Continuous monitoring of mixed venous (SvO2) and central venous (ScO2) oxygen saturation was compared in 7 critically-ill patients (Apache II score: 19 +/- 2.1) to determine whether or not information derived from ScO2 were reliable in clinical practice. Patients were catheterized with both a pulmonary artery (PA) and a central venous (CV) catheter, each of them mounted with fiberoptic sensors (Opticath PA Catheter P7110 and Opticath CV Catheter U440, Abbott). A total of 580 comparative measurements were obtained during periods without and with therapeutic interventions (drug-titration, bronchial suction, use of PEEP, changes in FiO2 ...). The systematic error between the 2 measurement techniques was 0.6% and 0.3% in periods with and without therapeutic interventions, respectively. The variability between the 2 techniques was 10% for both periods. Differences between the values were greater than or equal to 5% in 49% of values during periods of stability and in 50% of values during periods with therapeutic interventions. There were poor correlations between the values during periods without (r = 0.48) and with therapeutic interventions (r = 0.62). Better, but still less than ideal, correlations were obtained with changes in SvO2 and ScO2 during periods without (r = 0.70) and with therapeutic interventions (r = 0.77). Although there is a need to develop a simple technique to monitor mixed venous oxygen saturation, the present study indicates that ScO2 monitoring was not reliable in the study patients.