Anesth Analg 2004;98:884-890
© 2004 International Anesthesia Research Society
doi: 10.1213/01.ANE.0000105923.09732.93
CARDIOVASCULAR ANESTHESIA
Neither the Accuracy nor the Precision of Thermal Dilution Cardiac Output Measurements Is Altered by Acute Tricuspid Regurgitation in Pigs
Charles W. Buffington, MD*, and
Elisabet U. M. Nystrom, MD, PhD
Section Editor
*Department of Anesthesiology, University of Pittsburgh, Pittsburgh, Pennsylvania, and the
Department of Anesthesiology, Creighton University, Omaha, Nebraska
Address correspondence and reprint requests to Charles W. Buffington, MD, MUH N-463, 200 Lothrop Street, Pittsburgh, PA 15213. Address email to buffingtoncw{at}anes.upmc.edu
 |
Abstract
|
|---|
Whether measurement of cardiac output using the thermal dilution technique (TDCO) is valid in the presence of tricuspid regurgitation (TR) is controversial. We assessed the accuracy and precision of the technique in pigs by comparison with data from an electromagnetic flowmeter on the aorta (EMCO). TR was created with sutures that immobilized the free-wall leaflets of the tricuspid valve, and cardiac output was adjusted with dobutamine to give values comparable to control measurements. TR reduced forward stroke volume from 17.2 to 12.6 mL/beat and caused the right atrium to dilate and pulse in synchrony with the right ventricle. Acute TR did not affect the linear regression relation between TDCO and EMCO and did not alter the correlation coefficient (r = 0.94 during both control and TR). These data demonstrate that acute TR does not affect the accuracy or precision of TDCO in pigs.
IMPLICATIONS: Cardiac output is a valuable measurement that guides the medical care of patients with heart and lung disease. This study demonstrates that the thermal dilution technique of determining cardiac output is valid when acute tricuspid valve regurgitation is present in pigs.
|
Introduction
|
|---|
The thermal dilution (TD) technique provides easy and, for the most part, accurate measurements of cardiac output (CO) in clinical settings. There is controversy, however, regarding whether the technique is reliable in the presence of tricuspid valve regurgitation (TR). Authors of early reviews put forth, without the support of experimental data, the idea that TR invalidates the technique (1–4). Subsequently, reports of underestimation (5–8), overestimation (9,10), and no change (11–14) appeared.
Because TR generally decreases cardiac efficiency and CO, one of the persistent problems in these studies has been a limited overlap of data obtained with and without TR (7,8,10,11,13). This limits the statistical power of the analysis and can lead to a false conclusion of "difference." Our strategy to minimize this problem was to use a dobutamine infusion to adjust cardiac output so that values measured during TR would be similar to those measured at baseline. A second problem is that investigators have ignored the exponential nature of the relation between "true" CO and that measured by TD. As a result of slow transit times and indicator loss, TD measurements at very low CO should overestimate actual CO, yielding a curve that is best fit by a polynomial equation. Boerboom et al. (7) demonstrated this phenomenon in a careful study but other authors have not addressed the issue. Finally, some authors have concluded that TD CO measurements are "unreliable" in the presence of TR (8,9,15), implying a lack of precision that would make clinical use of the technique risky. This is an issue worthy of investigation because a consistent small change in accuracy could be factored into clinical decision-making, but deterioration in precision would be of serious concern.
Thus the goal of this study was to determine whether acute TR affects the accuracy and precision of TD CO measurements in an animal model using dobutamine to match values obtained in the presence and absence of TR.
|
Methods
|
|---|
Five farm-bred pigs weighing 23–26 Kg were sedated with ketamine (16 mg/kg given IM) and anesthetized with halothane by mask. After skin infiltration with 2% lidocaine, a tracheotomy was performed and the trachea was intubated. The lungs were ventilated with oxygen in air delivered by a positive-pressure ventilator (Harvard, S. Natick, MA) with 5 cm H2O positive end-expiratory pressure to maintain an end-tidal carbon dioxide level of approximately 30 mm Hg. Anesthesia was maintained with halothane (0.55%–0.8% ET). Airway CO2 and halothane were measured (Datex Ultima, Helsinki, Finland). Body temperature was maintained at 39°C with a heating pad. The protocol was approved by the Animal Care and Use Committee of the University of Pittsburgh.
