- Email: [email protected]
Esophageal Doppler monitor determinations of cardiac output and preload during cardiac operations
Esophageal Doppler Monitor Determinations of Cardiac Output and Preload During Cardiac Operations Charles J. DiCorte, MD, Paige Latham, MD, Phillip E. Greilich, MD, Mary V. Cooley, MT, Paul A. Grayburn, MD, and Michael E. Jessen, MD Division of Thoracic and Cardiovascular Surgery, Department of Surgery, Department of Anesthesiology and Pain Management, and Division of Cardiology, Department of Internal Medicine, The University of Texas Southwestern Medical Center at Dallas, and the Dallas Veterans Affairs Medical Center, Dallas, Texas
Background. Perioperative management of cardiac surgical patients frequently mandates measurements of cardiac output and left ventricular filling. This study compared cardiac output and left ventricular filling measured by pulmonary artery (PA) catheter and esophageal Doppler monitor (EDM). Methods. Thirty-four patients undergoing coronary artery bypass grafting were prepared by implanting a PA catheter, an EDM, and a transit-time ultrasonic flow probe around the ascending aorta. In 20 patients, left ventricular end-diastolic short-axis area (EDA) was measured by transesophageal echocardiography. At five time points, cardiac output was measured from the flow probe, the EDM, and the PA catheter (by thermodilution), and left ventricular filling was assessed from the PA catheter (as PA diastolic pressure), the EDM (corrected flow time), and the EDA. For cardiac output, concordance correlations relating EDM to flow probe and PA catheter to flow probe were calculated, transformed (Fisher’s z transfor-
mation), and compared by Student’s t test. For left ventricular filling, regression coefficients were created between corrected flow time and EDA and between PA diastolic pressure and EDA. Spearman correlations were compared by Wilcoxon rank sum test. Results. The EDM and the PA catheter exhibited similar relationships to the flow probe (concordance correlations, 0.55 ⴞ 0.35 [mean ⴞ standard deviation] and 0.49 ⴞ 0.34, respectively; p ⴝ 0.088). The correlation between corrected flow time and EDA was better than the correlation between PA diastolic pressure and EDA (concordance correlations, 0.49 ⴞ 0.55 versus 0.10 ⴞ 0.43, respectively; p < 0.01). Conclusions. These data suggest that the EDM may offer a less invasive technique for evaluating cardiac output and a more accurate estimate for preload compared with the PA catheter. (Ann Thorac Surg 2000;69:1782– 6) © 2000 by The Society of Thoracic Surgeons
A
management of some patients in the early postoperative period, but PA catheters carry risks of major complications [3] and add substantial cost. The value of PA catheters has not been well defined in cardiac surgical patients. However, use of these devices may be associated with an increased mortality risk in patients with acute myocardial infarction [4] and in critically ill patients in the intensive care unit [5]. In cardiac surgical patients, the accuracy of PA catheters has also been questioned, as measurement of CO by thermodilution can be subject to error induced by rapid temperature shifts after separation from cardiopulmonary bypass (CPB) [6, 7]. An alternative method for monitoring cardiac performance uses Doppler signals acquired by an esophageal Doppler monitor (EDM). With this technique, a pulsed, competitive-frequency continuous wave Doppler signal is emitted from a probe placed in the distal esophagus and directed at the descending thoracic aorta. The reflected ultrasound frequencies are recorded from a transducer on the probe and are processed to display a continuously monitored blood velocity signal. These velocity data can then be converted to flow by applying an equation that includes the estimated cross-sectional di-
s cardiac surgical procedures are more frequently performed in patients at high risk for postoperative complications, systems for monitoring cardiac performance assume greater importance. It was identified in 1969 that adult patients who failed to achieve a cardiac index of 2.0 L 䡠 min⫺1 䡠 m⫺2 in the early postoperative period were at greater risk for death and complications [1]. As a result, most surgical teams use monitoring techniques to quantify cardiac output (CO) and filling conditions in the perioperative period in some fraction of cardiac surgical patients. The most common tool in use is the pulmonary artery (PA) catheter [2]. This device provides a calculated CO determination from thermodilution, an indirect assessment of cardiac filling from PA diastolic or wedge pressures, and can supply a measurement of mixed venous oxygen saturation. These data can be important in the Accepted for publication Nov 24, 1999. Address reprint requests to Dr Jessen, Division of Thoracic and Cardiovascular Surgery, Department of Surgery, The University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd, Dallas, TX 75235-8879; e-mail: [email protected].
© 2000 by The Society of Thoracic Surgeons Published by Elsevier Science Inc
0003-4975/00/$20.00 PII S0003-4975(00)01129-2
Ann Thorac Surg 2000;69:1782– 6
ameter of the descending aorta, which is derived from nomograms [8]. The Doppler flow velocity waveform can also be analyzed to identify the flow time or duration of left ventricular ejection. This can be corrected for heart rate (cycle time) to yield an index of cardiac filling [9]. The device is simple to insert and, in initial studies, appears to yield CO measurements that correlate well with thermodilution values [10]. However, the technique is dependent on estimation of aortic diameter, its use can be affected by peripheral vascular disease, and it has seen limited application in patients undergoing cardiac operations [11]. The cardiac surgical patient provides a unique opportunity to compare these two monitoring devices. The surgical incision allows access to the ascending aorta for placement of a flow probe to directly measure left ventricular flow. Flow data can be compared with CO values simultaneously acquired from the thermodilution PA catheter and the EDM. In addition, transesophageal echocardiographic images can be obtained at the same setting to quantify left ventricular dimensions. These data can then be correlated with other indices of cardiac filling: PA diastolic pressure (PAD) measurements and Doppler-derived corrected flow time (FTc). This study was undertaken to compare CO measurements and cardiac filling variables immediately before and after CPB in patients undergoing coronary artery bypass grafting.
Material and Methods Thirty-six patients undergoing elective primary coronary artery bypass grafting at the Dallas Veterans Affairs Medical Center between July 1 and December 31, 1997, were studied under a protocol approved by the institutional review board. No patient had any clinical or echocardiographic evidence of valvular heart disease or aortic disease. In 2 of the 36 patients, an aortic flow probe of appropriate size was not available, and these patients were excluded from the study. The demographic characteristics of the remaining 34 patients are shown in Table 1.
