Reliable cardiac output (CO) monitoring is particularly useful in the cirrhotic patient undergoing liver transplant because this group of patients has an altered pattern of circulation, which is characterized by increased CO, decreased peripheral vascular resistance, and reduced ventricular response to physiological, pharmacological, and surgical stressors.1–3 The contractility of the ventricle in cirrhosis is impaired, which is tolerated even though the ejection fraction and CO are elevated because of the low peripheral resistance. However, during surgery, the cirrhotic patient can decompensate because of the physiological changes and the stress of surgery. Thus, the standard CO monitoring technique in these patients during major surgery has for many years been thermodilution using a pulmonary artery catheter (PAC), but recently, this method has come under scrutiny.4,5 Newer and less invasive methods are now being adopted and may replace thermodilution. One such device is the FloTrac/ Vigileo™ (Edwards Lifesciences, Irvine, CA), a self-calibrating arterial pulse contour CO monitoring system.
Unlike previous FloTrac validation studies in cardiac surgery patients,6–10 our previous and others’ data from liver transplant patients with cirrhosis11–13 showed poor agreement and failure of the system in low peripheral resistance states. Subsequently, Edwards Lifesciences has introduced a new third-generation software algorithm (version 3.02), incorporating so-called Dynamic Tone Technology. The FloTrac algorithm is based on the physiological principle that pulse pressure is proportional to stroke volume. Because the vasculature is changing continually, continuous assessment of vascular tone should be included in any pressure-based assessment of stroke volume. Thus, Edwards Lifesciences has now incorporated a new proprietary correction factor “Khi” in their algorithm to automatically adjust for changes in vascular tone so that a wider range of peripheral resistances can be accommodated.a
Thus, we have repeated our previous study protocol11 using this new third-generation software. Our hypothesis was that the modifications to the software would improve the ability of the FloTrac to measure changes in CO accurately in hyperdynamic liver transplant patients.
Patients and Anaesthesia
After obtaining ethics committee approval and informed written consent, 21 cirrhotic patients scheduled for liver transplantation were enrolled. Anesthesia was similar in all cases. It was induced with IV thiopental and fentanyl. Muscle relaxation was provided by cisatracurium bolus followed by continuous infusion. The trachea was intubated. The lungs were ventilated using a low-flow circuit with 50% oxygen in air with a tidal volume of 8 to 10 mL · kg−1. The end-tidal carbon dioxide was maintained between 4.5 and 5.3 kPa. Anesthesia was maintained with inhaled sevofluorane and a remifentanil infusion titrated to patient response.11 Patient monitoring included a radial artery pressure reading and a thermodilution PAC (Edwards Lifesciences). The radial artery catheter was connected to a FloTrac sensor that in turn was connected to the Vigileo monitor (Edwards Lifesciences) that calculated CO from the arterial pressure wave trace using the new third-generation software. Circulatory patient management during surgery and in the intensive care unit (ICU) was guided by thermodilution CO readings and followed standard practice for this type of case at our institution.14 Thermodilution CO and systemic vascular resistance ([mean arterial blood pressure − central venous pressure] × 80/CO) readings were indexed to body mass index (BMI) (cardiac index [CI], systemic vascular resistance index [SVRI]). Severity of the liver disease was assessed by the Mayo Clinic Model End Stage Liver Disease (MELD) score (range 0 to 40), UNOS modification, as calculated through a Website calculator.b
All measurements were performed by a senior staff anesthesiologist (G.B., L.M.B., M.E., M.B., L.M., or R.M.). Datasets were recorded at the following stages during the surgery and intensive care management: T1, on abdominal incision; T2, immediately before the start of the bypass; T3, to 30 minutes later; T4, to 5 minutes after graft reperfusion; T5, on abdominal closure; and T6, to 1 hour, T7 to 6 hours, T8 to 12 hours, T9 to 18 hours, and T10 to 24 hours after ICU admission.
Simultaneous CI measurements were made by single-bolus thermodilution and the Vigileo. For each thermodilution reading (CITD), 4 consecutive measurements were made by injecting 10 mL of ice-cold saline at random times during the respiratory cycle. The consistency of each washout curve was judged visually on a Vigilance CO monitor (Edwards Lifesciences). If readings differed by >10% the measurements were repeated. The average of these measurements was used. Pulse contour readings (mean CI value over the same period) were computed by the Viligeo monitor (CIV).
