Perioperative monitoring of cardiac

Transcription

1 Continuous cardiac output monitoring with pulse contour analysis: A comparison with lithium indicator dilution cardiac output measurement James Pittman, MBBS, FRCA; Shahar Bar-Yosef, MD; John SumPing, MBChB, FRCA; Matthew Sherwood, BS; Jonathan Mark, MD Objective: Pulse contour analysis can be used to provide beat-to-beat cardiac output (CO) measurement. The current study sought to evaluate this technique by comparing its results with lithium dilution CO (LiCO) measurements. Design: Prospective, observational study. Setting: Surgical intensive care unit. Patients: Twenty-two patients after cardiac or major noncardiac surgery. Measurements: After initial calibration of the pulse contour CO (PCO) method, CO was measured by PCO and by LiCO methods at 4, 8, 16, and 24 hrs. Recalibration of PCO was performed every 8 hrs. The systemic vascular resistance and dynamic response characteristics of the arterial catheter transducer system were measured at each time point to determine whether these influenced the agreement between PCO and LiCO methods. Main Results: There was an excellent correlation between methods (r.94). Bias was small ( L/min), and clinically acceptable limits of agreement were demonstrated between techniques. Although many catheter-transducer systems had poor dynamic response characteristics, this did not influence the level of agreement between the two techniques. An increase in systemic vascular resistance between two time points did tend to cause overestimation of LiCO by the PCO. Conclusions: PCO measurement compared well with the lithium dilution method and can be considered an accurate technique for measuring beat-to-beat CO with limited risk to the patient. (Crit Care Med 2005; 33: ) KEY WORDS: pulse contour analysis; cardiac output; lithium; indicator dilution techniques; physiologic monitoring Perioperative monitoring of cardiac output (CO) allows early recognition of occult hemodynamic abnormalities and can guide resuscitation, aiming to reduce morbidity and mortality in high-risk surgical patients (1, 2). In clinical practice, physicians appear unable to estimate CO accurately using physical examination or standard monitoring methods (3, 4). Consequently, the introduction of a simple, accurate, minimally invasive continuous CO monitoring technique would be From Duke University Medical Center and Durham Veterans Affairs Medical Center, Durham, NC. Current affiliation for Dr. Pittman: Royal Devon and Exeter NHS Hospital, Devon, UK. Current affiliation for Dr. SumPing: VA North Texas Health Care System and UT Southwestern Medical Center. Current affiliation for Mr. Sherwood: Emory University School of Medicine. Supported by LiDCO, Ltd. (London, United Kingdom), to the Department of Anesthesiology, Duke University Medical Center, Durham, NC. The authors have no financial interests to disclose. Copyright 2005 by the Society of Critical Care Medicine and Lippincott Williams & Wilkins DOI: /01.CCM F a valuable addition in the management of patients. Several methods for CO measurement are clinically available, but all have drawbacks that limit their usefulness. Thermodilution is currently the most widely applied technique, but it requires pulmonary artery catheterization with known attendant risks (5, 6). An alternative approach to CO measurement has focused on analysis of the arterial blood pressure waveform (7, 8), and initial results have been encouraging (8 15). Although arterial pulse contour analysis requires peripheral arterial catheterization, this widely used procedure has a low incidence of serious complications and is frequently used in patients who would benefit from CO monitoring. In the current study, we evaluated a new CO monitor that provides a unique approach to pulse contour analysis. In order for the new method to be an acceptable CO measurement technique, it must track CO accurately for hours at a time, and it must not be unusually sensitive to the characteristics of the arterial pressure waveform or catheter-transducer systems. Hence, the primary objective of this study was to investigate the accuracy and drift of this continuous, pulse contour CO (PCO) monitoring technique over a 24-hr period by comparison with a standard indicator dilution method of CO measurement (16 18). The secondary objective was to determine the effect of the dynamic response characteristics of the arterial blood pressure monitoring system and of changes in systemic vascular resistance (SVR) on the accuracy of pulse contour measurements. MATERIALS AND METHODS After approval by the institutional review board, written informed consent was obtained from 22 patients undergoing cardiac (12 patients), vascular (five patients), urologic (four patients), or thoracic (1 patient) surgery who were scheduled for postoperative admission to the intensive care unit. Patients underwent central and peripheral intravenous and arterial catheterization as standard care for their operative procedures. All patients had direct arterial pressure measurement via a 20-gauge arterial catheter (21 radial and one femoral) connected to a 150-cm-long pressure tubing 2015

