OBJECTIVES: This is a cost-utility analysis of saxagliptin (treatment group) vs. thiazolidinediones (control group) as add-on therapy in type 2 diabeties (T2D) patients not achieving appropriate glycaemic control with metformin, from the Brazilian private health system (PHS) perspective METHODS: A discrete event simulation model based on UKPDS68 study was developed in order to simulate 40 years for a cohort of 1000 patients. Safety and efficacy data were obtained from a systematic review and meta-analysis of published literature. Epidemiological and costing data were obtained from DIAPS79, an outcome study of the treatment patterns and costs of T2D patients in the Brazilian PHS. Pharmaceutical costs were based on Brazilian official factory price. Insulin plus metformin was defined as rescue therapy. An annual discount rate of 5% was applied to both costs and benefits. Deterministic and probabilistic sensitivity analyses were conducted to assess the robustness of the results. RESULTS: According to the model, the lipid profile benefits from thiazolidinediones did not translate into long-term vascular benefits when compared to saxagliptin (vascular fatal events risk reduction of -0.0034 vs. pioglitazone and -0.0053 vs. rosiglitazone). Saxagliptin was dominant when compared to both pioglitazone and rosiglitazone as the add-on therapy of choice to metformin (costs savings per patient of R$3.874 vs. rosiglitazone and R$3.996 vs. pioglitazone; incremental 0.13 QALY per patient vs. pioglitazone and 0.14 QALY per patient vs. rosiglitazone). In the deterministic sensitivity analysis, HbA1c level was the most impactful parameter in the model, but saxagliptin remained the dominant option in all cases. In the probabilisitc sensitivity analysis, saxagliptin had a greater than 90% probability of being cost-effective for a willingness-to-pay of zero. CONCLUSIONS: Saxagliptin is associated with lower costs and increased quality-adjusted life expectancy compared to thiazolidinediones as add-on therapy in T2D patients failing to achieve adequate glycaemic control on metformin monotherapy.