Data Analysis and Statistics
Results are presented as mean (SD) unless otherwise stated. P < 0.05 was considered significant. Demographic data were compared using 2-sided unequal variance Student t test and chi-squared test. A paired t test was used to examine the differences in CI. We used Pearson’s correlation coefficient and linear regression considering patient cluster when appropriate and Bland and Altman analysis, which was presented as the bias, 95% (or 2SD) limits of agreement and the percentage error.15,16 The percentage error was calculated from 2 times the SD of the bias over the mean CI for the analysis.
We adjusted for the effects of repeated measurements within each subject in our Bland and Altman analysis using the method suggested by Myles and Cui.17 The random effect was chosen to reflect the different intercept and slope for each individual with respect to their change of measures over time.17 Thus, the random effects model estimates the within-subject variation (SD) after the between-subjects variation.17 The 95% limits of agreement were estimated by deriving the residual variance for thermodilution and for Vigileo and the method by patient interaction variance.18 A random effects model for repeated-measures data was analyzed using the “xtmixed” function in Stata version 10 (StataCorp, College Station, TX).
Analysis of trending ability was performed using 3 separate statistical techniques:
- (i) By determining the correlation coefficients between CIV and CITD for serial data in each patient, the proportion of patients in whom the correlation was significant was determined. The difference in proportion and the 95% confidence interval was estimated using the Wilson score method.
- (ii) By a modified Bland and Altman analysis using ΔCI data.17 Percentage error was calculated using the mean thermodilution CO for the study.
- (iii) By plotting Δ CIV against Δ CITD on a fo4ur-quadrant plot of serial changes in CO, where ΔCI was the change in CI between sequential readings (i.e., (T1CI) − (T2CI), etc). Trending was assessed by direction of change analysis. The concordance, or agreement, of the direction of change between consecutive readings from the thermodilution and the Vigileo was scored as a percentage of the total count that agreed. We excluded ΔCI data of <0.5 to 1.0 L · min−1 · m−2, because in the 4-quadrant trend plot, data points that lie close to the center of the plot (0,0) represent small-only changes in CO, and these data points introduce excessive statistical noise (random variation) into the calculation of concordance that reduces the usefulness of the test.19 Exclusion zones of 0.5 and 1.0 L · min−1 · m−2 were chosen because previous studies on similar datasets using receiver operator characteristic (ROC) curve analysis had shown that these zones provided optimal sensitivity and specificity.19 On the basis of previous reports, the concordance rate should be >90% to 95% to confirm good trending, but only when suitable exclusion zones have been applied.20–22
Data showing the performance of the new software collected in this study were also compared with historical data collected in our previous study11 by examining the 95% confidence intervals, and these results are also reported.
Data from 21 patients were analyzed (7 females, 14 males). Mean (SD) age was 50.7 (7.5) years (range, 32–60 years) and BMI was 24.1 (2.0) kg m−2. The underlying diseases necessitating liver transplantation were 15 cases of postviral liver cirrhosis, 5 cases of cirrhosis from alcohol abuse, and 1 case of primary biliary cirrhosis. The mean MELD score of the study population was 23 (4). Two patients needed postoperative ventilation on the ICU longer than 24 hours, and 4 patients needed postoperative cardiovascular drug support (noradrenaline, dopamine, or both).
Two hundred ten paired readings were collected. The overall mean CITD was 4.7 ± 1.1 L · min−1 · m−2 (range, 2.3 to 8.5), which was slightly higher than the overall mean CIV, which was 4.3 ± 1.0 L · min−1 · m−2 (range, 2.0 to 7.8) (P < 0.001). The scatter plot and Bland and Altman plot of these paired readings are presented in Figure 1. After considering repeated measurements of the patients, the regression line was CIV =(0.64) x CITD + 1.31, and the correlation coefficient was moderate (r =0.67, 95% confidence interval, 0.40 to 0.86). The bias was 0.39 L · min−1 · m−2 with 95% limits of agreement of −1.29 to 2.07 L · min−1 · m−2. The percentage error was 37% (Fig. 1).
