A comparison of hemodynamic measurement methods during orthotopic liver transplantation: evaluating agreement and trending ability of PiCCO versus pulmonary artery catheter techniques

Abstract

Background: Significant hemodynamic changes occur during liver transplantation, emphasizing the importance of precious and continuous monitoring of cardiac output, cardiac index, and other parameters. Although the monitoring of cardiac output by pulse indicator continuous cardiac output (PiCCO) was statistically homogeneous compared to the clinical gold standard pulmonary artery catheterization (PAC) in previous studies of liver transplantation, there are fewer statistical methods for the assessment of its conclusions, and a lack of comparisons of other hemodynamic parameters (e.g., SVRI, systemic vascular resistance index). Some studies have also concluded that the agreement between PiCCO and PAC is not good enough. Overall, there are no uniform conclusions regarding the agreement between PiCCO and PAC in previous studies. This study evaluates the agreement and trending ability of relevant hemodynamic parameters obtained with PiCCO compared to the clinical gold standard PAC from multiple perspectives, employing various statistical methods.

Methods: Fifty-two liver transplantation patients were included. Cardiac output (CO), cardiac index (CI), SVRI and stroke volume index (SVI) values were monitored at eight time points using both PiCCO and PAC. The results were analyzed by Bland-Altman analysis, Passing-bablok regression, intra-class correlation coefficient (ICC), 4-quadrant plot, polar plot, and trend interchangeability method (TIM).

Results: The Bland-Altman analysis revealed high percentage errors for PiCCO: 54.06% for CO, 52.70% for CI, 62.18% for SVRI, and 51.97% for SVI, indicating poor accuracy. While Passing-Bablok plots showed favorable agreement for SVRI overall and during various phases, the agreement for other parameters was less satisfactory. The ICC results confirmed good overall agreement between the two devices across most parameters, except for SVRI during the new liver phase, which showed poor agreement. Additionally, four-quadrant and polar plot analyses indicated that all agreement rate values fell below the clinically acceptable threshold of over 90%, and all angular deviation values exceeded ± 5°, demonstrating that PiCCO is unable to meet the acceptable trends. Using the TIM, the interchangeability rates were found to be quite low: 20% for CO and CI, 16% for SVRI, and 13% for SVI.

Conclusions: Our study revealed notable disparities in absolute values of CO, CI, SVRI and SVI between PiCCO and PAC in intraoperative liver transplant settings, notably during the neohepatic phase where errors were particularly pronounced. Consequently, these findings highlight the need for careful consideration of PiCCO's advantages and disadvantages in liver transplantation scenarios, including its multiple parameters (such as the encompassing extravascular lung water index), against its limited correlation with PAC.

Keywords: Agreement analysis; Hemodynamic monitoring; Liver transplantation; PiCCO; Pulmonary artery catheterization; Trending ability..

Fig. 1

Shows eligible patients’ selection diagram

Fig. 2

Agreement was assessed using Bland-Altman analysis. (a) Bland-Altman analysis comparing the CO measured using PiCCO with that using PAC. (b) Bland-Altman analysis comparing the CI measured using PiCCO with that using PAC. (c) Bland-Altman analysis comparing the SVI measured using PiCCO with that using PAC. (d) Bland-Altman analysis comparing the SVRI measured using PiCCO with that using PAC. The blue line indicates the mean bias, and the dashed lines indicate the 95% limits of agreement in each analysis. SD, standard deviation

Fig. 3

Agreement was assessed using Passing-Bablok regression (PBR). PBR between PiCCO and PAC for CO (a), for CI (b), for SVI (c), for SVRI (d)

Fig. 4

Trending ability was assessed using the four quadrant plots. Four-quadrant plot corrected for repeated measurements shows changes in CO(a), CI (b), SVI (c), SVRI (d). The exclusion rates (red squares) for the central region were both set to 10% of the parameter mean

Fig. 5

Trending ability was assessed using polar plot. The distance from the center of the plot represents the mean change in cardiac output (△CO) and the angle with the horizontal (0-degree radial) axis represents agreement (a), the exclusion zones of 10% (b). The distance from the center of the plot represents the mean change in CI (c), the exclusion zones of 10% (d). The distance from the center of the plot represents the mean change in SVI (e), the exclusion zones of 10% (f). The distance from the center of the plot represents the mean change in SVRI (g), the exclusion zones of 10%(h). The radial agreement limit was taken as -30 to + 30°, and after excluding data from the central area, a compliance rate above 95% was considered good trend ability, 90% ~ 95% was borderline, and below 90% was poor trend ability

Fig. 6

(a) Four-quadrant graphical representation between changes in absolute values of CO measured by PiCCO and PAC ( 364 pairs of data points) according to the trend interchangeability method(TIM). (b) Four-quadrant graphical representation between changes in absolute values of CI measured by PiCCO and PAC according to the TIM. (c) Four-quadrant graphical representation between changes in absolute values of SVI measured by PiCCO and PAC according to the TIM. (d) Four-quadrant graphical representation between changes in absolute values of SVRI measured by PiCCO and PAC according to the TIM. A specific colour is applied to each change: uninterpretable (blue), non-interchangeable (red), in the gray zone of interpretation (orange), and interchangeable (green)