One method of continuous cardiac output monitoring by analysis of the radial pulse contour (Qcp) relates left ventricular stroke volume and systolic blood pressure by calculating the impedence characteristic of the aorta (Zao). It was assessed during haemodynamic monitoring by comparing it with the thermodilation method in the pulmonary artery (Qtd) in 20 patients with cardiac failure due to dilated cardiomyopathy (6 cases) and ischaemic cardiomyopathy (14 cases) treated by inotropic agents or vasodilators. Over an average monitoring period of 35 hours 159 measurements of cardiac output were performed by the two methods. There was an excellent correlation between the two methods (r = 0.90; p < 0.001; Qcp = 0.97 Qtd). The systematic error (bias) between the two methods was about 2.5%. The accuracy of Qcp compared with Qtd was 12.5%. During infusion with a vasoactive agent (Piroximone), the method based on pulse contour analysis did not reflect sudden variations in cardiac output. The systematic error between the two methods rose to 19% of the value measured, reflecting the lack of adaptation of parameters of correction in this situation and which necessitated recalibration of Zao at least once after injection of the drug.