This was a prospective, nonrandomized, noninterventional, multicenter trial performed in Chinese patients undergoing open-chest cardiothoracic surgery in accordance with the principles of the Declaration of Helsinki and National Medical Products Administration (NMPA) Good Clinical Practice Decree No. 25 (China, 2016). The study was registered at clinicaltrials.gov (NCT03807622). Ethical approval was obtained from the three participating hospitals in China: Fuwai Cardiovascular Disease Hospital in Beijing, Tianjin Chest Hospital in Tianjin and Wuhan Asia Heart Hospital in Wuhan. All patients provided written informed consent before they participated in the study. Prior to the start of the trial, the research teams were trained on the study protocol, the use of the equipment and performing the measurements.
Adult patients (≥ 18 years old) were recruited from the operating schedule if they were to undergo elective open-chest cardiothoracic surgery and indicated for CO measurements using a PAC catheter. Reasons to exclude patients from the trial were aortic or tricuspid valve regurgitation, aortic valve stenosis, aortic aneurysms, rhythm disorders, intracardiac shunt, or treatment with an intra-aortic balloon pump. Insufficient perfusion of the finger compromises the reliability of the ClearSight measurements. Therefore, we excluded patients with extreme contraction of the smooth muscle in the arteries or arterioles in the lower arm and hand, such as may be present in patients with Raynaud’s disease or Buerger’s disease or with extremely cold hands. Finally, known pregnancy and the inability to place the finger cuff appropriately due to subject anatomy or condition were also reasons to exclude patients.
Noninvasive measurement of BP by the ClearSight system is performed with a finger cuff having a photoplethysmographic system to measure arterial volume and an inflatable bladder to put pressure on the finger arteries. In order to keep the arterial volume constant throughout the cardiac cycle, the cuff pressure has to be adjusted with high frequency. Arterial blood volume can only be kept constant when the cuff pressure continuously matches the arterial BP. As such, the continuous arterial pressure wave can be derived from the cuff pressure (Penaz, 1975; Wesseling, 1995).
The arteries have to be clamped to their “unloaded” volume, i.e., the volume where internal arterial pressure and externally applied pressure are the same. The Physiocal method is used to establish the unloaded volume (Wesseling et al., 1995). It analyzes the sharpness and the curvature of the plethysmogram during short periods of constant cuff pressure. This automatic calibration is repeated regularly because the unloaded volume may change as a function of arterial wall smooth muscle tone. When the measurement remains stable, the Physiocal interval is automatically increased to maximally 70 beats. The measurement is considered stable when the length of the calibration interval is 30 beats or more.
Using a proprietary pulse wave analysis method, beat to beat stroke volume (SV) and CO are determined (Truijen et al., 2012). This method divides the area under the systolic part of the reconstructed BP curve by the aortic input impedance as determined from a 3 element Windkessel model that includes characteristic impedance, arterial compliance and peripheral resistance. For each patient, the model is individualized by using age, gender height, and weight. A more extensive description of this model can be found in the paper by Truijen et al. (Truijen et al., 2012).
A Swan-Ganz pulmonary artery catheter (Edwards Lifesciences, Irvine, USA) connected to a Vigilance II monitor (Edwards Lifesciences, Irvine, USA) was used to measure bolus thermodilution CO. A 10-ml sample of iced glucose solution (5%) was drawn through an iced injectate container (CO-SET+, Edwards Lifesciences, Irvine, USA, or any other container meeting the requirements, as judged by the investigator) and injected in a steady manner within 4 s. Four TD CO determinations, with at least 70 s between 2 injections, were averaged to obtain one single CO value. Each thermodilution curve was visually checked before acceptance. The dilution curve was automatically corrected for the temperature of the blood and of the injectate measured at the entrance of the catheter lumen.
The ClearSight finger cuff was connected to an EV1000 monitor (Edwards Lifesciences, Irvine, USA). Beat to beat hemodynamic data were stored on this monitor as well as markers indicating the exact timing of the thermodilution injections. This enabled determination of the start of a 30-s period where CO measured with ClearSight was averaged. For each patient, a single data pair was collected for the comparison, allowing for a within-group analysis only. A data pair included the average of three or four accepted TDCO values and the average of the matching CSCO values. Measurements were performed after anesthesia induction, but prior to the start of the surgery.
The accuracy of the bolus thermodilution method depends on hemodynamic stability during the measurement (Jansen et al., 1990; Truijen et al., 2018). After the measurement, the following acceptance criteria for each TD CO measurement were applied (Jansen, 1995; Harms et al., 1999; Jansen et al., 2001):
Mean arterial pressure and heart rate variation was less than 15%
If one of the four TD CO measurements deviated more than 15% from the average of the four measurements, this CO value was rejected
Similarly, the acceptance criterion for ClearSight data was a Physiocal interval of 30 beats or more.
If less than three hemodynamically stable measurements were available, the results of a patient were not included in the analysis of the primary objective. Calculation of the bias and standard deviation (SD) was performed using the method proposed by Bland and Altman (Bland & Altman, 1986). Percentage error was calculated as twice the SD of the bias divided by the mean of the CO values from TDCO and CSCO. In addition, a linear regression analysis was performed to assess agreement between TDCO and CSCO. Analyses were performed using Excel (Microsoft, Redmond, WA).
The sample size calculation was based on comparing a mean in one-sample equivalence testing (Z-test). In order to capture equivalence between the two CO methods, this study accepted a 90% power, a significance level of 0.05 and a bi-directional 0.5 l/min cutoff. Assuming a true mean bias of 0.0 and a SD of 1.4 l/min, demonstration of equivalence required 87 subjects to reach 90% power for the primary hypothesis. To allow for rejection of data, up to 125 subjects could be enrolled.