2 and pressure transducer (Baxter Healthcare). Seventeen patients underwent central venous catheterization of the right internal jugular vein with insertion of an 8.5-Fr introducer sheath and a 16-gauge central venous catheter. The remaining five patients had a 16- gauge peripheral intravenous catheter in the right or left forearm or antecubital fossa. After surgery, patients were brought to the surgical intensive care unit for continued care. PCO Measurement. PCO monitoring was performed using a commercially available device (PulseCO, LiDCO, London, UK). This monitor has been developed in conjunction with the lithium dilution CO (LiCO) method to give a beat-to-beat estimate of stroke volume and CO derived from the arterial pressure waveform. Stroke volume is calculated using a proprietary algorithm using beat duration, ejection duration, mean arterial pressure, and the modulus and phase of the first waveform harmonic (8). Similar to other pulse contour techniques, an initial calibration against another independent CO measurement technique is required. In the current study, we used LiCO for calibration. Figure 1. Regression analysis of cardiac output as measured by pulse contour cardiac output (PCO) and lithium dilution cardiac output (LiCO) in 21 patients. Each point represents the mean value of two or three readings, and data are aggregated from all measurement points across the 24-hr study period. PCO was averaged over 1 min during two separate minutes at each measurement time point. If the values differed by 10%, a third measurement over another minute was performed. The average of the two (or three) measurements was then used for data analysis. The PCO device was calibrated at baseline and every 8 hrs thereafter (i.e., at t 0, 8, 16, and 24 hrs), as recommended by the manufacturer, based on CO measured by lithium dilution. LiCO Measurement. LiCO measurement is a commercially available CO measurement technique utilizing indicator dilution principles. The major elements are described briefly below, and complete details have been previously reported (16, 18). Lithium chloride (0.3 mmol [2 ml]) is used as the indicator, and a disposable lithium-selective electrode, positioned in the arterial pressure catheter tubing, serves as the sensor. For each LiCO measurement, a lithium bolus is given through a central or peripheral intravenous catheter, while a batterypowered roller pump draws arterial blood through the lithium sensor. A lithium concentration washout curve is recorded, from which the LiCO computer derives the CO. For each time point in the current study, two consecutive LiCO measurements were made, and if these differed by 10%, a third measurement was performed. The average of these two (or three) measurements was then used in the analysis. Dynamic Response Characterization. A semisquare pressure waveform was produced by momentarily activating the flush system of the pressure transducer, and a computer printout was used to measure the dynamic response characteristics. Natural frequency, amplitude ratio, and damping coefficient were calculated according to the method described by Gardner (19). The dynamic response of the arterial catheter system was then categorized as good if it fell within the minimum requirements as found by Gardner (19); otherwise, it was categorized as poor. Experimental Procedures. The study was started when the patients were hemodynamically stable, 2 hrs after completion of surgery, and was continued over the next 24 hrs. At 4, 8, 16, and 24 hrs, both LiCO and PCO measurements were recorded for comparison. After recording the PCO values, the PCO device was recalibrated at t 8, 16, and 24 hrs based on the LiCO values. Before each measurement, the arterial pressure catheter and transducer were assessed subjectively by noting the ease of aspiration of 5 ml of blood, the ease of injection of 5 ml of saline, and the investigators visual impression of the arterial waveform contour. All catheters were flushed and aspirated easily, and the waveform was optimized if necessary by extending the wrist and manipulating the catheter. The arterial pressure monitoring system was then assessed objectively using the flush test, as described above. SVR was calculated at each