A catheter was inserted into the proximal aorta via the right carotid artery to measure systemic blood pressure. Heart rate was derived from the arterial pressure waveform with a cardiotachometer. Normal saline (5 mL · kg-1 · h-1 IV) was given via an ear vein. A balloon-tipped, multi-lumen catheter (#132–5F; Edwards, Irvine, CA) was inserted via the right external jugular vein and its distal port was maneuvered into the pulmonary artery by pressure waveform analysis. This catheter is designed for use in pediatric patients, and the injection port is located 15 cm proximal to the tip. Pulmonary artery and central venous and systemic arterial pressures were measured with saline-filled transducers (Gould, Cleveland, OH) and recorded on an oscillograph (Gould 3800). All transducers were zeroed to atmospheric pressure at the level of the right atrium with the animal supine.
A median sternotomy was made, the pericardium was opened, and the heart was suspended in a pericardial cradle. The aorta was dissected free from the pulmonary artery. A snug-fitting electromagnetic flow probe was placed on the ascending aorta after being zeroed in saline. The probe was connected to a flowmeter (Zepeda; Seattle, WA). Phasic aortic flow was recorded to check for baseline drift. The flow signal was also electronically averaged (time constant of 2 s) and recorded. At the end of the experiment, the aorta was excised and the flow probe calibrated by timed collection of the pig’s own blood. A length of dialysis tubing was put through the aorta at the site of the probe and connected to wide-bore tubing and a gravity-fed calibration setup capable of delivering flows from 0 to 5 L/min. The calibration curves were linear with correlation coefficients of 0.97–0.99.
TDCOs were obtained in triplicate at each measuring interval by use of an analog computer (Model 9520-A; Baxter-Edwards). Five mL of iced normal saline was injected into the right atrial port of the pulmonary artery catheter, and temperature deviation from baseline was recorded from a thermistor located in the pulmonary artery. Injection was done during apnea at end-expiration. Temperature versus time curves were recorded on the oscillograph to check for baseline drift and appropriate shape. Injections were continued until 3 similar curves and outputs were obtained (usually 3 or 4 injections were sufficient).
Prolene sutures (3.0, 8842, Ethicon, Somerville, NJ) were placed through the right ventricular wall in two locations to snare the two free-wall leaflets of the tricuspid valve. Traction on these sutures prevented normal closure of the tricuspid valve and dilated the tricuspid ring, producing regurgitation. The sutures were attached with rubber bands to an external frame in an attempt to produce maximal regurgitation without damage to the heart or interference with its normal motion during contraction. During traction, the right atrium dilated and pulsed in synchrony with the right ventricle. The right atrial pressure trace demonstrated "giant" V waves. Correct placement of the sutures was verified at autopsy.
Dobutamine HCl (Lilly, Indianapolis, IN) was diluted with normal saline to a concentration of 2 mg/mL and infused with a variable speed infusion pump (Harvard, S. Natick, MA) into the peripheral vein at doses of 2–15 µg · kg-1 · min-1 to increase CO at various times during the experiment.
The strategy of the experiment was to match COs during control conditions with those during TR by use of dobutamine. All measurements were made during transient hemodynamic steady states as indicated by a stable arterial blood pressure, heart rate, and aortic flow. Triplicate measurements of TDCO were made at each point. Measurements were made at baseline, then dobutamine was infused to increase CO 50%–100% and measurements were repeated. Another set of values was obtained at a higher dobutamine level and baseline measurements were repeated after the infusion was stopped. At this point, the sutures through the tricuspid valve were tightened to produce TR that was evident by inspection of the heart and oscillograph tracings. Measurements of TDCO and electromagnetic cardiac output (EMCO) at baseline and during dobutamine were then repeated. The rate of dobutamine infusion was adjusted to produce approximately the same levels of aortic flow obtained during control conditions.
