Continuous cardiac index (CI) monitoring is frequently used in critically ill patients. Few studies have compared the pulse contour-based device FloTrac/Vigileo to pulmonary artery thermodilution (PAC) in terms of accuracy for CI monitoring in septic shock. The aim of our study was to compare the third-generation FloTrac/Vigileo to PAC in septic shock. Eighteen patients with septic shock requiring monitoring by PAC were included in this study. We monitored CI using both FloTrac/Vigileo and continuous thermodilution (PAC-CI). Hemodynamic data were recorded every hour or every 2 min during fluid challenges. The primary endpoint was the global agreement of all CI-paired measurements determined using the Bland-Altman method adapted to replicated data. We tested the linearity of the bias by regression analysis, and compared the reactivity of the 2 techniques during fluid challenges. A receiver operating characteristic (ROC) curve analysis tested the ability of FloTrac/Vigileo to detect concordant and significative CI changes, using PAC-CI as the reference method. Overall, 1,201 paired CI measurements were recorded. The Bland-Altman analysis for global agreement of the 2 techniques showed a bias of -0.1 ± 2.1 L min(-1) m(-2) and a percentage error of 64 %. The overall correlation coefficient between PAC-CI and FloTrac/Vigileo CI was 0.47 (p < 0.01), with r(2) = 0.22. The area under the curve of the ROC curve for detecting concordant and significant changes in CI was 0.72 (0.53; 0.87). In our study, third-generation Flowtrac-Vigileo appears to be too inaccurate to be recommended for CI monitoring in septic shock.