measurement time point from the mean arterial pressure, central venous pressure, and LiCO using a standard formula. At each time point, the change in SVR was calculated as the difference in SVR from the previous time point. Statistical Analysis. The mean values for LiCO and PCO were compared at 4, 8, 16, and 24 hrs. Using the method described by Bland and Altman (20) for assessing agree- Table 1. Mean cardiac output, bias, upper and lower limits of agreement, and percentage error between pulse contour cardiac output (PCO) and lithium dilution cardiac output (LiCO) measurement of cardiac output (CO) Finding All Combined t 4 hrs t 8 hrs t 16 hrs t 24 hrs LiCO 6.00 (1.26) 5.72 (1.06) 5.82 (0.88) 6.32 (1.29) 6.16 (1.66) PCO 5.98 (1.59) 5.52 (1.37) 5.76 (1.31) 6.43 (1.63) 6.31 (1.94) Bias (PCO-LiCO) (0.82) 0.21 (0.79) 0.06 (0.9) 0.11 (0.90) 0.15 (0.68) Limits of agreement 1.65 to to to to to 1.51 Percentage error (%) Unless stated otherwise, data are mean (SD) in L/min. 2016

3 Figure 2. Bland Altman bias plots of pulse contour cardiac output (PCO) vs. lithium dilution cardiac output (LiCO) at(a) 4,(b) 8,(c) 16, and (d) 24 hrs in 21 patients. The mean bias (PCO-LiCO) is shown as a solid line, and the limits of agreement (bias 2 SD) are shown as a dotted line. ment between measurement techniques, differences between PCO and LiCO were plotted against the mean values for these pairs at 4, 8, 16, and 24 hrs. The bias (mean difference) and limits of agreement (bias 2 SD) were determined and used to summarize the level of agreement between methods. The percentage error at each measurement time point was calculated as the ratio between the limits of agreement (i.e., 2 SD of the bias) divided by the CO (calculated as the mean of both methods). The effect of time on bias was analyzed using a repeatedmeasures analysis of variance. The relationship between CO measurement bias and the dynamic response characteristics of the arterial pressure monitoring system was analyzed using a Student s t-test for the categorical classification of dynamic response behavior ( good vs. poor ) and using linear regression analysis for the continuous variables of natural frequency and damping coefficient. A regression analysis was also used to analyze the relationship between the changes in SVR and CO measurement bias. Data are presented as the mean 1 SD unless specified otherwise. p.05 was considered significant. RESULTS One patient was excluded from the study owing to a persistent postoperative arrhythmia that made PCO measurement impossible. The results for the remaining 21 patients are presented here. All patients were male; 15 were white, and six were African American. The patients average age, height, and weight were yrs, cm, and kg, respectively. There was a considerable interpatient variability in heart rate (range, beats/min), mean arterial pressure (range, mm Hg), and central venous pressure (range, 0 18 mm Hg). All patients were studied during hemodynamically stable periods as indicated by the small changes in heart rate and mean arterial pressure during the measurement periods. Ninety seven percent of patients had a 15% change in heart rate, and 93% had a 15% change in mean arterial pressure. A total of 83 paired measurements of LiCO and PCO were available for comparison, and all were included in the analysis. The PCO (mean, 5.98 L/min; range, L/min) and LiCO (mean, 6.00 L/min; range, L/min) were highly correlated (Pearson r ) (Fig. 1). A comparison of PCO and LiCO at 4, 8, 16, and 24 hrs showed that the two techniques provided similar results during this interval (Table 1). The SD around the mean CO was slightly greater for PCO than for LiCO at each time point. Bland Altman plots comparing the PCO and LiCO methods at 4, 8, 16, and 24 hrs are shown in Figure 2. The bias between PCO and LiCO was small ( L/min), and the limits of agreement were comparable at all time points (Table 1). 2017

4 Figure 3. Arterial pressure waveforms as recorded for six illustrative patients at 4, 8, 16, and 24 hrs demonstrating the wide intra- and interpatient morphologic differences in the pressure waveform over time. All waveforms are printed to standard scale. The amplitude of the waveform represents the pulse pressure. Overall percentage error was 27% and ranged from 22% to 31% at individual time points. Repeated-measures analysis showed no effect of time on the measurement bias (p.49). Individual arterial pressure waveforms and the dynamic response of the arterial pressure monitoring systems were recorded at each measurement point. All arterial catheters were subjectively deemed to be working acceptably at the time of CO measurement. A wide variety of arterial waveform morphologies