Near continuous cardiac output by thermodilution

NEAR CONTINUOUS CARDIAC OUTPUT BY THERMODILUTION Jos R. C. Jansen, PhD,1 Royce W. Johnson, PhD,2 JohnY. Yan, MSc,2 and Piet D.Verdouw, PhD 3 Jansen JRC, Johnson RW, Yan JY, Verdouw PD. Near continuous cardiac output by thermodilution. J Clin Monit 1997; 13: 233^239 ABSTRACT. A new thermodilution method for frequent (near continuous) estimation of cardiac output, without manual injection of £uid into the blood, was tested. The method utilizes a pulmonary artery catheter equipped with a £uid ¢lled heat exchanger. The technique is based on cyclic cooling of the blood in the right atrium and measurement of the temperature changes in the pulmonary artery. Using this technique, a new estimate of cardiac output can be obtained every 32 s. Cardiac output estimates, obtained for a running mean of three measurements with this method, were compared to the mean of three conventional thermodilution measurements. The measurements were obtained during short periods of stable respiration and circulation. In six pigs, we made 46 paired measurements of conventional thermodilution (TD) and near continous (TDc) thermodilution. The cardiac output (COTD ) ranged from 2.4^ 13.7 l/min (mean 5.4 l/min). The best linear ¢t through the paired data points was COTDc = ÿ0.57 + 1.01 COTD . The mean di¡erence between the methods was ÿ0.50 l/min (S.D. = 0.39). The mean coe¤cient of variation of repeated measurements with the near continuous thermodilution was 3.6%. Considering changes of more than 0.25 l/min to be signi¢cant, all changes in cardiac output measured by conventional thermodilution were followed by the running mean of three near continuous thermodilution estimates. This study demonstrates the feasibility of the new method to monitor cardiac output, and to detect all changes greater than 0.25 l/min. KEY WORDS. Thermodilution, continuous cardiac output, mechanical ventilation. INTRODUCTION From the 1 Pathophysiological Laboratory, Department of Pulmonary Diseases, Erasmus University, Rotterdam, The Netherlands, 2 Ohmeda Medical Devices, 100 Mountain Ave, Murray Hill, NJ07974, U.S.A., 3 Experimental Cardiology, Department of Cardiology, Erasmus University, Rotterdam, The Netherlands. Received Jun 26, 1996. Accepted for publication Mar 4, 1997. Address correspondence to Dr J. R. C. Jansen, Dept. of Pulmonary Diseases, Erasmus University, room Ee2251, P.O. Box 1738, 3000 DR Rotterdam, The Netherlands. Journal of Clinical Monitoring 13: 233^239, 1997. ß 1997 Kluwer Academic Publishers. Printed in the Netherlands. A continuous estimate of cardiac output, in hemodynamically unstable patients, is highly desirable to detect acute changes since it would permit intervention before cardio-vascular function deteriorates severely. Many methods have been evaluated [3, 12, 15], but none of these methods have reached general acceptance. The conventional thermodilution has been generally accepted as su¤ciently accurate but it is highly intermittent in nature. This technique also requires repeated £uid injections, and this limits the method because of the risk of infection, volume overload, and user errors as well as labor-intensive usage. Recently two new thermodilution methods have become available, which do not require the injection of £uid into the patient [4, 15]. The ¢rst method is based on the delivery of electrically generated heat to the blood near the right atrium [15]. To prevent damage to the blood and surrounding 234 Journal of Clinical Monitoring Vol 13 No 4 July 1997 tissue the amount of delivered heat is limited. This low amount of heat adversely a¡ects the quality of the resulting temperature change in the pulmonary artery. To overcome this problem a signal processing technique of pseudo-random heating and cross correlation of the heating signal and temperature response is used [15]. However, this technique requires collection of data over approximately 5^15 min. Consequently, the displayed cardiac output value is an average value over the last 5^15 min. In the present study we used a catheter-based cold £uid heat exchanger, which can produce a larger thermal energy transient to the