Association of two key variants mapping to the MTHFR gene (C677T (rs1801133) and A1298C (rs1801131)) with response to methotrexate (MTX) remains controversial. We investigated these and other markers spanning the gene as predictors of MTX efficacy and adverse events in a UK rheumatoid arthritis (RA) patient cohort and performed a meta-analysis of the two key variants using all published data. The tagging single nucleotide polymorphisms (SNPs) were genotyped in 309 patients with well-defined outcomes to MTX treatment and 17 studies were included in the meta-analysis. No association of the SNPs tested was detected with MTX efficacy or toxicity in our UK cohort. After combining our data with previous studies by meta-analysis, the random effects pooled odds ratios (OR) for both C677T and A1298C showed no association with efficacy or toxicity for either of the SNPs (efficacy: OR=1.05 (95% confidence interval (CI) 0.83-1.32) and OR=0.81 (95% CI 0.53-1.24), respectively; toxicity: OR=1.38 (95% CI 0.90-2.12) and OR=1.19 (95% CI 0.80-1.78), respectively). The available evidence suggests that the MTHFR C677T and A1298C gene polymorphisms are not reliable predictors of response to MTX treatment in RA patients.