In our study, the continuous thermodilution method was utilized to measure CO instead of the gold standard using the intermittent thermodilution technique, in which the accuracy can be affected by the timing of the injection within the respiratory cycle, the change of pulmonary artery blood temperature, the injectate, the speed of the injection and the placement of the catheter [
13]. Under hemodynamically stable condition, a good correction, accuracy and precision between continuous and intermittent cardiac output measurement during has been shown in the literature [
27‐
29]. The limitation agreement between the continuous thermodilution method and the intermittent thermodilution technique is clinically acceptable. Bottiger BW et al. [
30] has demonstrated significant correlation between intermittent and continuous CO measurements (
r = 0.87,
p < 0.0001), accompanied with a bias of 0.240 L/min during orthotopic liver transplantation. They also revealed that the changes in the pulmonary artery blood temperature would influence the CO measurements more by intermittent thermodilution than by continuous thermodilution during reperfusion. The continuous CO monitor was therefore used to determine the reliability of EV-based CO measurements. We found that CO
Ev showed limited accuracy when compared to continuous thermodilution CO assessment.
This study demonstrates that the Aesculon™ system using the EV formula is not interchangeable with the established automatic thermodilution method using a PAC in patients undergoing LDLT. The percentage error was 60% and was not clinically acceptable. In our study, CO
Ev was generally higher than CO
PAC. The mean bias between CO
PAC and CO
Ev was − 1.26 L/min. Several factors could have contributed to such poor interchangeability between the Aesculon™ system and thermodilution using a PAC during LDLT. First, in the equivalent of CO
Ev, the VEPT (mL) is related to the patient’s body weight [
31]. VEPT could be miscalculated in the presence of a large volume of ascites. As ascites is one of the clinical manifestations in patients with liver cirrhosis requiring LDLT, the presence of ascites may cause an overestimation of the patient’s body weight, leading to an overestimation of CO by the Aesculon™ monitor. In our study, the amount of ascites varied greatly among the patients, ranging from 0 ml to 12,000 ml. In the extreme situation, the ascites could make up to 20% of the body weight overestimation, which in turn may give a greater value CO
Ev than CO
PAC. Secondly, the massive ascites could displace the upward the diaphragm upwards, causing geometric changes that could alter the conductivity of the thorax and thus affect the accuracy of the EV method. Thirdly, surgical manipulation of the upper abdomen could have an impact on the CO
Ev. Surgical interventions to the upper abdomen could cause a shift in the bioimpedance cardiac output index readings by > 1 L/min/m
2, and the direction of the shift was unpredictable [
32]. The use of surgical retractor in addition to abdominal wall compression may additionally contribute to a change in conductivity. A poor correlation between those two methods was noted in this study. The correction coefficient between CO
PAC and CO
Ev was 0.415 with
p < 0.01. The subgroup analysis showed that correlation coefficients were 0.553 in the dissection phase, 0.276 in the anhepatic phase and 0.376 in the reperfusion phase. The relatively stable hemodynamics during dissection may be the reason why this phase was shown to have a better correlation coefficient. Manipulation of the great vessels during the anhepatic phase could cause a decrease in the preload, and renal vein congestion may cause hemodynamic instability [
33]. Reperfusion syndrome may also lead to low blood pressure during the reperfusion phase [
34,
35]. The reason why hemodynamic stability affected the correction coefficient was due to the measurement interval of those two techniques. The CO
Ev measurement was based on a change in conductivity and measured CO in seconds. However, the continuous CO
PAC measurement needs several minutes [
36]. Even though the readings on the CO
PAC monitor refreshed CO every 30 s, the CO displayed was the average of CO over the preceding 3–6 min [
30]. When the CO changes, the CO
PAC machine needed more time to determine the real CO. With hemodynamic disturbances, CO
Ev could be assumed to be closer to real-time CO than CO
PAC, which could be minutes behind. The best intragroup correction coefficient was achieved immediately after the inferior vena cava partial clamp was released, possibly as a result of the relatively hemodynamic stability and minimal interruption of the electrical signal during vessel anastomosis. It takes minutes to impact the hemodynamic after off-clamp. Due to the relatively slow response of the continuous thermodilution measurement, the correction coefficients were much worse during hemodynamic instability.
The concordance rate was 56.5% when the central exclusion zone was ±1 L/min, much lower than the clinically acceptable concordance rate of 92%. This may also be attributed by the lag in CO measurement using the continuous thermodilution technique.
This study has several limitations. First, the EV used four electrodes to detect the signal and calculate the CO. However, electro-coagulation and cutting could interfere with these signals during surgery. With every interruption, the EV machine required approximately at least 30 s to reacquire the conduction signal. The artefacts overlaying the recorded signal may in fact cause an over- or underestimation of CO, in particularly during the dissection phase. In extreme situations, the EV machine could only acquire one data point per hour due to interference from electro-coagulation. EV interference could also be due to mechanical compression of the thoracic electrodes by the surgeon. Second, the small sample size warrants further studies with a larger population size. Third, the trending ability of the EV monitor was only partially surveyed. During LDLT, fluid challenge, inotropic agent usage and great vessel clamping could cause hemodynamic changes.