Objective： Certain epidemiological studies suggested that selenium might associate with diabetes mellitus (DM). However, conflicting results have been reported in different studies. Therefore we performed a comprehensive meta-analysis to the impact of selenium on DM. Methods： Studies of the association of selenium with DM were included from PubMed databases (inception to February 2017). Results： Thirteen eligible articles and one unpublished data from NHANES 2010-2014 incorporating 45,067 participants were included. In meta-analysis, selenium was significantly associated with DM (pooled odd ratios [OR], 1.84; 95%CI, 1.32-2.58). However, significant heterogeneity was found (12 values, 85%). Subgroup analysis was performed based on measurement methods of selenium in each study. Among the studies with blood, individuals with high concentration of selenium showed significantly high risk of DM than those with low concentration (OR, 2.21; 95%CI, 1.46-2.21; 12 values, 78%). In sensitivity analysis to identify heterogeneity, the pooled ORs changed little after omitting each study. Because analysis with the funnel plot was asymmetric, the trim-and fill method to adjust for publication bias was conducted by adding one estimated missing studies and still showed significant results (OR, 1.92; 95% Cl, 1.23-1.9) Conclusion： Our result demonstrates that selenium is associated with an increased risk of DM.