Br J 1990;63:93-7 93 Non-invasive determination of by Doppler and electrical bioimpedance

David B Northridge, Iain N Findlay, John Wilson, Esther Henderson, Henry J Dargie

Abstract None of the previous studies of electrical Cardiac output measured by thermodilu- bioimpedance and only one study of Doppler tion in 25 patients within 24 hours ofacute included patients with acute myocardial infarc- myocardial infarction was compared tion.3 with cardiac output measured by Doppler A new bioimpedance monitor (BoMed echocardiography (24 patients) and elec- NCCOM3) has recently been developed that trical bioimpedance (25 patients). The simplifies the measurement of cardiac output. mean (range) cardiac outputs measured We wished to compare the accuracy ofthis new by Doppler (4 03 (22-6 0) I/min) and elec- monitor with in trical bioimpedance (3-79 (1-16-2) I/min) acute myocardial infarction. It is not possible to were similar to the mean thermodilution extrapolate from results obtained in normal value (3 95 (2-1-62) I/min). Both non- volunteers during exercise or patients in gen- invasive techniques agreed closely with eral intensive care because patients with thermodilution in most patients. None myocardial infarction may have much lower the less, three results with each method cardiac outputs. disagreed with thermodilution by more The aims of this study were to assess the than 1 l/min. feasibility, reproducibility, and ease of use of Both non-invasive techniques were both non-invasive techniques in patients with reproducible and accurate in most recent ( < 24 hours) myocardial infarction and patients with acute myocardial infarc- to compare cardiac output determined by tion. Doppler echocardiography was time Doppler echocardiography and electrical consuming and technically demanding. bioimpedance with that obtained by thermo- Electrical bioimpedance was simple to dilution in the same patients. use and had the additional advantage of allowing continuous monitoring of the Patients and methods cardiac output. PATIENTS We studied men and women aged < 75 years admitted to our coronary care unit with a Knowledge of the cardiac output is valuable in diagnosis of acute myocardial infarction. All the management ofpatients with acute myocar- patients were studied within 24 hours of the dial infarction because it identifies patients at onset of chest pain. Patients were included in- greatly increased risk and also provides a rapid the study if there was a clinical indication for means of monitoring the response to treat- Swan-Ganz catheterisation (early heart failure, ment.' Most patients, however, are managed hypotension, or poor urine output). without measurement of cardiac out-put-by Patients were excluded ifthey had any condi- relying on clinical signs. Unfortunately these tion known to interfere with the bioimpedance can be misleading in this group of patients. signal such as obesity or pleural effusion. An Arterial blood pressure may be maintained in initial screening echocardiographic study was the face of a low cardiac output by raised performed to exclude patients with significant systemic vascular resistance. A fall in urine disease, tricuspid incompetence, or output is an indication of reduced renal blood pericardial effusion. flow, but this may be due to activation of the sympathetic nervous system and does not SWAN-GANZ CATHETERISATION necessarily indicate a low cardiac output.2 The A 7F, triple lumen, Edwards Laboratories chest x ray is a sensitive indicator of raised left Swan-Ganz catheter was advanced from an atrial pressure,2 but measurement of cardiac antecubital to Department of the under Cardiology, Western output has until recently required right heart fluoroscopic control. Intracardiac and pul- Infirmary, Glasgow catheterisation. An accurate non-invasive monary artery pressures were recorded and D B Northridge method of measuring cardiac output would were then returned to I N Findlay patients coronary care. J Wilson therefore have many attractions: two of the Thermodilution cardiac output was measured E Henderson most promising techniques are Doppler echo- by an Edwards Laboratories cardiac output H J Dargie cardiography and electrical bioimpedance. computer with 10 ml iced