We developed a noninvasive computer-based system for estimating continuous cardiac output by a modified pulse contour method using a finger pressure waveform. The method requires no individual patient calibration or baseline cardiac output. First, we calibrated the system in a "learn" group of 20 patients. The computer-based cardiac output was then compared with thermodilution cardiac output in 27 patients undergoing coronary artery bypass surgery. A total of 94 cardiac outputs were performed (three averaged per determination) at four predetermined time periods: preinduction, postinduction, prebypass, and postbypass. During determination of each thermodilution cardiac output, the pulse wave data were simultaneously recorded on cassette tape. The patients had cardiac outputs ranging from 2.9 to 6.4 L/min. The correlation coefficient was 0.75. The average thermodilution cardiac output was 4.50 (+/- 0.83 SD) L/min, while the cardiac output derived from the finger pressure wave was 4.48 (+/- 0.7 SD) L/min (95% confidence interval [CI] of difference, 0-3.2%). The mean difference between the two methods was 0.02 (+/- 0.55 SD) L/min. The 95% CI for the bias was 0.0001 to 0.036 L/min. The 95% CI for the lower limit of agreement was -1.12 to -1.06 L/min; the upper limit for the 95% CI was 1.09 to 1.16 L/min. The program demonstrated that information about cardiac output can be obtained by using the Finapres device (Ohmeda, Boulder, CO). The cardiac output values obtained by this continuous noninvasive technique were within +/- 20% of the simultaneous thermodilution values 87% of the time. This was true over the narrow range of cardiac outputs (2.9 to 6.4 L/min) and wide range of heart rates (45 to 140 beats/min).