44 . For each study, the point estimate of risk, the OR and the corresponding 95% CIs of MTX responders versus nonresponders were calculated. Then, the overall pooled OR and corresponding 95% CIs were estimated using the Mantel-Haenszel method, and the fixed effect was the absence of moderate inconsistency (> 25%) across studies 3 . A fixed effect framework assumes that the effect of allele frequency is constant across studies and between-study variations are caused by chance or random variation. The random effects model was used when heterogeneity > = 25% and the fixed effect model was used when heterogeneity < 25%, and it assumes different underlying effects, considers both within-and between-study variation and is advantageous because it accommodates diversity between studies and provides a more conservative estimate. The odds ratio (OR) was pooled using inverse variance methods to generate a summary OR and 95% confidence interval (CI). We assessed the heterogeneity between the included studies using the χ 2-based Cochran's Q statistic. The percentage of across-study variability attributable to heterogeneity beyond chance was estimated using the I 2 statistic. Differences in the pooled ORs were compared using a Z test. Potential publication bias was assessed with the Egger's test and represented graphically with Begg's funnel plots of the natural log of the OR versus its standard error. A two-sided P value of less than 0.05 was considered significant for all analyses. All statistical meta-analyses were completed with STATA (version 13.0; Stata Corp, College Station, TX, USA) 45 .