Introduction
Among the different devices that are available today for estimating cardiac output, the Nexfin technology (BMeye, Amsterdam, The Netherlands) is particularly original. This device provides a non-invasive estimation of cardiac output in two steps. First, this device allows a continuous estimation of the arterial pressure curve through the volume-clamp method [
1]. For this purpose, the device includes an inflatable cuff that is wrapped around a finger. It also includes a photoplethysmographic device that measures the diameter of the finger arteries. At each systole, the photoplethysmographic device senses the increase of the finger arteries' diameter. A fast servo-controlled system immediately inflates the cuff in order to keep the arteries' diameter constant. Therefore, cuff pressure reflects the arterial pressure. Its continuous measurement allows estimation of the arterial pressure curve [
2]. The second step is to estimate cardiac output from the non-invasive arterial pressure curve. For this purpose, the Nexfin device includes pulse contour analysis software that computes cardiac output from the arterial pressure curve [
3].
The Nexfin has been quite largely validated for measuring arterial pressure [
4‐
10]. However, its reliability to measure cardiac index (CI) has been mainly investigated in non-critically ill patients [
5,
11‐
15]. Our aim was to assess whether the Nexfin estimation of CI was able to reflect CI in critically ill patients and to track its changes during a fluid challenge.
Discussion
The main finding of this study was that the non-invasive estimation of cardiac output provided by the Nexfin method was not reliable compared to transpulmonary thermodilution in critically ill patients even though, during volume expansion, CInoninv and CInoninv often changed in the same direction.
In the recent years, many efforts have been made to develop new techniques for monitoring cardiac output, with the particular aim of reducing their invasiveness. In this regard, the Nexfin technique seems particularly seductive since it only requires a finger pneumatic cuff. Nevertheless in the present study, we found that the Nexfin device was not reliable to estimate the absolute value of cardiac output. The Bland-Altman analysis showed a wide range for limits of agreement and the percentage error, which must be below 30% if transpulmonary thermodilution is used as the reference technique [
27,
29], was much higher. More than in a given value of cardiac output, clinicians might be more interested in cardiac output changes. In this regard also, the Nexfin estimation of cardiac output was not reliable. Indeed, the correlation between CI
noninv and CI
inv was only weak. If analyzing this trending ability by a four-quadrant approach, as proposed by Critchley
et al. [
28], we found that the concordance rate was 84% when excluding from analysis changes in cardiac output that were too small to have any technological significance. It means that, in a majority of cases, CI
inv and CI
noninv changed in the same direction, which would suggest a good trending ability of CI
noninv. Nonetheless, this four-quadrant analysis only takes into account the direction of cardiac output changes and not their magnitude, that is, it only allows a rough evaluation. For clinical practice, not only the direction but also the amplitude of CI changes might be important. For instance, the response to a fluid challenge is assessed by precisely measuring the relative change in CI that it induces. In this regard, we report a poor ability of CI
noninv to detect a positive response to fluid administration, in particular with a low sensitivity.
This poor reliability of the Nexfin device to estimate cardiac output is in discrepancy with some previous results obtained with the same technique [
5,
11‐
15,
30]. A recent publication also reported a good ability of the Nexfin device to estimate cardiac output measured by transpulmonary thermodilution as well as its changes observed over time or during a passive leg-raising test [
6]. The most plausible hypothesis explaining the discrepancy of such previous results with our findings is the difference in the population of interest. The studies that demonstrated a better reliability of the Nexfin were conducted in the operating theatre [
11,
12], in cardiac surgery patients after discontinuation of mechanical ventilation and inotropes [
5], in patients undergoing resynchronization therapy [
14], in an echocardiography laboratory [
13] or in healthy subjects [
15]. By contrast, in the present study, we included critically ill patients, a majority suffering from septic shock. Interestingly, our results are in agreement with another study reporting a poor reliability of the Nexfin and that was also conducted in intensive care unit patients [
31]. In fact, such patients are much more likely to have an impaired finger perfusion, due either to endogenous sympathetic stimulation, to norepinephrine administration or to sepsis-related microcirculatory abnormalities. In this regard, the fact that the reliability of the Nexfin device was not different between patients with and without norepinephrine suggests that exogenous amines administered might not be the only factor explaining a poor finger perfusion in such critically ill patients. Poor finger perfusion very likely impedes a correct assessment of the finger pressure curve by the volume-clamp method, which includes an analysis of the finger photoplethysmographic signal. In this regard, we observed that the non-invasive arterial pressure curve was not obtainable in a relatively large proportion of our patients. We made a similar observation with another volume-clamp device in a similar population [
32]. This suggests that such a limitation is inherent to all these techniques of non-invasive arterial pressure assessment.