When Bland and Altman analysis data were corrected for repeated measures using random effects modeling,18 the mean bias (CITD− CIV) became 0.38 L · min−1 · m−2 (95% confidence interval, 0.23 to 0.54 L · min−1 · m−2), and the adjusted limits of agreement became −1.94 to 2.71. The adjusted percentage error became 52% (95% confidence interval, 49% to 55%) (Fig. 1). There were no demonstrable relationships between gender and MELD scores or any of the study hemodynamic variables.
Populations and data from the present study are compared with data from a previous study in Tables 1 to 3.11 The mean CI was similar between the studies (P =0.46). The mean bias was significantly less using the third-generation software, and the amount of variance accounted for increased to 45%.
Mean (standard error of the mean) arterial blood pressure, CIs, and SVRI measurements for each stage of the procedure were plotted (Fig. 2: left). The CIV now followed changes in CITD with a fair degree of consistency (left middle). Comparisons are made with data from our previous study using second-generation software version 01.10 (right).11 In our previous study a larger discrepancy between CIV and CITD measurements was present (right middle). In the present study, CIV was more aligned with CITD (left middle), reflecting improved trending.
The relationship between the bias (CITD− CIV) and peripheral resistance (SVRI) is shown by a log-scale plot (Fig. 3). The present study data (upper panel) using third-generation software were influenced only slightly by low SVRI states. However, previous study data (lower panel) using second-generation software incurred a significant bias when SVRI values were low.
Assessment of Trending Ability
Regression Analysis from Individual Patients
Significant correlation between data pairs over the course of surgery and the first 24 hours in the ICU (i.e., T1 to 10) could be shown in 11 of the 21 patients using the new third-generation software. The median (interquartile range) for the correlation coefficient was r =0.65 (0.54 to 0.78). Data from our previous study using second generation software were 6 of 29 patients and r =0.41 (0.22 to 0.64). The difference in proportion of patients with significant correlation of data points between the studies was significant (32%; 95% confidence interval, 5% to 54%).
Bland and Altman Using ΔCI
The Bland and Altman plots for repeated measures of the ΔCI were drawn (Fig. 4). The third-generation software data (upper-right plot) had a bias of 0.02 (95% confidence interval, −0.23 to 0.18) L · min−1 · m−2 and limits of agreement from −2.86 to +2.81 L · min−1 · m−2. The percentage error of the trend was 62% (Table 2), where mean CI was 4.5 L · min−1 · m−2. Similar data analysis for the second-generation software data (lower-right plot) could not be completed because random effects modeling to correct for repeated measures failed to converge.
We analyzed 189 ΔCI data pairs in the present study with third-generation software, in comparison with 261 data pairs for the previous study using second-generation software. Four-quadrant plots of the ΔCI data were drawn (Fig. 4) and exclusion zones applied. Data collected from the present study (upper-left plot) were better correlated (r =0.67) than were data from the previous study (lower-left plot) (r =0.25) (Table 2). Concordance rates after applying exclusion zones of 0.5 and 1.0 L · min−1 · m−2 were marginally better using the new third-generation software, 72% and 74%, in comparison with previous study data using the second-generation software, 68% and 67% (Table 2). However, all of these rates were below the 90%–95% acceptance threshold that we previously set,11 implying that trending ability was still poor when using the third generation of software in this patient group.
This is the first study to evaluate the FloTrac/Vigileo system against single-bolus PAC thermodilution using the new third-generation software (version 3.02) in cirrhotic patients undergoing liver transplantation. The percentage error from our Bland and Altman analysis was 37% and increasing to 52% when adjusted for repeated measures (Fig. 1).17 The concordance rate was 72% when central zone data (e.g., CI <0.5 L · min−1 · m−2) were excluded (Fig. 4). These values remain outside 30% 16 and 90%–95% 11,19 benchmarks for acceptable agreement and trending. Studies using albeit previous software versions (e.g., version 1.10) by Biais et al. (e.g., percentage error 43%) and more recently by Krejci et al. (e.g., precision error 69% and concordance rate 72%),12,13 also in liver transplant patients, confirm previous findings that previous versions of the FloTrac lacked precision and trending ability. In the present study, we made comparisons between the previous generation of FloTrac software (e.g., version 01.10) using our own historical data11 and the newly collected third-generation FloTrac software data. In cirrhotic patients undergoing liver transplantation, the new third-generation software (a) reduced the adverse effect on CO reading bias of low peripheral resistance states and (b) improved the overall precision and trending ability of the system.
