Abstract
Background
This study compared the cardiac output (CO) obtained from PiCCO with that obtained from the noninvasive NICOM method.
Methods
Twenty-one cirrhotic patients receiving liver transplantation were enrolled. During the operation, their CO was measured by the PiCCO system via the thermodilution method as the standard and by the NICOM method. Two parameters including cardiac index (CI) and stroke volume index (SVI) were collected simultaneously at three phases during the surgery including the dissection phase (T1), the anhepatic phase (T2), and the reperfusion phase (T3). Correlation, Bland and Altman methods, and linear mixed model were used to evaluate the monitoring ability of both systems.
Results
Poor correlation was noted between the data measured by NICOM and PiCCO; the correlation coefficients for CI and SVI measured between the two systems were 0.32 and 0.39, respectively. Bland and Altman analysis showed the percentage error of CI as 63.7%, and that of SVI as 66.6% for NICOM compared to PiCCO. Using the linear mixed model, the CI and SVI measured using NICOM were significantly higher than those using PiCCO (estimated regression coefficient 0.92 and 10.77, both p < 0.001). Mixed model analysis showed no differences between the trends of CI and SVI measured by the two methods.
Conclusions
NICOM provided a comparable CI and SVI trend when compared to the gold standard PiCCO, but it raises concerns as an effective CO monitor because of its tendency to overestimate CI and SVI especially during the state of high cardiac output.
Keywords
bioreactance ; cardiac output monitoring ; liver transplantation ; thermodilution;
1. Background
A subset of patients with cirrhosis develops ascites, volume overload, and signs of hyperdynamic circulation including high cardiac output (CO) and low systemic vascular resistance.1 Intraoperatively, this altered hemodynamics is also further complicated by the relative hypovolemia due to peripheral autonomic neuropathy, surgical manipulation of the major vessels, hemorrhage, and reperfusion.2 These patients therefore present a significant management challenge for anesthesiologists during liver transplantation. Thus, accurate cardiovascular monitoring to guide the delicate fluid management and optimize hemodynamic stability during liver transplantation is critical to maintain adequate systemic and graft perfusion.
There are at the present several forms of CO monitoring available, and they can be divided into invasive techniques and noninvasive techniques.3 The gold standard in CO measurement is the invasive thermodilution technique through the pulmonary artery catheter (PAC). However, the clinical necessity of PACs has been challenged because of its invasiveness and the possibilities of serious complications.4, 5Therefore, less invasive techniques such as the PiCCO System (Pulsion Medical Systems AG, Munich, Germany) have emerged commercially in lieu of the more invasive monitor. Many studies had shown that the PiCCO system, which uses transpulmonary thermodilution for the calibration procedure, has a high accuracy and is comparable to that of PAC.6, 7, 8, 9, 10 This less invasive PiCCO system is currently used routinely on patients undergoing liver transplantation in the National Taiwan University Hospital.
The NICOM system (Cheetah Medical Inc., Portland, OR, USA) uses four electrodes placed over the chest wall, which provide a measure of bioreactance, the analysis of frequency shifting. Compared with the traditional bioimpedance method, NICOM yields a 100-fold greater signal/noise ratio. Recent studies have shown that NICOM also provides a continuous and completely noninvasive measurement of CO for patients receiving outpatient hemodialysis therapy.11, 12 However, the accuracy of NICOM in the surgical setting is not well established.
The aim of this study is to determine the accuracy of CO measurements obtained from the noninvasive NICOM system and to compare the data with the current standard PiCCO monitoring system during liver transplantation surgery.
2. Methods
2.1. Patients and anesthesia
This study was approved by the institutional review board of National Taiwan University Hospital (trial registration: NCT01922557), and written informed consent was obtained from all patients. Cirrhotic patients scheduled for liver transplantation were enrolled from December 2010 to June 2012. Patients with history of aortic regurgitation were excluded. All patients received general anesthesia with endotracheal tube intubation. The anesthesia was induced with intravenous etomidate (18–20 mg) and fentanyl 2–4 (μg/kg). Muscle relaxation was provided by a bolus cisatracurium, followed by 3 μg/kg/min infusion. Inhaled desflurance was used as the intraoperative maintenance anesthesia. Tidal volume of the ventilator was set at 8–10 mL/kg using volume control mode with 40–60% of low flow oxygen. End tidal CO2was maintained at 35–45 mmHg. Intraoperative analgesia was given with fentanyl boluses, which was titrated according to the patients' reaction.
Aside from noninvasive monitors as suggested by the American Society of Anesthesiologists, patients also received invasive radial or brachial artery pressure and central venous pressure measured via the internal jugular vein as standard invasive monitors. All patients had a standard adult PiCCO catheter inserted through the right femoral artery with one end of the catheter connected to the PiCCO System. The CO data measured by PiCCO were acquired via the thermodilution method. The alternative device to measure CO was NICOM. All patients were attached to the NICOM electrodes after the induction of anesthesia. The anesthetic management was guided by routine monitors, and the CO value was measured by PiCCO. The measured CO values and stroke volume values were indexed by each patient's estimated body surface area.