Experimental Protocol All patients received premedication with midazolam hydrochloride and Reglan (metoclopramide hydrochloride) and were taken to the operating room in the fasted state. General anesthesia was induced with intravenous etomidate (0.2 to 0.4 mg/kg), rocuronium bromide (0.1 mg/kg), and fentanyl (5 to 10 g/kg). Anesthesia was maintained throughout the surgical procedure with fentanyl, midazolam, and isoflurane (0.2% to 2.0%). After induction, monitoring lines were inserted and consisted of a radial artery pressure catheter, Foley catheter, electrocardiographic electrodes, and an 8F introducer sheath placed in the right internal jugular vein. An Oximetrix PA catheter (Mountain View, CA) was advanced through the sheath into the PA until a wedge tracing was obtained. The balloon was left deflated for the remainder of the study. An EDM (Deltex Medical, Irving, TX) was inserted through the oropharynx into the distal esophagus ap-
DiCORTE ET AL ESOPHAGEAL DOPPLER MONITOR
1783
Table 1. Demographic Data for 34 Patientsa Variable Age (y) Sex (male:female) Height (cm) Weight (kg) BSA (m2) ABI No. of grafts Use of LITA (%) Preoperative LVEF Preoperative LVEDP (mm Hg) Aortic cross-clamp time (min) CPB time (min) Lowest CPB temperature (°C) a
Value 62 ⫾ 7 34:0 184 ⫾ 13 92 ⫾ 16 2.1 ⫾ 0.2 1.1 ⫾ 0.3 3.2 ⫾ 1.0 94 0.55 ⫾ 0.13 19 ⫾ 6 70 ⫾ 20 117 ⫾ 32 28 ⫾ 1
Where applicable, data are shown as the mean ⫾ the standard deviation.
ABI ⫽ ankle-brachial blood pressure index; BSA ⫽ body surface area; CPB ⫽ cardiopulmonary bypass; LITA ⫽ left internal thoracic artery; LVEDP ⫽ left ventricular end-diastolic pressure; LVEF ⫽ left ventricular ejection fraction.
proximately 35 to 40 cm from the incisors. The probe was manipulated until an optimal signal was obtained. In 20 of the patients, a 5-MHz esophageal echocardiographic probe (Sonos 2500; Hewlett-Packard, Andover, MA) was also inserted into the midesophagus, and the heart was imaged. A median sternotomy was performed, and saphenous vein and the left internal thoracic artery were harvested. The pericardium was incised and fashioned into a cradle. Each patient was administered heparin sodium in a dose of 300 U/kg intravenously, and an aortic cannula was placed in the distal ascending aorta and connected to the CPB circuit. A sterile aortic flow probe (Transonic Systems, Inc, Ithaca, NY) matched to the size of the ascending aorta (28 to 36 mm) was selected and placed around the ascending aorta proximal to the cannula. The probe was connected to a Transonic Systems flow computer and calibrated. The heart, aorta, and flow probe were immersed in warm, sterile saline solution. All pressure transducers were zeroed, and internal calibration of the mixed venous oxygen saturation monitor was completed. The first set of data was then acquired. Heart rate, systemic and PA blood pressures, central venous pressure, and mixed venous oxygen saturation were recorded from their respective monitors. The EDM signal was optimized and the tracing frozen, and values for CO (COEDM) and FTc were recorded. The flow value from the aortic flow probe (COFP) was also recorded. A short-axis echocardiographic image of the left ventricle was obtained at the midpapillary level and recorded. Three separate thermodilution measurements of CO were obtained over 3 minutes by injection of 10 mL of room temperature saline solution into the proximal port of the PA catheter. If an unstable baseline message appeared, the resulting thermodilution value was discarded, and the process was repeated until a satisfactory measurement was obtained. After the three thermodilution measurements were made, a repeat measurement of COFP
1784
DiCORTE ET AL ESOPHAGEAL DOPPLER MONITOR
was recorded, and repeat values of COEDM and FTc were obtained. Thermodilution injections and CO determinations were made by a senior anesthesiologist. The EDM data were acquired by a representative from the manufacturer (Deltex Medical). Echocardiographic images were acquired by a separate anesthesiologist or cardiologist familiar with the technique. All data were recorded by an individual investigator (MVC), and results of any one technique were blinded to investigators acquiring data by another method. Flow probe, EDM, and thermodilution measurements were averaged at each time point. After initial data acquisition, volume from the CPB circuit (primed with 1,500 mL of lactated Ringer’s solution, 200 mL of 20% mannitol, 100 mL of 25% albumin, and 10,000 units of heparin) was infused through the aortic cannula until an increase in COFP of at least 1 L/min was observed. A repeat set of all variables was then measured and recorded. After this, additional volume was infused to achieve a further increment in flow, and all variables were measured and recorded again. The aortic flow probe was removed, cannulation was completed, and the patient was placed on CPB and cooled to 28°C. Coronary artery bypass grafting was performed, and the patient was rewarmed during the construction of the proximal anastomoses. The lungs were reinflated, and the patient was separated from CPB. The aortic flow probe was repositioned, and signals from all devices were optimized. Another set of measurements was recorded by the identical protocols 5 minutes after separation from CPB. Volume was then infused as judged appropriate by the surgical team who were observing the heart directly and monitoring systemic blood pressure. Fifteen minutes after separation from bypass, a final set of measurements was recorded. Thus, data were measured at a total of five time points for each patient. No protamine sulfate was administered until after the final time point. The aortic flow probe was removed and the operation completed.
Data Analysis For each patient at each time point, averages of CO measured from the two aortic flow probe recordings (COFP), the two EDM recordings (COEDM), and the three thermodilution recordings (COTD) were used. Each of the 34 subjects had five observations of CO (one from each time point) by the three different techniques. The direct measurement of aortic flow from the transit-time ultrasonic flow probe, COFP, was considered the gold standard. For each patient, a concordance correlation was computed for the relationship between COEDM and COFP and between COTD and COFP. Each concordance correlation was transformed to statistical normality by the Fisher z transformation. The 34 pairs of transformed concordance correlations were compared by a paired t test, and a p value of less than 0.05 was considered significant. The comparison of indices of cardiac filling required a slightly modified analysis. The echocardiographic images were analyzed by tracing the left ventricular endocardium at end-diastole at the midpapillary level in the
Ann Thorac Surg 2000;69:1782– 6
short-axis view to obtain an end-diastolic area (EDA). This measurement was considered the gold standard for ventricular filling [12]. Variables from the EDM (FTc) and the PA catheter (PAD) use indirect methods (time and pressure, respectively) to imply volume conditions. Therefore, unlike CO comparisons, correlations between these variables with different units cannot be analyzed with the assumption that values should be equal to the gold standard. Instead, filling measurements should rise or fall in proportion to changes observed in the EDA. For each of the 20 patients with transesophageal echocardiographic data, Spearman correlations were computed for FTc versus EDA and for PAD versus EDA. These correlations examine how well the data fit the linear regression relating the two variables over the five time points as opposed to how closely they fit the equality line ( y ⫽ x). The Spearman correlations of these regressions were compared by Wilcoxon rank sum test, and p values of less than 0.05 were considered significant.
