Because of their simplicity, uncalibrated pulse contour (UPC) methods have been introduced into clinical practice in critical care but are often validated with a femoral arterial waveform. We aimed to test the accuracy of cardiac index (CI) measurements and trending ability from a radial artery with one UPC. An observational study. Tertiary care mixed-surgical ICU. Data were obtained from April 2015 to July 2016. We studied 20 critically ill mechanically ventilated patients monitored by UPC (PulsioFlex; Pulsion Medical Systems SE, Feldkirchen, Germany). We used transpulmonary thermodilution (PiCCO2) as a reference. Bland-Altman-analyses with percentage errors were calculated to assess the accuracy of CI values from radial pulse contour analysis (CIRAD), autocalibration (CIAC) and femoral pulse contour analysis (CIFEM). All were compared with a reference (CITD) at 4-h intervals for 24 h. Trending ability was assessed by polar-plots and four-quadrant-plots. CI is given in l min m. Bland-Altman-analyses: for CIRAD, the mean bias was -0.1 with limits of agreement (LOA) of -2.9 to 2.7 and a percentage error of 70%; for CIAC, the mean bias was 0 with LOA -2.8 to 2.7 and a percentage error of 70%; for CIFEM, the mean bias was 0 with LOA -1.2 to 1.2 and a percentage error of 30%, respectively. Polar plots for trending: for CIRAD, the angular bias was 12° with radial LOA of 39°, a polar concordance rate of 73% and a concordance rate of 67% in the four-quadrant-plot; for CIAC, the angular bias was 4° with radial LOA of 41°, polar concordance rate of 79% and a concordance rate of 74% in the four quadrant plot; for CIFEM, the angular bias was -2° with radial LOA of 50°, polar concordance rate of 74% and a concordance rate of 81%. In critically ill patients, the PulsioFlex system connected to a radial arterial catheter is inaccurate for CI measurements and does not track changes in CI adequately. We therefore recommend using validated thermodilution techniques for monitoring in the critical care setting.