Figures 2 and 3 showed how the bias between thermodilution and FloTrac readings was reduced, and much less affected by changes in peripheral resistance. Regression and Bland and Altman analysis and plots (Fig. 1 and Table 1) showed that the correlation coefficient increased from r =0.39 to r =0.67, bias decreased from 1.3 to 0.4 L · min−1 · m−2, and percentage error was decreased from 60% to 52%, all significant improvements in performances when the new third-generation software was compared with the previous second-generation software. Trending ability when using the new software was also improved (Fig. 4 and Table 3). However, despite these improvements in performance, the third-generation software cannot yet be said to provide clinically acceptable precision and trending in cirrhotic liver transplant patients.
Our patient cohort had a MELD score of 23/40, which represents reasonably well controlled liver failure with a 3-month mortality rate of 20% (see Footnote b).Thus, it could be argued that our patient cohort was in relatively good health, in comparison with some institutions, to properly test the technology. However, in the 3 previous studies evaluating the FloTrac in cirrhotic liver transplant patients, the MELD scores were even less abnormal, 21/40 in our previous study, 16/40 in the Biais et al. study, and 18/40 in the Krejci et al. study.11–13 However, precision errors in these studies were still worse than seen in many FloTrac validation studies from cardiac surgery cohorts,6–10 suggesting that liver transplant surgery is a very exacting and acceptable setting for testing pulse contour technology.
Liver cirrhosis is frequently associated with a special pattern of altered hemodynamics known as cirrhotic cardiomyopathy, which is characterized by a hyperdynamic circulation, increased CO, impaired ventricular contractility, and decreased peripheral vascular resistance.1–3 Because the association between cirrhosis and these hemodynamic features is so variable, it is a significant challenge to the attending anesthesiologist, particularly during liver transplant surgery.1 Therefore, reliable hemodynamic monitoring is of great importance and needs to be properly tested. Whereas the current third-generation FloTrac software does not appear sufficiently reliable to be used in hemodynamically unstable patients in whom wide variations in peripheral resistance are expected, such as those with liver cirrhosis and septic patients taking vasopressors, it may still have a clinical role in more hemodynamically stable situations such as after cardiac surgery, and the results of validation studies in these patient groups should be more favorable.
Our 2 cohorts of patients were from the same liver transplant center in Italy and although separated by time were operated on by the same team of surgeons, anesthesiologists, and theater nurses. Patients were similar in age, weight, BMI, and MELD score (Table 3), but there were some noticeable differences in hemodynamics (Fig. 2). In particular, the CI was higher and SVRI lower during the initial stages of surgery at time points T1 to 2 when using the older second-generation software. However, it is doubtful whether these differences in hemodynamics were sufficient to account for our study outcomes; improvements in the software algorithm are the more likely explanation. The presence of significant correlation in the 10 time interval pairs (T1 to T10) occurred in only 11 of 21 cases, about 50%. In the previous study it was 6 of 29 cases, which is near 20%. As one of the indicators of trending, it shows that the 10 points line up in only about half the cases. We included these data for the sake of completeness, because individual regression analyses are an older and less refined indicator of trending ability used in our previous paper. Combined data from different patients and the strength of the correlation were other indicators that also supported our conclusion that the third-generation software (r =0.67) is better than the second-generation software (r =0.25) but that improvement is still needed.