2.2. Data collection
Two primary parameters including cardiac index (CI) and stroke volume index (SVI) were collected simultaneously at three time points including dissection phase (T1), anhepatic phase (T2), and reperfusion phase (T3) with at least two measurements made in each phase. Each of these phases was defined as follows: T1 as 30 minutes after skin incision, T2 as 5 minutes after the recipient's liver being removed, and T3 as 30 minutes after reperfusion. CI and SVI were recorded simultaneously from both PiCCO and NICOM systems, with the results abbreviated as PCI (CI of PiCCO), NCI (CI of NICOM), PSVI (SVI of PiCCO), and NSVI (SVI of NICOM). PCI was measured using the transpulmonary thermodilution technique, after injecting 15–20 mL boluses of cold normal saline via a central venous catheter. If the variation among the two readings is more than 10%, one additional measurement was taken for an average reading.
2.3. Statistical analysis
Sample size was calculated with http://biostat.mc.vanderbilt.edu/twiki/bin/view/Main/PowerSampleSize, where α = 0.05 was set. A power of 80% results in a sample size of 51 pairs of dataset. With three phases of measurement for each patient, a minimum of 17 patients was determined to be necessary. Demographic data were expressed as mean values and standard deviation (SD) for continuous variable or percentage for categorical variable. Pearson's correlation and linear regression of CI and SVI measured by different techniques were assessed. The mean, bias (mean difference), limits of agreement (bias ± 2SD), and percentage error were all calculated for both CI and SVI. Percentage error was calculated as the ratio of 2SD of the bias to mean CI or SVI of the study. From these results, the degree of agreement was analyzed using Bland–Altman analysis, where the differences between two different values of NICOM and PiCCO are plotted against the average values of these two measurements. The same analysis was repeated for both CI and SVI.
Paired t test was used to examine the differences in CI and SVI by PiCCO and NICOM. The trend analysis of PCI and NCI and PSVI and NSVI were performed by linear mixed model for the group × time interaction effect, the time effect, and the group effect. As time points were not equidistant, time was considered a categorical variable. A p value <0.05 was considered to be statistically significant. All statistical analyses were performed with SAS version 9.2 (SAS Institute Inc., Cary, NC, USA).
3. Results
Twenty-one patients were enrolled in the study (9 female and 12 male). The demographic characteristics are summarized in Table 1. The indications for liver transplantation were hepatocellular carcinoma related liver cirrhosis (8), hepatitis B or C related liver cirrhosis (8), hepatitis B exacerbation (2), primary biliary cirrhosis (1), biliary atresia status after Kasai operation (1), and alcoholic related liver cirrhosis (1).
Download full-size image
A total of 130 paired observations were collected. Data for CI and SVI were both measured by NICOM system (NCI and NSVI) and PiCCO system (PCI and PSVI). The overall mean (SD) of PCI and NCI was 4.0 L/min/m2 (1.0 L/min/m2) and 5.2 L/min/m2 (1.6 L/min/m2), respectively (p < 0.0001). The overall mean of PSVI was 44.9 mL/m2 (12.1 mL/m2) and the overall mean of NSVI was 58.5 mL/m2 (18.8 mL/m2) (p < 0.0001). The scattered plots and regression line of CI and SVI (Figure 1) showed poor correlation between the data measured by NICOM and PiCCO systems. The correlation coefficients for CI (Figure 1A) and SVI (Figure 1B) were 0.32 and 0.39 (p = 0.0004 and p < 0.0001), respectively. The slope of the regression line was less than 1, with most of the dots being distributed over the left lower side of Figure 1.
Download full-size image
The result of Bland–Altman analysis is shown in Figure 2. Figure 2A illustrates the CIs measured by NICOM and PiCCO. The mean difference (bias) of CI (NCI – PCI) was 1.22 L/min/m2 with 95% limits of agreement of −1.89 to 4.32 L/min/m2. Figure 2B illustrates the SVIs measured by NICOM and PiCCO. The mean bias of SVI (NSVI – PSVI) was 13.60 mL/m2 with 95% limits of agreement of −21.59 to 48.78 mL/m2. The percentage error of CI was 63.7%, and that of SVI was 66.6%. The scatter plots in Figures 2A and 2B both revealed that the bias increased as CI or SVI increased.
Download full-size image
When we investigated the CI and SVI at different phases, the trends for CI and SVI are shown in Figure 3. The group effect in the linear mixed model revealed that both CI and SVI measured with NICOM were significant higher than those with PiCCO, with estimated regression coefficient of 0.92 (p = 0.0009) and 10.77 (p = 0.0013), respectively. The trend analysis for both CI and SVI using mixed model analysis with repeated measures showed no differences between the trend of NCI and PCI, with estimated regression coefficient for T1 to T2 as 0.54 (p = 0.18), and estimated regression coefficient for T2 to T3 as −0.12 (p = 0.77). Similar results were found for NSVI and PSVI, with estimated regression coefficient for T1 to T2 as 4.87 (p = 0.33), and estimated regression coefficient for T2 to T3 as −0.71 (p = 0.89).