Results Cardiac output data were successfully acquired by all three techniques in the 34 patients in this study. In 20 patients, additional data were obtained from transesophageal echocardiography to assess filling conditions of the left ventricle. All patients were in hemodynamically stable condition throughout the prebypass period and remained in stable condition during the early postbypass interval. The mean volumes of CPB circuit prime administered between the first and second measurements and between the second and third measurements were 261 ⫾ 93 mL and 251 ⫾ 146 mL (mean ⫾ SD), respectively. These volume manipulations created a substantial alteration in both filling conditions and CO. After separation from bypass, intravenous nitroglycerin was routinely used to control blood pressure. No vasopressor agents were used in these patients except to modulate blood pressure during CPB. No inotropic support was used at any time during the study. The relationship between COEDM and COFP for all patients at all time points is shown in Figure 1. In general, the correlation between COs measured by these two techniques was good. The equation of the linear regression of this relationship was calculated as: COEDM ⫽ 1.027 ⫻ COFP ⫹ 0.571. The relationship between COTD and COFP was also linear over the range studied. This relationship is depicted in Figure 2 and is described by the equation: COTD ⫽ 0.874 ⫻ COFP ⫹ 1.625. To compare the relative accuracy of COEDM and COTD with the data acquired from the aortic flow probe (or COFP), concordance correlations were created for each patient as already described. Concordance correlations (mean ⫾ standard deviation) were as follows: COEDM versus COFP, 0.55 ⫾ 0.35, and COTD versus COFP, 0.49 ⫾ 0.34. Although COEDM exhibited a closer affinity to COFP than did COTD, this difference was not significant by paired t test ( p ⫽ 0.088). For comparison of the indices of left ventricular filling, regressions were also created between FTc from the EDM
Ann Thorac Surg 2000;69:1782– 6
Fig 1. Relationship between cardiac output as recorded by esophageal Doppler monitor (COEDM) and by aortic flow probe (COFP) for all patients at all time points (n ⫽ 160). The linear regression line and 95% confidence limits are plotted. This relationship is described by the equation: COEDM ⫽ 1.027 ⫻ COFP ⫹ 0.571 (r ⫽ 0.765).
and EDA from transesophageal echocardiographic information. Spearman correlation statistics from these regressions were compared with those calculated from the regression of PAD (from the PA catheter) and EDA. Regression coefficients (mean ⫾ standard deviation) for FTc versus EDA and PAD versus EDA were 0.49 ⫾ 0.55 and 0.10 ⫾ 0.43, respectively. The regression coefficients for FTc were significantly higher than those for PAD by Wilcoxon rank sum test ( p ⬍ 0.01).
Comment The immediate perioperative period is the most critical time of most cardiac surgical procedures and efforts to optimize hemodynamics during this interval can be critical to patient outcomes. Most adverse outcomes after coronary artery bypass grafting are related to either inadequate early cardiac performance or late multisystem organ failure [13]. These problems can be reduced if hemodynamic inadequacies can be identified and treated early in their course. In high-risk patients, monitoring of CO may allow the surgical team to identify situations where organ perfusion is in jeopardy and make interventions to avoid end-organ dysfunction. When CO or blood pressure is low, accurate measurement of filling conditions can be critical to the analysis of the hemodynamic problem. In general (assuming a stable rhythm), the first goal is to optimize cardiac filling. Increases in CO and ventricular performance derived from increases in preload are more efficient in terms of myocardial oxygen consumption than are increases produced by augmenting myocardial contractility with inotropic agents [14]. Thus, measurements of CO and ventricular filling can be very important in managing patients after cardiac surgical procedures. Measurement of CO can be accomplished by a variety of techniques, each with unique advantages and disadvantages. Direct measurement of left ventricular output by an aortic flow probe is done by comparisons of antegrade and retrograde ultrasound transit-time and may offer the most accurate determination [15, 16]. However, this method requires invasive instrumentation that is not easily adapted to the postoperative arena, does not account for the portion of CO that supplies the coronary
DiCORTE ET AL ESOPHAGEAL DOPPLER MONITOR
1785
circulation, and can be inaccurate if the probe size does not match the ascending aorta precisely. Thermodilution CO determinations mandate intravascular and intracardiac line placement and can be affected by valvular abnormalities (especially tricuspid insufficiency [17]) and inaccuracies in injectate volume or temperature [18]. Thermal variations in patients recently rewarmed after CPB can also cause error [6, 7]. Thermodilution CO can be obtained continuously by some catheters currently available, but the initial measurements are not displayed until several minutes after separation from CPB, and therefore, standard injection methods were used in this study. The EDM is less invasive and simpler to insert, although esophageal instrumentation can entail some risks. The technique depends on the ability of the user to accurately focus the signal on the descending aorta and makes assumptions on aortic diameter based on algorithms from patient size. In this study, EDM measurements of CO were at least as accurate as thermodilution determinations when compared with aortic flow measured directly. This suggests that the less invasive EDM device can play a useful role in monitoring the postoperative cardiac surgical patient. Thermodilution measurements were slightly, but not significantly, less accurate. The thermodilution CO measurements may have been adversely affected by the thermal changes induced by systemic rewarming after CPB. However, repeating the analysis with only prebypass observations did not materially alter this relationship. Using left ventricular dimensions measured by transesophageal echocardiography as the reference standard, this study also found that cardiac filling was better assessed by FTc than by PAD. Correlation coefficients between PAD and EDA were very poor. The correlation coefficients between FTc and EDA, although significantly better, were not especially strong by some criteria. In part, this finding may reflect the uncertainties caused by indirect methods of estimating chamber volume from time and pressure information. The superiority of FTc as an index of left ventricular filling suggests that the EDM offers advantages over the
Fig 2. Relationship between cardiac output as recorded by thermodilution technique (COTD) and by aortic flow probe (COFP) for all patients at all time points (n ⫽ 160). The linear regression line and 95% confidence limits are plotted. This relationship is described by the equation: COTD ⫽ 0.874 ⫻ COFP ⫹ 1.625 (r ⫽ 0.748).