Data were reduced from the oscillograph record manually and entered into a computer. Two measurements during TR were obtained during an unsteady hemodynamic state and were excluded from analysis. A standard statistical package (SPSSPC Version 1.1, SPSS, Chicago, IL) was used for analysis. Two-way analysis of variance (ANOVA) was used to determine the effect of dobutamine and TR on hemodynamics and CO values. For ANOVA only, baseline values before and after dobutamine and data at the two doses of dobutamine were grouped. Linear regression analysis of raw and transformed TDCO and EMCO was used to find the best-fit model and test the hypothesis that TR altered the relation. Data were also analyzed by the method of Bland and Altman (16) to determine the agreement between TDCO and EMCO measurements.
|
Results
|
|---|
Halothane anesthesia produced a stable cardiovascular state in these open-chest pigs although arterial blood pressure, heart rate and CO were less than expected in an awake animal (Table 1). Dobutamine had predictable effects, increasing systemic arterial pressure, heart rate, and CO but not right atrial pressure. TR increased right atrial pressure and heart rate but had little effect on arterial blood pressure and no effect on CO. However, calculated forward stroke volume decreased from 17.2 ± 4.7 to 12.6 ± 1.8 mL/beat (P < 0.001), indicating significant regurgitant flow during TR.
Autopsy results revealed that the traction sutures immobilized the central segment of each of the two free-wall leaflets of the tricuspid valve effectively.
A plot of the relation between CO determined by TD and the EM flowmeter is shown in Figure 1. Stepwise regression analysis using raw and a number of transformed values found that the square of EMCO was the best predictor of TDCO with a correlation coefficient of 0.948. Raw EMCO predicted TDCO with almost equal precision, demonstrated by a very similar correlation coefficient of 0.944. As transformations alter the weighting of individual data points and the correlations were so similar, raw data were used for the next step in the analysis.
View larger version (16K):
[in this window]
[in a new window]
|
Figure 1. Thermal dilution cardiac output (TDCO) versus electromagnetic cardiac output (EMCO) from a flow probe on the ascending aorta. Filled circles and solid line represent data during control conditions. Open circles and dashed line show data during acute tricuspid regurgitation (TR). Dobutamine was used to increase CO to comparable levels. TR had no statistical effect on the relation between TDCO and EMCO. The correlation coefficient (r = 0.94) during control conditions was identical to that found during TR.
|
|
Multivariable regression was used to determine if TR affects the relation between TDCO and EMCO. A dummy variable coded 0 for control and 1 for TR was entered along with the raw data into a stepwise regression model. The model selected EMCO as the primary predictor of TDCO but did not add the dummy variable to the equation (T = -1.45; P = 0.155 for inclusion), indicating that TR does not affect the relation between TDCO and EMCO. Thus, the linear regression equation for control conditions (TDCO = -0.40 + 1.32 EMCO, r = 0.943, P < 0.0001) was not different from that obtained in the presence of TR (TDCO = -0.31 + 1.22 EMCO, r = 0.945, P < 0.0001).
Analysis of agreement between the two techniques for measuring CO by the method of Bland and Altman yielded the data shown in Figures 2 and 3. The mean difference (bias) during control conditions was 0.21 ± 0.25 L/min (1 SD). This variance establishes the limits of agreement, defined as ± 2 SD. Similarly, the mean difference during TR was 0.08 ± 0.15 L/min (NS). These data indicate that TR degrades neither the accuracy nor the precision of TDCO measurements.
View larger version (15K):
[in this window]
[in a new window]
|
Figure 2. Bland-Altman plot of the difference in thermal dilution cardiac output (TDCO) and electromagnetic cardiac output (EMCO) versus the average values during control conditions. TDCO overestimated this average value by 0.21 ± 0.25 L/min (bias). The limits of agreement (±2 SD) are shown. The upward trend of data in this plot is the result of having a slope greater than 1.0 in the basic relationship between TDCO and EMCO.
|
|
View larger version (17K):
[in this window]
[in a new window]
|
Figure 3. Bland-Altman plot for measurements done during acute tricuspid regurgitation (TR). See Figure 2 for details. The bias (0.08 ± 0.15 L/min) and scatter are similar to those obtained during control measurements.
|
|
|
Discussion
|
|---|
The primary finding of this study is that acute TR does not affect the precision or accuracy of TDCO measurements.
