Calculated values of oxygen consumption have been used to calculate a Fick cardiac output when thermodilution measurements are unreliable and when oxygen consumption measurements are unavailable. To determine the accuracy of these calculations, we measured cardiac output in 20 patients by four methods: (1) a reference Fick cardiac output calculated from metabolic oxygen consumption measurements and arterial-venous oxygen content difference (COmet); (2) thermodilution cardiac output (COtherm), (3) an estimated Fick cardiac output based on calculated oxygen consumption using standard equations (COcalc), and (4) an estimated Fick cardiac output using a bedside measurement of expired carbon dioxide production (COexp). The mean difference ± 95% limits of agreement between COtherm and COmet was 1.71 ± 5 liters/min. The mean difference between COcalc and COmet was -0.04 ± 3.33 liters/min. The mean difference between COexp and COmet was 0.31 ± 3.01 liters/min. On the basis of these wide confidence intervals, we conclude that (1) thermodilution and metabolic measurements of cardiac output frequently differ in critically ill patients, and (2) estimates of oxygen consumption, based on either standard equations or on expired carbon dioxide production measurements, are poor substitutes for metabolic measurements of oxygen consumption in critically ill subjects and may provide inaccurate estimates of cardiac output.