1786
DiCORTE ET AL ESOPHAGEAL DOPPLER MONITOR
PA catheter for cardiac monitoring. For PAD to increase, the left ventricle must be loaded sufficiently to raise left ventricular end-diastolic pressure. In a ventricle with normal diastolic compliance, major volume changes may not produce substantial pressure changes, thus limiting the usefulness of PAD as a marker for filling. Also, ventricular compliance can change after cardioplegic arrest, hypothermia, or ischemia, thereby making comparisons with prebypass values hazardous. In contrast, the FTc value implies the amount of time required to eject the stroke volume, corrected for heart rate. As the ventricle is filled, it requires more time to eject the greater stroke volume, even in ventricles with normal compliance. This may explain the superiority of FTc as an index of preload. Studies in volunteers subjected to maneuvers that alter preload, afterload, and contractility have documented good correlation of FTc with preload changes and minimal effect of alterations in inotropic state. Afterload reduction influences FTc and may need to be considered when assessing the volume status of a patient with esophageal Doppler techniques [19]. There are limitations to this study. To our knowledge, it is the first to compare EDM and PA catheter techniques, with a transit-time ultrasonic flow probe and transesophageal echocardiograms providing reference values. However, this additional instrumentation mandated that all measurements be made with an open chest and open pericardium. Sternal closure and positivepressure ventilation may influence these results. The use of PA wedge pressure might prove to be a better reflection of left ventricular volume changes than PAD. We did not measure PA capillary wedge pressure because of concern about PA injury by cold catheters in the early period after bypass. The study size is small (34 patients), and this could limit the ability to detect differences in some variables, particularly CO. In addition, patients with severely impaired ventricles were not included in this study, in part because of our reluctance to distend such hearts during the prebypass loading phase. Therefore, it may not be possible to generalize our results to this subpopulation. Patients included did not have severe peripheral vascular disease, aortic atherosclerosis, or valvular heart defects. The influence of these factors is not defined in this study. Finally, all data were collected by experienced personnel. Whether these results, particularly the EDM findings, can be replicated by observers less familiar with the technology remains to be established. Other investigators [20] have found EDM information to be reproducible after a brief learning curve. Although many centers limit the use of any form of invasive monitoring to reduce costs or complications after cardiac operations, there is a group of patients in whom hemodynamic monitoring is important. On the basis of this study, the EDM may offer a less invasive technique for evaluating CO and a more accurate estimate of preload compared with the PA catheter. Further investigation of this technology appears warranted.
Ann Thorac Surg 2000;69:1782– 6
References 1. Dietzman RH, Ersek RA, Lillehei CW, Castañeda AR, Lillehei RC. Low output syndrome. Recognition and treatment. J Thorac Cardiovasc Surg 1969;57:138–50. 2. Civetta JM, Gabel JC. Flow-directed pulmonary artery catheterization in surgical patients. Ann Surg 1972;176:753– 6. 3. Boyd KD, Thomas SJ, Gold J, Boyd AD. A prospective study of complications of pulmonary artery catheterizations in 500 consecutive patients. Chest 1983;84:245–50. 4. Zion MM, Balkin J, Rosenmann D, et al. Use of pulmonary artery catheters in patients with acute myocardial infarction. Analysis of experience in 5,841 patients in the SPRINT Registry. SPRINT Study Group. Chest 1990;98:1331–5. 5. Connors AF, Speroff T, Dawson NV, et al. The effectiveness of right heart catheterization in the initial care of critically ill patients. JAMA 1996;276:889–97. 6. Bazaral MG, Petre J, Novoa R. Errors in thermodilution cardiac output measurements caused by rapid pulmonary artery temperature decreases after cardiopulmonary bypass. Anesthesiology 1992;77:31–7. 7. Latson TW, Whitten CW, O’Flaherty D. Ventilation, thermal noise, and errors in cardiac output measurements after cardiopulmonary bypass. Anesthesiology 1993;79:1233– 43. 8. Payen D. Oesophageal Doppler monitoring: history, physical principles, and clinical applications. Int Proc J 1994;1:3–9. 9. Wallmeyer K, Wann LS, Sagar KB, et al. The influence of preload and heart rate on Doppler echocardiographic indexes of left ventricular performance: comparison with invasive indexes in an experimental preparation. Circulation 1986;74:181– 6. 10. Valtier B, Mottlin D, Nolland B, Molis T, Payen D. Cardiac output measurement by four methods in intensive care units: interest of oesophageal Doppler device [Abstract]. Am Rev Respir Dis 1992;145:A781. 11. Fratacci M-D, Payen D, Beloucif S, Laborde F. Doppler cardiac output and left ventricular performance after cardiac surgery. Chest 1992;102:380– 6. 12. Clements FM, Harpole DH, Quill T, Jones RH, McCann RL. Estimation of left ventricular volume and ejection fraction by two-dimensional transoesophageal echocardiography: comparison of short axis imaging and simultaneous radionuclide angiography. Br J Anaesth 1990;64:331– 6. 13. Cosgrove DM, Loop FD, Lytle BW, et al. Primary myocardial revascularization. Trends in surgical mortality. J Thorac Cardiovasc Surg 1984;88:673– 84. 14. Lekven J, Brunsting LA, Jessen ME, Abd-Elfattah AS, Doherty NE 3d, Wechsler AS. Myocardial oxygen use during epinephrine administration to ischemically injured canine hearts. Circulation 1988;78(5 Pt 2):III125–36. 15. Lundell A, Bergqvist D, Mattsson E, Nilsson B. Volume blood flow measurements with a transit time flowmeter: an in vivo and in vitro variability and validation study. Clin Physiol 1993;13:547–57. 16. Dean DA, Cabreriza SE, Jia C-X, et al. Validation study of a new transit-time ultrasonic flow probe for continuous measurements on the great vessels [Abstract]. ASAIO J 1996;42:26. 17. Cigarroa RG, Lange RA, Williams RH, et al. Underestimation of cardiac output by thermodilution in patients with tricuspid regurgitation. Am J Med 1989;86:417–20. 18. Renner LE, Morton MJ, Sakuma GY. Indicator amount, temperature, and intrinsic cardiac output affect thermodilution cardiac output accuracy and reproducibility. Crit Care Med 1993;21:586–97. 19. Singer M, Allen MJ, Webb AR, Bennett ED. Effects of alterations in left ventricular filling, contractility, and systemic vascular resistance on the ascending aortic blood velocity waveform in normal subjects. Crit Care Med 1991; 19:1138– 45. 20. Lefrant JY, Bruelle P, Aya AG, et al. Training is required to improve reliability of esophageal Doppler to measure cardiac output in critically ill patients. Intensive Care Med 1998; 24:347–52.