The present study has a few obvious limitations. The trial design of comparing pairs of readings at set time points during surgery and aftercare in the ICU is in keeping with most recent studies and recommendations.19 Furthermore, the statistical analysis of the data (i) using Bland and Altman for repeated measures,17 (ii) quoting percentage errors,16 and (iii) assessing trending using concordance analysis on 4-quadrant plot with central zone data exclusion are also all in keeping with the latest recommendations.11,19 One could criticize the small study size with only 21 patients, because most clinical CO validation studies enroll 29 (20–42) subjects (shown as median [quartiles]).20 Ideally, one also should know the precision of the reference method when analyzing precision errors,23 but this was not done because of the inherent difficulties involved.24 Single-bolus thermodilution CO measurements using Edwards Lifesciences catheters and a Vigilance monitor were used; thus the precision of the reference method should have been reasonably reliable with a precision of approximately 20%.
Liver transplant surgery in cirrhotic patients represents a probative setting during which testing the potential of new cardiovascular monitoring techniques can be performed. Our data showed that the new third-generation Vigileo/ FloTrac algorithm software provided a substantial improvement over previous versions with better overall precision and trending ability. However, the percentage error was still above the benchmark limits of 30%.16 One hopes that further algorithm refinements, backed up by reliable validation studies, will increase this technology’s reliability to be extensively used in the highly complex setting of liver transplantation in cirrhotic patients.
Name: Gianni Biancofiore, MD.
Contribution: This author helped design the study, conduct the study, and write the manuscript.
Attestation: Gianni Biancofiore approved the final manuscript.
Name: Lester AH Critchley, MD.
Contribution: This author helped analyze the data and write the manuscript.
Attestation: Lester A. H. Critchley approved the final manuscript.
Name: Anna Lee, PhD.
Contribution: This author helped analyze the data and write the manuscript.
Attestation: Anna Lee approved the final manuscript.
Name: Xiao-Xing Yang, PhD.
Contribution: This author helped analyze the data.
Attestation: Xiao-Xing Yang approved the final manuscript.
Name: Lucia M. Bindi, MD.
Contribution: This author helped design the study and conduct the study.
Attestation: Lucia M. Bindi approved the final manuscript.
Name: Massimo Esposito, MD.
Contribution: This author helped conduct the study.
Attestation: Massimo Esposito approved the final manuscript.
Name: Massimo Bisà, MD.
Contribution: This author helped conduct the study.
Attestation: Massimo Bisà approved the final manuscript.
Name: Luca Meacci, MD.
Contribution: This author helped conduct the study.
Attestation: Luca Meacci approved the final manuscript.
Name: Roberto Mozzo, MD.
Contribution: This author helped conduct the study.
Attestation: Roberto Mozzo approved the final manuscript.
Name: Franco Filipponi, MD.
Contribution: This author helped design the study.
Attestation: Franco Filipponi approved the final manuscript.
a Available from http://www.edwards.com/sitecollectionimages/products/mininvasive/ar04099.pdf.
b Website calculator available at http://www.mayoclinic.org/meld/mayomodel6.html. Accessed on September 2010.
1. De Wolf AM. Perioperative assessment of the cardiovascular system in ESLD and transplantation. Int Anesthesiol Clin 2006; 44: 59–78
2. Milani A, Zaccaria R, Bombardieri R, Gasbarrini A, Pola G. Cirrhotic cardiomiopathy. Dig Liver Dis 2007; 39: 507–15
3. Lee RF, Glenn TK, Lee SS. Cardiac dysfunction in cirrhosis. Best Pract Res Clin Gastroenterol 2007; 21: 125–40
4. Sandham JD, Hull RD, Brant RF, Knox L, Pineo GF, Doig CJ, LaPOrta DP, Viner S, Passerini L, David H, Kirby A, Jack M. A randomized, controlled trial of the use of pulmonary-artery catheters in high-risk surgical patients. N Engl J Med 2003; 348: 5–14
