Monitoring of cardiac index (CI) by uncalibrated pulse contour (PC) methods has been shown to be inaccurate in critically ill patients. We tested accuracy and trending of a new pulse contour method and a modified Fick method using central venous oxygen saturation. We studied 21 critically ill and mechanically ventilated patients (age 20-86 years) monitored by PC (PulsioFlex®) and transpulmonary thermodilution (TPTD, PiCCO2®) as reference. At baseline, reference and PC-derived CI (CIPC) were recorded and CI obtained by Fick's method (FM, CIFICK). After four hours, measurements were performed analogously for trending analysis. CI are given in l/min/m2 as mean±standard deviation. At baseline CITPTD was 3.7±0.7, CIPC 3.8±0.7 and CIFICK 5.2±1.8. After 4 hours, CITPTD was 3.5±0.6, CIPC 3.8±1.2 and CIFICK 4.8±1.7. Mean bias for PC at baseline was -0.1 (limits of agreement [LOA] -1.4 to 1.2) and -0.4 (LOA -2.6 to 1.9) after four hours. Percentage errors (PE) were 34% and 60% respectively. FM revealed a bias of -1.5 (LOA -4.8 to 1.8, PE 74%) at baseline and -1.5 (LOA -4.5 to 1.4, PE 68%) at four hours. With an exclusion window of 10% of mean cardiac index, trending analysis by polar plots showed an angular bias of 5° (radial LOA±57°) for PC and 16° (radial LOA±51°) for FM. Although PC values at baseline were marginally acceptable, both methods fail to yield clinically acceptable absolute values. Likewise, trending ability is not adequate for both methods to be used in critically ill patients.