Background. Perioperative management of cardiac surgical patients frequently mandates measurements of cardiac output and left ventricular filling. This study compared cardiac output and left ventricular filling measured by pulmonary artery (PA) catheter and esophageal Doppler monitor (EDM). Methods. Thirty-four patients undergoing coronary artery bypass grafting were prepared by implanting a PA catheter, an EDM, and a transit-time ultrasonic flow probe around the ascending aorta. In 20 patients, left ventricular end-diastolic short-axis area (EDA) was measured by transesophageal echocardiography. At five time points, cardiac output was measured from the flow probe, the EDM, and the PA catheter (by thermodilution), and left ventricular filling was assessed from the PA catheter (as PA diastolic pressure), the EDM (corrected flow time), and the EDA. For cardiac output, concordance correlations relating EDM to flow probe and PA catheter to flow probe were calculated, transformed (Fisher’s z transfor-
mation), and compared by Student’s t test. For left ventricular filling, regression coefficients were created between corrected flow time and EDA and between PA diastolic pressure and EDA. Spearman correlations were compared by Wilcoxon rank sum test. Results. The EDM and the PA catheter exhibited similar relationships to the flow probe (concordance correlations, 0.55 ⴞ 0.35 [mean ⴞ standard deviation] and 0.49 ⴞ 0.34, respectively; p ⴝ 0.088). The correlation between corrected flow time and EDA was better than the correlation between PA diastolic pressure and EDA (concordance correlations, 0.49 ⴞ 0.55 versus 0.10 ⴞ 0.43, respectively; p < 0.01). Conclusions. These data suggest that the EDM may offer a less invasive technique for evaluating cardiac output and a more accurate estimate for preload compared with the PA catheter. (Ann Thorac Surg 2000;69:1782– 6) © 2000 by The Society of Thoracic Surgeons
A
management of some patients in the early postoperative period, but PA catheters carry risks of major complications [3] and add substantial cost. The value of PA catheters has not been well defined in cardiac surgical patients. However, use of these devices may be associated with an increased mortality risk in patients with acute myocardial infarction [4] and in critically ill patients in the intensive care unit [5]. In cardiac surgical patients, the accuracy of PA catheters has also been questioned, as measurement of CO by thermodilution can be subject to error induced by rapid temperature shifts after separation from cardiopulmonary bypass (CPB) [6, 7]. An alternative method for monitoring cardiac performance uses Doppler signals acquired by an esophageal Doppler monitor (EDM). With this technique, a pulsed, competitive-frequency continuous wave Doppler signal is emitted from a probe placed in the distal esophagus and directed at the descending thoracic aorta. The reflected ultrasound frequencies are recorded from a transducer on the probe and are processed to display a continuously monitored blood velocity signal. These velocity data can then be converted to flow by applying an equation that includes the estimated cross-sectional di-
s cardiac surgical procedures are more frequently performed in patients at high risk for postoperative complications, systems for monitoring cardiac performance assume greater importance. It was identified in 1969 that adult patients who failed to achieve a cardiac index of 2.0 L 䡠 min⫺1 䡠 m⫺2 in the early postoperative period were at greater risk for death and complications [1]. As a result, most surgical teams use monitoring techniques to quantify cardiac output (CO) and filling conditions in the perioperative period in some fraction of cardiac surgical patients. The most common tool in use is the pulmonary artery (PA) catheter [2]. This device provides a calculated CO determination from thermodilution, an indirect assessment of cardiac filling from PA diastolic or wedge pressures, and can supply a measurement of mixed venous oxygen saturation. These data can be important in the Accepted for publication Nov 24, 1999. Address reprint requests to Dr Jessen, Division of Thoracic and Cardiovascular Surgery, Department of Surgery, The University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd, Dallas, TX 75235-8879; e-mail: [email protected].
© 2000 by The Society of Thoracic Surgeons Published by Elsevier Science Inc
0003-4975/00/$20.00 PII S0003-4975(00)01129-2
Ann Thorac Surg 2000;69:1782– 6
ameter of the descending aorta, which is derived from nomograms [8]. The Doppler flow velocity waveform can also be analyzed to identify the flow time or duration of left ventricular ejection. This can be corrected for heart rate (cycle time) to yield an index of cardiac filling [9]. The device is simple to insert and, in initial studies, appears to yield CO measurements that correlate well with thermodilution values [10]. However, the technique is dependent on estimation of aortic diameter, its use can be affected by peripheral vascular disease, and it has seen limited application in patients undergoing cardiac operations [11]. The cardiac surgical patient provides a unique opportunity to compare these two monitoring devices. The surgical incision allows access to the ascending aorta for placement of a flow probe to directly measure left ventricular flow. Flow data can be compared with CO values simultaneously acquired from the thermodilution PA catheter and the EDM. In addition, transesophageal echocardiographic images can be obtained at the same setting to quantify left ventricular dimensions. These data can then be correlated with other indices of cardiac filling: PA diastolic pressure (PAD) measurements and Doppler-derived corrected flow time (FTc). This study was undertaken to compare CO measurements and cardiac filling variables immediately before and after CPB in patients undergoing coronary artery bypass grafting.
Material and Methods Thirty-six patients undergoing elective primary coronary artery bypass grafting at the Dallas Veterans Affairs Medical Center between July 1 and December 31, 1997, were studied under a protocol approved by the institutional review board. No patient had any clinical or echocardiographic evidence of valvular heart disease or aortic disease. In 2 of the 36 patients, an aortic flow probe of appropriate size was not available, and these patients were excluded from the study. The demographic characteristics of the remaining 34 patients are shown in Table 1.
