- Email: [email protected]
Comparison of transesophageal echocardiographic, fick, and thermodilution cardiac output in critically ill patients
Journal of Critical Care SEPTEMBER 1996
VOL 11, NO 3
Comparison of Transesophageal Echocardiographic, Fick, and Thermodilution Cardiac Output in Critically Ill Patients Olivier
Axler,
Claude
Tousignant,
Christopher R. Thompson, Josette Dall’ava-Santucci, James A. Russell, and Keith R. Walley
P. Terry Phang,
Purpose: Recent observations have highlighted errors inermodilution technique of measuring cardiac output. Thus, cardiac output measurements using transesophageal echocardiography and the Fick method were compared with simultaneous thermodilution measurements. Methods: In 13 mechanically ventilated critically ill patients. cardiac output was determined simultaneously using (1) transesophageal echocardiography (CO&, (2) the Fick method (COuck), and (3) thermodilution (CO,) immediately before and after a rapid infusion of 500 mL of saline. Left ventricular enddiastolic and end-systolic areas were measured using the transesophageal echocardiographic transgastric short axis view, and CqEs was calculated from the corresponding volumes. Absolute cardiac output values and the changes from before to after saline infusion (ACO) were compared using analysis of variance, linear regression, and the Bland and Altman method. Results: There were no significant differences between COmr (8.0 f 3.4), COrrcx (8.4 f 3.3), and COrD
(8.3 ? 3.0) or between ACOTEE, ACO~~K, and AC&D using analysis of variance. However, correlations between Cqsr and COro (r* = 0.48; P < 90001). COrlcx and CqO (r2 = 0.48; P < .OOOOl), and Chs and COrlck (r* = 0.42; P < .OOOOl) were only moderately good. Using the method of Bland and Altman, the mean difference (+2 standard deviations) between CO~EE and COru was 0.3 + 4.3 L/min, between COrrcx and Cqc was -1.0 f 3.8 L/min, and between COrss and COrlcx was 0.8 f 5.8 L/min, whereas the difference between ACOTEE and ACOT~ was 0% f 25%. between ACO~~K and ACOTO was 9% + 45%, and between ACOne and ACOF,~~ was 8% f 39%. Conclusions: There are substantial differences in cardiac output as measured by these three methods, best demonstrated using the method of Bland and Altman. The variability of cardiac output and its derivatives (eg, oxygen delivery) should be borne in mind when making clinical decisions on individual patients. Copyright o 1996 by W.B. Saunders Company
T
to substantial errors, particularly in the intensive care unit, where pneumonia and other conditions result in significant oxygen consump-
HERMODILUTION CARDIAC output (COru) is the clinical standard used to measure cardiac output in critically ill patients. However, the thermodilution method of measuring cardiac output is a frequent and often underestimated source of errors.12J6J9,20These errors lead to inappropriate therapeutic interventions because of incorrect hemodynamic assessment of cardiac output values or incorrect calculation of oxygen consumption and oxygen delivery changes. 23 Thus, in the clinical setting there is a need for more accurate cardiac output measurement. The Fick method is the original reference method used to validate thermodilution.4,7J However, measurement of cardiac output by the Fick method (COrick) is also subject Journalof
CriticalCare,
Vol 11, No 3 (September),
1996: pp 109-l 16
From the Divisions of Critical Care Medicine and Cardiology, St Paul’s Hospital, University of British Columbia, Vancouver, EC, Canada; and Set-vice de Physiologic Respiratoire, Hopital Cochin, Universite Paris, Paris, France. 0. Axler was supported in part by a grant from the American Hospital in Parts and General Electric France. This study was supported by the Heart and Stroke Foundation of British Columbia and Yukon. Keith Walley is a Scholar qf the Heart and Stroke Foundation of Canada. Address reprint requests to Olivier Axler, MD, Senice de Reanimation Polyvalente, Clinique La Francilienne, 16!18 Av de I’Hotel de Ville, 77340 Pontault-Combault, France. Copyright 0 1996 by W. B. Saunders Company 0883-944119611103-0001$05.00i0 109
110
AXLER
tion by the lungs. l3 Transesophageal echocardiography is a safe and accurate technique to visualize the heart directly and assess ventricular area in critically ill patients, particularly during mechanical ventilation.21 Thus, it may represent a useful method of determining cardiac output. Considering the frequency and importance of measurements of cardiac output and the potential consequences of errors,23 we compared simultaneous transesophageal echocardiographic cardiac output (CO,,) and COFICK with COTD. To date, no comparison of these three methods of measuring cardiac output has been made in critically ill, mechanically ventilated patients, in whom errors in these techniques are most prominent and have the greatest impact on patient outcome. Different measures of cardiac output have previously been compared using analysis of variance to detect differenceslO and using regression analysis to show a correlation between methods.2,28 More recently, Bland and Altmanl have pointed out that analysis of variance and regression analysis may not show fully the differences between measurement techniques. Therefore, we compared COTEE, COFICK, and COTD using analysis of variance, regression analysis, and the method of Bland and Altman. Because changes in cardiac output rather than single measures of absolute cardiac output are often used in clinical decision making, we also compared changes in cardiac output (ACO) measured by these three techniques by repeating cardiac output measurement after rapid infusion of 500 mL of saline. METHODS The study was approved by the ethics committees of St Paul’s Hospital and the University of British Columbia. Signed informed consent was obtained from each patient or next of kin.
weaning inotropicivasoactive drug infusions. Patients were excluded if the arterial POTwas less than 55 mm Hg on an FrOz of 60% or greater to avoid the risk of worsening gas exchange. Patients were also excluded if there was increased risk of esophageal and/or gastric injury with the transesophageal probe (esophagitis, varices, ulcer, gastritis, recent upper gastrointestinal surgery). In one patient, COmcx could not be used for technical reasons; therefore, only 12 patients were compared using the Fick method. With these criteria, we selected 13 patients, 10 of whom were male, with an average age of 58 years. The principal diagnoses of the patients were septic shock (n = S), pneumonia (n = 3), post-cardiac surgery (n = l), and cocaine overdose (n = 1). These patients were mechanically ventilated using the assist control or pressure control mode. Seven patients were receiving continuous infusions of dopamine, dobutamine, and/or epinephrine. Ten patients were alive 3 weeks after the study and discharged from the intensive care unit. The main hemodynamic features of the patients are summarized in Table 1.
