Patients and anesthesia
After obtaining approval from the Ethics Committee of our hospital (MEC 2014-210) and written informed consent from all patients, 25 patients were enrolled from August 2014 to February 2016. This prospective, non-interventional study was performed at the Departments of Anesthesiology and Intensive Care Medicine of the Erasmus Medical Center (Rotterdam, the Netherlands). Patients were included if they were scheduled for distal ascending aorta or aortic arch replacement and deep hypothermic circulatory arrest was expected to be necessary during surgery. Prolonged cardiopulmonary bypass time and deep hypothermic circulatory arrest causes more hemodynamic fluctuations in the postoperative period at the intensive care unit. We hypothesized the clinician would have most benefit of extra cardiac output monitoring in this specific patient group. Dual arterial pressure monitoring in this type of cardiac surgery is standard care. Exclusion criteria were age < 18 years old, emergency surgery, intracardiac shunts and a contra-indication for femoral artery catheterization.
Anesthesia, cardiac surgery and postoperative intensive care management were performed according to institutional standards. Before anesthesia was induced, an arterial catheter (Flowswitch 20G, BD, USA or Leadercath 20G, Vygon, UK) was inserted into the right radial artery. If this artery was not accessible, a catheter was placed in the right brachial artery. Thereafter anesthesia was induced and a 16-cm 9,5 F five-lumen central venous catheter (Arrow International Inc, Teleflex Medical, Reading, USA) was inserted into the right internal jugular vene. A PiCCO cold injectate delivery system (Pulsion, Maquet Getinge Group, Germany) used for intermittent transpulmonary thermodilution was connected to the proximal lumen and a pressure transducer was connected to measure central venous pressure. A 20-cm 5 F thermistor-tipped arterial catheter (Pulsiocath, Maquet Getinge Group, Germany) was introduced via the left femoral artery. This catheter was connected to a pressure transducer, which was connected to the PiCCO monitor as well as to another transducer system that was connected to a Pulsioflex monitor (Pulsion, Maquet Getinge Group, Germany). The arterial pressure catheter in the radial artery was connected to a second Pulsioflex monitor.
Hemodynamic interventions, such as fluid loading and cardiovascular drug support, were given at the discretion of the attending anesthesiologist, in particular based on intra-operative transesophageal echocardiography and by the intensivist in the postoperative period.
ProAQT® system
The ProAQT® system estimates CO with arterial pressure waveform analysis. It needs the input of patient data (age, gender, weight, height). To calculate CO, the ProAQT® software uses the algorithm derived from the PiCCO Pulse-Contour-Analysis technology. The algorithm is based on the relationship between area under the arterial pressure curve and stroke volume and a value quantifying vessel compliance and peripheral resistance. Vessel compliance is estimated based on demographic patient data and arterial resistance is calculated from arterial waveform characteristics [
15,
16]. The system collects data with a high sampling frequency of 250 Hz to provide beat-to-beat CO monitoring.
Study protocol
At each time point, three CO measurements were recorded simultaneously: COtd (using thermodilution technique with the PiCCO system), COpR (using the ProAQT® system connected to the radial artery) and COpF (using the ProAQT® system connected to the femoral artery). The thermodilution measurements (COtd) were performed in triplicate with a sodium chloride 0.9% solution (20 mL, 4–6 °C) via the central venous catheter [
18]. The average of three measurements was used for analysis. If the global end-diastolic volume readings varied by more than 10%, one more thermodilution measurement was performed and the extreme one was excluded. During thermodilution measurements, COpR and COpF were registered by a dedicated observer, before and after activation of the internal auto-calibration modus.
Measurements were performed at pre-defined time points: after induction of anesthesia just before incision (T0), after extracorporeal circulation and stabilized hemodynamics (T1), after sternal closure (T2), on arrival to ICU (T3), before a fluid challenge within 6 h after admission on the ICU (T4), immediately after this fluid challenge (T5), 8 h after arrival on the ICU (T6) and 16 h after arrival on the ICU (T7). The fluid challenge was defined as a bolus of more than 250 mL of fluid administered in less than 15 min. All study measurements were performed under stable hemodynamic conditions; therefore no hemodynamic interventions were performed during the CO collection period.
