The reliability of non-invasive determination of cardiac output using first-pass radionuclide cardiography at rest and during exercise in the upright position was evaluated in 20 patients with coronary artery disease. Cardiac output values ranged from 2.97 to 5.99 l/min at rest and from 5.08 to 10.82 l/min during exercise. Cardiac output results obtained by the radionuclide method were compared with those derived from the thermodilution technique performed simultaneously. The mean difference between the two techniques was 0.02 l/min at rest and -0.34 l/min during exercise; the limits of agreement (mean +/- 1.96 SD) were -1.29 to 1.33 l/min and -1.97 to 1.29 l/min, respectively, indicating an acceptable level of agreement. A high reproducibility of the radionuclide technique was found, with a mean difference between determinations by two observers of 0.03 l/min at rest and 0.21 l/min during exercise, the corresponding limits of agreement being -0.75 to 0.81 l/min and -0.79 to 1.21 l/min, respectively. With the aid of a variance component analysis of two determinations by each of four observers, 95% confidence intervals of +/- 10% at rest and +/- 12% during exercise were computed for the radionuclide cardiac output measurements. The observer variation was most pronounced for the part of the cardiac output determination related to measurement of left ventricular equilibrium activity during exercise. First-pass radionuclide cardiography is a reliable method for determination of cardiac output in cardiac patients at rest and during exercise in the upright position.