were recorded over the 24-hr measurement period (Fig. 3). Mean natural frequency and damping coefficient of the arterial catheters were 13.6 ( 3.7) Hz and 0.28 ( 0.09), respectively. Although 68% of the catheter-transducer systems had poor dynamic response characteristics (Fig. 4), this did not influence the level of agreement between PCO and LiCO measurement: poor dynamic response catheter monitoring systems shared the same bias for CO monitoring ( L/min) as good monitoring systems ( L/min) (p.82). Similarly, linear regression analysis showed no relationship between CO measurement bias and either the natural frequency of the monitoring system (p.2) or the damping coefficient (p.75). Seventeen patients had a total of 64 measurements where central venous pressure data were available and SVR could be calculated. Although the mean change in SVR between successive measurements was small ( 26.5 dyne/sec/ cm 5 ), the range was wide ( 511 to 495 dyne/sec/cm 5 ). CO measurement bias was positively correlated with the change in SVR (r 2.06, p.04; Fig. 5): the larger the increase in SVR between two consecutive time points, the more PCO tended to overestimate the LiCO values, and conversely, the larger the decrease in SVR, the more PCO underestimated LiCO. DISCUSSION PCO measurement showed good agreement with LiCO measurement over a 24-hr period in postoperative surgical patients. Analysis at each time point showed that the mean difference between LiCO and PCO was consistently 0.2 L/min. The limits of agreement were similar across the study period and within acceptable clinical limits of 30%, as reported by other investigators (21). This new technology is minimally invasive, needing only radial artery and intravenous cannulation for CO measurement, and therefore has advantages when compared with more invasive methods such as pulmonary artery catheterization. The continuous measurement of CO allows detection of true real-time beat-to-beat changes in stroke volume and CO. To our knowledge, the accuracy of the PCO algorithm used in this study has been previously evaluated only once in a clinical trial, with results similar to ours (8). Several researchers have reported on the performance of other PCO measuring systems in various clinical situations (Table 2) (9 15). These systems used the three-element Windkessel algorithm, based on a method described by Wesseling et al (7). The arterial pressure wave was recorded in these studies from the aorta using a femoral arterial catheter, and calibration values were obtained by transpulmonary thermodilution. In general, these studies have confirmed that the theory behind pulse contour analysis is robust, and it can provide an acceptable method of continually measuring CO. It is not yet known if the different algorithm used in the current study provides a better and more reliable measurement of CO than those based on the threeelement Windkessel theory (7, 8). It does have the clinical advantage of not requiring femoral artery catheterization. All pulse contour technologies require calibration by a reference method for measuring CO, which itself may increase the risk to the patient and reduce the perceived benefits of pulse contour analysis. The PCO device reported herein uses the lithium indicator dilution method for calibration. Although traditionally CO has been measured using the pulmonary thermodilution technique through a pulmonary artery catheter, the LiCO measurement is integrated into the PCO monitor and is quickly and easily performed. Furthermore, it requires only peripheral intravenous and arterial vascular access, both of which are commonly placed for perioperative care (22). In several studies, lithium indicator dilution has been shown to be a reliable and accurate measure of CO and compares well with the thermodilution technique of CO measurement (17, 22). As a calibration technique, the LiCO measurement does have a few technical limitations, including a requirement for accurate measurement of serum sodium and hemoglobin concentrations as well as a limitation on the number of calibration measurements 2018

