To evaluate the in vivo performance of a continuous intra-arterial blood gas monitor as compared with in vitro arterial blood gases for measurements of PaO2, PaCO2, and arterial pH. Consecutive patient enrollment. Medical intensive care unit of a county teaching hospital. Five critically ill patients. All patients had a fiberoptic sensor placed through a 20-gauge cannula inserted into the radial artery. Sensor and arterial blood gas measurements were monitored up to 68 hrs. Arterial blood gases were analyzed on two blood gas analyzers. A total of 104 arterial blood gases were obtained for comparison of sensor measurements with blood gas analyzer values. Comparison of the sensor values with the blood gas analyzer values showed bias and precision values of -0.021 and 0.037 for arterial pH, 1.74 and 6.06 torr (0.23 and 0.81 kPa) for PaCO2, and -5.89 and 13.19 torr (-0.79 and 1.76 kPa) for PaO2, respectively. Comparison of the two blood gas analyzer measurements showed bias and precision values of -0.030 and 0.010 for arterial pH, 1.96 and 2.55 torr (0.26 and 0.34 kPa) for PaCO2, and -5.77 and 17.15 torr (-0.77 and 2.29 kPa) for PaO2, respectively. No complications attributable to the sensor were detected. The performance of this fiberoptic continuous intra-arterial blood gas monitor is comparable to that of blood gas analyzers and compares favorably with previously reported studies utilizing other sensors in reliably and reproducibly approximating PaO2, PaCO2, and arterial pH values. This monitoring capability was accomplished with no patient morbidity. Further study is indicated to confirm these initial results and to establish the role of a continuous intra-arterial blood gas monitor in critically ill patients.