BACKGROUND: The transpulmonary thermodilution (TPTD) technique for measuring cardiac output (CO) has never been validated in the presence of a left-to-right shunt. METHODS: In this experimental, paediatric animal model, nine lambs with a surgically constructed aorta-pulmonary left-to-right shunt were studied under various haemodynamic conditions. CO was measured with closed and open shunt using the TPTD technique (CO(TPTD)) with central venous injections of ice-cold saline. An ultrasound transit time perivascular flow probe around the main pulmonary artery served as the standard reference measurement (CO(MPA)). RESULTS: Seven lambs were eligible for further analysis. Mean (sd) weight was 6.6 (1.6) kg. The mean CO(MPA) was 1.21 litre min(-1) (range 0.61-2.06 l min(-1)) with closed shunt and 0.93 litre min(-1) (range 0.48-1.45 litre min(-1)) with open shunt. The open shunt resulted in a mean Q(p)/Q(s) ratio of 1.8 (range 1.6-2.4). The bias between the two CO methods was 0.17 litre min(-1) [limits of agreement (LOA) of 0.27 litre min(-1)] with closed shunt and 0.14 litre min(-1) (LOA of 0.32 litre min(-1)) with open shunt. The percentage errors were 22% with closed shunt and 34% with open shunt. The correlation (r) between the two methods was 0.93 (P