5. Harvey S, Harrison DA, Singer M, Ashcroft J, Jones CM, Elbourne D, Brampton W, Williams D, Young D, Rowan K. Assessment of the clinical effectiveness of pulmonary artery catheters in the management of patients in intensive care (PAC-man): a randomized controlled trial. Lancet 2005; 366: 472–7
6. Mehta Y, Chand RK, Sawhney R, Bhise M, Singh A, Trehan N. Cardiac output monitoring: comparison of a new arterial pressure waveform analysis to the bolus thermodilution technique in patients undergoing off-pump coronary artery bypass surgery. J Cardiothorac Vasc Anesth 2008; 22: 394–9
7. Staier K, Wiesenack C, Günkel L, Keyl C. Cardiac output determination by thermodilution and arterial pulse waveform analysis in patients undergoing aortic valve replacement. Can J Anaesth 2008; 55: 22–8
8. Prasser C, Trabold B, Schwab A, Keyl C, Ziegler S, Wiesenack C. Evaluation of an improved algorithm for arterial pressure– based cardiac output assessment without external calibration. Intensive Care Med 2007; 33: 2223–5
9. Mayer J, Boldt J, Wolf MW, Lang J, Suttner S. Cardiac output derived from arterial pressure waveform analysis in patients undergoing cardiac surgery: validity of a second generation device. Anesth Analg 2008; 106: 867–72
10. Senn A, Button D, Zollinger A, Hofer CK. Assessment of cardiac output changes using a modified FloTrac/Vigileo algorithm in cardiac surgery patients. Crit Care 2009; 13: R32
11. Biancofiore G, Critchley LA, Lee A, Bindi L, Bisà M, Esposito M, Meacci L, Mozzo R, DeSimone P, Urbani L, Filipponi F. Evaluation of an uncalibrated arterial pulse contour cardiac output monitoring system in cirrhotic patients undergoing liver surgery. Br J Anaesth 2009; 102: 47–54
12. Biais M, Nouette-Gaulain K, Cottenceau V, Vallet A, Cochard JF, Revel P, Sztark F. Cardiac output measurement in patients undergoing liver transplantation: pulmonary artery catheter versus uncalibrated arterial pressure waveform analysis. Anesth Analg 2008; 106: 1480–6
13. Krejci V, Vannucci A, Abbas A, Chapman W, Kangrga IM. Comparison of calibrated and uncalibrated arterial pressure– based cardiac output monitors during orthotopic liver transplantation. Liver Transpl 2010; 16: 773–82
14. Biancofiore G, Bindi ML, Romanelli A, Boldrini A, Bisà M, Esposito M, Urbani L, Catalano G, Mosca F, Filipponi F. Fast track in liver transplantation: 5 years’ experience. Eur J Anaesthesiol 2005; 22: 584–90
15. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1: 307–10
16. Critchley LA, Critchley JA. A meta-analysis of studies using bias and precision statistics to compare cardiac output measurement techniques. J Clin Monit Comput 1999; 15: 85–91
17. Myles PS, Cui J. Using the Bland–Altman method to measure agreement with repeated measures. Br J Anaesth 2007; 99: 309–11
18. Carstensen B, Simpson J, Gurrin LC. Statistical models for assessing agreement in method comparison studies with replicate measurements. Int J Biostat 2008; 4 (1) article 16
19. Critchley LA, Lee A, Ho AM. A critical review of the ability of continuous cardiac output monitors to measure trends in cardiac output. Anesth Analg 2010; 111: 1180–92
20. Perrino AC, Harris SN, Luther MA. Intraoperative determination of cardiac output using multiplane transesophageal echocardiography: a comparison to thermodilution. Anesthesiology 1998; 89: 350–7
21. Jansen JR, Schreuder JJ, Mulier JP, Smith NT, Settels JJ, Wesseling KH. A comparison of cardiac output derived from the arterial pressure wave against thermodilution in cardiac surgery patients. Br J Anaesth 2001; 87: 212–22
22. Osthaus WA, Huber D, Beck C, Winterhalter M, Boethig D, Wessel A, Sümpelmann R. Comparison of electrical velocimetry and transpulmonary thermodilution for measuring cardiac output in piglets. Paediatr Anaesth 2007; 17: 749–55
23. Cecconi M, Rhodes A, Poloniecki J, Della Rocca G, Grounds RM. Bench-to-bedside review: the importance of the precision of the reference technique in method comparison studies—with specific reference to the measurement of cardiac output. Crit Care 2009; 13: 201
© 2011 International Anesthesia Research Society
24. Yang XX, Critchley LA, Joynt GM. Determination of the precision error of the pulmonary artery thermodilution catheter using an in vitro continuous flow test rig. Anesth Analg 2011; 112: 70–7