Protocol Patients received additional midazolam or diazepam and fentanyl or morphine as ordered by the attending physician for sedation and analgesia 30 minutes before transesophageal echocardiography. All other medications were continued as ordered by the attending physician. Enteral nutrition was stopped 2 to 3 hours before introduction of the transesophageal echocardiography probe, and the nasogastric tube was removed. A metabolic monitor (Deltatrac Metabolic Monitor; Datex Instrumentarium Corp, Helsinki, Finland) was connected to the expiratory port of the mechanical ventilator for minute-by-minute measurement of oxygen consumption. The recorded oxygen consumption was the average value of 10 measurements, obtained one each minute at the time of other cardiac output measurements. The oxygen consumption measurements were very stable during the procedure. COFICKwas then calculated as oxygen consumption divided by the difference between arterial and mixed venous oxygen content. Two operators collected data in this study. The first step of this protocol was to simultaneously collect the echocardiographic view, oxygen consumption data, blood gas measure-
Table
1. Baseline
Thirteen critically ill patients were studied prospectively in a mixed medical and surgical intensive care unit. Patients were included if they were mechanically ventilated and had a Swan-Ganz catheter already in place as part of their clinical treatment as well as a radial or femoral artery catheter. Seven patients required inotropic/vasoactive drugs to maintain circulatory stability. Thus, at the time of study all patients were hemodynamically stable. Patients were studied when the attending physician planned a volume infusion to detect potential hypovolemia and to aid in
Hemodynamics in 13 Critically
Mean arterial
Patients
ETAL
pressure
(mm
and Ventricular
Hg)
Heart rate (beats/min) Mean pulmonary artery pressure (mm Hg) Pulmonary artery occlusion pressure (mm Hg) COTD (L/min) COFICK (Llmin) COTEE (L/min) Left ventricular
end-diastolic
Let? ventricular Left ventricular Left ventricular
end-systolic end-diastolic end-systolic
NOTE.
All results
area (cm*) volume (mL) volume (mL) as mean
79.7 2 11.6 1022 25 27.7 13.1 8.0 8.4
2 f 'k
5.6 3.7 3.4 3.3
8.3 2 3.0 16.4 2 3.3 6.7 2 3.0
area (cm*)
are expressed
Mechanics
III Patients
123230 41 e 24 k SD.
COMPARISON
OF CARDIAC
OUTPUT
111
METHOD
ments, and hemodynamic measurements at baseline. These simultaneous measurements were repeated completely three times at 5-minute intervals at baseline to allow us to estimate the coefficient of variation of cardiac output measurement for each technique. Then 500 mL of normal saline was infused as rapidly as possible (5 to 10 minutes), and a single repeat set of identical measurements was made immediately after volume infusion. The hemodynamic measurements included systemic arterial pressure, heart rate, pulmonary arterial pressure, right atria1 pressure, pulmonary artery occlusion pressure, and cardiac output in triplicate using thermodilution with 10 mL of room-temperature normal saline per injection. Paired arterial and mixed venous blood samples were collected to measure arterial and mixed venous blood gases (ABL-30; Radiometer, Copenhagen, Denmark), hemoglobin levels, and oxygen saturations (Cooximeter IL 482; Instrumentation Laboratories, Lexington, MA).
Transesophageal Echocardiographic Measurements We obtained a two-dimensional short-axis view of the left ventricle at the midpapillary level using a Hewlett-Packard Sonos 1000 or 1500 sonograph with a ~-MHZ transesophageal echocardiographic probe (Hewlett Packard, Andover, MA). Care was taken to maintain a constant view with respect to internal cardiac landmarks. No clinically significant tricuspid regurgitation was observed throughout the respiratory cycle, and no change was noted with volume loading. Images were recorded on an SVHS videotape. The endocardium was outlined using the trackball of the echocardiograph at end-expiration. From this, we measured the left ventricular end-systolic area and the left ventricular enddiastolic area using the planimetry method by delineation of the innermost endocardial border, excluding the papillary muscles.6 The end-systolic and end-diastolic views of each cardiac cycle were defined as the smallest and largest short axis views, respectively. The end-diastolic and end-systolic views chosen were verified by comparison with the peak of the R wave or the T wave, respectively, on a synchronous electrocardiographic signal. The systolic and diastolic left ventricular volumes (LVESV and LVEDV) were calculated according to the previously validated formula: V = (36 + SAX,,,)/(12 + SAX,,,) x (SAX,,,)3~2,18~z8 where SAX,, is the short-axis area at the papillary muscle level. This method is a modified ellipsoid model using singleplane data at the level of the papillary muscle.1x~28Cardiac output (COrnr) was then calculated as the difference between LVEDV and LVESV multiplied by heart rate.
Accuracy and Reproducibility of Measurements The coefficient of variation of each cardiac output measurement technique was calculated as the standard deviation divided by the mean of three measurements repeated at 5-minute intervals at baseline. In addition, the reliability of the echocardiographic measurements was assessed by repeated measurements at a stable physiological state.*i The interobserver variability between the two echocardiographic readers was calculated as the ratio of the difference between the two readers’ values compared with the average of the
two readers’ values. The intraobserver variability was calculated as the coefficient of variation for each of the two echocardiographic readers. A sample of 30 to 40 measurements was used for these calculations.
Hemodynamic Measurements All patients had thermodilution Swan-Ganz catheters and radial or femoral arterial catheters in place before the study. A four-channel paper recorder was used to record echocardiographic data, systemic and pulmonary pressures, and Con,. A continuous analog respiratory line from the ventilator (Puritan Bennett 7200 Series; Puritan-Bennet Co, Los Angeles, CA) to the sonograph allowed assessment of phase of respiration when the areas were measured. From the paper trace we determined heart rate; systolic, diastolic, and mean arterial pressure; right atria1 pressure; systolic, diastolic, and mean arterial pulmonary pressure; and pulmonary artery occlusion pressure. We calculated stroke volume as cardiac output divided by heart rate.
Statistical Analysis Initial cardiac output values and AC0 determined by these three techniques were compared in three ways. First, initial cardiac output values and AC0 were compared using repeated-measures analysis of variance (ANOVA). We chose P < .05 as statistically significant. Second, linear regression analysis was performed and the Pearson correlation coefficient was calculated. Finally, the three techniques of measuring cardiac output and AC0 were compared using the method of Bland and Ahman’ by plotting means of the differences as a function of the average values of the different methods. RESULTS
There was no significant difference between the three methods in initial cardiac output or in AC0 after saline infusion (ANOVA). Table 2 shows the means and standard deviations for the three methods of measuring cardiac output, which appear to be very close. Furthermore, cardiac output measurements using the three techniques were significantly correlated (Fig 1). Correlations between COTEn and Corn (r = 0.68; r* = 0.46; P < .OOOOl), COFICK and = 0.68; r* = 0.46; P < .OOOOl), and COTD cr and COFrCK (r = 0.65; r* = 0.42; COTEE P < .OOOOl) were statistically significant, but substantial discrepancies can be seen (Fig 1). Table
Preinfusion Postinfusion NOTE.
