The purpose of the study was to compare cardiac output (CO) measurement by continuous (CTD) with that by conventional thermodilution (TD) in critically ill patients. In 19 of 20 critically ill patients requiring a pulmonary artery catheterism, 105 paired CO measurements were performed by both CTD and TD. Regression analysis showed that: CTD CO = 1.18 TD CO - 0.47. Correlation coefficient was 0.96. Bias and limit of agreement were -0.8 and 2.4 L.min-1, respectively. When a Bland and Altman diagram was constructed according to cardiac index ranges, biases were -0.2 and -0.3 and -0.8 L.min-1.m-2 and limits of agreement were 0.3, 0.7 and 1.6 L.min-1.m-2 for low (< 2.5 L.min-1.m-2), normal (between 2.5 and 4.5 L.min-1.m-2) and high (> 4.5 L.min-1.m-2) cardiac indexes, respectively. It is concluded that CTD, compared with TD, is a reliable method of measuring CO, especially when cardiac index is < or = 4.5 L.min-1.m-2.