The main findings of the present study conducted in adult cardiac surgical patients are that: 1) the ECOM device, although less sensitive and not interchangeable with calibrated pulse contour analysis, provides consistent continuous measurements of cardiac index under dynamic conditions; and 2) changes in CIECOM during PLR predict fluid responsiveness with a good discrimination and could be a valuable alternative to calibrated pulse contour analysis in postoperative cardiac surgery patients.
Numerous previous published studies report a poor correlation and lack of agreement between cardiac output measured by usual thoracic electrical bioimpedance and a reference technique (most often thermodilution) in various subgroups of subjects [18, 23–25]. Because of the anatomical proximity of the trachea and the ascending aorta, results could be improved markedly with the ECOM device, as initially demonstrated by Wallace et al. . A poor correlation and lack of agreement between ECOM and pulmonary artery thermodilution at different intraoperative time points were, however, recently reported in patients undergoing cardiac surgery [12, 13]. We further demonstrated that neither ECOM nor calibrated pulse contour analysis were interchangeable with transpulmonary thermodilution in the cardiac surgical setting . In the present study, we found a weak positive relationship between both absolute values and percent changes in cardiac index when simultaneously using ECOM and pulse contour analysis in patients receiving volume loading. Even if an acceptable bias was evidenced, the limits of agreement were large and the percentage error as high as 45%. In their recent meta-analysis, Peyton and Chong  showed that none of the four alternative tested methods (i.e., pulse contour analysis, esophageal Doppler, partial PCO2 rebreathing, and thoracic electrical bioimpedance) achieved agreement with bolus thermodilution, which meets the expected 30% limits. They raise questions about the appropriateness of imposing arbitrary limits on the acceptability of accuracy and precision of cardiac index measurement, suggesting that the percentage error of agreement was only one marker of acceptability of a method. Thus, a more dynamic approach could be more interesting for clinical practice  and the efficacy of a clinical cardiac output monitor involves many factors other than its absolute accuracy and includes safety, convenience, and adaptability, which are characteristics that could be attributed to the ECOM system. Finally, a real-time tracking of the direction of changes in cardiac index could be more important than the ability of the monitor to deliver a highly accurate single measurement under stable conditions [28–30].
Numerous limitations that may decrease the applicability of PPV and/or SVV in daily clinical practice have been described in critically ill patients. This last point emphasizes the specific interest of PLR to predict fluid responsiveness in ICU [5–7], especially under conditions where heart-lung interaction indices cannot be interpretable . Its clinical application requires a continuous and dynamic assessment of cardiac output. Changes in cardiac output during PLR have been found recently to predict fluid responsiveness accurately in postoperative cardiac surgery patients when uncalibrated pulse contour analysis was used . We partially confirm these results, as an increase by 10% on average in cardiac index during PLR predicted fluid responsiveness with a moderate discrimination (ROCAUC = 0.72). Minor changes in cardiac index during PLR when ECOM was used simultaneously predicted a positive response to fluid challenge with a ROCAUC above 0.8. An explanation could be that PLR induces not only a brief and completely reversible self-volume challenge corresponding to the transfer of approximately 150 mL of blood  but also significant acute changes in arterial compliance that could lead to an increase in pulse contour-related stroke volume calculation. The reliability of the pulse contour method in detecting true variations in cardiac output would be negatively impacted by PLR itself. In contrast, ECOM, although clearly less sensitive than pulse contour analysis to detect changes in cardiac index during PLR, could be more reliable to predict fluid responsiveness. These last results are consistent with a previous report conducted in patients undergoing cardiac surgery and showing that the bioreactance-based NICOM system was clinically valid to predict fluid responsiveness from changes in cardiac output during PLR . The cutoff value of 3% must be taken with caution, because we did not assess the reproducibility of ECOM in detecting changes in cardiac index after fluid challenge. To date, such a small value is lower than the reproducibility of all existing cardiac output measurement techniques. Thus, even if the present results suggest that ECOM could be a helpful monitor to conduct perioperative hemodynamic goal-oriented therapy in cardiac surgery patients requiring initial intubation, clinical utility/outcome (phase 3) studies are mandatory to further evaluate the ECOM device and definitely validate its clinical interest for routine practice .
Some comments are necessary concerning the limitations of the current study. First, we only investigated a small cohort of patients with a narrow range of cardiac index values. Other studies conducted in different subgroups of critical care patients with larger ranges of cardiac index values and testing different clinical approaches of variations in cardiac index are mandatory to validate our results externally. More severe patients with acute circulatory failure or shock states should probably be included in these studies. A greater sensitivity of ECOM could indeed be observed in these patients as previous reports using thoracic electrical bioimpedance suggest that the more the changes in preload are important, the more the magnitude of dZ/dt max (and subsequently CI values) is increased. Second, we only investigated sedated and ventilated patients with stable postoperative sinus rhythm. Future studies should evaluate the ability of ECOM to predict fluid responsiveness from changes in cardiac output during PLR in spontaneously breathing patients and/or in patients with cardiac arrhythmias, i.e., under clinical conditions where heart-lung interaction indices cannot be interpretable. Third, ECOM measurements, in their current form, are heavily dependent upon the fidelity of the arterial line tracing. Many patients after cardiac surgery have dampened arterial line waveforms. Even if the ECOM signal quality index was excellent at any time in all patients, indicating that measurements of cardiac index were valuable for analysis, we cannot exclude that this could affect the accuracy of the measurements from the ECOM device. Last, we used the first version of the ECOM software. Upgraded versions could give better results in the future.