Results
2. Cardiac
Output
Measurements
COTS (L/min)
COW (L/min)
COW (L/mid
8.0 ? 3.4 8.5 t 3.6
8.3 2 3.0 8.8 t 2.8
8.4 k 3.3 8.3 + 3.5
are expressed
There are no statistically using repeated-measures
significant ANOVA.
as mean
2 standard
differences
when
deviation. compared
112
AXLER
0
2
5 8 10 CO TD (l/m,)
12
iS
0
2
5 8 10 CO TD (I/mu)
ET AL
12
15
16 14 1+
PT
%
lo8-
I
642
.,‘,‘,.,‘,.,‘,‘~ 2 4 6
8 10 12 14 CO FICK (Umn)
16
18
The differences between the techniques are best illustrated using the Bland and Altman technique (Figs 2 and 3). The mean difference (+2 standard deviations) between COTEE and COru was 0.3 2 4.3 L/min, between COricx and COT,, was -1.0 ? 3.8 L/min, and between COTEE and COFICK was 0.6 ? 5.6 L/min (Fig 2). The difference between ACOTEE and ACOrn was 0% f 26%, between ACOFICK and ACOru was 9% + 46%, and between ACOmE and was 8% + 39% (Fig 3). These confiACOFICK dence intervals are wide, indicating poor agreement among the three techniques. The coefficients of variation of three cardiac output determinations repeated at Sminute intervals at baseline were 4.8% + 2.9% for COro, 6.7% 2 4.6% for COrick, and 8.6% ? 4.1% for COTEE. There was no statistically significant difference between these coefficient of variation values. For the transesophageal echocardiographic measurements of left ventricular end-diastolic area, we found an interob-
Fig 1. Cardiac output (CO) measured using three different techniques: thermodilution (TD), Fick, and transesophageal echocardiography (TEE) were significantly correlated (P < .OOOOl), but a number of points lie a substantial distance from the line of unity.
server variability of 9.6% ? 5.8% and intraobserver coefficients of variation of 3.4% & 1.3% and 2.9% ? 4.3% for each observer. DISCUSSION
The major finding of this study is that there are substantial differences in cardiac output measurement using transesophageal echocardiography, the Fick method, and thermodilution in a population of critically ill patients with respiratory and circulatory failure. Important variation in all techniques is likely because the comparison between COTEE and COrick is no better than the comparison of COro with either of the other two techniques. The Bland and Altman technique’ most clearly shows the differences among the three techniques of measuring cardiac output. Exactly how wide a confidence interval (Fig 2) should be considered “small” or “large” depends on what is clinically acceptable.’ We think that in some patients a difference in cardiac output of 0.5 to 1
COMPARISON
OF CARDIAC
OUTPUT
113
METHOD
0 5
n
-2sD
e 2SD
MCWl
2SD
-
TD+CO
FICK)/2
I
(Umn)
u
-8l (CO
8
.4
-“-f *b6 2
“-14 (CO
Mean
FICK)/2
4
(CO
I
6
TD+CO
8
TEE)/2
10
12
14
(Vmn)
---
II 1
TEE+CO
ZSD q
(I/mn)
L/mm may alter interpretation of the physiological state. In almost all patients, a difference in cardiac output of 2 to 4 L/min would substantially alter physiological interpretation and patient treatment. In this light, we interpret the confidence intervals in Fig 2 of more than 4 Limin as indicating poor agreement between the techniques. A number of studies have shown good correlations between COTD and COFICK. However, several issues should be considered when interpreting these studies. First, extrapolation of results from relatively healthy subjects during spontaneous ventilation to mechanically ventilated patients3,24 may not be accurate. An additional problem is that data from a number of patients have been pooled in some studies,
Fig 2. Comparison of cardiac output (CO) measurements using the Bland and Altman method shows poor agreement between the three techniques.
potentially leading to erroneous relationships when none really exist.’ A further problem illustrated by our results is that different statistical treatment of data can lead to different results. Indeed, the regression method used by most studies often finds a good correlation between different measures of cardiac output. The approach described by Bland and Altman’ illustrates that “agreement” between measurements of cardiac output often is not good. These three techniques of measuring cardiac output assume different things and measure different aspects of the circulation, as discussed by Pinsky.i9 These differences account for some of the discrepancy we noted between methods. The Fick method measures a time-averaged signal that varies little with ventilation, although
114
AXLER
60
4
30 25 ,:
+2SD
50
q
:
0
30-
Bqu
0
zo-
% ’ O-0 E K
+2SD
15-
1
Q
”
20 -
40
lo-
8
-lO-
4.
-2O-
ET AL
q
MeaIl
E
lo-
c
:-
$
-10 -5- -
8
-15 -
o
0
nu 0 0 cl
Meall
a
0 q
0
0
B
-2O-
-30 .2SD -40
; , . , . , -20 .15 -10 -5 (AC0
,
, 0
.
, 5
TD + AC0 FICK)d
,
, . , . 10 15 :
-2SD
-30
. -20
(76))
. -15
, .lO
.
* -5
,
.
0
(AC0 TD + AC0 TEE)/2
I ”
-40 71 -30
-25 ,
.
5
, 10
. 15
20
(+)
+2SD
-25
.20
.15
-10
-5
0
(AC0 TEE+ AC0 FICK)/2
5
10
15
20
(96)
pulmonary arterial flow may vary greatly.19 Pulmonary oxygen uptake alters this measurement13; however, this should not alter the ability to detect changes in cardiac output accurately if the rate of pulmonary oxygen consumption is constant. Thermodilution techniques measure pulmonary arterial blood flow, which is highly variable depending on the point within the ventilatory cycle at which the measurement is taken.9 Variability in this measurement can be decreased by making measurements at only one point in the respiratory cycle; however, an average value of cardiac output is best determined by averaging measurements taken throughout the respiratory cycle.9 The transesophageal method estimates cardiac output from left ventricular stroke volume and R-R interval. Respiratory changes in stroke volume, in transducer position, in identifying the endo-
Fig 3. Comparison of AC0 method shows poor agreement
using the Bland and Altman between the three techniques.
cardial border, and in the R-R interval will influence estimates of cardiac output. There are a number of problems with each technique of measuring cardiac output. As a result, there is no gold standard.25 There are now multiple studies validating metabolic carts,22,27suggesting that the Fick method may be a reasonably good measure of cardiac output. However, reliability of the measurements of oxygen consumption in critically ill patients may be decreased by metabolic instability, lack of synchronization with the ventilator, high F102, high airway pressures, circuit leaks, etc.16 Compact metabolic monitors are now available making the measurement of oxygen consumption in the intensive care unit easier. Nevertheless, these compact devices remain somewhat difficult to use, particularly during mechanical ventilation, and metabolic monitors must be
COMPARISON
OF CARDIAC
OUTPUT
115
METHOD
validated regularly. 17,22 The reliability of thermodilution is estimated to be approximately 15%,12J” and numerous sources of error are possible.7,‘2,16,20,29Th ese include some intracardiac conditions such as low cardiac output,29 tricuspid regurgitation,10,26 left-to-right shunt, and major dysrythmias. l1 Although these problems have been well identified, they are not always taken into account in clinical practice in the intensive care unit. Transesophageal echocardiography is reasonably easy to perform and reproducible, and it is particularly suitable for critically ill patients, especially those who are mechanically ventilated. Measurement of the left ventricular areas by the short-axis transgastric view, as well as the derived volumes, has been validated.6J8,28 Nevertheless, there are limitations associated with the echocardiographic method of measuring cardiac output. A single short-axis view of the heart at the level of the papillary muscles for the purpose of monitoring the filling volume, ejection fraction, and therefore cardiac output has been used previou~ly,~J~~*~but the limitations of single-plane measurements are well recognized when the heart is dilated or has wall motion abnormalities. In addition, some authors have found it difficult to obtain a true short-axis view from the esophagus of patients with a large or horizontal heart,r4 although this was not our experience. We were always able to find good anatomic views that enabled clear definition of the inner endocardial border. A main limitation of the echocardiographic method is extrapolating a volume from a planar measurement. Methods using more planes or even three-dimensional reconstructions of the left ventricle’” are much more difficult to perform, particularly in the intensive care unit, where fast and frequent transesophageal echocardiographic examinations may be required. Echocardiographic Doppler methods of calculating cardiac output are well validatedi but have substantial limitations when performed by transesophageal echocardiography.is We observed a coefficient of variation of Corn (4.8% ? 2.9%) similar to the value of 4.0% 2 2.9% reported by Ronco et al.2’ These investigators also report a coefficient of variation of oxygen consumption of 4.8% and a coefficient of variation of oxygen content of
