Until recently, continuous cardiac output (CO) monitoring required an invasive method, via a pulmonary artery catheter for thermodilution. During the past decade, several less invasive methods have been proposed 12. Among these techniques, the FloTrac-Vigileo™ which uses arterial pressure signal monitoring to assess stroke volume, has given interesting preliminary results, but still requires an arterial catheterization 3. A totally Non Invasive CO Monitoring (NICOM™) device, based on chest bioreactance, has been used in the majority of patients after cardiac surgery and could be useful in monitoring critically ill patients 45. We designed this study to compare the clinical acceptability of the Vigileo™ and NICOM™ devices in critically ill patients, using semi-continuous thermodilution CO (PAC-CCO) as a reference method.

We studied 29 patients (26 men and 3 women), with a mean age of 63.2 ± 10.7 years, and a mean SAPS II of 36 ± 10. Surgical procedures consisted of 12 valves replacements, 12 coronary grafts, and 5 mixed-procedure operations. All patients were under mechanical ventilation at the start of the protocol, five patients received inotropic support, three received vasopressors, and five patients received vasodilators. Continuous recording of CO data was performed over 1210 minutes per patient (ranged from 1013 to 1454 minutes), allowing 12,099 simultaneous measurements to be obtained for each device (417 ± 107 per patient). PAC-CCO measurements, considered as the reference values, ranged from 2.10 to 12.80 L/minute (mean 4.86 ± 1.13 L/minute; Table 1). Comparison results between the three methods are displayed in Table 1 for global results.

Although assessment of oxygen consumption (VO₂) is limited by numerous difficulties, rapid adaptation of VO₂ to metabolic needs remains conceptually one of the major objectives of hemodynamic resuscitation 91011. By neglecting soluble blood gases and considering the hemoglobin blood level as normal and stable, VO₂ is a direct function of only three variables: cardiac output (CO), arterial oxygen saturation (SaO₂), and mixed venous oxygen saturation (SvO₂). The goal of hemodynamic care can therefore be schematically described as reaching and maintaining a specific combination of SaO₂, SvO₂, and CO values to meet estimated metabolic needs. Continuous and accurate monitoring of these three variables is consequently of major interest for early detection of acute events in any patient with or at risk for a compromised hemodynamic situation.

This study shows that a totally non-invasive method of CO monitoring can have the same performance as a moderately invasive tool in postoperative high-risk patients. Our results identify a negligible bias that is acceptable in 79% of individual cases for both NICOM™ and Vigileo™ according to our own restrictive tolerance criteria 5. We considered a ± 20% tolerance because it is approximately the variability of the reference method 121314. Critchley and Critchley suggested that a ± 30% limit of agreement was acceptable for CO measurements 15. However, this recommendation is based on limits of agreements, on a central value from the average of the two-tested technologies assuming that none of them is considered as a reference. Then, the real difference between the two-tested technologies may be more than 30%. Taking into consideration this level of tolerance assumes that PAC-CCO is not a reference and increases the accuracy of the Vigileo™ and NICOM™ systems to 91% and 94%, respectively.

Even if controversial, PAC-CCO was taken as reference because it remains the most widely used device for continuous CO monitoring in many settings 8161718. Fick 19 or bolus thermodilution methods 14 could be considered as more robust references for CO snapshot measurements but we were interested in comparing these new automatic and continuous monitoring tools with a real equivalent monitoring reference. Using Fick or bolus thermodilution, it would have been impossible to compare precision and responsiveness because they require too much time due to manual data acquisition and averaging of several measurements. In addition, we considered PAC-CCO as a reference for accuracy when the PAC-CCO trend line slope was nearly flat and when the fluctuation of measurements around this trend line slope was small, indicating periods of CO stability. In such circumstances, the standard error of the mean (SEM) is given by the formula SEM = SD/√n. Even if PAC-CCO SD was 20%, the SEM was 2% when 100 points are averaged during a period of stable CO 20. Then the lack of precision of the PAC-CCO can be compensated by the time during which stable CO values are averaged.

When studying a monitoring tool, precision and responsiveness may be of greater clinical importance than accuracy. The precision of both Vigileo™ and NICOM™, the two-tested devices, was similar and always clinically acceptable. The responsiveness of both devices was faster than continuous thermodilution and the amplitude responsiveness was not significantly different. Finally, sensitivity and specificity for detecting clinically relevant CO changes were good for both NICOM™ and Vigileo™ by comparison with PAC-CCO.

A significant difference with the PAC-CCO trend line slope was found in 5 to 13% of the cases for the NICOM™ and in 0 to 17% for the Vigileo™. In 21% of the patients, the bias was more than 20% for both NICOM™ and Vigileo™. It is obvious that several factors have artificially increased these proportions of unacceptable response. First, even when using the 'STAT' button (that provides a quicker re-assessment of CO), the PAC-CCO value is not really the averaged of one minute of measurement but takes into considerations the past five minutes. It results in a smoothing of acute CO changes and could have impacted our results. Second, even when CO is globally stable, the lag-time difference between NICOM™, Vigileo™, and thermodilution may have created transient disagreements in the minute-to-minute comparison. Third, results of the NICOM™ and Vigileo™ are more likely to be transiently altered by artifacts resulting from nurses' interventions and/or from patient movements. Fourth, the software that was included in the Vigileo™ device used in this study was the 1.01 version; this software should be upgraded in the future, leading to a potential improvement in its performances that will obviously require further clinical assessment. Finally, our study was performed on a selected population of postoperative cardiac patients. Our results cannot be translated to a wider range of clinical CO values, especially in high CO values and hyperdynamic states such as sepsis.

In this study, the clinical acceptability of CO monitoring using a completely noninvasive technique (NICOM™) was equivalent to the performance of a minimally invasive technique (Vigileo™). The data was collected from postoperative cardiac surgery patients, limited by the high proportion of males to females, but included a wide range of CO values. According to our predetermined criteria, accuracy was acceptable in a large proportion of patients, precision was always clinically acceptable, and responsiveness was faster than thermodilution. We believe that the noninvasive bioreactance technology, considering its performance, should be added to the array of CO monitoring tools in selected patients.

• NICOM™, a noninvasive cardiac output monitoring system based on chest bioreactance, is equivalent to FloTrac-Vigileo™ in terms of accuracy, precision, time, and amplitude responsiveness.

CO: cardiac output; FloTrac-Vigileo™: cardiac output monitoring system based on arterial pressure signal; ICC: Intra Class Correlation; ICU: intensive care unit; NICOM™: Non Invasive Cardiac Output Monitoring system based on chest bioreactance; NS: not significant; PAC-CCO: continuous thermodilution using a pulmonary artery catheter; r: Pearson correlation coefficient; SaO₂: arterial oxygen saturation; SAPS: Simplified Acute Physiologic Score; SD: standard deviation; SEM: standard error of the mean; SvO₂: mixed venous oxygen saturation; VO₂: oxygen consumption.

AC, JCD, and PS are consultants for Edwards life Sciences. JDC and PS are members of the scientific advisory board of Cheetah-med.

SM collected the data and drafted the manuscript. AC, JCD, and PS conceived of the study, performed the statistical analysis, and collaborated to finalize the manuscript. All authors read and approved the final manuscript.