Download full-size image
4. Discussion
This is the first study to directly compare the cardiac monitoring ability of NICOM system with that of the transpulmonary thermodilution PiCCO system in intraoperative CO monitoring of cirrhotic patients receiving liver transplantation.
In the study, an overestimation of CI and SVI by the NICOM system compared to the PiCCO system is seen, as shown by the slope of the regression line in the scatter plot, and paired t test. Further analysis using the Bland–Altman plot of CI showed that the percentage error of both CI and SVI were more than 60%. In a meta-analysis performed by Critchley and Critchley13 regarding the reliability of CO measurements, they concluded that the percentage error should be less than 30% for a new measurement to be clinically acceptable. Thus, the measurement of CO by NICOM was not an acceptable for evaluating CO during liver transplantation surgery.
Our Bland–Altman plot analysis showed a sloping relationship between the two methods. Ideally, most data points should fall along the reference line at the zero difference and the slope should be near zero. However, in this study, the bias increased as CI increased, and the bias decreased as CI decreased. This indicates that as the sample means increases, the amount of disagreement increases between the two techniques. The findings further undermine the reliability of the NICOM system.
In addition to quantitative findings, it is also important to evaluate the ability of NICOM to monitor and track the change in CO over time. Although there was a significant difference between the absolute values of the parameters measured by the two monitor systems, there were no significant differences between the trends (slopes) of the measurements obtained within the three time points. Similarly, there were no significant differences between the sequential changes in SVI measured by the two monitoring systems.
In terms of practical considerations, NICOM provides a completely noninvasive method to measure CO and can be set up in a really short time. It has the potential to be used regularly in high-risk patients in operating rooms and in ICUs. Without other invasive monitors, the CO value measured by NICOM alone can provide anesthesiologists and the critical care specialists information about the trend of cardiac performance. Despite its tendency for overestimation, the trend of CO measured by NICOM is relatively reliable based on the findings of a number of prior studies14, 15, 16 and by the present study.
The NICOM system involves placement of two dual electrodes on either side of the thorax. Sine-wave high-frequency (75 kHz) current is transmitted into the body through one electrode, and the other electrode is used by the voltage input amplifier. The mean of the two will give the final value.17 Major limitations such as interference with electrocautery, proper electrode placement, variation in the thoracic blood volume because of respiration, patient's movements, and arrhythmia may affect its accuracy.18, 19 Our study showed an overestimation of CO by the bioreactance technique compared with the current gold standard CO measurement (i.e., thermodilution). There may be several explanations. First, in patients with high CO, the inherent bias by bioreactance methods tends to overestimate CO. A study was performed on healthy individuals undergoing graded cardiopulmonary exercise test, and it examined different CO estimates between bioreactance and bioimpedance techniques.17 In that study, Jakovljevic et al17 showed that although the CO estimates were not significantly different between the two methods at resting state, the estimates were higher in the bioreactance method during peak exercise. Second, the bioreactance technique is an unacceptable option to measure CO during major abdominal surgeries including liver transplantation.20, 21 During major abdominal surgeries, massive fluid shifting and tissue retraction affect the accuracy of bioreactance technique. In our study during liver transplantation, extensive abdominal wall retraction may change the relative distance of the electrodes, thus leading to the overestimation of CO. Third, the algorithm of bioreactance technique may not be applicable to patients with hyperdynamic status and severe liver cirrhosis. Further verification is needed to develop an algorithm for liver transplant recipients.21
There are a number of limitations in our current study. First of all, there was no consensus on the best method of evaluating trending ability. Many studies used the correlation or Bland–Altman methods,15, 16, 17 whereas some followed the recommendations by Myles and Cui.22, 23 All three methods were incorporated in our study and showed similar results. In addition, a few conditions during liver transplantation may affect the bioreactance measurements, including electrocautery, rapid fluid shift, and opening the abdomen. Although this may potentially affect the reliability of CO measurements by NICOM, these potential confounding conditions are unavoidable in liver transplantation surgery and may be present to varying degrees in other types of surgery. Nonetheless, further studies are needed to evaluate the effects of operation techniques on the bioreactance methods.
In conclusion, perioperative patient CO monitors are used to provide the medical staff with accurate assessments of the circulatory status. Not only should these measured parameters represent the changes in intraoperative hemodynamic conditions, but they should also provide a better-than-acceptable accuracy level. Although the NICOM system provided a reliable CI and SVI trend compared with the PiCCO system, it did not provide accurate measures of CO and tends to overestimate CI and SVI especially during the state of high CO.
Conflicts of interest
The authors declare that they have no conflicts of interest.
Acknowledgments
We thank the staff of the Department of Anesthesiology and the Department of General Surgery, National Taiwan University Hospital, for helpful assistance.