DNA mismatch repair, known as a fundamentally biological pathway, plays key roles in maintaining genomic stability, eliminating mismatch bases and preventing both mutagenesis in the short term and cancerogenesis in the long term. Polymorphisms of MLH1 in individuals may have an effect on the DNA repair capacity and therefore on cancer risk. Recently, emerging studies have been done to evaluate the association between MLH1 -93 G/A polymorphism and cancer risk in diverse populations. However, the results remain conflicting rather than conclusive. In this meta-analysis, we assessed reported studies of association between the MLH1 -93 G/A polymorphism and cancer risk including 13 691 cancer cases and 14 068 controls from 17 published studies. A borderline significant association between the MLH1 -93 G/A polymorphism and cancer risk was observed in overall analysis [heterozygote: odds ratio (OR) = 1.15; 95% confidence interval (CI) 1.05-1.26; homozygote: OR = 1.21; 95% CI, 1.04-1.40; dominant model: OR = 1.13; 95% CI 1.01-1.26; recessive model: OR = 1.21; 95% CI 1.07-1.35, respectively]. In subgroup analysis by ethnicity, significantly increased risks were found in Asian population and mixed population but not in Caucasian population. After stratified analysis according to the quality of literature, increased cancer risks were observed in the studies of lower quality but not in the studies of higher quality. Similarly, elevated cancer risks were observed in hospital-based studies but not in population-based studies. These findings showed no persuasive evidence that MLH1 -93 G/A polymorphism was associated with an increased risk of cancer. On the conservative standpoint, well-designed population-based studies with larger sample size in different ethnic groups should be performed to further confirm these results.