We assumed that the EM flowmeter yielded accurate, reproducible, and precise measurements of aortic flow, providing a standard value for comparison with TD measurements. The Zepeda flow probe and flowmeter used have excellent zero and calibration stability and are not sensitive to changes in hematocrit. The flowmeter was applied acutely, which can lead to problems with signal stability. However, care was taken to ensure a snug fit around the aorta and the area was kept submerged in saline to provide the necessary electrical contact. Phasic aortic flow was monitored closely throughout the experiments for changes in diastolic aortic flow that would indicate baseline drift. We neglected retrograde (coronary) flow; thus "EMCO" measurements underestimate true CO by 3%–5%, contributing to the difference between the observed slopes (1.32 and 1.22) and 1.0 (the expected value).
We assumed that our suture immobilization of the tricuspid valve leaflets was effective in producing acute valvular regurgitation but we did not quantify the actual severity of TR by a technique such as 2-dimensional echocardiography or use of an additional flow probe on the inferior vena cava. Right atrial pressure (RAP) increased on average from 4.5 to 7.4 mm Hg (Table 1) when traction was applied to the sutures while the right atrium dilated and developed a pulse synchronous with the right ventricle. A similar technique was used by Kashtan et al. (11), who found TR increased RAP from 4.9 to 7.2 mm Hg. Otake and Lust (17) produced severe TR resulting from annular dilation by making multiple incisions in the fibrous annular ring and found that RAP increased only from 3 to 7 mm Hg. However, other studies have documented more dramatic increases in RAP; immobilization of 2 leaflets increased RAP from 6 to 14 mm Hg in Heerdt et al.’s study (10). The modest increase in RAP may signify modest TR in our study, but it is also possible that the right atrium of our animals was unstressed at the low filling pressures observed at baseline and may have accommodated a large regurgitant flow without a dramatic increase in pressure.
For any measurement technique to have clinical utility it must provide data that approximate the true value (accuracy) and do so in a reproducible fashion (precision). Two estimates of accuracy are found in the present data, the slope of the regression relation between TDCO and EMCO and the mean difference between the techniques (bias) from the Bland-Altman analysis. For reasons that are not entirely clear, the control slope was 1.32, indicating the TDCO values were approximately 30% more than EMCO. Part of this difference probably occurs because the aortic flow probe is located above the coronary ostia and thus does not include a measure of coronary blood flow, whereas the TD method does. This difference in slopes translates to a mean difference of 0.21 L/min between the techniques under control conditions. Such discrepancies appear in previous studies. Kashtan et al. (11), for example, found that TDCO was approximately 15% higher than pulmonary blood flow. In contrast, an informal review of 6 of the referenced studies shows an average slope of 0.85 for TDCO versus CO measured by a number of other techniques (6–8,11–13), demonstrating relatively lower TDCO values.
In the presence of acute TR in our study, the slope of the relation between TDCO and EMCO was 1.22 and the mean bias was 0.08 L/min. The similarity of these values to control indicates that TR does not change the accuracy of the measurement. In the 6 studies cited above, the slope increased in 3 and decreased in 3, averaging 0.90. In contrast, a well-done study by Cigarroa et al. (5) found a mean bias of -0.04 L/min during control and -0.77 L/min during TR, which led to the conclusion that TDCO underestimated actual CO by approximately 15%. Cigarroa et al.’s study used an older-model analog computer (model DTCCO-06/V2212; Electronics for Medicine, Sudbury, MA) to calculate area under the temperature versus time curve (AUC). The algorithm used by this device to calculate the AUC is not stated. It is possible that the Baxter-Edwards algorithm used in more modern computers provides more accurate data in the presence of TR and low CO. The Baxter-Edwards algorithm "cuts off" at 30% of peak height (and increases AUC by 22% as an estimate of the tail area). If the Electronics for Medicine algorithm cuts off at a lower value, the calculated AUC in the study of Cigarroa et al. may have been augmented by recirculation, leading to a falsely decreased CO.