0.3% to 1.1%,2’ which would lead to a total coefficient of variation of COFICK consistent with our measured value of 6.7% + 4.6%. We found that the coefficient of variation was greatest for COTEn and least for Corn, although the difference was not statistically significant when correction for multiple comparisons was made (P = .09). However, it is important to note that the most common cardiac output measurement method, Corn, is certainly no worse than COrick or COrEE. The choice of measurement method in individual patients should depend on the specific clinical question and the relative merits and disadvantages of each of the three techniques, summarized in Table 3. In conclusion, individual measures of cardiac output by any given technique are highly variable, and discrepancies between individual measurements by different techniques are large. Table
3. Advantages
and Disadvantages
of Measuring
of Three
Techniques
Output
Advantages
Method
COTD
Cardiac
Disadvantages
Easy
Invasive
Fast Can be repeated frequently
catheter Substantial
pulmonary
artery
operator
dependency Underestimation
with tri-
cuspid regurgitation, right to left shunt, and major dysrythmias Overestimation with low COFiCK Easy Minimal
operator
dependency
cardiac output Invasive pulmonary artery catheter and expensive oxygen monitor Approximately 30 minutes’ set-up time Stable O2 consumption necessary Oxygen monitor F102 > 0.6 Affected
COTEE
errors
by lung oxygen
Less invasive than a pulmonary artery catheter
consumption Expensive echo required
Significantly more data are derived so that the cause of low or high car-
Substantial operator dependency Requires experienced
disc output mined
is also deter-
for
machine
operator, although technique is easily learned Off-line measuring required to calculate cardiac output Not as easy to repeat frequently
116
AXLER
This is attributable to the combination of measurement and clinical errors introduced into each technique and the fact that each technique measures a somewhat different physiological
ET AL
entity. The variability in estimates of cardiac output and its derivatives (eg, oxygen delivery) should be borne in mind when making clinical decisions on individual patients.
REFERENCES 1. Bland JM, Altman DG: Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1:307-310, 1986 2. Chandraratna PA, Nanna M, McKay C, et al: Determination of cardiac output by transcutaneous continuouswave ultrasonic Doppler computer. Am J Cardiol 53:234237,1984 3. Chappell TR, Rubin LJ, Markham RV, et al: Independence of oxygen consumption and systemic oxygen transport in patients with either stable pulmonary hypertension or refractory left ventricular failure. Am Rev Respir Dis 128:30-33, 1983 4. Fegler G: Measurement of cardiac output in anaesthetized animals by a thermodilution method. Q J Exp Physiol 39:153-164, 1954 5. Feldman HA: Families of lines: Random effects in linear regression analysis. J Appl Physiol64:1721-1732,1988 6. Force TL, Folland ED, Aebischer N, et al: Echocardiographic assessment of ventricular function, in Marcus ML, Schelbert HR, et al (eds): Cardiac Imaging. Philadelphia, PA, Saunders, 1991, pp 376-401 7. Grossman W: Blood Flow Measurement: The Cardiac Output, in Grossman W (ed): Cardiac Catheterisation and Angiography. Philadelphia, PA, Lea & Febiger, 1986, pp 101-117 8. Hamilton WF: Measurement of cardiac output, in American Physiological Society: Handbook of Physiology. Baltimore, MD, Williams &Wilkins, 1962, pp 551-584 9. Jansen JRC, Schreuder JJ, Settels JJ, et al: An adequate strategy for the thermodilution technique in patients during mechanical ventilation. Intensive Care Med 16:422-425, 1990 10. 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 70:538-539,1992 11. Leatherman JW, Marini J: Clinical use of the pulmonary artery catheter, in Hall JB, Schmidt GA, Wood LDH (eds): Principles of Critical Care. New York, NY, McGrawHill, 1992, pp 323-342 12. Levett JM, Replogue RL: Thermodilution cardiac output: A critical analysis and review of the literature. J Surg Res 27:392-404,1979 13. Light RB: Intrapulmonary oxygen consumption in experimental pneumococcal pneumonia. J Appl Physiol 64:2490-2495,1988 14. Martin RW, Bashein G: Measurement of stroke volume with three-dimensional transesophageal ultrasonic scanning: Comparison with thermodilution measurement. Anesthesiology 70:470-476,1989 15. Muhiudeen IA, Kuecherer HF, Lee E, et al: Intraop-
erative estimation of cardiac output by transesophageal pulsed doppler echocardiography. Anesthesiology 749-14, 1991 16. Nishikawa T, Dohi S: Errors in the measurements of cardiac output by thermodilution. Can J Anaesth 40:142153,1993 17. Nunn JF, Makita K, Royston B: Validation of oxygen measurements during artificial ventilation. J Appl Physiol 67:2129-2134,1989 18. Parisi AF, Moynihan PF, Feldman CL, et al: Approaches to determination of left ventricular volume and ejection fraction by real-time two-dimensional echocardiography. Clin Cardiol2:257-263,1979 19. Pinsky MR: The meaning of cardiac output. Intensive Care Med 16:415-417,199O 20. Putterman C: Swan-Ganz catheter: A decade of hemodynamic monitoring. J Crit Care 2:127-146,1989 21. Reich DL, Konstadt SN, Nejat M, et al: Intraoperative transesophageal echocardiography for the detection of cardiac preload changes induced by transfusion and phlebotomy in pediatric patients. Anesthesiology 79:10-15, 1993 22. Ronco JJ, Phang T: Validation of an indirect calorimeter to measure oxygen consumption in critically ill patients. J Crit Care 1:36-41,199l 23. Ronco JJ, Phang PT, Walley KR, et al: Oxygen consumption is independent of changes in oxygen delivery in severe adult respiratory distress syndrome. Am Rev Respir Dis 143:1267-1273,199l 24. Rubin SA, Siemienczuk D, Nathan MD, et al: Accuracy of cardiac output, oxygen uptake, and arteriovenous oxygen difference at rest, during exercise, and after vasodilator therapy in patients with severe, chronic heart failure. Am J Cardiol50:973-978,1982 25. Schuster AH, Nanda NC: Doppler echocardiographic measurement of cardiac output: Comparison with a non-golden-standard. Am J Cardiol53:257-259,1984 26. Spinale FG, Mukherjee R, Tanaka R, et al: The effects of valvular regurgitation on thermodilution ejection fraction measurements. Chest 101:723-731,1992 27. Takala J, Keinlnen 0, Vaisanen P, et al: Measurement of gas exchange in intensive care: Laboratory and clinical validation of a new device. Crit Care Med 20:10411047,1989 28. Thys D, Hillel Z, Goldman ME, et al: A comparison of hemodynamic indices derived by invasive monitoring and two-dimensional echocardiography. Anesthesiology 67:630634,1987 29. Van Grondelle A, Ditchey RV, Groves BM, et al: Thermodilution method overestimates low cardiac output in human. Am J Physiol245:H690-H692,1983
VOL 11, NO 3
Comparison of Transesophageal Echocardiographic, Fick, and Thermodilution Cardiac Output in Critically Ill Patients Olivier
Axler,
Claude
Tousignant,
Christopher R. Thompson, Josette Dall’ava-Santucci, James A. Russell, and Keith R. Walley
P. Terry Phang,
