Cardiac index was measured using thoracic bioimpedance (CIbi) and thermodilution (CItd) in 19 patients with proven sepsis, undergoing artificial ventilation of the lungs. There was a poor correlation between the techniques (r = 0.36, 242 data sets, regression line CIbi = 0.16 CItd + 2.56 litre min-1 m-2). The overall bias (CItd-CIbi) was 1.69 litre min-1 m-2 with limits of agreement (precision) of +4.17 to -0.79 litre min-1 m-2. In individual patients the bias was from -0.46 to 4.56 litre min-1 m-2 with the limits of agreement from +/- 0.29 to +/- 2.55 litre min-1 m-2 around the bias values. The two techniques cannot be used interchangeably in this group of patients.