Aims/hypothesis Opening of ATP-sensitive potassium (K ATP ) channels during myocardial ischaemia shortens action potential duration and is believed to be an adaptive, energy-sparing response. Thiazolidinedione drugs block K ATP channels in non-cardiac cells in vitro. This study determined whether thiazolidinedione drugs block cardiac K ATP channels in vivo. Methods Experiments in 68 anaesthetised pigs determined: (1) effects of inert vehicle, troglitazone (10 mg/kg i.v.) or rosiglitazone (0.1 or 1.0 mg/kg i.v.) on epicardial monophasic action potential (MAP) during 90 min low-flow ischaemia; (2) effects of troglitazone, rosiglitazone or pioglitazone (1 mg/kg i.v.) on response of MAP to intracoronary infusion of a K ATP channel opener, levcromakalim; and (3) effects of inert vehicle, rosiglitazone (1 mg/kg i.v.) or the sarcolemmal K ATP blocker HMR-1098 on time to onset of ventricular fibrillation following complete coronary occlusion. Results With vehicle, epicardial MAP shortened by 44±9 ms during ischaemia. This effect was attenuated to 12±8 ms with troglitazone and 6±6 ms with rosiglitazone (p<0.01 for both vs vehicle), suggesting K ATP blockade. Intracoronary levcromakalim shortened MAP by 38±10 ms, an effect attenuated to 12±8, 13±4 and 9±5 ms during co-treatment with troglitazone, rosiglitazone or pioglitazone (p<0.05 for each), confirming K ATP blockade. During coronary occlusion, median time to ventricular fibrillation was 29 min in pigs treated with vehicle and 6 min in pigs treated with rosiglitazone or HMR-1098 (p<0.05 for both vs vehicle), indicating that K ATP blockade promotes ischaemic ventricular fibrillation in this model. Conclusions/interpretation Thiazolidinedione drugs block cardiac K ATP channels at clinically relevant doses and promote onset of ventricular fibrillation during severe ischaemia.