0260: 3D-transeosphageal echocardiography usefulness for assessment of cardiac output in intensive care unit: an ultrasound versus thermo-dilution comparative study for patients under mechanical ventilation

Abstract

Hemodynamic evaluation is a major tool for management of intensive care unit (ICU) patients. Due to insufficient echogenicity, transthoracic echocardiography could be noninformative (dorsal decubitus, mechanical ventilation). Three-dimensional trans-oesophageal echocardiography (3D-TEE) is a new non-invasive ultrasound modality for quantitative and semi-quantitative assessment of cardiac output. Only few validation data are available in this indication. To evaluate feasibility and diagnostic performance of 3D-TEE for assessment of cardiac output in ICU. Intermittent thermodilution measurement via transpulmonary method was used as benchmark. Fifteen patients under mechanical ventilation, without any significant valvular disease or mechanical hemodynamic support were prospectively included. Cardiac output was calculated with transpulmonary thermodilution (PICCO monitoring). 3D-TEE (Philips, IE33) was performed just after invasive measure. Left ventricular volume loops were recorded then semi-automatic analysis of 3D-loops were performed off-line and blinded to thermodilution values. We used correlation coefficient and Bland-Altman method to compared these two modalities. Thirty invasive measures were recorded for fifteen patients under mechanical ventilation. 29 (97%) 3D-TEE were usable for semi-automatic analysis of left ventricular volume and cardiac output. Correlation coefficient between invasive and non-invasive methods was 0.78. Cardiac output estimation with 3D-TEE were associated with a mean bias of 0.35 l/min with 95% limits of agreement between -2.8 et 2.2 l/min. Mean duration of 3D-TEE semi-automatic analysis was 5 minutes. Cardiac output assessment with 3D-TEE is feasible with ICU patients under mechanical ventilation. Data obtained with this new non-invasive ultrasound modality have a good correlation with thermodilution values. Bias seems to be acceptable but 95% limits of agreement of both methods are quite broad.