dextrose. The mean Correspondence to Although a high degree of accuracy has been ofthree injections, all within 10% ofeach other Dr David B Northridge, Department of Cardiology, claimed for both methods,37 all ofthese studies and excluding the first injection, was recorded. Western Infirmary of were conducted in normal volunteers subjected It was not possible to perform all three tech- Glasgow, Dumbarton Road, Glasgow GI 6NT. to various artificial means of altering cardiac niques simultaneously because the bio- Accepted for publication output or in patients in intensive care with a impedance monitor interferes with the Doppler 8 November 1989 wide range of haemodynamic disturbances. signal. Therefore, Doppler recordings ofaortic 94 Northridge, Findlay, Wilson, Henderson, Dargie

Figure 1 midaxillary lines and the cervical sensing elec- Echocardiographic trodes were positioned laterally at the base of measurement of the diameter of the left the neck as close as possible to the clavicles. ventricular outflow tract. The "current injecting" electrodes delivered Ao, ; LA, left an alternating current at 70 kHz and were ; LV, left ; RV, right ventricle. placed with one pair 5 cm above the cervical sensing electrodes and the other pair 5 cm below the thoracic sensing electrodes. The BoMed NCCOM3 estimated from the impedance signal recorded from the inner pairs ofelectrodes using the Stramek et al equation.7 SV = VEPT x VET x (dz/dt)max/ZO Where VEPT = volume of electrically parti- cipating tissue (a constant derived from body height and weight); VET =. ventricular ejec- tion time; dz/dt = rate of change of impedance flow were recorded immediately before or during systole; and ZO = basal thoracic immediately after the other two techniques. impedance. The monitor automatically averages stroke DOPPLER ECHOCARDIOGRAPHY volume over twelve cardiac cycles and displays The diameter of the left ventricular outflow the cardiac output continuously. Electrical tract was used to calculate the cross sectional bioimpedance cardiac output was taken simul- area, assuming a circular profile. Measure- taneously with thermodilution measurements ments were made by cross sectional echo- as the mean of 10 displayed values (that is the cardiography in the parasternal long axis view mean of 120 cardiac cycles). To investigate the (Hewlett Packard model 77020A with a 2-5 accuracy ofbioimpedance measurements with- MHz phased array transducer). With the image out the influence of body size or , we frozen in midsystole, the diameter of the left also compared the stroke volume index ventricular outflow tract was measured im- measured by electrical bioimpedance with that mediately proximal to the aortic valve leaflets measured by thermodilution.8 as described by Lewis et al.5 We calculated the diameter as the mean of five measurements REPRODUCIBILITY using the trailing-edge to leading-edge method A pilot study was performed to measure the (fig 1). reproducibility of each non-invasive tech- We measured aortic blood velocity by con- nique. Ten patients were studied who were all tinuous wave Doppler ultrasound using a small haemodynamically stable in the recovery phase 1 9 MHz probe and the same Hewlett Packard after myocardial infarction. The bioimpedance system. Care was taken to obtain the clearest monitor was applied as described above, and spectral profile from both the suprasternal and cross sectional echocardiography and Doppler apical windows. When both windows were studies were recorded on videotape. After 30 successful the site giving the highest aortic flow minutes' rest (to allow time for the electrode velocities was used for analysis, as described by skin marks to disappear) the bioimpedance Bouchard et al.6 No attempt was made to monitor was reapplied and a second cardiac measure and correct for the angle between the output was recorded. Doppler echocardio- continuous wave Doppler ultrasound beam and graphy was performed a second time, by the the direction of blood flow. We took hard same observer, and recorded on a separate copies on a strip chart recorder and analysed videotape. The two tapes were analysed four them off line using a Kontron computer and weeks apart. magnetised digitising tablet. Stroke