The correlation coefficient of the regression equation and the standard deviation of the Bland-Altman mean difference in measurement values are 2 estimates of precision. Our correlation coefficient was 0.94 under control conditions, similar to that found in 6 previous studies (r = 0.91) (6–8,11–13), and was unchanged by TR (r = 0.94). In the 6 studies, TR produced no change or an increase in the correlation coefficient in 3 and a decrease in 3, all studies averaging 0.83. Taken together, these data support the conclusion that TR does not alter the precision of TDCO measurements. However, several studies do show significant scatter in the data obtained during TR (5,8) and several reports (9,15) document wildly erroneous values during TR. It is conceivable that technical problems such as having the tip of the catheter wedged during injection account for some of these extreme values. Keeping a careful eye on the temperature versus time curves produced will allow clinicians to discard spurious determinations.
Our study was done partially in response to work published by Heerdt et al. (10). They studied dogs and concluded that "acute TR produced underestimation of cardiac output by thermal dilution when flow is relatively high, produces overestimation when flow is relatively low, or has minimal effect when flow is in the midrange." A problem with that study is that CO was considerably lower in the presence of TR than during control; TR values ranged from 0.6 to 2.7 L/min compared with control values from 1.7 to 3.8 L/min. Thus, overlap occurred in approximately one-third of the range, making statistical comparison difficult. We achieved a much better overlap of control and TR values by titrating dobutamine, giving added assurance to our conclusion of "no difference." The authors also assumed a linear relation despite an apparent upward bend in their data at low COs. This overestimation at low CO can be predicted on theoretical grounds (loss of indicator) and led to a polynomial best fit in the study by Boerboom et al. (7) and an exponential best-fit in our data. Unfortunately, there is no way to know if use of a polynomial fit would change the conclusions of Heerdt et al. Our linear fit was virtually identical to the exponential fit, so our conclusion of "no difference" should be robust.
We studied acute TR, and it is possible that different results would be obtained in the presence of chronic TR, such as occurs in the presence of congestive heart failure or pulmonary hypertension. In these settings, right atrial and ventricular dilation could further delay washout of indicator. This theoretic concern may not be a real problem, however. Hoeper et al. (14) found no effect of TR or low CO on TDCO versus FickCO in patients with chronic pulmonary hypertension, and Hamilton et al. (12) demonstrated a lack of effect of TR on the relation in patients with severe congestive heart failure who had an average left ventricular ejection fraction of only 16%.
This study does not test the hypothesis that atrial or ventricular arrhythmias that can accompany chronic TR affect TDCO measurement accuracy or precision. A varying R-R interval could alter right ventricular stroke volume sufficiently to "confuse" the CO computer. The resulting irregular curve would alert clinicians to a spurious measurement. An irregular R-R interval would likely have a more profound effect on right ventricular ejection fraction measurements done with TD methodology than on TDCO. The present study used animals with normal cardiac filling pressures, and different results might have been obtained in the presence of acute hypervolemia, an issue that has not been studied.
Our methods presumably produced moderate TR, and different results may have been obtained with a more severe degree of TR. In this regard, Spinale et al. (6) found significant linear correlations between TDCO and actual flow with regurgitant fractions as high as 40%. The correlation decreased in a linear fashion from r = 0.94 at baseline to r = 0.87 with the most severe regurgitation. Finally, the present study consisted of 5 experimental animals, and different results may have been obtained if more animals were studied.
The TD technique has inherent errors. If these errors were smaller, it would probably have been possible in the current experiment to statistically separate an effect of TR from the background noise. The magnitude of the difference in TDCO caused by TR we observed would not likely have changed, however; thus the clinical conclusion would be the same.
