The performance of a new gas exchange monitor was assessed both in laboratory simulation and in ICU patients. Laboratory simulation using N2 and CO2 injections resulted in a mean error of 2 +/- 2% in CO2 production (VCO2) and 4 +/- 4% in oxygen consumption (VO2) in respirator measurements (n = 55) and in a mean error of 3 +/- 2% in VCO2 and 4 +/- 2% in VO2 in canopy measurements (n = 25). The mean error in RQ during ethanol burning was 2 +/- 2% in respirator measurements (n = 45) and 1 +/- 1% in canopy measurements. FIO2 had little effect on the accuracy of VCO2, whereas the accuracy on high rates of VO2 (VO2 = 400 ml/min) was reduced, when FIO2 increased: the error ranged from 1 +/- 1% to 6 +/- 1%, except at VO2 400 ml/min during FIO2 0.8, where the error was 16 +/- 3%. Neither peak airway pressure (+13 to +63 cm H2O) nor PEEP (0 to +20 cm H2O) had an effect on the accuracy. The highest level of minute ventilation studied (22.5 L/min) reduced the accuracy slightly (mean error of VCO2 4 +/- 1% and VO2 7 +/- 2%). In patients during controlled mechanical ventilation, increasing FIO2 from 0.4 to 0.6 had no effect on the results. VO2 was consistently higher by gas exchange than by the Fick principle: 16 +/- 9% during controlled ventilation (n = 20), 21 +/- 8% on synchronized intermittent mandatory ventilation (n = 10) and 25 +/- 8% during spontaneous breathing. We conclude that the device proved to be accurate for gas exchange measurements in the ICU.