5 Figure 4. Natural frequencies and damping coefficients of arterial pressure monitoring systems. Monitoring systems were defined as either good (shaded area) or poor depending on these characteristics. Figure 5. Cardiac output measurement bias (i.e., the difference between pulse contour cardiac output [PCO] and lithium dilution cardiac output [LiCO]) as a function of change in systemic vascular resistance (SVR) between two consecutive time points. Lines represent the fitted regression line and 95% confidence limits for the regression line. (based on the maximal daily recommended dose of lithium of 3 mm). Lithium calibration cannot be performed in patients who have recently (15 30 mins) received neuromuscular blockers because these drugs react with the lithium sensor. In addition, the LiCO measurement cannot be used in patients receiving lithium therapy. Other potential limitations of this study should be noted (23). The mean PCO in this study was 5.98 L/min (range, L/min). We do not know if PCO remains accurate at lower or higher than normal CO, although other researchers have reported no loss of agreement between pulse contour analysis and other measurement techniques at high and low COs after pharmacologic manipulation (10, 13, 15). Additional possible causes for decreased peripheral perfusion that were not tested in our study include severe hypotension and hypothermia. Further analysis of PCO over a wide physiologic range of COs is needed to determine whether PCO, as implemented in the device studied here, remains accurate under these conditions. Arrhythmias may make pulse contour analysis unreliable: we excluded one patient from the current study because atrial fibrillation made the arterial pressure signal too unstable for PCO measurements. The heart rate can also be miscalculated when a very deep dicrotic notch is seen in the pressure waveform. New PCO device software allows the physician to alter the electrical gain on the monitor in an attempt to overcome this problem. Significant fluctuations in the compliance of the arterial vascular system may change the morphology of the arterial pressure waveform and affect the accuracy of pulse contour analysis. Such changes are likely to occur in the operating room or Table 2. Previously reported studies comparing pulse contour cardiac output measurement with the bolus thermodilution technique Study No. of Patients Population Data Points Duration (hr) Bias (L/min) Limits of Agreement (L/min) Weissman et al. (9) 11 Neurosurgery 119 Intraop Jansen et al. (10) 54 Cardiac surgery 490 Intraop Gratz et al. (11) 27 Cardiac surgery 94 Intraop Rodig et al. (12) 26 Cardiac surgery 308 Intraop 2 hrs 0.15 a 1.96 a postop Goedje et al. (13) 24 ICU Zollner et al. (14) 19 ICU Buhre et al. (15) 12 Cardiac surgery 36 Intraop Intraop, intraoperative; postop, postoperative; ICU, intensive care unit. a These are approximate values calculated from the data presented by Rodig et al (12). 2019

6 Pulse contour cardiac output measurement compared well with the lithium dilution method and can be considered an accurate technique for measuring beat-to-beat cardiac output with limited risk to the patient. intensive care unit after surgery. The alterations in arterial waveform shape over time in a sample of six patients are shown in Figure 3. Although some of these changes may be due to real fluctuations in CO, they are as likely to represent changes in arterial compliance and resistance. Frequent recalibration provides a simple solution but is potentially time consuming. On the basis of our data, calibration every 8 hrs is sufficient for accurate continuous PCO monitoring in the intensive care unit setting. The pressure waveform measured in the large arteries is a combination of the forward pressure wave produced by the contraction of the heart and the backward pressure wave reflected from arterial arteriolar junctions (8). The magnitude of wave reflection is affected by changes in arterial resistance and compliance, and ignoring this factor may explain the susceptibility of the traditional three-element Windkessel algorithm to increased measurement bias induced by changes in SVR (12). In contrast, Linton and Linton (8) have developed a different algorithm for PCO measurement that incorporates wave reflectance and have claimed a preserved accuracy in the setting of changes in SVR. In the current study, employing the same algorithm used by Linton and Linton, we found a weak association between SVR change and the CO measurement bias: an increase in SVR tended to increase the PCO measurement relative to the LiCO value, while a decrease in SVR tended to decrease the PCO relative to the LiCO. This effect, although statistically significant, was small: a change in SVR of 600 dyne/ sec/cm 5 was predictive of a change in the measurement bias of 1 L/min. However, whenever changes in PCO prompt major alteration in patient management, recalibration would be a prudent step. Schwid (24) has demonstrated that arterial pressure monitoring systems in clinical practice often achieve suboptimal dynamic response characteristics, with one half of them operating in the underdamped region. Our findings were similar (Fig. 4). It was not previously known whether the performance of the arterial catheter transducer system and the effect on