1 London, UK 2 Rome, Italy
Declaration of interest. The authors have previously worked for LiDCO Ltd (manufacturer of monitoring equipment), and RL has an equity interest in this company.
EditorIn the last few years, there have been several papers describing continuous cardiac output (CO) estimation by analysis of the arterial pressure waveform. These pulse contour cardiac output (PCO) methods require calibration for each patient using a method such as indicator dilution.
After calibration, the purpose of PCO methods is to track ensuing changes in CO. As part of a recent report in this journal,1 a PCO method was calibrated with transpulmonary thermodilution. Changes in CO were analysed but there was no corresponding graph and no discussion of these results. Unfortunately, many reports evaluating PCO do not provide any statistical comparison of changes in CO, assessed by the different methods. Instead they provide a BlandAltman analysis comparing calibrated PCO with the reference method: this can be misleading.
To demonstrate the problem, consider a hypothetical study with 50 patients comparing a PCO method with a reference cardiac output (refCO) method. After the initial calibration, CO is measured five times in each patient. The results, as typically presented, are shown in Figure 1.
|
PCO methods are designed to track changes in CO. Therefore, only changes in CO should be pooled from different patients. This is important from a clinical perspective as well as from a statistical one; a moderate CO that is falling may require more urgent attention than a low CO that is stable. Preferably the data should be analysed after logarithmic transformation because proportionate errors (rather than absolute errors) are clinically important. Bland and Altman provide further details for using logarithmic data.2
In Figure 2, the percentage change in CO following calibration was calculated for each data point. This reveals the lack of agreement between PCO and refCO. Currently, thermodilution cardiac output (TDCO) is frequently used as the reference method. Typically, the error of TDCO measurements is 10 (2 SD)% and so the error of calculated changes in TDCO is 2x10=14%. In order to demonstrate agreement between PCO and TDCO, the actual CO must change by more than these measurement errors.
|
N. W. F. Linton
R. A. F. Linton
London, UK
EditorWe appreciate the interest of Drs Linton and Linton in our article,1 and the opportunity to reply. We agree with them over the necessity for consensus about data analysis to compare the results of different studies. In our article we used the COpa technique, which involves bolus thermodilution in the pulmonary artery, as the reference method for all comparisons, because it is the current clinical standard. For statistical analysis, we used the method described by Bland and Altman.2 Additionally, correlation analysis was performed between methods. As stated in our article, changes () with each technique in COpa, aortic transpulmonary thermodilution (COart), continuous thermodilution (CCO) and PCCO were analysed using BlandAltman plots and linear regression, and the results were reported in Table 7.1 Changes in the variables were calculating by subtracting the first measurement from the second one (
1=T2T1), and the second from the third (
2=T3T2).
We now report these changes diagrammatically in Figures 3 and 4 (linear regression) and Figures 5 and 6 (bias±2 SD). No significant difference in the agreement and precision of PCCO, CCO, or COart against COpa was found during the study period. As previously reported,3 during the clamping and reperfusion phases we observed the development of a non-significant inaccuracy in these measurements. We agree that changes in CO and the accuracy of these measurements when they are changing rapidly are clinically more important than the accuracy of stable values. For this reason we compared the differences as the bias±2 SD and gave the correlation coefficients (Table 6).1 In our study the analysis of changes () confirmed the efficacy of these techniques even in stressful and extreme haemodynamic conditions, albeit with a lesser degree of agreement and precision (Figs 5 and 6). This may be because of the corresponding lack of accuracy during the clamping and reperfusion phases.
|
|
|
|
G. Della Rocca
M. G. Costa
Rome, Italy
References
1 Della Rocca G, Costa MG, Pompei L, Coccia C, Pietropaoli P. Continuous and intermittent cardiac output measurement: pulmonary artery catheter versus aortic transpulmonary technique. Br J Anaesth 2002; 88: 3506
2 Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurements. Lancet 1986; 1: 30710
3 Bottiger BW, Soder M, Rauch H, et al. Semi-continuous versus injectate cardiac output measurement in intensive care patients after cardiac surgery. Intensive Care Med 1996; 22: 3128
4 Rödig G, Prasser C, Keyl C, Liebold A, Hobbhahn J. Continuous cardiac output measurement: pulse contour analysis vs thermodilution technique in cardiac surgical patients. Br J Anaesth 1999; 82: 52530
5 Gödje O, Höeke K, Lichtwarck-Aschoff M, Faltchauser A, Lamm P, Reichart B. Continuous cardiac output by femoral arterial thermodilution calibrated pulse contour analysis: comparison with pulmonary arterial thermodilution Critical Care Med 1999; 27: 240712
6 Sakka SG, Reinhart K, Meier-Hellmann A. Comparison of pulmonary artery and arterial thermodilution cardiac output in critically ill patients. Intensive Care Med 1999; 25: 8436
7 Burhe W, Weyland A, Kazmaier S, et al. Comparison of cardiac output by pulse-contour analysis and thermodilution in patients undergoing minimally invasive direct coronary artery bypass grafting. J Cardiothorac Vasc Anesth 1999; 13: 43740