OBJECTIVES: An internal study showed the relative risk of stroke was 0.800 and of MI was 0.621 for PIO based therapies versus non-TZD. Reduced MI and stroke events may drive reduced healthcare costs. Our objective was to compare healthcare costs in T2DM patients treated with PIO based versus non-TZD based therapies who had no prior MI or stroke events. METHODS: A retrospective cohort analysis was conducted using US health plan data (I3Innovus) from 1/1/2003 to 6/30/2006. Initial ICD-9 codes for T2DM (250.x0 and 250.x2) defined patient index date. The PIO group included patients treated with PIO combined with any other antidiabetic agents (excluding rosiglitazone). The non-TZD group included patients treated with any antidiabetic treatment, but not TZDs. Inclusion criteria were: age ³45 years, continuous enrollment for at least 6 months before and 1 month after the index date, no history of stroke or MI in the last six months and documented medical claims after the index date. Stroke and MI were defined by appropriate ICD-9 codes. We fitted a generalized linear model with log link and gamma distribution to the monthly average cost in [1] stroke and [2] MI cohorts, controlling for age, hypertension, and hypercholesteremia. RESULTS: In the stroke cohort, mean total healthcare cost per patient per month was $1,839 in the PIO group versus $2115 in the non-TZD group (p<0.0001). In the MI cohort, mean total health care cost per patient per month was $1843 in the PIO group versus $2,191 in the non-TZD group (p<0.0001). Mean medical cost per patient per month in the stroke and MI cohorts was 20% and 23% less, respectively, in the PIO group compared to the non-TZD group. CONCLUSION: PIO based therapies are associated with significantly lower medical and total healthcare costs compared to those not treated with TZDs in MI and stroke cohorts.