With the Nexfin device in the present study, we could not determine whether the poor reliability of estimation of cardiac output was related to a poor assessment of the arterial pressure curve or to a failure of the pulse contour analysis process since this proprietary process was unknown to us. Nevertheless, we hypothesize that the pulse contour analysis did not account a lot for the results. Indeed, the therapeutic intervention we used, that is, volume expansion, is known to alter the pulse contour analysis to a lesser extent than treatments modifying the properties of the arterial tree, such as vasoactive agents [
33‐
35].
The estimation of arterial pressure by the Nexfin technique is based on the old volume-clamp method [
1] but also includes an individual calibration technique (Physiocal) developed by Wesseling
et al. [
36]. Compared with the previous Finapress device, the Nexfin device has benefited from many technological improvements [
7]. It has been reported to reliably estimate arterial pressure [
4‐
10]. In the present study, the assessment of such reliability was based only on the comparison of mean arterial pressures provided by the Nexfin vs. the arterial line. Indeed, due to the pulse wave amplification phenomenon, the systolic and diastolic values of arterial pressure are physiologically different between the brachial pressure, estimated by the Nexfin device, and the femoral pressure that we invasively measured. By contrast, the mean arterial pressure is almost constant along the arterial tree and we used it for assessing the reliability of the Nexfin arterial pressure estimation. The bias between non-invasive and invasive mean arterial pressures was low, while the limits of agreement were relatively high. Previous studies reported better results for this device [
4‐
10]. Like for the estimation of cardiac output, these different results might be explained by the fact that we included patients with septic shock receiving vasopressors, in whom the analysis of the non-invasive arterial pressure curve might be difficult due to a poor finger perfusion. The fact that the Nexfin device was of poorer reliability in critically ill patients than reported in the perioperative setting [
5,
11,
12] suggests that such a device is more suitable for the operating theatre than for intensive care units.
A first limitation of our study is that we could not investigate the ability of the Nexfin device to estimate pulse pressure variation, as we did for a similar device in a previous study [
32]. This was due to the fact that we included a majority of patients in whom pulse pressure variation is of difficult measurement and interpretation, that is, spontaneous breathing activity and/or cardiac arrhythmias and/or acute respiratory distress syndrome [
37,
38]. Also, although the manufacturer recommends that the finger cuff should be applied to the middle phalanx of the second or third fingers, it was not the case in all the patients of the study since the position of the cuff was changed when the arterial pressure could not be obtained from the third finger. In addition, we cannot exclude that when the patient's hand was rewarmed, this could have distorted the Nexfin measurement by changing the microcirculation. Finally, although we hypothesize that the discrepancy between the present study and previous publications was related to the type of included patients, we did not directly compare the device in similar conditions between critically ill and perioperative patients. Such a comparison remains to be performed.
Competing interests
Profs Jean-Louis Teboul and Xavier Monnet are members of the Medical Advisory Board of Pulsion Medical Systems. The other authors have no financial interest to disclose.
Authors' contributions
XM conceived the study, performed analysis and interpretation of the data and drafted the manuscript. FP performed the collection of data, contributed to analysis and interpretation of the data and to drafting of the manuscript. EL performed the collection of data, contributed to analysis and interpretation of the data and to drafting of the manuscript. MM performed the collection of data, contributed to analysis and interpretation of the data and to drafting of the manuscript. JD conceived the study, participated in its design, contributed to analysis and interpretation of the data and helped to draft the manuscript. CR participated in the design of the study, contributed to analysis and interpretation of the data and helped to draft the manuscript. J-LT conceived the study, participated in its design, contributed to analysis and interpretation of the data and helped to draft the manuscript. All authors read and approved the final manuscript.