blood without damaging the blood or cardiac tissues compared to the method above. The method is similar to the standard bolus injection approach, with correction factors that take into account the loss of thermal indicator, during and after injection of cold via the cold £uid/blood heat exchanger. With this system frequent thermodilution measurements can be performed without volume loading the patient or animal. We evaluated whether this new thermodilution method can monitor cardiac output accurately by comparing the results to those obtained with our improved conventional thermodilution technique [7^9]. METHODS All experiments were performed in accordance with the NIH ``Guide for Care and Use of Laboratory Animals'' [DHEW Publication No. (NIH) 80^23, Revised 1978, O¤ce of Science and Health Reports, DRR/NIH, Bethesda, MD 20205] and under the regulations of the Animal Care Committee of the Erasmus University, Rotterdam, The Netherlands. After an overnight fast, six crossbred Landrace Yorkshire pigs (HVC, Hedel, The Netherlands) of either sex (40^60 kg) were sedated with intramuscular ketamine 20 mg/kg (AUV, Cuijk, The Netherlands), anesthetized with intravenous pentoarbital sodium 3^4 mg kgÿ1 hÿ1 (pentobarbitone sodium; Sano¢, Paris), intubated, and connected to a ventilator (Servo 900B, Siemens, Sweden) for intermittent positive pressure ventilation with a mixture of O2 and N2 (1 : 2, vol : vol). The ventilation (l/min) was adjusted to keep arterial blood gases within the normal range: 7.35 < pH < 7.45, 35 mmHg < PCO2 < 45 mmHg, and 100 mmHg < PO2 < 160 mmHg. The respiratory rate was 10^20 per min and tidal volume was adjusted. A catheter with double lumen was positioned in the aortic arch for: measurement of arterial pressure, determination of arterial blood gases, the administration of dobutamine, the infusion of dextran solution and bleeding of the animal. Using £uoroscopy another, special pulmonary artery catheter (PAC) was positioned in the pulmonary artery for measurement of pulmonary artery pressure, determination of cardiac output by the conventional thermodilution technique, and near continuous estimation of cardiac output by the new thermodilution technique. Instrumentation The system consisted of a cardiac output computer (Viggo-Spectramed SP1475), a £uid pump (ViggoSpectramed SP6275) and the special multi-lumen PAC (Viggo-Spectramed SP5708HP, 8.5 Fr, dead-space volume of catheter tubing and heat exchanger 2.5 ml) (Figure 1). The multi-lumen catheter was equipped with inlet and outlet lumens connected to a heat exchanger mounted on the catheter. The heat exchanger is an 8 cm long thin polymer sheath that tightly hugs the PAC between the 20 and 30 cm marks (i.e., positioned in the right atrium). Cold saline from a closed-loop system is pumped through the heat exchanger using a motor-driven syringe. This closed-loop system allows for continuous cycling. For each cycle 16 ml saline is pumped through the heat exchanger for a period of 12 s, followed by a stop period of 20 s during which the syringe is ¢lled again. Each measurement cycle lasts 32 s. The total cycle time of 32 s allows for complete curve washout under normal circumstances, and an acceptable coolant £ow rate through the catheter without excessively high coolant pressure (i.e., <800 mmHg). The inlet and outlet temperatures of the cooling sleeve (Figure 1) are measured with thermistors located proximally to the PAC. The blood temperature is measured in the pulmonary artery with a third thermistor mounted near the end of the catheter. With the assumption of constant blood £ow, com- Fig. 1. Schematic diagram of the modi¢ed pulmonary artery catheter. Jansen et al: Continuous Cardiac Output plete mixing with blood via the heat exchanger and no addition or loss of cold to heart and vessels, we can write the following expression for cardiac output (see Appendix): COTDc ˆ R 2 Tso ÿTsi †dt ÿA R p b Cpb 0
Tb dt s Cps Fs 1 1† in which the symbols are depicted as CO: cardiac output; s and b : speci¢c densities of saline and blood; Cps and Cpb : speci¢c heat of saline and blood; Fs : saline volume £ow through heat exchanger; Tsi and Tso: inlet and outlet saline temperature on the heat exchanger;
