OBJECTIVE Comparison of the bias and the limits of agreement (LOA; 2 SD) of the central venous saturation (S(cv)O(2)) before, during and after coronary artery bypass graft surgery with a simultaneous measurement of the mixed venous saturation (S(v)O(2)). METHODS Prospective controlled study in a university hospital department of anaesthesiology. METHODS 60 patients with coronary artery bypass surgery, 300 paired measurements of S(v)O(2) and S(cv)O(2). RESULTS S(cv)O(2) and S(v)O(2) were analysed after induction of anaesthesia 15 min after cardiopulmonary bypass and 1, 6 and 18 h after admission to the intensive care unit. Regression analysis for the pooled measurements of S(cv)O(2) and S(v)O(2) showed a correlation R (2) = 0.52. After induction of anaesthesia 15 min after weaning from cardiopulmonary bypass and 6 h after admission to the intensive care unit the correlation coefficient was R (2) = 0.46, on admission to the intensive care unit it was R (2) = 0.42, and at 18 h it was R (2) = 0.38. Bland-Altman analysis for the measurements of S(cv)O(2) and S(v)O(2) showed a mean bias and LOA of 0.3% and -11.9 to +12.4%. In patients with a low S(cv)O(2) there was a trend to overestimate the S(v)O(2) by using the S(cv)O(2). The only factor that influenced the DeltaS(v)O(2) - S(cv)O(2) was the oxygen extraction rate (R (2) = 0.16). In patients with S(cv)O(2) below 70% this association was more pronounced (R (2) = 0.60). CONCLUSIONS Our findings demonstrate that oxygen extraction rate is the major factor in the difference between S(v)O(2) and S(cv)O(2). Under certain circumstances S(cv)O(2) differed substantially from S(v)O(2). Therefore in selected patients both parameters should be monitored to exclude general or focal hypoperfusion.