distance The coefficient of variation for repeat was taken as the mean of 10 measurements of measurements by the same observer using the area under the time velocity curve. Doppler electrical bioimpedance was 207%. The vari- cardiac output was calculated as: ability of echocardiographic measurement of CO = SD x CSA x HR aortic diameter was 1-5% and of the Doppler where SD was stroke distance and CSA was the stroke distance 3-0%. Thus the variability of cross sectional area of left ventricular outflow Doppler echocardiographic measurements was tract. greater than electrical bioimpedance, since the The stroke distance was compared with square of the aortic dimension is multiplied by stroke volume index to determine whether the stroke distance in the calculation of cardiac measurement of the cross sectional area was output. necessary (SVI = thermodilution CO/body surface area). STATISTICS Cardiac outputs measured by the different ELECTRICAL BIOIMPEDANCE techniques were compared by the method of After skin preparation the bioimpedance mon- Bland and Altman.9 The intra-observer itor was connected to the patients via four pairs variability of the non-invasive methods was of low contact impedance electrodes. The calculated as the coefficient ofvariation. Means lower thoracic voltage sensing electrodes were were compared by Student's paired t tests, and placed at the level ofthe xiphoid sternum in the a p value < 0-05 was regarded as significant. Non-invasive determination of cardiac output by Doppler echocardiography and electrical bioimpedance 95

Results c PATIENTS During the 10 month study, 29 patients had a 8 95s. limit of Swan-Ganz catheter inserted within 24 hours c 1.0- agreement of myocardial infarction. Four were excluded from the study because of aortic valve disease 0 E 0.5 0 (two patients), severe obesity (one patient), or E~~~05-~~~ inadequate visualisation of left ventricular out- 8°- * * s *Mean flow tract on screening echocardiographic * study (one patient). Therefore 25 patients (22 ~0. difference men), mean age 59 (range 47-72) years, were a-0.5* 0 Ei 0~~~~~ recruited and an attempt was made to measure E °5 -1 *0- X cardiac output by all three techniques in every D patient. This was successful in all but one o X95X. limit of patient, who was not suitable for Doppler -1.5 o agreement echocardiographic examination (see below). w The mean interval from the onset of pain to 0 1 2 3 4 5 6 7 inclusion in the study was 16 hours 24 minutes. Mean of measurements (i/min) Acute myocardial infarction was confirmed Figure 3 Difference between electrical bioimpedance and thermodilution measurements of cardiac output plotted subsequently by typical electrocardiographic against the mean value. changes and a rise in cardiac enzymes in all patients. DOPPLER AND ELECTRICAL BIOIMPEDANCE DOPPLER CARDIAC OUTPUT VERSUS MEASUREMENTS OF STROKE VOLUME INDEX THERMODILUTION The mean (range) stroke distance by Doppler Cardiac output was measured by Doppler was 14-3 (4-3-21-7) cm. When compared with echocardiography in 24 patients; in one patient thermodilution measurements of 3troke it was not possible to obtain a clear recording of volume index the correlation coefficient was aortic flow from either the apical or the supra- 0-82; SD = 0-43 x SVI + 3-38. Stroke volume sternal window. The mean (range) cardiac index was also calculated from the bio- output by Doppler was 4 03 (2 2-6 0) 1/min, impedance measurements. The 95% limits of which was not significantly different from the agreement with thermodilution were -8 1 to mean cardiac output by thermodilution, 3.95 + 6-7 ml/m2. (21-6.2) 1/min (p = 0 72). Analvsis of the results bv the method of Bland and Altman9 (fig 2) gave th4e 95% limits D.iscussion of agreement as - 123 to + 1 -32 1/min(-31% To our knowledge this is the first reported to + 33% of the mean cardiac output). In three study of the accuracy of electrical bio- patients the Doppler result diiffered from impedance in patients with acute myocardial thermodilution by more than 1 1/nnin. infarction, and only the second study of Doppler echocardiography.3 Previous reports ELECTRICAL BIOIMPEDANCE VERSUS compared Doppler determination of cardiac THERMODILUTION output with an invasive technique in