Tb: change in pulmonary artery blood temperature due to the injection of saline through the heat exchanger; 1 and 2: injection start and stop time; p : integration time per measurement cycle; t : time; A: loss of indicator estimated by model computations and £ow bench tests. Measurements In all six pigs, we monitored ECG, aortic pressure, pulmonary artery pressure, and central venous pressure. The Viggo-Spectramed system was also used to estimate cardiac output by thermodilution in the conventional bolus mode. To improve the accuracy of the conventional thermodilution estimates, we used ventilatory phase controlled injections [7, 8, 11]. A Siemens model 900B servo ventilator provided the time pulses to phasecontrol the injection of 5 ml saline at room temperature with a power injector. We did these timed thermodilution measurements at the phases of 33, 66 and 99% of the ventilatory cycle respectively. The interval between each of the measurements was approximately 30 s. The average of each of those series of three timed conventional measurements was compared to the average of the last three near continuous cardiac output estimates. In this way the in£uence of mechanical ventilation, causing cyclic changes in venous return, on the estimation of mean cardiac output was reduced [8]. Experimental protocol After completion of surgery and instrumentation, a stabilization period of 20 min as allowed; the near continuous thermodilution measurement system was then switched on. A number of di¡erent stable hemodynamic conditions were created: normovolemia I, hypervolemia plus dobutamine (7.7 mg kgÿ1 minÿ1 ), hypervolemia, normovolemia II, and hypovolemia. 235 Hypervolemia was created by infusing the animal with macrodex 40% (10 ml/kg), normovolemia II by bleeding of 10 ml/kg, and hypovolemia by bleeding again of 10 ml/kg. After stabilizing at each of these conditions, the near continuous cardiac output measurements were interrupted for approximately 5 min. During these periods a series of three timed conventional thermodilution measurements were carried out. During hypervolemia, normovolemia II, and hypovolemia the measurements were done at a ventilatory frequency of both 10 and 20 per min. Statistical analyses The conventional thermodilution and near continuous thermodilution measurements were evaluated by using linear regression and, according to Bland and Altman [1], by the computation of bias (mean), standard deviation (S.D.) and coe¤cient of variation (C.V.) (C.V. = 100% S.D./mean) for the di¡erence. The signi¢cance of di¡erences between measurements within the same pig was calculated according to a paired Student t-test. Di¡erences were considered signi¢cant when p was lower then 0.05. RESULTS A typical sequence of ¢ve cooling cycles by the near continuous thermodilution method is given in Figure 2. The blood temperature, Tb, periodically decreased in response to the cold saline £ushed through the heat exchanger and increased toward body temperature during the re¢lling phase of the injector. In this ¢gure a highly repeatable value for the estimated cardiac output can be observed. An example of the comparison of the two methods in one pig is presented in Figure 3. Cardiac output measurements using near continuous thermodilution approximate those found with conventional thermodilution. It is noticeable that the direction of change in cardiac output is the same for all comparisons. In six pigs, 46 series of three determinations were done with the near continuous thermodilution method (TDc) and with the conventional thermodilution method (TD). The mean coe¤cient of variation (C.V.) of all series was 3.6% with the TDc method and 6.6% with the TD method. The C.V. was slightly but signi¢cantly lower at a respiratory rate of 20 minÿ1 compared to that at 10 minÿ1 . There was no signi¢cant di¡erence between the mean cardiac output of the two series at the two di¡erent respiratory rates. The cardiac output 236 Journal of Clinical Monitoring Vol 13 No 4 July 1997 Fig. 2. Typical temperature pro¢les: saline inlet temperature (Tsi ) (dashed line) and outlet temperature (Tso ) (solid