OBJECTIVE: To compare cardiac outputs estimated from the classical oxygen Fick and modified CO2 Fick methods with thermodilution cardiac output. The modified CO2 Fick cardiac output was obtained by replacing the oxygen uptake (VO2) in the Fick equation with the CO2 production (VCO2) divided by either an assumed or measured value of the respiratory exchange ratio or with an independently determined constant (Crit Care Med 1991; 19:1270-1277). DESIGN: Criterion standard study. SETTING: The medical and surgical intensive care unit (ICU) in a Veterans Affairs Medical Center. PATIENTS: A total of 17 patients (26 studies) and 11 surgical patients (13 studies), predominantly mechanically ventilated using the intermittent mandatory ventilation mode, were studied over a period of 4.3 hrs. MEASUREMENTS: A respiratory gas exchange monitor was used to measure VO2, VCO2, and respiratory exchange ratio at 3-min intervals. Calculations were performed with arterial and venous oxygen saturations measured with both a laboratory cooximeter and bedside pulse and venous reflectance oximeters. In the oxygen Fick method, cardiac output was calculated from VO2 together with arterial and venous oxygen saturations. In the modified CO2 Fick methods, cardiac output values were calculated from arterial and venous oxygen saturations with VCO2, divided by either: a) an assumed value of the respiratory exchange ratio equal to 0.8 for all patients (method 1); b) the patient's measured value of the respiratory exchange ratio (method 2); or c) a constant, determined from an initial, simultaneous measurement of thermodilution cardiac output, VCO2, and oximetry saturations. Data were examined by linear regression analysis and bias and precision calculations. MAIN RESULTS: Thermodilution cardiac output was more related to cardiac outputs calculated with the 3 modified CO2 Fick methods than to the oxygen Fick cardiac output. Thermodilution cardiac output was closely related to the modified CO2 Fick cardiac output calculated via method 3. For this method, with pulse and venous reflectance oximetry saturations, linear regression yielded an r2 = .85, a standard error of the estimate of 0.88 L/min (n = 111) and a bias and precision of 0.11 and 0.97 L/min, respectively. Thermodilution cardiac output was less closely related to oxygen Fick cardiac output, which, when calculated with pulse and venous reflectance oximetry saturations, yielded an r2 = .50, a standard error of the estimate of 1.47 L/min (n = 128), and a bias and precision of 0.01 and 1.85 L/min, respectively. CONCLUSIONS: We conclude from this study that thermodilution cardiac output is more closely related to cardiac output calculated from modified CO2 Fick methods than to oxygen Fick cardiac output. Since cardiac output calculated with the modified CO2 Fick method 3 obviates the difficulties associated with measuring VO2 accurately and requires neither an assumption of nor measurement of the respiratory exchange ratio, method 3 may prove to be clinically useful for continuous cardiac output monitoring via oximetry in ICU patients.