Statistical analysis
For data collection and statistical analysis, we used SPSS (version 21; SPSS Inc., Chicago, IL, USA) and an Excel spreadsheet (Microsoft Office Excel 2007, Microsoft Corp, USA). To generate the plots, we used the open source statistical program RStudio using packages (
http://cran.r-project.org/) and Prism 6.0 (Graphpad Software, La Jolla, California, USA). Based on the data distribution visually inspected and tested with the Shapiro-Wilk test, results were expressed as mean (standard deviation) or median (interquartile range). Repeated hemodynamic data were analysed by a test of contrasts with a paired t test or the Wilcoxon signed rank test, depending on data distribution, with a Bonferroni correction for repeated testing.
To evaluate the accuracy of the ProAQT® device we described the agreement between the paired data of COp and COtd using the Bland-Altman analysis with correction for multiple measurements per subject [
19]. Subsequently, the percentage error (PE) was calculated as 1.96xSD of the bias of the tested method divided by the mean CO of the tested and the reference method × 100% [
20]. An acceptable PE and acceptable agreement is recently discussed [
21‐
23]. It depends on the error of the reference method and the precision that is clinically acceptable. The true error of the thermodilution reference method is unknown. We agreed with an acceptable PE of maximal 30% set by Critchley et al. based on an estimated precision of 20% for the thermodilution reference method and based on a clinically acceptable precision of less than 1 L/min for a mean cardiac output of around 5 L/min [
24].
To investigate the ability of the ProAQT® device to detect serial changes in CO (i.e. trending ability), we used the four quadrant plot and the polar plot approach [
25,
26]. In the four-quadrant plots, the ProAQT® ∆CO is plotted against the reference ∆CO. The four quadrant concordance rate is defined as the percentage of the number of data points that fall into 1 of the 2 quadrants of agreement. At the center of the plot, data tend to be randomly and not depending on the degree of agreement between the reference and test method. Therefore, exclusion zones for central data points are advised. In our study we chose an exclusion zone of 0.75 L/min, which is 12,5% of the mean CO [
25]. A limitation in all CO comparison studies is the lack of well-defined cutoff values for good, acceptable and poor trending ability based on concordance rates [
27,
28]. As a guide, we agreed with the best available cut off value which stated a concordance rate with the use of an exclusion zone of more than 92% indicates good trending ability [
25,
26].
The polar plot analysis describes the vector of CO change as an angle to the line of identity (x = y). In case this angle is zero or 180°, the agreement of the two CO readings is 100%. The magnitude of change in CO is represented as the average of the reference (PiCCO) and the test (ProAQT®) cardiac output. We used the recommended smaller exclusion zone of 10% in the polar plots [
25]. In our study this was set on 0.5 L/min. Two statistical variables were derived from the polar data: (1) the mean polar angle or angular bias, which is the average angle between all polar data points and the horizontal polar axis. The angular bias provides an assessment of how well the test method is calibrated compared with the reference method. Good calibration can be considered to exist when the angular bias is no greater than 5% [
25]. (2) The polar concordance rate, which is based on the proportion of data points that lie within 30° of the polar axis. A polar concordance rate of more than 95% indicates good trending [
25]. Because exclusion zones are still a matter of debate and might hinder good comparison between different studies, we computed both the concordance rates with and without excluding central zone data.
Because Stroke Volume (SV) is a more pure predictor of fluid responsiveness instead of CO we visualized changes in SV before and after a fluid challenge. We deliberately did not use the above described statistical methods to describe fluid induced changes because of the small sample size (
n = 22). Comparison of fluid-induced changes in SV between thermodilution and the ProAQT® measurement was made by the Spearman coefficient (chosen based on distribution of SV data). Correlation coefficients were compared by using the Fisher transformation. The change in SV in the ProAQT® system was calculated by using the auto-calibrated SV data before fluid bolus and both the auto- and non-calibrated SV data after the fluid bolus. We defined fluid responsiveness as an increase in SV of more than 10% [
29].