This study aimed to compare O2 consumption (VO2) determination by the gas-exchange (VO2GE) and Fick (VO2F) methods in cardiac surgical patients. A total of 10 mechanically ventilated postoperative patients were studied prospectively. Thermodilution was performed using three randomly applied techniques: room temperature saline injected at end expiration, room temperature saline randomly injected in the respiratory cycle, and iced saline injected at end expiration. The influence of the number of thermodilution determinations was assessed by comparing results from 2 and 10 injections. The variability of VO2F was greater than that of VO2GE. There was no bias between VO2GE and VO2F values using injectate at room temperature. Accuracy and precision were not improved by increasing the number of cardiac output determinations from 2 to 10. A significant bias was observed using ice-cold injectate, VO2F being 18.0 +/- 15.4 ml/min/m2 lower than VO2GE (p = 0.001). Published results when comparing VO2F and VO2GE are discrepant. However, a significant bias was found in all studies using cold injectate, with lower VO2F values. We conclude that iced injectate should not be used to assess VO2 in critically ill patients.