other The mean (range) cardiac outputt by electical clinical settings. These studies, however, bioimpedance in 25 patients was 3*79 (1 1-6-2) varied greatly in the numbers and types of 1/min, which did not differ signifficantly from patients studied, the number of observations thermodilution at 3-95 (2-1-6-2) 1/min (p = per subject, the invasive technique used for and the method used. 0-22). The 950% limits of agreemerit (fig 3) were comparison, Doppler Good results were obtained with both - 1-43 to + 1-11 l/min (- 36% to + 28°/ ofthe pulsed mean cardiac output). In three patients the and continuous wave Doppler, with sampling electrical bioimpedance result dliffered from of flows in the aortic root, ascending aorta, thermodilution by more than 1 hlnnin. descending aorta (including transoesophageal approach) and across the mitral, tricuspid, or Figure 2 Difference 1.5- 95.h limitof pulmonary valves,""'2 but the results of dif- between Doppler and agreemt ferent studies were variable. Three important thermodilution ._I . measurements of cardiac E 1.0- studies compared the accuracy of these various output plotted against the * Doppler sites in the same patients, and found mean value. 0 u * the ascending aorta and left ventricular outflow 0.5- methods to be the most accurate.5 1112 Similar 4L,a . _w a Mean comparisons showed that of the many aortic 0- difference diameters that can be used in the calculations *b the aortic annulus or left ventricular outflow

-10- * . * tract gives the best results, and when the 0 ascending aortic root is used there is considera- 0. 0 ble overestimation ofcardiac output.5613 It was 0--1 -O- 0 BA 95%. limit of possible to obtain measurements of aortic flow agreement from either the suprasternal or apical window -1.5- in > 9000 ofpatients in a previous large study." 5 6 7 We also found that this was the case in 24 ofour 0 1 2 3 4 Mean of measurements (1,/min) 25 patients, and when both sites were success- 96 Northridge, Findlay, Wilson, Henderson, Dargie

fully studied the apical window gave the high- to compare our results with previous reports est peak velocity in every case. because they have all relied on correlation In the only previous study of Doppler coefficients to describe the accuracy of bioim- measurements in acute myocardial infarction pedance. None the less, our results seem to be the methods used were comparable with our broadly in keeping with previous experien- study, and the technique was feasible in all ce.202' We found that the cardiac output by patients. It was also thought to be accurate, thermodilution is likely to lie within a range with a reported correlation coefficient of 095.3 from 1 1 I/min below to 1-4 1/min above a These workers, in common with most previous bioimpedance estimate. The high degree of studies, investigated the agreement between reproducibility of electrical bioimpedance the Doppler and invasive results solely by using measurements in this study was also reported the correlation coefficient. This is not a good by others.'9 In the largest series described so method of determining the accuracy of a test or far, electrical bioimpedance could be measured of comparing the results of one study with in all of 58 critically ill patients after general another.9"' It can be considerably affected by a surgery.22 Only one previous study has com- few reasonably accurate results at the extremes pared both Doppler and bioimpedance of a range of measurements and it is dependent measurements with thermodilution, and this upon the range under study (when cardiac was conducted in an intensive care unit.20 Both output is studied the range depends upon the techniques were feasible in > 85%0 of patients types of patients included). This may explain and, in keeping with our findings, bio- why studies in which cardiac outputs have impedance was slightly more accurate than varied by up to 20 fold have produced co- Doppler though the difference was not statis- efficients as high as 0 98.'5 We used the method tically significant. of Bland and Altman to assess the accuracy of Cardiac output is the product of stroke the technique.91416 This statistical method is volume and heart rate and under normal condi- not affected by the range under