line) at the top panel, and blood temperature in the artery pulmonalis (Tb ) together with near continuous cardiac output (CCO, &) at the bottom panel. Five cycles are shown. In cycle two the period of coolant saline £ow (inj.) and no £ow (stop) are indicated. Fig. 4. Scatter diagram plotting the di¡erence between the estimates of cardiac output by conventional and near continuous thermodilution versus their mean. The level of the mean value for the di¡erence as well as 2 S.D. are indicated. Fig. 3. Tracing of the cardiac output with the near continuous thermodilution method (solid line) and the values obtained with conventional thermodilution ( & ). The dashed area shows the period of ¢lling the animals with ``cold'' Macrodex (10 ml/kg). measurements by TD showed a range of 2.4^13.7 l/min. Mean cardiac output for TD was 5.43 l/min. The regression equation of all 46 paired data is COTDc = ÿ0.57 + 1.01 COTD , with a correlation coe¤cient of 0.99. Analysis of the di¡erence between the two methods showed a mean bias of 0.50 l/min and a S.D. of 0.39 l/min or C.V. of 7.2% (Figure 4). Trending capability is demonstrated in Figure 5. In this ¢gure we plotted the di¡erence between sequential values for conventional and for near continuous thermodilution measurements against each other. The change in cardiac output estimated by TD and that by TDc correlated well (COTDc = ÿ0.01 + 0.94 COTD , Fig. 5. The feasibility of the near continuous thermodilution method to follow changes in cardiac output is given by plotting the di¡erence between two sequential estimates, for both conventional and near continuous thermodilution. The square around the origins indicates the chosen signi¢cance level of 0.25 l/min. correlation coe¤cient 0.98). If we consider a change of 0.25 l/min or less of no relevance (i.e., the blocked area in Figure 5) then in all cases the direction of sequential changes are the same for both methods. Jansen et al: Continuous Cardiac Output DISCUSSION In this study, the averaged bias and the agreement between the near continuous thermodilution technique and conventional thermodilution method has been studied over a large range of cardiac output values. High cardiac outputs were created by loading the animals with macrodex and the infusion of dobutamine. Low cardiac output conditions were obtained by bleeding the animals. We were confronted with a bias of 0.5 l/min. This bias may ¢nd its origin in using a wrong correction factor for the loss of indicator (A, Equation 1). This correction factor is based on the human anatomy instead of the anatomy of our small pig. Due to the relative small size of the right atrium of our animals compared to the length of the heat exchanger, part of the heat exchanger is positioned in the vena cava. Thus, cold is transported also to the blood in the vena cava. In contrast to the heart, the amount of unidirectional loss of cold indicator in the vena cava is signi¢cant and £ow dependent [2]. Using larger pigs resulted in a large reduction of the bias [4], because a larger part of the heat exchanger is positioned in the right atrium. Use of a neural network appears to reduce it further [4]. The low standard deviation of the di¡erence between the methods supports the application of the near continuous thermodilution technique for monitoring cardiac output. The feasibility to monitor cardiac output by the near continuous thermodilution method was examined by comparing sequential changes in cardiac output for both methods. It revealed that, if we de¢ne a change of 0.25 l/min to be irrelevant, the new near continuous themodilution method always followed the conventional thermodilution method. This level is considerable lower than the level derived from the analysis of the di¡erence between the two methods (2  S.D. = 0.78 l/min, or 14.2%). We used the averaged value of three measurements of both the near continuous and the conventional thermodilution technique to eliminate the in£uence of cyclic modulation of £ow caused by mechanical ventilation on the estimation of mean cardiac output. The duration of ``injection'' of 12 s for the near continuous thermodilution method in combination with a ventilatory cycle time of 6 s or 3 s, however, seems to eliminate the need to average three measurements (Figure 2). In this case the near continuous method delivers a new cardiac output value every 32 s, and an averaging technique is not needed. Therefore, a truly fast response of the system to changes in cardiac output appears possible. This is in contrast to the heater method [15], which has a response time of 12^15 min [5, 6]. 