OBJECTIVE: To assess the cost-effectiveness of pioglitazone (PIO) in combination therapy versus usual care for patients with type 2 diabetes. METHODS: A published, validated model for type1 diabetes mellitus developed by the Institute of Medical Informatics and Biostatistics was adapted to simulate long-term management, health outcomes, resource utilisation and treatment costs of patients with type2 diabetes. The model accounts for most complications occurring in diabetes patients: nephropathy, retinopathy, acute myocardial infarction, angina pectoris, stroke and amputation. The analysis was done from a third-party-payer perspective and costs figured relative to the year 2000. A 3% discount rate was applied to costs and outcomes and sensitivity analysis was performed to test the results. RESULTS: PIO 30 mg in combination with metformin (MF) was associated with a longer life expectancy (14.06 years) than sulphonylureas (SU)/MF (13.49 years) or rosiglitazone (RSG) 8 mg/MF (13.98 years). PIO-based combinations were associated with the lowest number of complications and deaths. For every 38 patients treated with PIO 30 mg/MF rather than SU/MF or every 37 patients, respectively, for PIO 15 mg/SU rather than MEF/SU, one complication was avoided. Also, for every 39 patients treated with PIO 30 mg/MF rather than SU/MF, one death was avoided (after 15 years of treatment). PIO was more expensive, but this was in part offset by a reduction in complications. The average undiscounted patient lifetime additional cost of treatment with PIO combinations over their alternatives ranges from 9,421 Denmark Kroner (DKK) to a maximum of 74,687DKK. The incremental cost per life year gained of PIO 30 mg/MF relative to SU/MF, and RSG 8 mg/MF is 208,657DKK and 167,060DKK, respectively. CONCLUSION: This model suggests that combined treatments with pioglitazone improve survival and reduce complications in patients with type 2 diabetes and represents a cost-effective use of scarce resources when judged against other therapeutic interventions. It is necessary to confirm the results of this theoretical model once long-term effectiveness data with the compared alternatives are available.