Purpose: Recent observations have highlighted errors inermodilution technique of measuring cardiac output. Thus, cardiac output measurements using transesophageal echocardiography and the Fick method were compared with simultaneous thermodilution measurements. Methods: In 13 mechanically ventilated critically ill patients. cardiac output was determined simultaneously using (1) transesophageal echocardiography (CO&, (2) the Fick method (COuck), and (3) thermodilution (CO,) immediately before and after a rapid infusion of 500 mL of saline. Left ventricular enddiastolic and end-systolic areas were measured using the transesophageal echocardiographic transgastric short axis view, and CqEs was calculated from the corresponding volumes. Absolute cardiac output values and the changes from before to after saline infusion (ACO) were compared using analysis of variance, linear regression, and the Bland and Altman method. Results: There were no significant differences between COmr (8.0 f 3.4), COrrcx (8.4 f 3.3), and COrD
(8.3 ? 3.0) or between ACOTEE, ACO~~K, and AC&D using analysis of variance. However, correlations between Cqsr and COro (r* = 0.48; P < 90001). COrlcx and CqO (r2 = 0.48; P < .OOOOl), and Chs and COrlck (r* = 0.42; P < .OOOOl) were only moderately good. Using the method of Bland and Altman, the mean difference (+2 standard deviations) between CO~EE and COru was 0.3 + 4.3 L/min, between COrrcx and Cqc was -1.0 f 3.8 L/min, and between COrss and COrlcx was 0.8 f 5.8 L/min, whereas the difference between ACOTEE and ACOT~ was 0% f 25%. between ACO~~K and ACOTO was 9% + 45%, and between ACOne and ACOF,~~ was 8% f 39%. Conclusions: There are substantial differences in cardiac output as measured by these three methods, best demonstrated using the method of Bland and Altman. The variability of cardiac output and its derivatives (eg, oxygen delivery) should be borne in mind when making clinical decisions on individual patients. Copyright o 1996 by W.B. Saunders Company
T
to substantial errors, particularly in the intensive care unit, where pneumonia and other conditions result in significant oxygen consump-
HERMODILUTION CARDIAC output (COru) is the clinical standard used to measure cardiac output in critically ill patients. However, the thermodilution method of measuring cardiac output is a frequent and often underestimated source of errors.12J6J9,20These errors lead to inappropriate therapeutic interventions because of incorrect hemodynamic assessment of cardiac output values or incorrect calculation of oxygen consumption and oxygen delivery changes. 23 Thus, in the clinical setting there is a need for more accurate cardiac output measurement. The Fick method is the original reference method used to validate thermodilution.4,7J However, measurement of cardiac output by the Fick method (COrick) is also subject Journalof
CriticalCare,
Vol 11, No 3 (September),
1996: pp 109-l 16
From the Divisions of Critical Care Medicine and Cardiology, St Paul’s Hospital, University of British Columbia, Vancouver, EC, Canada; and Set-vice de Physiologic Respiratoire, Hopital Cochin, Universite Paris, Paris, France. 0. Axler was supported in part by a grant from the American Hospital in Parts and General Electric France. This study was supported by the Heart and Stroke Foundation of British Columbia and Yukon. Keith Walley is a Scholar qf the Heart and Stroke Foundation of Canada. Address reprint requests to Olivier Axler, MD, Senice de Reanimation Polyvalente, Clinique La Francilienne, 16!18 Av de I’Hotel de Ville, 77340 Pontault-Combault, France. Copyright 0 1996 by W. B. Saunders Company 0883-944119611103-0001$05.00i0 109
110
AXLER
tion by the lungs. l3 Transesophageal echocardiography is a safe and accurate technique to visualize the heart directly and assess ventricular area in critically ill patients, particularly during mechanical ventilation.21 Thus, it may represent a useful method of determining cardiac output. Considering the frequency and importance of measurements of cardiac output and the potential consequences of errors,23 we compared simultaneous transesophageal echocardiographic cardiac output (CO,,) and COFICK with COTD. To date, no comparison of these three methods of measuring cardiac output has been made in critically ill, mechanically ventilated patients, in whom errors in these techniques are most prominent and have the greatest impact on patient outcome. Different measures of cardiac output have previously been compared using analysis of variance to detect differenceslO and using regression analysis to show a correlation between methods.2,28 More recently, Bland and Altmanl have pointed out that analysis of variance and regression analysis may not show fully the differences between measurement techniques. Therefore, we compared COTEE, COFICK, and COTD using analysis of variance, regression analysis, and the method of Bland and Altman. Because changes in cardiac output rather than single measures of absolute cardiac output are often used in clinical decision making, we also compared changes in cardiac output (ACO) measured by these three techniques by repeating cardiac output measurement after rapid infusion of 500 mL of saline. METHODS The study was approved by the ethics committees of St Paul’s Hospital and the University of British Columbia. Signed informed consent was obtained from each patient or next of kin.
weaning inotropicivasoactive drug infusions. Patients were excluded if the arterial POTwas less than 55 mm Hg on an FrOz of 60% or greater to avoid the risk of worsening gas exchange. Patients were also excluded if there was increased risk of esophageal and/or gastric injury with the transesophageal probe (esophagitis, varices, ulcer, gastritis, recent upper gastrointestinal surgery). In one patient, COmcx could not be used for technical reasons; therefore, only 12 patients were compared using the Fick method. With these criteria, we selected 13 patients, 10 of whom were male, with an average age of 58 years. The principal diagnoses of the patients were septic shock (n = S), pneumonia (n = 3), post-cardiac surgery (n = l), and cocaine overdose (n = 1). These patients were mechanically ventilated using the assist control or pressure control mode. Seven patients were receiving continuous infusions of dopamine, dobutamine, and/or epinephrine. Ten patients were alive 3 weeks after the study and discharged from the intensive care unit. The main hemodynamic features of the patients are summarized in Table 1.
Protocol Patients received additional midazolam or diazepam and fentanyl or morphine as ordered by the attending physician for sedation and analgesia 30 minutes before transesophageal echocardiography. All other medications were continued as ordered by the attending physician. Enteral nutrition was stopped 2 to 3 hours before introduction of the transesophageal echocardiography probe, and the nasogastric tube was removed. A metabolic monitor (Deltatrac Metabolic Monitor; Datex Instrumentarium Corp, Helsinki, Finland) was connected to the expiratory port of the mechanical ventilator for minute-by-minute measurement of oxygen consumption. The recorded oxygen consumption was the average value of 10 measurements, obtained one each minute at the time of other cardiac output measurements. The oxygen consumption measurements were very stable during the procedure. COFICKwas then calculated as oxygen consumption divided by the difference between arterial and mixed venous oxygen content. Two operators collected data in this study. The first step of this protocol was to simultaneously collect the echocardiographic view, oxygen consumption data, blood gas measure-
Table
1. Baseline
Thirteen critically ill patients were studied prospectively in a mixed medical and surgical intensive care unit. Patients were included if they were mechanically ventilated and had a Swan-Ganz catheter already in place as part of their clinical treatment as well as a radial or femoral artery catheter. Seven patients required inotropic/vasoactive drugs to maintain circulatory stability. Thus, at the time of study all patients were hemodynamically stable. Patients were studied when the attending physician planned a volume infusion to detect potential hypovolemia and to aid in
Hemodynamics in 13 Critically
Mean arterial
Patients
ETAL
pressure
(mm
and Ventricular
Hg)
Heart rate (beats/min) Mean pulmonary artery pressure (mm Hg) Pulmonary artery occlusion pressure (mm Hg) COTD (L/min) COFICK (Llmin) COTEE (L/min) Left ventricular
end-diastolic
Let? ventricular Left ventricular Left ventricular
end-systolic end-diastolic end-systolic
NOTE.