237 To estimate cardiac output clinically, a generally accepted method is the averaging of multiple random measurements [13, 14]. Repeated measurements at a chosen ¢xed moment in the ventilatory cycle decreases the random errors, but may introduce systematic errors [7]. The choice of one instant in the ventilatory cycle for cardiac output estimation will give di¡erent values for the di¡erent pigs, because of the inter-individual di¡erences in modulation pattern of the cardiac output estimates with ventilation. In a former animal study [7] we showed that under hemodynamically stable conditions, highly repeatable estimates of mean cardiac output (C.V. = 3.5%) were found by averaging three thermodilution measurements initiated at moments equally spread (not randomly!) over the ventilatory cycle. The results of this study [7] also showed values for mean cardiac output equal to those obtained by the Fick method for O2 . Therefore, we accepted a mean of three of those thermodilution measurements as an accurate reference of mean cardiac output, and used it to test the new near continuous thermodilution method. In this study, the C.V. of 7.2% of the di¡erence between the methods combines three types of errors: (1) errors in the reference mean cardiac output by the series of three thermodilution measurements; (2) errors of the near continuous thermodilution method; and (3) errors due to hemodynamic instability. In a previous study [7], the error in a repeated series of three measurements equally spread over the ventilatory cycle under stable circumstances was 3.5%. The error of repeated measurements with the near continuous cardiac output method was 3.6%. CONCLUSIONS With the new near continuous thermodilution method it was possible to estimate cardiac output every 32 s. The new method was capable of tracking changes in cardiac output with acceptable accuracy. Advances in the algorithm may be required to reduce the bias observed in small experimental animals. APPENDIX The energy calculation follows the ¢rst law of thermodynamics to a ¢xed control volume that includes the catheter with heat exchanger, blood volume and tissue surrounding the catheter (Figure A1). The result is equivalent to the Stewart^Hamilton equation, but for delivery of energy via a closed loop heat exchanger. The rate of energy loss to the heart or vessel walls, qb2b, is 238 Journal of Clinical Monitoring Vol 13 No 4 July 1997 We then obtain the cardiac output estimate after rearrangement: CO ˆ Fig. A1. Schematic diagram of the heat exchange catheter. For explanation of symbols see Appendix. equal to the rate of energy stored into saline via the heat exchanger, qs, plus the rate of loss of energy in the blood, qb, plus the rate of energy delivered by the heart, qh . qb2b ˆ qs ‡ qb ‡ qh Writing the balance with heat or temperature the equation becomes; qb2b ˆ Cps T 0so ÿ T 0si †m_ s ‡ Cpb Tbi ÿ Tbo †m_ b ‡ Z Cp cv @T dv @t 2† where m_ s and m_ b are the mass £ow rate of blood (b) and saline (s ), Cp is the heat capacity, T 0so and T 0si are the temperatures at the outlet and inlet of the heat exchanger, Tbi and Tbo are R the temperatures at the instream and outstream, and cv Cp @T @t dv, is the rate of energy from the heart into the control volume. If we assume a steady £ow, the stored energy inside the ¢xed control volume remains constant with time. Therefore, the volume integral in Equation 2 is zero. Integrating over one forced cooling cycle, we get Z p qb2b dt ˆ 0 Z 2 Cps T 0so ÿ T 0si †m_ s dt ‡ 1 Z p Cpb Tbi ÿTbo †m_ b dt 0 p qb2b dt ˆ Z 2 1 Cps T 0so ÿT 0si †s Fs dt ‡ R 2 1 Additional correction terms are needed to accommodate for transient behavior of the forced cooling cycle. One term, Qm2s, accounts for heating or