OBJECTIVE: We examined cardiovascular risk by comparing pioglitazone (PIO) therapy with insulin (INS) therapy in a large database in which medical, drug, and laboratory information was collected using electronic case report forms. METHODS: Adult patients with type 2 diabetes mellitus were included if active in the database after 1999, and if no cardiovascular events were present in the history before baseline. Patients on monotherapy (PIO or INS) or in combination with sulfonylureas were included. To avoid selection bias and increase precision on the estimated treatment effect, we used propensity scoring, stratified matching methods, and logistic regression analysis. Baseline demographics and clinical characteristics such as disease duration, comorbidities, medical therapies, and treatment duration were used to calculate the propensity score. RESULTS: A total of 515 patients taking PIO alone or with sulfonylureas were compared with 2554 patients taking INS alone or in combination with sulfonylureas. The treatment period ranged from 6-36 months. The crude cardiac event rate in the PIO group was 5.44%, compared with 10.96% in the INS group (P <0.003), and the hazard ratio was 0.499 for PIO (95% confidence interval [CI]: 0.315, 0.791; P <0.003). When patients on monotherapy alone were compared, the crude event rates were 3.86%, compared with 11.32% in the INS group (P <0.002), and the hazard ratio was 0.346 for PIO (95% CI: 0.172, 0.694; P <0.003). The significant risk reduction in the PIO groups could not be explained by baseline clinical or laboratory measurements. CONCLUSION: In a retrospective propensity-matched cohort analysis in patients with type 2 diabetes, patients taking PIO had a significantly lower hazard for a cardiovascular event over a period of 6-36 months than those taking insulin.