All results
area (cm*) volume (mL) volume (mL) as mean
79.7 2 11.6 1022 25 27.7 13.1 8.0 8.4
2 f 'k
5.6 3.7 3.4 3.3
8.3 2 3.0 16.4 2 3.3 6.7 2 3.0
area (cm*)
are expressed
Mechanics
III Patients
123230 41 e 24 k SD.
COMPARISON
OF CARDIAC
OUTPUT
111
METHOD
ments, and hemodynamic measurements at baseline. These simultaneous measurements were repeated completely three times at 5-minute intervals at baseline to allow us to estimate the coefficient of variation of cardiac output measurement for each technique. Then 500 mL of normal saline was infused as rapidly as possible (5 to 10 minutes), and a single repeat set of identical measurements was made immediately after volume infusion. The hemodynamic measurements included systemic arterial pressure, heart rate, pulmonary arterial pressure, right atria1 pressure, pulmonary artery occlusion pressure, and cardiac output in triplicate using thermodilution with 10 mL of room-temperature normal saline per injection. Paired arterial and mixed venous blood samples were collected to measure arterial and mixed venous blood gases (ABL-30; Radiometer, Copenhagen, Denmark), hemoglobin levels, and oxygen saturations (Cooximeter IL 482; Instrumentation Laboratories, Lexington, MA).
Transesophageal Echocardiographic Measurements We obtained a two-dimensional short-axis view of the left ventricle at the midpapillary level using a Hewlett-Packard Sonos 1000 or 1500 sonograph with a ~-MHZ transesophageal echocardiographic probe (Hewlett Packard, Andover, MA). Care was taken to maintain a constant view with respect to internal cardiac landmarks. No clinically significant tricuspid regurgitation was observed throughout the respiratory cycle, and no change was noted with volume loading. Images were recorded on an SVHS videotape. The endocardium was outlined using the trackball of the echocardiograph at end-expiration. From this, we measured the left ventricular end-systolic area and the left ventricular enddiastolic area using the planimetry method by delineation of the innermost endocardial border, excluding the papillary muscles.6 The end-systolic and end-diastolic views of each cardiac cycle were defined as the smallest and largest short axis views, respectively. The end-diastolic and end-systolic views chosen were verified by comparison with the peak of the R wave or the T wave, respectively, on a synchronous electrocardiographic signal. The systolic and diastolic left ventricular volumes (LVESV and LVEDV) were calculated according to the previously validated formula: V = (36 + SAX,,,)/(12 + SAX,,,) x (SAX,,,)3~2,18~z8 where SAX,, is the short-axis area at the papillary muscle level. This method is a modified ellipsoid model using singleplane data at the level of the papillary muscle.1x~28Cardiac output (COrnr) was then calculated as the difference between LVEDV and LVESV multiplied by heart rate.
Accuracy and Reproducibility of Measurements The coefficient of variation of each cardiac output measurement technique was calculated as the standard deviation divided by the mean of three measurements repeated at 5-minute intervals at baseline. In addition, the reliability of the echocardiographic measurements was assessed by repeated measurements at a stable physiological state.*i The interobserver variability between the two echocardiographic readers was calculated as the ratio of the difference between the two readers’ values compared with the average of the
two readers’ values. The intraobserver variability was calculated as the coefficient of variation for each of the two echocardiographic readers. A sample of 30 to 40 measurements was used for these calculations.
Hemodynamic Measurements All patients had thermodilution Swan-Ganz catheters and radial or femoral arterial catheters in place before the study. A four-channel paper recorder was used to record echocardiographic data, systemic and pulmonary pressures, and Con,. A continuous analog respiratory line from the ventilator (Puritan Bennett 7200 Series; Puritan-Bennet Co, Los Angeles, CA) to the sonograph allowed assessment of phase of respiration when the areas were measured. From the paper trace we determined heart rate; systolic, diastolic, and mean arterial pressure; right atria1 pressure; systolic, diastolic, and mean arterial pulmonary pressure; and pulmonary artery occlusion pressure. We calculated stroke volume as cardiac output divided by heart rate.
Statistical Analysis Initial cardiac output values and AC0 determined by these three techniques were compared in three ways. First, initial cardiac output values and AC0 were compared using repeated-measures analysis of variance (ANOVA). We chose P < .05 as statistically significant. Second, linear regression analysis was performed and the Pearson correlation coefficient was calculated. Finally, the three techniques of measuring cardiac output and AC0 were compared using the method of Bland and Ahman’ by plotting means of the differences as a function of the average values of the different methods. RESULTS
There was no significant difference between the three methods in initial cardiac output or in AC0 after saline infusion (ANOVA). Table 2 shows the means and standard deviations for the three methods of measuring cardiac output, which appear to be very close. Furthermore, cardiac output measurements using the three techniques were significantly correlated (Fig 1). Correlations between COTEn and Corn (r = 0.68; r* = 0.46; P < .OOOOl), COFICK and = 0.68; r* = 0.46; P < .OOOOl), and COTD cr and COFrCK (r = 0.65; r* = 0.42; COTEE P < .OOOOl) were statistically significant, but substantial discrepancies can be seen (Fig 1). Table
Preinfusion Postinfusion NOTE.
Results
2. Cardiac
Output
Measurements
COTS (L/min)
COW (L/min)
COW (L/mid
8.0 ? 3.4 8.5 t 3.6
8.3 2 3.0 8.8 t 2.8
8.4 k 3.3 8.3 + 3.5
are expressed
There are no statistically using repeated-measures
significant ANOVA.