the pressure signal could influence the PCO measurements. One of the important findings in this study was that the dynamic response characteristics of the arterial pressure measurement system had little effect on the agreement between PCO and LiCO. This would suggest that the characteristics of the arterial pressure transducer systems used in intensive care units do not usually have a significant effect on the accuracy of PCO, provided a clinically acceptable pressure waveform is present. CONCLUSIONS Continuous PCO measurement with intermittent calibration every 8 hrs agrees well with indicator dilution CO in intensive care patients. This technology provides a promising opportunity for true beat-to-beat CO measurement and can be used in patients with only a peripheral arterial catheter and intravenous access. REFERENCES 1. Sinclair S, James S, Singer M: Intraoperative volume optimization and length of stay after repair of proximal femoral fracture: Randomized controlled trial. BMJ 1997; 315: Mythen MG, Webb AR: Intra-operative gut mucosal hypoperfusion is associated with increased post-operative complications and cost. Intensive Care Med 1994; 20: Eisenberg PR, Jaffe AS, Schuster DP: Clinical evaluation compared to pulmonary artery catheterization in the hemodynamic assessment of critically ill patients. Crit Care Med 1984; 12: Linton RA, Linton NW, Kelly F: Is clinical assessment of the circulation reliable in postoperative cardiac surgical patients? J Cardiothorac Vasc Anesth 2002; 16: Connors AF Jr, Speroff T, Dawson NV, et al: The effectiveness of right heart catheterization in the initial care of critically ill patients. JAMA 1996; 276: Polanczyk CA, Rohde LE, Goldman L, et al: Right heart catheterization and cardiac complications in patients undergoing noncardiac surgery: An observational study. JAMA 2001; 286: Wesseling KH, Jansen JRC, Settels JJ, et al: Computation of aortic flow from pressure in humans using a nonlinear, three-element model. J Appl Physiol 1993; 74: Linton NWF, Linton RAF: Estimation of changes in cardiac output from the arterial blood pressure waveform in the upper limb. Br J Anaesth 2001; 86: Weissman C, Ornstein E, Young W: Arterial pulse contour analysis trending of cardiac output: Hemodynamic manipulations during cerebral arteriovenous malformation resection. J Clin Monit 1993; 9: Jansen JRC, Wesseling KH, Settels JJ, et al: Continuous cardiac output monitoring by pulse contour during cardiac surgery. Eur Heart J 1990; 11(Suppl): Gratz I, Karidin J, Jacobi AG, et al: Continuous noninvasive cardiac output as estimated from the pulse contour curve. J Clin Monit 1992; 8: Rodig G, Prasser C, Keyl C, et al: Continuous cardiac output measurement: Pulse contour analysis vs thermodilution technique in cardiac surgical patients. Br J Anaesth 1999; 82: Goedje O, Hoeke K, Lichtwarck-Aschoff M, et al: Continuous cardiac output by femoral arterial thermodilution calibrated pulse contour analysis: Comparison with pulmonary artery thermodilution. Crit Care Med 1999; 27: Zollner C, Haller M, Weis M, et al: Beat-tobeat measurement of cardiac output by intravascular pulse contour analysis: A prospective criterion standard study in patients after cardiac surgery. J Cardiothorac Vasc Anesth 2000; 14: Buhre W, Weyland A, Kazmaier S, et al: Comparison of cardiac output assessed by pulsecontour analysis and thermodilution in patients undergoing minimally invasive direct coronary artery bypass grafting. J Cardiothorac Vasc Anesth 1999; 13: Linton RA, Band DM, Haire KM: A new method of measuring cardiac output using lithium dilution. Br J Anaesth 1993; 71: Linton RA, Band DM, O Brien TK, et al: Lithium dilution cardiac output measurement: A comparison with thermodilution. Crit Care Med 1997; 25: Band DM, Linton RAF, O Brien TK, et al: The shape of indicator dilution curves for cardiac output measurement in man. J Physiol 1997; 498: Gardner R: Direct blood pressure measurement Dynamic response requirements. Anesthesiology 1981; 54: Bland JM, Altman DG: Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1:

7 21. Critchley LA, Critchley JAH: A metaanalysis of studies using bias and precision statistics to compare cardiac output measurement techniques. J Clin Monit 1999; 15: Garcia-Rodriguez C, Pittman JA, Cassell CH, et al: Lithium dilution cardiac output measurement: A clinical assessment of central venous and peripheral venous indicator injection. Crit Care Med 2002; 30: Van Lieshout JJ, Wesseling KH: Continuous cardiac output by pulse contour analysis? Br J Anaesth 2001; 86: Schwid HA: Frequency response evaluation of radial artery catheter-manometer systems: Sinusoidal frequency analysis rather than flush method. J Clin Monit 1988; 4: