OBJECTIVES: To assess overall and diabetes-related medical service costs associated with pioglitazone (PIO) and sulfonylureas (SU) treatment among T2DM patients. METHODS: This is a retrospective cohort study based on electronic medical records (January 1, 2004-January 31, 2009) from the Geisinger Clinic in the Northeastern region of the US. The date of the initial prescription for PIO or SU was denoted as the index date. Patients were required to be aged 18 years or older and prescribed an oral antidiabetic treatment in the 1 year prior to index. Patients with type 1 or gestational diabetes and prior insulin use were excluded, as were those who had prescriptions for the index drug in the 90 days prior. Propensity score 1:1 matching and a second stage of generalized linear regression were employed to assess overall and diabetes-related medical service costs (pharmacy costs were not available in the database) in the 2 years following the index date, adjusting for patient demographics, baseline comorbidities, medication use, and healthcare resource utilization  RESULTS: A total of 2758 patients, 1379 each in the PIO and SU cohorts, were analyzed. For both cohorts, mean age was 62 years, 46% were male, and 96% were Caucasian. The two cohorts were similar in terms of current smoking status and diabetes-related comorbidities. The unadjusted total and diabetes-related medical costs were $1258 and $705 higher for SU versus PIO patients. After adjusting for covariates, the overall and diabetes-related medical service costs remained higher for patients receiving SU versus PIO ($8360 vs. $7400 for overall, and $5577 vs. $5238 for diabetes-related costs, P<0.05 for both comparisons).  CONCLUSIONS: Over a two-year follow-up, patients with T2DM initiated on PIO therapy in an integrated system incurred lower overall and diabetes-related medical service costs than patients initiated on SU. Further studies describing clinical and humanistic aspects of PIO versus SU are warranted.