as mean
2 standard
differences
when
deviation. compared
112
AXLER
0
2
5 8 10 CO TD (l/m,)
12
iS
0
2
5 8 10 CO TD (I/mu)
ET AL
12
15
16 14 1+
PT
%
lo8-
I
642
.,‘,‘,.,‘,.,‘,‘~ 2 4 6
8 10 12 14 CO FICK (Umn)
16
18
The differences between the techniques are best illustrated using the Bland and Altman technique (Figs 2 and 3). The mean difference (+2 standard deviations) between COTEE and COru was 0.3 2 4.3 L/min, between COricx and COT,, was -1.0 ? 3.8 L/min, and between COTEE and COFICK was 0.6 ? 5.6 L/min (Fig 2). The difference between ACOTEE and ACOrn was 0% f 26%, between ACOFICK and ACOru was 9% + 46%, and between ACOmE and was 8% + 39% (Fig 3). These confiACOFICK dence intervals are wide, indicating poor agreement among the three techniques. The coefficients of variation of three cardiac output determinations repeated at Sminute intervals at baseline were 4.8% + 2.9% for COro, 6.7% 2 4.6% for COrick, and 8.6% ? 4.1% for COTEE. There was no statistically significant difference between these coefficient of variation values. For the transesophageal echocardiographic measurements of left ventricular end-diastolic area, we found an interob-
Fig 1. Cardiac output (CO) measured using three different techniques: thermodilution (TD), Fick, and transesophageal echocardiography (TEE) were significantly correlated (P < .OOOOl), but a number of points lie a substantial distance from the line of unity.
server variability of 9.6% ? 5.8% and intraobserver coefficients of variation of 3.4% & 1.3% and 2.9% ? 4.3% for each observer. DISCUSSION
The major finding of this study is that there are substantial differences in cardiac output measurement using transesophageal echocardiography, the Fick method, and thermodilution in a population of critically ill patients with respiratory and circulatory failure. Important variation in all techniques is likely because the comparison between COTEE and COrick is no better than the comparison of COro with either of the other two techniques. The Bland and Altman technique’ most clearly shows the differences among the three techniques of measuring cardiac output. Exactly how wide a confidence interval (Fig 2) should be considered “small” or “large” depends on what is clinically acceptable.’ We think that in some patients a difference in cardiac output of 0.5 to 1
COMPARISON
OF CARDIAC
OUTPUT
113
METHOD
0 5
n
-2sD
e 2SD
MCWl
2SD
-
TD+CO
FICK)/2
I
(Umn)
u
-8l (CO
8
.4
-“-f *b6 2
“-14 (CO
Mean
FICK)/2
4
(CO
I
6
TD+CO
8
TEE)/2
10
12
14
(Vmn)
---
II 1
TEE+CO
ZSD q
(I/mn)
L/mm may alter interpretation of the physiological state. In almost all patients, a difference in cardiac output of 2 to 4 L/min would substantially alter physiological interpretation and patient treatment. In this light, we interpret the confidence intervals in Fig 2 of more than 4 Limin as indicating poor agreement between the techniques. A number of studies have shown good correlations between COTD and COFICK. However, several issues should be considered when interpreting these studies. First, extrapolation of results from relatively healthy subjects during spontaneous ventilation to mechanically ventilated patients3,24 may not be accurate. An additional problem is that data from a number of patients have been pooled in some studies,
Fig 2. Comparison of cardiac output (CO) measurements using the Bland and Altman method shows poor agreement between the three techniques.
potentially leading to erroneous relationships when none really exist.’ A further problem illustrated by our results is that different statistical treatment of data can lead to different results. Indeed, the regression method used by most studies often finds a good correlation between different measures of cardiac output. The approach described by Bland and Altman’ illustrates that “agreement” between measurements of cardiac output often is not good. These three techniques of measuring cardiac output assume different things and measure different aspects of the circulation, as discussed by Pinsky.i9 These differences account for some of the discrepancy we noted between methods. The Fick method measures a time-averaged signal that varies little with ventilation, although
114
AXLER
60
4
30 25 ,:
+2SD
50
q
:
0
30-
Bqu
0
zo-
% ’ O-0 E K
+2SD
15-
1
Q
”
20 -
40
lo-
8
-lO-
4.
-2O-
ET AL
q
MeaIl
E
lo-
c
:-
$
-10 -5- -
8
-15 -
o
0
nu 0 0 cl
Meall
a
0 q
0
0
B
-2O-
-30 .2SD -40
; , . , . , -20 .15 -10 -5 (AC0
,
, 0
.
, 5
TD + AC0 FICK)d
,
, . , . 10 15 :
-2SD
-30
. -20
(76))
. -15
, .lO
.
* -5
,
.
0
(AC0 TD + AC0 TEE)/2
I ”
-40 71 -30
-25 ,
.
5
, 10
. 15
20
(+)
+2SD
-25
.20
.15
-10
-5
0
(AC0 TEE+ AC0 FICK)/2
5
10
15
20
(96)
pulmonary arterial flow may vary greatly.19 Pulmonary oxygen uptake alters this measurement13; however, this should not alter the ability to detect changes in cardiac output accurately if the rate of pulmonary oxygen consumption is constant. Thermodilution techniques measure pulmonary arterial blood flow, which is highly variable depending on the point within the ventilatory cycle at which the measurement is taken.9 Variability in this measurement can be decreased by making measurements at only one point in the respiratory cycle; however, an average value of cardiac output is best determined by averaging measurements taken throughout the respiratory cycle.9 The transesophageal method estimates cardiac output from left ventricular stroke volume and R-R interval. Respiratory changes in stroke volume, in transducer position, in identifying the endo-
Fig 3. Comparison of AC0 method shows poor agreement
using the Bland and Altman between the three techniques.
cardial border, and in the R-R interval will influence estimates of cardiac output. There are a number of problems with each technique of measuring cardiac output. As a result, there is no gold standard.25 There are now multiple studies validating metabolic carts,22,27suggesting that the Fick method may be a reasonably good measure of cardiac output. However, reliability of the measurements of oxygen consumption in critically ill patients may be decreased by metabolic instability, lack of synchronization with the ventilator, high F102, high airway pressures, circuit leaks, etc.16 Compact metabolic monitors are now available making the measurement of oxygen consumption in the intensive care unit easier. Nevertheless, these compact devices remain somewhat difficult to use, particularly during mechanical ventilation, and metabolic monitors must be
COMPARISON
OF CARDIAC
OUTPUT
115
METHOD
validated regularly. 17,22 The reliability of thermodilution is estimated to be approximately 15%,12J” and numerous sources of error are possible.7,‘2,16,20,29Th ese include some intracardiac conditions such as low cardiac output,29 tricuspid regurgitation,10,26 left-to-right shunt, and major dysrythmias. l1 Although these problems have been well identified, they are not always taken into account in clinical practice in the intensive care unit. Transesophageal echocardiography is reasonably easy to perform and reproducible, and it is particularly suitable for critically ill patients, especially those who are mechanically ventilated. Measurement of the left ventricular areas by the short-axis transgastric view, as well as the derived volumes, has been validated.6J8,28 Nevertheless, there are limitations associated with the echocardiographic method of measuring cardiac output. A single short-axis view of the heart at the level of the papillary muscles for the purpose of monitoring the filling volume, ejection fraction, and therefore cardiac output has been used previou~ly,~J~~*~but the limitations of single-plane measurements are well recognized when the heart is dilated or has wall motion abnormalities. In addition, some authors have found it difficult to obtain a true short-axis view from the esophagus of patients with a large or horizontal heart,r4 although this was not our experience. We were always able to find good anatomic views that enabled clear definition of the inner endocardial border. A main limitation of the echocardiographic method is extrapolating a volume from a planar measurement. Methods using more planes or even three-dimensional reconstructions of the left ventricle’” are much more difficult to perform, particularly in the intensive care unit, where fast and frequent transesophageal echocardiographic examinations may be required. Echocardiographic Doppler methods of calculating cardiac output are well validatedi but have substantial limitations when performed by transesophageal echocardiography.is We observed a coefficient of variation of Corn (4.8% ? 2.9%) similar to the value of 4.0% 2 2.9% reported by Ronco et al.2’ These investigators also report a coefficient of variation of oxygen consumption of 4.8% and a coefficient of variation of oxygen content of
