To compare the clinical performance of three pulmonary artery oximetry catheters (Oximetrix 3, SAT-2, and HEMOPRO2) in intensive care unit (ICU) patients. Unblinded comparison of performance over 24 h using an IL-282 CO-oximeter as a criterion standard. Multispecialty adult ICU at a university teaching hospital. Thirty critically ill patients selected from those requiring hemodynamic monitoring for medical management. By all measures, performance of the Oximetrix 3 and SAT-2 systems were comparable; bias +/- precision were -1.98 +/- 3.07 and +1.80 +/- 3.49, respectively, vs -2.28 +/- 5.24 for the HEMOPRO2. The Oximetrix 3 and SAT-2 systems demonstrated consistent performance over the range of saturations tested, though Oximetrix 3 tended to underestimate and SAT-2 tended to overestimate the CO-oximeter value. The HEMOPRO2 underestimated the CO-oximetry-derived saturation, although this was not constant across the range of values tested. The 95 percent confidence limits based on intrasubject variability were similar (+/- 4.59, +/- 5.66, and +/- 6.56 for the Oximetrix 3, SAT-2, and HEMOPRO2, respectively); however, the 95 percent confidence limits based on total variability, while similar for Oximetrix 3 (+/- 6.03) and SAT-2 (+/- 6.86), were larger for the HEMOPRO2 (+/- 10.30). The expected SD was similar for the three systems (2.03, 2.50, and 2.90 for the Oximetrix 3, SAT-2, and HEMOPRO2 systems, respectively). None of the systems equaled or exceeded (p greater than 0.05) the manufacturers' published specifications, which, in all cases, are listed as +/- 2 percent (saturation; 1 SD) when compared with bench oximetry. Although each system measures mixed venous oxygen saturation, the Oximetrix 3 and SAT-2 systems demonstrate closer agreement with CO-oximetry. However, none of these catheters provided statistically significant evidence that they would perform within +/- 2 percent of CO-oximetry. As a continuous monitor used to detect changes or trends, any of the three may be acceptable.