A new monitor (CDI 2000) that brings blood gas measurements to the patient's bedside has been developed. To measure blood gases, blood is drawn into the patient's arterial pressure-monitoring line past in-line fluorescent-based sensors. After measurement, the blood is returned to the patient, avoiding blood loss and delays in sample turnaround and reducing the risk of infection to both patient and operator. We assessed this system's performance in vitro with tonometered bovine blood. Bias (mean difference between monitor and tonometered gas or measured pH values) +/- the standard deviation (SD) were 0.01 +/- 0.02 at pH = 7.39; 0.0 +/- 0.7 mm Hg at Pco2 = 39 mm Hg; and 2.4 +/- 3.2 mm Hg at a Po2 = 100 mm Hg (n = 54). Changes in hematocrit, blood temperature, or serum sodium concentration did not have clinically significant effects on system performance. Studies in normal volunteers, in whom large changes in blood gases were induced, showed a bias (mean difference between monitor and IL 1306 values) +/- SD of 0.00 +/- 0.02 for pH, -0.4 +/- 2.0 mm Hg for Pco2, and -3.6 +/- 7.7 mm Hg for Po2 (n = 69). We conclude from the present study that the performance of this system is comparable to that of conventional blood gas analyzers.