Experimental Protocol All patients received premedication with midazolam hydrochloride and Reglan (metoclopramide hydrochloride) and were taken to the operating room in the fasted state. General anesthesia was induced with intravenous etomidate (0.2 to 0.4 mg/kg), rocuronium bromide (0.1 mg/kg), and fentanyl (5 to 10 g/kg). Anesthesia was maintained throughout the surgical procedure with fentanyl, midazolam, and isoflurane (0.2% to 2.0%). After induction, monitoring lines were inserted and consisted of a radial artery pressure catheter, Foley catheter, electrocardiographic electrodes, and an 8F introducer sheath placed in the right internal jugular vein. An Oximetrix PA catheter (Mountain View, CA) was advanced through the sheath into the PA until a wedge tracing was obtained. The balloon was left deflated for the remainder of the study. An EDM (Deltex Medical, Irving, TX) was inserted through the oropharynx into the distal esophagus ap-
DiCORTE ET AL ESOPHAGEAL DOPPLER MONITOR
1783
Table 1. Demographic Data for 34 Patientsa Variable Age (y) Sex (male:female) Height (cm) Weight (kg) BSA (m2) ABI No. of grafts Use of LITA (%) Preoperative LVEF Preoperative LVEDP (mm Hg) Aortic cross-clamp time (min) CPB time (min) Lowest CPB temperature (°C) a
Value 62 ⫾ 7 34:0 184 ⫾ 13 92 ⫾ 16 2.1 ⫾ 0.2 1.1 ⫾ 0.3 3.2 ⫾ 1.0 94 0.55 ⫾ 0.13 19 ⫾ 6 70 ⫾ 20 117 ⫾ 32 28 ⫾ 1
Where applicable, data are shown as the mean ⫾ the standard deviation.
ABI ⫽ ankle-brachial blood pressure index; BSA ⫽ body surface area; CPB ⫽ cardiopulmonary bypass; LITA ⫽ left internal thoracic artery; LVEDP ⫽ left ventricular end-diastolic pressure; LVEF ⫽ left ventricular ejection fraction.
proximately 35 to 40 cm from the incisors. The probe was manipulated until an optimal signal was obtained. In 20 of the patients, a 5-MHz esophageal echocardiographic probe (Sonos 2500; Hewlett-Packard, Andover, MA) was also inserted into the midesophagus, and the heart was imaged. A median sternotomy was performed, and saphenous vein and the left internal thoracic artery were harvested. The pericardium was incised and fashioned into a cradle. Each patient was administered heparin sodium in a dose of 300 U/kg intravenously, and an aortic cannula was placed in the distal ascending aorta and connected to the CPB circuit. A sterile aortic flow probe (Transonic Systems, Inc, Ithaca, NY) matched to the size of the ascending aorta (28 to 36 mm) was selected and placed around the ascending aorta proximal to the cannula. The probe was connected to a Transonic Systems flow computer and calibrated. The heart, aorta, and flow probe were immersed in warm, sterile saline solution. All pressure transducers were zeroed, and internal calibration of the mixed venous oxygen saturation monitor was completed. The first set of data was then acquired. Heart rate, systemic and PA blood pressures, central venous pressure, and mixed venous oxygen saturation were recorded from their respective monitors. The EDM signal was optimized and the tracing frozen, and values for CO (COEDM) and FTc were recorded. The flow value from the aortic flow probe (COFP) was also recorded. A short-axis echocardiographic image of the left ventricle was obtained at the midpapillary level and recorded. Three separate thermodilution measurements of CO were obtained over 3 minutes by injection of 10 mL of room temperature saline solution into the proximal port of the PA catheter. If an unstable baseline message appeared, the resulting thermodilution value was discarded, and the process was repeated until a satisfactory measurement was obtained. After the three thermodilution measurements were made, a repeat measurement of COFP
1784
DiCORTE ET AL ESOPHAGEAL DOPPLER MONITOR
was recorded, and repeat values of COEDM and FTc were obtained. Thermodilution injections and CO determinations were made by a senior anesthesiologist. The EDM data were acquired by a representative from the manufacturer (Deltex Medical). Echocardiographic images were acquired by a separate anesthesiologist or cardiologist familiar with the technique. All data were recorded by an individual investigator (MVC), and results of any one technique were blinded to investigators acquiring data by another method. Flow probe, EDM, and thermodilution measurements were averaged at each time point. After initial data acquisition, volume from the CPB circuit (primed with 1,500 mL of lactated Ringer’s solution, 200 mL of 20% mannitol, 100 mL of 25% albumin, and 10,000 units of heparin) was infused through the aortic cannula until an increase in COFP of at least 1 L/min was observed. A repeat set of all variables was then measured and recorded. After this, additional volume was infused to achieve a further increment in flow, and all variables were measured and recorded again. The aortic flow probe was removed, cannulation was completed, and the patient was placed on CPB and cooled to 28°C. Coronary artery bypass grafting was performed, and the patient was rewarmed during the construction of the proximal anastomoses. The lungs were reinflated, and the patient was separated from CPB. The aortic flow probe was repositioned, and signals from all devices were optimized. Another set of measurements was recorded by the identical protocols 5 minutes after separation from CPB. Volume was then infused as judged appropriate by the surgical team who were observing the heart directly and monitoring systemic blood pressure. Fifteen minutes after separation from bypass, a final set of measurements was recorded. Thus, data were measured at a total of five time points for each patient. No protamine sulfate was administered until after the final time point. The aortic flow probe was removed and the operation completed.
Data Analysis For each patient at each time point, averages of CO measured from the two aortic flow probe recordings (COFP), the two EDM recordings (COEDM), and the three thermodilution recordings (COTD) were used. Each of the 34 subjects had five observations of CO (one from each time point) by the three different techniques. The direct measurement of aortic flow from the transit-time ultrasonic flow probe, COFP, was considered the gold standard. For each patient, a concordance correlation was computed for the relationship between COEDM and COFP and between COTD and COFP. Each concordance correlation was transformed to statistical normality by the Fisher z transformation. The 34 pairs of transformed concordance correlations were compared by a paired t test, and a p value of less than 0.05 was considered significant. The comparison of indices of cardiac filling required a slightly modified analysis. The echocardiographic images were analyzed by tracing the left ventricular endocardium at end-diastole at the midpapillary level in the
Ann Thorac Surg 2000;69:1782– 6
short-axis view to obtain an end-diastolic area (EDA). This measurement was considered the gold standard for ventricular filling [12]. Variables from the EDM (FTc) and the PA catheter (PAD) use indirect methods (time and pressure, respectively) to imply volume conditions. Therefore, unlike CO comparisons, correlations between these variables with different units cannot be analyzed with the assumption that values should be equal to the gold standard. Instead, filling measurements should rise or fall in proportion to changes observed in the EDA. For each of the 20 patients with transesophageal echocardiographic data, Spearman correlations were computed for FTc versus EDA and for PAD versus EDA. These correlations examine how well the data fit the linear regression relating the two variables over the five time points as opposed to how closely they fit the equality line ( y ⫽ x). The Spearman correlations of these regressions were compared by Wilcoxon rank sum test, and p values of less than 0.05 were considered significant.
