BACKGROUND: The hematopoietically-expressed homeobox (HHEX) gene is identified as a promising candidate for type 2 diabetes by genome-wide association studies, triggering plenty of subsequent replications; however, the results are conflicting. We therefore conducted a meta-analysis of three widely-evaluated polymorphisms in HHEX gene and diabetes risk. METHODOLOGY/PRINCIPAL FINDINGS: A random-effects model was adopted irrespective of heterogeneity. Data and study quality were assessed in duplicate. There were 49 studies (cases/controls: 57931/74658) for rs1111875, 18 studies (18227/30366) for rs5015480 and 26 studies (25725/30579) for rs7923837, respectively. Overall analyses indicated that rs1111875-C allele (odds ratio [OR] = 1.16; 95% confidence interval [CI]: 1.13-1.2; P<0.0005),... rs5015480-C allele (OR = 1.16; 95% CI: 1.06-1.26; P = 0.001) and rs7923837-G allele (OR = 1.18; 95% CI: 1.12-1.24; P<0.0005) conferred significantly increased risk for type 2 diabetes, yet accompanying moderate to strong evidence of heterogeneity. Despite vast divergence in allele distributions, subgroup analyses by ethnicity showed comparable risk estimates between Asians and Caucasians for three examined polymorphisms. Moreover, results of studies with hospital-based controls deviated greatly from that of all qualified studies, especially for rs7923837-G allele carrying a doubled risk (OR = 1.37 versus 1.18). Furthermore, when only large studies (≥ 500 case-patients) were considered, risk effects were identical to the overall estimates for three examined polymorphisms. The Begg's funnel plot and Egger's test indicated low probability of publication bias. CONCLUSIONS: Our results provide clarification to the significant association of rs1111875, rs5015480 and rs7923837 in HHEX gene with type 2 diabetes.