study or values tions stroke volume is related to body surface at the extremes of the measurement, and it has area. Spurious associations can therefore be the added advantage ofexpressing the results in generated when cardiac output is calculated by a clinically meaningful manner. Using this a method, such as electrical bioimpedance, that method we showed that Doppler may over- incorporates both heart rate and body surface estimate the thermodilution cardiac output by area in its formula.8 Indeed older bioimpedance up to 1 3 1/min, or underestimate it by up to monitors were unable reliably to determine 1 2 1/min. Thus for 950% of measurements the stroke volume index,23 and estimated cardiac thermodilution result would lie between 1-3 1/ output mainly from the patient's size and heart min below and 1-2 1/min above a single Doppler rate. A recent study found that although car- estimate. The magnitude ofthese errors did not diac output determined by the BoMed depend upon the absolute cardiac output, so NCCOM3 correlated with indirect Fick they are potentially much more of a problem in measurements, this relation was no longer patients with low cardiac outputs. For instance, significant when stroke volume index was used a result of 3 1/min by either non-invasive instead of cardiac output.'9 We therefore com- technique may under or overestimate the pared stroke volume index measured by elec- thermodilution result by 37-47%, whereas for trical bioimpedance with thermodilution, and a result of 6 1/min the error lies between 1800 found that the 95% limits of agreement were and 23% - 8-1 to + 6-7 ml/m2 (- 32.4% to + 26 8% of Stroke distance measured by Doppler does the mean stroke volume index). Thus electrical not have the same units as stroke volume index bioimpedance measurement of stroke volume measured by thermodilution, and because dif- index was just as accurate as the measurement ferent variables were being compared we used of cardiac output. This study differs from the regression analysis rather than the method of previous one," in which healthy volunteers Bland and Altman. Regression analysis showed were studied. Large changes in cardiac output a close correlation (r = 0 82), which lends were induced by maximal exercise such that a support to the adoption of stroke distance as a correlation was found between the two non- measure of cardiovascular function, as pro- invasive techniques even though individual posed by Haites et al."7 It is also important in measurements varied widely. Because the in- practical terms because stroke distance can be crease in cardiac output that accompanies exer- measured by inexpensive "stand alone" Dop- cise is produced mainly by an increase in heart pler equipment without the need for imaging. rate,24 a relatively smaller range of stroke This is an interesting additional finding in a volumes were studied. It is, therefore, not study designed primarily to investigate the surprising that the correlation is less strong. A measurement of cardiac output not stroke dis- further problem with this study is that bio- tance. impedance was compared with indirect Fick Interest has been aroused in electrical bio- measurement of cardiac output rather than an impedance by the recent development of the invasive method. We used thermodilution as a new bioimpedance monitor (BoMed "gold standard," and although we accept that NCCOM3) used in this study.819 A review of this technique is far from perfect it is an the published reports showed nine previous established reference technique. studies in adults in which this monitor was We found that both non-invasive techniques compared with thermodilution. All were con- were feasible in patients with acute myocardial ducted in intensive care, usually on patients infarction. Doppler echocardiography after general surgery. Once again it is difficult required a lot of practice and the system we Non-invasive determination of cardiac output by Doppler echocardiography and electrical bioimpedance 97

