The association between diabetes and risk of prostate cancer has been investigated widely. However, study results remain inconsistent and contradictory. Using a meta-analytic approach, the present study explore the relationship incorporating more recent studies and provide more powerful evidence without the limitations of any individual study. Relevant studies were identified by searching Pubmed and the Cochrane Central Register of Controlled Trials through May 18, 2012. The strength of the relationship between diabetes mellitus and risk of prostate cancer was assessed using relative risk (RR). Either a fixed effects or random effects model was used to calculate the pooled RRs. Stratification analyses and sensitivity analyses were conducted, and publication bias was assessed by Egger's test and Begg's test. Twelve case-control studies involving 9,767 cases and 19,790 controls, and 25 cohort studies involving 118,825 cases were included. The person-years of follow-up ranged from 29,963 to 6,264,890 among included cohort studies. Diabetes was not significantly associated with incidence of prostate cancer in our analysis of case-control studies only (RR = 0.846, 95 % CI [0.710, 1.009]) or that of cohort studies only (RR = 0.925, 95 % CI [0.811, 1.054]). However, through subgroup analyses, statistically significant associations between diabetes and prostate cancer were found when considering population-based studies only (RR = 0.719, 95 % CI [0.637, 0.812]), cohort studies conducted in the United States (RR = 0.789, 95 % CI [0.727, 0.857]), and studies with follow-up of more than 5 years. Compared to risk of prostate cancer among people without diabetes, diabetic patients using insulin treatment experienced reduced incidence of prostate cancer in both case-control and cohort studies. The results suggest that diabetes mellitus is associated with decreased incidence of prostate cancer, specifically in the population of the United States. In addition, the time since onset of diabetes was positively associated with decreasing incidence of prostate cancer. The present conclusions should be considered carefully, however, and confirmed with further studies.