This study in pigs found no effect of acute TR on the relation between CO measured by TD and by an EM flow probe on the aorta. Correlation of these variables was good (r = 0.94) and was not affected by TR. This means that TR does not degrade the accuracy or precision of the measurement. We can have confidence in this conclusion because measurements were made at comparable levels of CO in the presence and absence of TR. This study and the literature support a conclusion that acute TR does not affect the accuracy or precision of TDCO.
|
Acknowledgments
|
|---|
Dr. Michael Pinsky provided valuable advice and the Baxter-Edwards cardiac output computer.
|
References
|
|---|
- Fischer AP, Benis AM, Jurado RA, et al. Analysis of errors in measurement of cardiac output by simultaneous dye and thermal dilution in cardiothoracic surgical patients. Cardiovas Res 1978; 12: 190–9.[ISI][Medline]
- Levett JM, Replogle RL. Thermodilution cardiac output: a critical analysis and review of the literature. J Surg Res 1979; 27: 392–404.[ISI][Medline]
- Goldenheim PD, Kazemi H. Cardiopulmonary monitoring of critically ill patients. N Engl J Med 1984; 311: 776–80.[ISI][Medline]
- Kadota LT. Theory and application of thermodilution cardiac output measurements. Heart Lung 1985; 14: 605–14.[ISI][Medline]
- 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.[ISI][Medline]
- Spinale FG, Mukherjee R, Tanake R, Zile MR. The effects of valvular regurgitation on thermodilution ejection fraction measurements. Chest 1992; 101: 723–31.[Abstract/Free Full Text]
- Boerboom LE, Kinney TE, Olinger GN, Hoffmann RG. Validity of cardiac output measurement by the thermodilution method in the presence of acute tricuspid regurgitation. J Thorac Cardiovasc Surg 1993; 106: 636–42.[Abstract]
- Balik M, Pachl J, Hendl J. Effect of the degree of tricuspid regurgitation on cardiac output measurements by thermodilution. Intensive Care med 2002; 28: 1117–21.[ISI][Medline]
- Lipkin DP, Poole-Wilson PA. Measurement of cardiac output during exercise by the thermodilituion and direct Fick techniques in patients with chronic congestive heart failure. Am J Cardiol 1985; 56: 321–4.[ISI][Medline]
- Heerdt PM, Blessios GA, Beach ML, Hogue CW. Flow dependency of error in thermodilution measurement of cardiac output during acute tricuspid regurgitation. J Cardiothoracic Vasc Anesth 2001; 15: 183–7.[Medline]
- Kashtan HI, Maitland A, Salerno TA, et al. Effects of tricuspid regurgitation on thermodilution cardiac output: studies in an animal model. Can J Anaesth 1987; 34: 246–51.[Abstract/Free Full Text]
- Hamilton MA, Stevenson LW, Woo M, et al. Effect of tricuspid regurgitation on the reliability of the thermodilution cardiac output technique in congestive heart failure. Am J Cardiol 1989; 64: 945–8.[Medline]
- Konishi T, Nakamura Y, Morii I, et al. Comparison of thermodilution and Fick methods for measurement of cardiac output in tricuspid regurgitation. Am J Cardiol 1992; 70: 538–9.[Medline]
- Hoeper MM, Maier R, Tongers J, et al. Determination of cardiac output by the Fick method, thermodilution and acetylene rebreathing in pulmonary hypertension. Am J Respir Crit Care Med 1999; 160: 535–41.[Abstract/Free Full Text]
- Heerdt PM, Pond CG, Blessios GA, Rosenbloom M. Inaccuracy of cardiac output by thermodilution during acute tricuspid regurgitation. Ann Thorac Surg 1992; 53: 706–8.[Abstract]
- Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; i: 307–10.
- Otake M, Lust RM. Modification of De Vega’s tricuspid annuloplasty for experimental tricuspid regurgitation. J Card Surg 1994; 9: 399–404.[Medline]
Accepted for publication October 23, 2003.
Top
Abstract
Introduction