Results Cardiac output data were successfully acquired by all three techniques in the 34 patients in this study. In 20 patients, additional data were obtained from transesophageal echocardiography to assess filling conditions of the left ventricle. All patients were in hemodynamically stable condition throughout the prebypass period and remained in stable condition during the early postbypass interval. The mean volumes of CPB circuit prime administered between the first and second measurements and between the second and third measurements were 261 ⫾ 93 mL and 251 ⫾ 146 mL (mean ⫾ SD), respectively. These volume manipulations created a substantial alteration in both filling conditions and CO. After separation from bypass, intravenous nitroglycerin was routinely used to control blood pressure. No vasopressor agents were used in these patients except to modulate blood pressure during CPB. No inotropic support was used at any time during the study. The relationship between COEDM and COFP for all patients at all time points is shown in Figure 1. In general, the correlation between COs measured by these two techniques was good. The equation of the linear regression of this relationship was calculated as: COEDM ⫽ 1.027 ⫻ COFP ⫹ 0.571. The relationship between COTD and COFP was also linear over the range studied. This relationship is depicted in Figure 2 and is described by the equation: COTD ⫽ 0.874 ⫻ COFP ⫹ 1.625. To compare the relative accuracy of COEDM and COTD with the data acquired from the aortic flow probe (or COFP), concordance correlations were created for each patient as already described. Concordance correlations (mean ⫾ standard deviation) were as follows: COEDM versus COFP, 0.55 ⫾ 0.35, and COTD versus COFP, 0.49 ⫾ 0.34. Although COEDM exhibited a closer affinity to COFP than did COTD, this difference was not significant by paired t test ( p ⫽ 0.088). For comparison of the indices of left ventricular filling, regressions were also created between FTc from the EDM
Ann Thorac Surg 2000;69:1782– 6
Fig 1. Relationship between cardiac output as recorded by esophageal Doppler monitor (COEDM) and by aortic flow probe (COFP) for all patients at all time points (n ⫽ 160). The linear regression line and 95% confidence limits are plotted. This relationship is described by the equation: COEDM ⫽ 1.027 ⫻ COFP ⫹ 0.571 (r ⫽ 0.765).
and EDA from transesophageal echocardiographic information. Spearman correlation statistics from these regressions were compared with those calculated from the regression of PAD (from the PA catheter) and EDA. Regression coefficients (mean ⫾ standard deviation) for FTc versus EDA and PAD versus EDA were 0.49 ⫾ 0.55 and 0.10 ⫾ 0.43, respectively. The regression coefficients for FTc were significantly higher than those for PAD by Wilcoxon rank sum test ( p ⬍ 0.01).
Comment The immediate perioperative period is the most critical time of most cardiac surgical procedures and efforts to optimize hemodynamics during this interval can be critical to patient outcomes. Most adverse outcomes after coronary artery bypass grafting are related to either inadequate early cardiac performance or late multisystem organ failure [13]. These problems can be reduced if hemodynamic inadequacies can be identified and treated early in their course. In high-risk patients, monitoring of CO may allow the surgical team to identify situations where organ perfusion is in jeopardy and make interventions to avoid end-organ dysfunction. When CO or blood pressure is low, accurate measurement of filling conditions can be critical to the analysis of the hemodynamic problem. In general (assuming a stable rhythm), the first goal is to optimize cardiac filling. Increases in CO and ventricular performance derived from increases in preload are more efficient in terms of myocardial oxygen consumption than are increases produced by augmenting myocardial contractility with inotropic agents [14]. Thus, measurements of CO and ventricular filling can be very important in managing patients after cardiac surgical procedures. Measurement of CO can be accomplished by a variety of techniques, each with unique advantages and disadvantages. Direct measurement of left ventricular output by an aortic flow probe is done by comparisons of antegrade and retrograde ultrasound transit-time and may offer the most accurate determination [15, 16]. However, this method requires invasive instrumentation that is not easily adapted to the postoperative arena, does not account for the portion of CO that supplies the coronary
DiCORTE ET AL ESOPHAGEAL DOPPLER MONITOR
1785
circulation, and can be inaccurate if the probe size does not match the ascending aorta precisely. Thermodilution CO determinations mandate intravascular and intracardiac line placement and can be affected by valvular abnormalities (especially tricuspid insufficiency [17]) and inaccuracies in injectate volume or temperature [18]. Thermal variations in patients recently rewarmed after CPB can also cause error [6, 7]. Thermodilution CO can be obtained continuously by some catheters currently available, but the initial measurements are not displayed until several minutes after separation from CPB, and therefore, standard injection methods were used in this study. The EDM is less invasive and simpler to insert, although esophageal instrumentation can entail some risks. The technique depends on the ability of the user to accurately focus the signal on the descending aorta and makes assumptions on aortic diameter based on algorithms from patient size. In this study, EDM measurements of CO were at least as accurate as thermodilution determinations when compared with aortic flow measured directly. This suggests that the less invasive EDM device can play a useful role in monitoring the postoperative cardiac surgical patient. Thermodilution measurements were slightly, but not significantly, less accurate. The thermodilution CO measurements may have been adversely affected by the thermal changes induced by systemic rewarming after CPB. However, repeating the analysis with only prebypass observations did not materially alter this relationship. Using left ventricular dimensions measured by transesophageal echocardiography as the reference standard, this study also found that cardiac filling was better assessed by FTc than by PAD. Correlation coefficients between PAD and EDA were very poor. The correlation coefficients between FTc and EDA, although significantly better, were not especially strong by some criteria. In part, this finding may reflect the uncertainties caused by indirect methods of estimating chamber volume from time and pressure information. The superiority of FTc as an index of left ventricular filling suggests that the EDM offers advantages over the
Fig 2. Relationship between cardiac output as recorded by thermodilution technique (COTD) and by aortic flow probe (COFP) for all patients at all time points (n ⫽ 160). The linear regression line and 95% confidence limits are plotted. This relationship is described by the equation: COTD ⫽ 0.874 ⫻ COFP ⫹ 1.625 (r ⫽ 0.748).