the mitral orifice method. Br Heart J 1985;53:130-6. used took approximately 30 minutes to analyse 11 Labovitz AJ, Buckingham TA, Habermehl K, Nelson J, each measurement. Electrical bioimpedance Kennedy HL, Williams GA. The effects of sampling site on the two-dimensional echo-Doppler determination of was highly reproducible, simple to use, and cardiac output. Am Heart J 1985;109:327-32. gave instant results. It has the additional 12 Nicolosi GL, Pungercic E, Cervesato E, et al. Feasibility and variability of six methods for the echocardiographic and advantage of allowing continuous monitoring Doppler determination of cardiac output. Br Heart J of the cardiac output, which can be helpful in 1988;59:299-303. 13 Ihlen H, Amlie J, Dale J, et al. Determination of cardiac assessing the response to treatment. output by Doppler echocardiography. Br Heart J 1984;51: 54-60. 14 Diamond GA. Correlation, causation and agreement. Am J Cardiol 1989;63:392. 1 George RJD, Winter RJD. The clinical value of measuring 15 Mehta N, Iyawe VI, Cummin AR, Bayley S, Saunders KB, cardiac output. Br J Hosp Med 1985;34:89-95. Bennett ED. Validation of a Doppler technique for beat- 2 Forrester JS, Diamond G, McHugh TJ, Swan HJC. Filling to-beat measurement of cardiac output. Clin Sci 1985; pressures in the right and left sides of the heart in acute 69:377-82. myocardial infarction. N EngI J Med 1971;285:190-3. 16 Ihlen H, Endresen K, Golf S, Nitter-Hauge S. Cardiac 3 Mehta N, Bennett DE. Impaired left ventricular function in stroke volume during exercise measured by Doppler acute myocardial infarction assessed by Doppler echocardiography: comparison with the thermodilution measurement of ascending aortic blood velocity and technique and evaluation of reproducibility. Br Heart J maximal acceleration. Am J Cardiol 1986;57:1052-8. 1987;58:455-9. 4 Appel PJ, Kram HB, MacKabeeJ, Fleming AW, Shoemaker 17 Haites NE, McLennan FM, Mowat DHR, Rawles JM. WC. Comparison of measurements of cardiac output by Assessment of cardiac output by the Doppler ultrasound bioimpedance and thermodilution in severely ill surgical technique alone. Br Heart J 1985;53:123-9. patients. Crit Care Med 1986;14:933-5. 18 Gibson DG. Stroke distance-an improved measure of 5 Lewis JF, Kuo LC, Nelson JG, Limacher MC, Quinones cardiovascular function? Br Heart J 1985;53:121-2. MA. Pulsed Doppler echocardiographic determination of 19 Smith SA, Russell AE, West MJ, Chalmers J. Automated stroke volume and cardiac output: clinical validation of non-invasive measurement of cardiac output: comparison two methods using the apical window. Circulation 1984; ofelectrical bioimpedance and carbon dioxide rebreathing 70:425-31. techniques. Br Heart J 1988;59:292-8. 6 Bouchard A, Blumlein S, Schiller NB, et al. Measurement of 20 Wong DH, Tremper KD, Trujillo RJ, Zaccari J. Non- left ventricular stroke volume using continuous wave invasive cardiac output: simultaneous comparisons of two Doppler echocardiography of the ascending aorta and M- methods with thermodilution [Abstract]. J Clin Monit mode echocardiography of the aortic valve. J Am Coll 1988;4:164. Cardiol 1987;9:75-83. 21 Shellock G, Jakubowski A. Measurement of cardiac output 7 Stramek BB, Rose DM, Miyamoto A. Stroke volume in critically ill patients by a new non-invasive bio- equation with a linear base impedance and its accuracy, as impedance system: comparison with the thermodilution compared to thermodilution and electromagnetic flow methods [Abstract]. Physiologist 1987;30:215. meter techniques in animals and humans. Proceedings of 22 Shoemaker WC, Appel PL, Kram HB, Nathan RC, the sixth Intemational Conference on Electrical Bio- Thompson JL. Multicomponent noninvasive physiologic impedance, Zadar, Yugoslavia. 1983:p38. monitoring of circulatory function. Crit Care Med 8 Smith SA. Pitfalls and limitations in the use of impedance 1988;16:482-90. cardiography [Letter]. Br Heart J 1989;61:129. 23 Keim HJ, Wallace JM, Thurston H, Case DB, Drayer JIM, 9 Bland JM, Altman DG. Statistical methods for assessing Laragh JH. Impedance cardiography for determination of agreement between two methods ofclinical measurement. stroke volume index. J Appi Physiol 1976;41:797-800. Lancet 1986;i:307-10. 24 Chapman CB, Fisher JN, Sproule BJ. Behaviour of stroke 10 Zhang Y, Nitter-Hauge S, Ihlen H, Myhre E. Doppler volume at rest and during exercise in human beings. J Clin echocardiographic measurement of cardiac output using Invest 1960;39:1208-13.