OBJECTIVE: To compare the association of congestive heart failure (CHF) risk between pioglitazone use and insulin use in type 2 diabetes patients. METHODS: Adults with type 2 diabetes who initiated pioglitazone or insulin (INS) treatment between January 1999 and December 2001 were identified and an "index date" assigned based on the first antidiabetic drug prescription. The sample was restricted to those continuously enrolled at least 12 months before and at least 3 months after therapy initiation. Patients with CHF diagnosis or digoxin use in the preindex period, troglitazone use during preindex or follow up, or any oral antidiabetic use other than metformin or a sulfonylurea (final 6 months of preindex only) was excluded. INS patients were matched (1:1) to pioglitazone patients based on a difference of no more than ±0.01 in the estimated propensity score for pioglitazone therapy; propensity was modeled using logistic regression; controlling for demographics, comorbidities, preindex utilization costs; and postindex treatment duration. CHF risk (based on >1 diagnoses during follow up) was examined using Cox proportional hazards models. Further, risk of inpatient hospitalization for CHF also was explored. RESULTS: A total of 1668 matched pairs were identified for pioglitazone vs. insulin comparison. Mean patient age was 51 years; 51% were male. Crude CHF incidence was higher in the INS group (4.0% vs. 2.0% for INS group at 2 years, p = 0.001). In Cox proportional hazards models controlling for age and preindex total Health care costs, 2-year CHF risk was significantly lower for the pioglitazone group (HR = 0.501, 95% CI = 0.331, 0.758; p = 0.001) than for the INS group; findings were similar for inpatient hospitalization for CHF (HR = 0.263, 95% CI = 0.135, 0.511; p = 0.0001). Differences were similar when pioglitazone monotherapy was compared with insulin. CONCLUSION: Pioglitazone use appears to be associated with lower CHF risk versus INS use in type 2 diabetes patients.