1786
DiCORTE ET AL ESOPHAGEAL DOPPLER MONITOR
PA catheter for cardiac monitoring. For PAD to increase, the left ventricle must be loaded sufficiently to raise left ventricular end-diastolic pressure. In a ventricle with normal diastolic compliance, major volume changes may not produce substantial pressure changes, thus limiting the usefulness of PAD as a marker for filling. Also, ventricular compliance can change after cardioplegic arrest, hypothermia, or ischemia, thereby making comparisons with prebypass values hazardous. In contrast, the FTc value implies the amount of time required to eject the stroke volume, corrected for heart rate. As the ventricle is filled, it requires more time to eject the greater stroke volume, even in ventricles with normal compliance. This may explain the superiority of FTc as an index of preload. Studies in volunteers subjected to maneuvers that alter preload, afterload, and contractility have documented good correlation of FTc with preload changes and minimal effect of alterations in inotropic state. Afterload reduction influences FTc and may need to be considered when assessing the volume status of a patient with esophageal Doppler techniques [19]. There are limitations to this study. To our knowledge, it is the first to compare EDM and PA catheter techniques, with a transit-time ultrasonic flow probe and transesophageal echocardiograms providing reference values. However, this additional instrumentation mandated that all measurements be made with an open chest and open pericardium. Sternal closure and positivepressure ventilation may influence these results. The use of PA wedge pressure might prove to be a better reflection of left ventricular volume changes than PAD. We did not measure PA capillary wedge pressure because of concern about PA injury by cold catheters in the early period after bypass. The study size is small (34 patients), and this could limit the ability to detect differences in some variables, particularly CO. In addition, patients with severely impaired ventricles were not included in this study, in part because of our reluctance to distend such hearts during the prebypass loading phase. Therefore, it may not be possible to generalize our results to this subpopulation. Patients included did not have severe peripheral vascular disease, aortic atherosclerosis, or valvular heart defects. The influence of these factors is not defined in this study. Finally, all data were collected by experienced personnel. Whether these results, particularly the EDM findings, can be replicated by observers less familiar with the technology remains to be established. Other investigators [20] have found EDM information to be reproducible after a brief learning curve. Although many centers limit the use of any form of invasive monitoring to reduce costs or complications after cardiac operations, there is a group of patients in whom hemodynamic monitoring is important. On the basis of this study, the EDM may offer a less invasive technique for evaluating CO and a more accurate estimate of preload compared with the PA catheter. Further investigation of this technology appears warranted.
Ann Thorac Surg 2000;69:1782– 6
References 1. Dietzman RH, Ersek RA, Lillehei CW, Castañeda AR, Lillehei RC. Low output syndrome. Recognition and treatment. J Thorac Cardiovasc Surg 1969;57:138–50. 2. Civetta JM, Gabel JC. Flow-directed pulmonary artery catheterization in surgical patients. Ann Surg 1972;176:753– 6. 3. Boyd KD, Thomas SJ, Gold J, Boyd AD. A prospective study of complications of pulmonary artery catheterizations in 500 consecutive patients. Chest 1983;84:245–50. 4. Zion MM, Balkin J, Rosenmann D, et al. Use of pulmonary artery catheters in patients with acute myocardial infarction. Analysis of experience in 5,841 patients in the SPRINT Registry. SPRINT Study Group. Chest 1990;98:1331–5. 5. Connors AF, Speroff T, Dawson NV, et al. The effectiveness of right heart catheterization in the initial care of critically ill patients. JAMA 1996;276:889–97. 6. Bazaral MG, Petre J, Novoa R. Errors in thermodilution cardiac output measurements caused by rapid pulmonary artery temperature decreases after cardiopulmonary bypass. Anesthesiology 1992;77:31–7. 7. Latson TW, Whitten CW, O’Flaherty D. Ventilation, thermal noise, and errors in cardiac output measurements after cardiopulmonary bypass. Anesthesiology 1993;79:1233– 43. 8. Payen D. Oesophageal Doppler monitoring: history, physical principles, and clinical applications. Int Proc J 1994;1:3–9. 9. Wallmeyer K, Wann LS, Sagar KB, et al. The influence of preload and heart rate on Doppler echocardiographic indexes of left ventricular performance: comparison with invasive indexes in an experimental preparation. Circulation 1986;74:181– 6. 10. Valtier B, Mottlin D, Nolland B, Molis T, Payen D. Cardiac output measurement by four methods in intensive care units: interest of oesophageal Doppler device [Abstract]. Am Rev Respir Dis 1992;145:A781. 11. Fratacci M-D, Payen D, Beloucif S, Laborde F. Doppler cardiac output and left ventricular performance after cardiac surgery. Chest 1992;102:380– 6. 12. Clements FM, Harpole DH, Quill T, Jones RH, McCann RL. Estimation of left ventricular volume and ejection fraction by two-dimensional transoesophageal echocardiography: comparison of short axis imaging and simultaneous radionuclide angiography. Br J Anaesth 1990;64:331– 6. 13. Cosgrove DM, Loop FD, Lytle BW, et al. Primary myocardial revascularization. Trends in surgical mortality. J Thorac Cardiovasc Surg 1984;88:673– 84. 14. Lekven J, Brunsting LA, Jessen ME, Abd-Elfattah AS, Doherty NE 3d, Wechsler AS. Myocardial oxygen use during epinephrine administration to ischemically injured canine hearts. Circulation 1988;78(5 Pt 2):III125–36. 15. Lundell A, Bergqvist D, Mattsson E, Nilsson B. Volume blood flow measurements with a transit time flowmeter: an in vivo and in vitro variability and validation study. Clin Physiol 1993;13:547–57. 16. Dean DA, Cabreriza SE, Jia C-X, et al. Validation study of a new transit-time ultrasonic flow probe for continuous measurements on the great vessels [Abstract]. ASAIO J 1996;42:26. 17. Cigarroa RG, Lange RA, Williams RH, et al. Underestimation of cardiac output by thermodilution in patients with tricuspid regurgitation. Am J Med 1989;86:417–20. 18. Renner LE, Morton MJ, Sakuma GY. Indicator amount, temperature, and intrinsic cardiac output affect thermodilution cardiac output accuracy and reproducibility. Crit Care Med 1993;21:586–97. 19. Singer M, Allen MJ, Webb AR, Bennett ED. Effects of alterations in left ventricular filling, contractility, and systemic vascular resistance on the ascending aortic blood velocity waveform in normal subjects. Crit Care Med 1991; 19:1138– 45. 20. Lefrant JY, Bruelle P, Aya AG, et al. Training is required to improve reliability of esophageal Doppler to measure cardiac output in critically ill patients. Intensive Care Med 1998; 24:347–52.