BackgroundContradictory results have been reported regarding the association between Pro12Ala polymorphism of PPARγ2 and coronary artery disease (CAD). We sought to estimate the inconsistent results by performing a comprehensive meta-analysis.MethodsStudies in English or Chinese publications were identified by screening MEDLINE, Embase, CNKI, Wanfang and CBM. 22 studies including 8948 cases and 14427 controls were selected. A random-effects model was applied to combine the divergent outcomes of the individual studies, while addressing between-study heterogeneity and publication bias.ResultsThe Pro12Ala polymorphism of control population followed Hardy-Weinberg equilibrium for all studies (P>0.05). Overall, a marginal increased risk of CAD under the recessive genetic model (AlaAla vs ProAla+ProPro: P = 0.04, OR = 1.31, 95%CI 1.01–1.69, Pheterogeneity = 0.67, I2 = 0%) and the homozygote comparison (AlaAla vs ProPro: P = 0.04,OR = 1.30, 95%CI 1.01–1.68, Pheterogeneity = 0.68, I2 = 0%) was observed. In the subgroup analysis by ethnicity, carriers of AlaAla homozygotes had a significant increased risk for CAD among Caucasians (AlaAla vs ProAla+ProPro: P = 0.01, OR = 1.45, 95%CI 1.08–1.96, Pheterogeneity = 0.48, I2 = 0%; AlaAla vs ProPro: P = 0.02,OR = 1.44, 95%CI 1.07–1.93, Pheterogeneity = 0.46, I2 = 0%). After dividing into population source, the CAD risk magnitude of hospital-based studies was distinctly strengthened under the recessive model (P = 0.03,OR = 1.85,95%CI 1.07–3.19, Pheterogeneity = 0.87,I2 = 0%) and the homozygote comparison (P = 0.03,OR = 1.83, 95%CI 1.06–3.16, Pheterogeneity = 0.88, I2 = 0%). There was no observable publication bias as reflected by funnel plot and Egger’s linear regression test (t = -0.12, P = 0.91).Conclusion:Our results demonstrated that the PPARγ2 Pro12Ala polymorphism might be risk-conferring locus for the progression of CAD among Caucasians, but not among Asians.