0.3% to 1.1%,2’ which would lead to a total coefficient of variation of COFICK consistent with our measured value of 6.7% + 4.6%. We found that the coefficient of variation was greatest for COTEn and least for Corn, although the difference was not statistically significant when correction for multiple comparisons was made (P = .09). However, it is important to note that the most common cardiac output measurement method, Corn, is certainly no worse than COrick or COrEE. The choice of measurement method in individual patients should depend on the specific clinical question and the relative merits and disadvantages of each of the three techniques, summarized in Table 3. In conclusion, individual measures of cardiac output by any given technique are highly variable, and discrepancies between individual measurements by different techniques are large. Table
3. Advantages
and Disadvantages
of Measuring
of Three
Techniques
Output
Advantages
Method
COTD
Cardiac
Disadvantages
Easy
Invasive
Fast Can be repeated frequently
catheter Substantial
pulmonary
artery
operator
dependency Underestimation
with tri-
cuspid regurgitation, right to left shunt, and major dysrythmias Overestimation with low COFiCK Easy Minimal
operator
dependency
cardiac output Invasive pulmonary artery catheter and expensive oxygen monitor Approximately 30 minutes’ set-up time Stable O2 consumption necessary Oxygen monitor F102 > 0.6 Affected
COTEE
errors
by lung oxygen
Less invasive than a pulmonary artery catheter
consumption Expensive echo required
Significantly more data are derived so that the cause of low or high car-
Substantial operator dependency Requires experienced
disc output mined
is also deter-
for
machine
operator, although technique is easily learned Off-line measuring required to calculate cardiac output Not as easy to repeat frequently
116
AXLER
This is attributable to the combination of measurement and clinical errors introduced into each technique and the fact that each technique measures a somewhat different physiological
ET AL
entity. The variability in estimates of cardiac output and its derivatives (eg, oxygen delivery) should be borne in mind when making clinical decisions on individual patients.
REFERENCES 1. Bland JM, Altman DG: Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1:307-310, 1986 2. Chandraratna PA, Nanna M, McKay C, et al: Determination of cardiac output by transcutaneous continuouswave ultrasonic Doppler computer. Am J Cardiol 53:234237,1984 3. Chappell TR, Rubin LJ, Markham RV, et al: Independence of oxygen consumption and systemic oxygen transport in patients with either stable pulmonary hypertension or refractory left ventricular failure. Am Rev Respir Dis 128:30-33, 1983 4. Fegler G: Measurement of cardiac output in anaesthetized animals by a thermodilution method. Q J Exp Physiol 39:153-164, 1954 5. Feldman HA: Families of lines: Random effects in linear regression analysis. J Appl Physiol64:1721-1732,1988 6. Force TL, Folland ED, Aebischer N, et al: Echocardiographic assessment of ventricular function, in Marcus ML, Schelbert HR, et al (eds): Cardiac Imaging. Philadelphia, PA, Saunders, 1991, pp 376-401 7. Grossman W: Blood Flow Measurement: The Cardiac Output, in Grossman W (ed): Cardiac Catheterisation and Angiography. Philadelphia, PA, Lea & Febiger, 1986, pp 101-117 8. Hamilton WF: Measurement of cardiac output, in American Physiological Society: Handbook of Physiology. Baltimore, MD, Williams &Wilkins, 1962, pp 551-584 9. Jansen JRC, Schreuder JJ, Settels JJ, et al: An adequate strategy for the thermodilution technique in patients during mechanical ventilation. Intensive Care Med 16:422-425, 1990 10. 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 70:538-539,1992 11. Leatherman JW, Marini J: Clinical use of the pulmonary artery catheter, in Hall JB, Schmidt GA, Wood LDH (eds): Principles of Critical Care. New York, NY, McGrawHill, 1992, pp 323-342 12. Levett JM, Replogue RL: Thermodilution cardiac output: A critical analysis and review of the literature. J Surg Res 27:392-404,1979 13. Light RB: Intrapulmonary oxygen consumption in experimental pneumococcal pneumonia. J Appl Physiol 64:2490-2495,1988 14. Martin RW, Bashein G: Measurement of stroke volume with three-dimensional transesophageal ultrasonic scanning: Comparison with thermodilution measurement. Anesthesiology 70:470-476,1989 15. Muhiudeen IA, Kuecherer HF, Lee E, et al: Intraop-
erative estimation of cardiac output by transesophageal pulsed doppler echocardiography. Anesthesiology 749-14, 1991 16. Nishikawa T, Dohi S: Errors in the measurements of cardiac output by thermodilution. Can J Anaesth 40:142153,1993 17. Nunn JF, Makita K, Royston B: Validation of oxygen measurements during artificial ventilation. J Appl Physiol 67:2129-2134,1989 18. Parisi AF, Moynihan PF, Feldman CL, et al: Approaches to determination of left ventricular volume and ejection fraction by real-time two-dimensional echocardiography. Clin Cardiol2:257-263,1979 19. Pinsky MR: The meaning of cardiac output. Intensive Care Med 16:415-417,199O 20. Putterman C: Swan-Ganz catheter: A decade of hemodynamic monitoring. J Crit Care 2:127-146,1989 21. Reich DL, Konstadt SN, Nejat M, et al: Intraoperative transesophageal echocardiography for the detection of cardiac preload changes induced by transfusion and phlebotomy in pediatric patients. Anesthesiology 79:10-15, 1993 22. Ronco JJ, Phang T: Validation of an indirect calorimeter to measure oxygen consumption in critically ill patients. J Crit Care 1:36-41,199l 23. Ronco JJ, Phang PT, Walley KR, et al: Oxygen consumption is independent of changes in oxygen delivery in severe adult respiratory distress syndrome. Am Rev Respir Dis 143:1267-1273,199l 24. Rubin SA, Siemienczuk D, Nathan MD, et al: Accuracy of cardiac output, oxygen uptake, and arteriovenous oxygen difference at rest, during exercise, and after vasodilator therapy in patients with severe, chronic heart failure. Am J Cardiol50:973-978,1982 25. Schuster AH, Nanda NC: Doppler echocardiographic measurement of cardiac output: Comparison with a non-golden-standard. Am J Cardiol53:257-259,1984 26. Spinale FG, Mukherjee R, Tanaka R, et al: The effects of valvular regurgitation on thermodilution ejection fraction measurements. Chest 101:723-731,1992 27. Takala J, Keinlnen 0, Vaisanen P, et al: Measurement of gas exchange in intensive care: Laboratory and clinical validation of a new device. Crit Care Med 20:10411047,1989 28. Thys D, Hillel Z, Goldman ME, et al: A comparison of hemodynamic indices derived by invasive monitoring and two-dimensional echocardiography. Anesthesiology 67:630634,1987 29. Van Grondelle A, Ditchey RV, Groves BM, et al: Thermodilution method overestimates low cardiac output in human. Am J Physiol245:H690-H692,1983