cooling by the catheter body and £uids in the peripheral lumens during aspiration and idle periods of the cycle. Another term, Qsa, accounts for heating or cooling by saline in the inlet and outlet lumens in the aspiration and idle periods. CO ˆ s Fs Cps R 2 1 R T 0so ÿT 0si †dt ÿ 0 p qb2b dt ÿ Qm2s ÿ Qsa R b Cpb 0 p Tbi ÿTbo †dt The amount of energy entering the control volume R from the heart and vessels ^ qb2b may be signi¢cant ^ is di¤cult to calculate due to biological variations but it can be considered to be constant during a measurement cycle. Combining the three constants to A, the ¢nal equation to estimate cardiac output is CO ˆ R 2 T 0so ÿ T 0si †dt ÿA R ÿb Cpb 0 p
Tb dt s Fs Cps 1 3† The energy delivered into the control volume is simply the numerator in this equation. A prerequisite for the validity of the equation is the measurement of the temperature of saline in the inlet and outlet of the heat exchanger, whereas these temperatures are measured at the inlet and outlet of the catheter. To overcome this problem a model was used to compute the loss of indicator during its transport to, from, and through the catheter. For our study we used the formula given above with the corrections for this loss of energy. 0 where p is period of cooling cycle, and 1 and 2 are start and end time of saline £ow. Substituting m_ b = b (CO ) and m_ s = s (Fs ), the above equation becomes: Z R T 0so ÿT 0si †dt ÿ 0 p qb2b dt R b Cpb 0 p Tbi ÿTbo†dt s Fs Cps Z p Cpb Tbi ÿTbo †b CO†dt 0 where CO is cardiac output, Fs is volumetric £ow rate of saline and s, b , are the densities of saline s and blood b. REFERENCES 1. Bland JM, Altman DG. Statistical methods for assessing agreement between methods of clinical measurements. Lancet 1986; 1: 307^310 2. Bryant GH, Cucinell SA, Barcia PJ. Determination of heat gain in the inferior vena cava during thermodilution measurements. J Surgical Res 1985; 39: 224^229 3. Ehlers KC, Mylrea KC, Waterson CK, Calkins JM. Cardiac output measurements. A review of current techniques and research. Ann Biomed Eng 1986; 14: 219^239 4. Gu S, Orr JA, Yan JY, Forster FK, Johnson RW. Neural network and physical model can accurately estimate con- Jansen et al: Continuous Cardiac Output tinuous cardiac output. Anesthesiology 1993; V79: A468 5. Haller M, Zo«llner C, Briegel J, Forst H. Evaluation of a new continuous thermodilution cardiac output monitor in critically ill patients: A prospective criterion standard study. Crit Care Med 1995; 23: 860^866 6. Hennessy MM, Siegel LC, Pearl RG. Delayed time response of the continuous cardiac output pulmonary artery catheter. Anesthesiology 1995; V83: A460 7. Jansen JRC, Versprille A. Improvement of cardiac output estimation by the thermodilution method during mechanical ventilation. Int Care Med 1986; 12: 71^79 8. Jansen JRC, Schreuder JJ, Settels JJ, Kloek JJ, Versprille A. An adequate strategy for the thermodilution technique in patients during mechanical ventilation. Int Care Med 1990; 16: 422^425 9. Jansen JRC, Wesseling KH, Settels JJ, Schreuder JJ. Continuous cardiac output monitoring by pulse contour during cardiac surgery. Eur Heart J 1990; 11: 26^32 10. Normann RA, Johnson RW, Messinger JE, Sohrab B. A continuous cardiac output computer based on thermodilution principles. Ann Biomed Eng 1989; 17: 61^73 11. Okamoto K, Komatsu T, Kumar V, Sanchala V, Kabul K, Bhalodia R, Shibutani K. E¡ects of intermittent positive-pressure ventilation on cardiac output measurements by thermodilution. Crit Care Med 1986; 14: 977^980 12. Siegel CL, Pearl RG. Noninvasive cardiac output measurement: Troubled technologies and troubled studies. Anest Analg 1992; 74: 790^792 13. Stetz CW, Miller RG, Kelly GE. Reliability of the thermodilution method in the determination of cardiac output in clinical practice. Am Rev Respir Dis 1982; 125: 1001^1004 14. Stevens JH, Ra¤n TA, Mihm FG, Rosenthal MH, Stetz CW. Thermodilution cardiac output measurement. Effect of the respiratory cycle on its reproducibility. JAMA 1985; 253: 2240^2242 15. Yelderman ML. Continuous measurement of cardiac output with the use of stochastic system identi¢cation techniques. J Clin Monit 1990; 6: 322^332 239