- Email: [email protected]
A comparison of thermodilution and pulsed Doppler cardiac output measurement in critically ill children
A comparison of thermodilution and pulsed Doppler cardiac output measurement in critically ill children Daniel A. N o t t e r m a n , MD, Frank V. Castello, MD, C h a r l o t t e S t e i n b e r g , MD, Bruce M. G r e e n w a l d , MD, John E. O'Loughlin, MD, a n d Jeffrey P. Gold, MD From the Divisions of Pediatric Critical Care Medicine and Pediatric Cardiology, Department of Pediatrics, Cornell University Medical College, New York To e v a l u a t e the pulsed Doppler c a r d i a c output method as a noninvasive means for determining c a r d i a c output in critically ill children, we performed paired pulsed Doppler and thermodilution c a r d i a c output determinations in 17 critically ill children. C o m m e r c i a l l y a v a i l a b l e equipment, specifically d e s i g n e d for this purpose, was e m p l o y e d . Forty paired thermodilution and pulsed Doppler determinations were made. There was a significant correlation b e t w e e n the two measurements (pulsed Doppler - 0.84 thermodilution + 0.39; r = 0.79, p <0.01). The ranges of thermodilution measurements (1.02 to 6.26 L/min; m e d i a n 2.77 L/ min) and pulsed Doppler measurements (1.13 to 6.35 L/min; median 2.57 L/min) were not different (p = 0.25). However, differences b e t w e e n individual paired thermodilution and pulsed Doppler measurements were large (-3.13 to 2.03 L/ min; median 0.12 L/min), and the p e r c e n t a g e difference b e t w e e n individual paired thermodilution and pulsed Doppler measurements r a n g e d from 0.41% to 102.5% (median 12.7%). A d i s c r e p a n c y of 15% or more b e t w e e n thermodilution and pulsed Doppler was e n c o u n t e r e d in 18 (45%) of 40 of paired measurements (95% c o n f i d e n c e interval: 29% to 61%), and one fourth of the paired measurements differed by more than 25%. We c o n c l u d e that, as e m p l o y e d in this study, pulsed Doppler c a r d i a c output determination is not sufficiently representative of the thermodilution output to be e m p l o y e d for h e m o d y n a m i c monitoring in critically ill children. (J PEDIATR1989;115:554-60)
Measurement of cardiac output by thermodilution is a standard method of evaluating hemodynamic function in critically ill children. 14 However, this technique requires insertion of a pulmonary artery catheter and is associated with measurable morbidity and mortality rates. 2-6 l~n addition, inserting a pulmonary artery catheter in a child may be difficult and time-consuming.7 A rapid, noninvasive method of accurately determining Cardiac output in critically ill children would be an important advance. Pulsed
Submitted for publication Dec. 9, I988; accepted May 23, 1989. Reprint requests: Daniel A. Notterman, MD, Divisionof Pediatric Critical Care Medicine,The New York Hospital 'Cornell Medical Center, 535 East 68th St., New York, NY 10021.
9/20/14217 554
Doppler measurement of aortic flow velocity has recently been proposed as a noninvasive approach for determining cardiac output. This technique has been described in detail in both the adult and pediatric populations8-15 and has been critically reviewed. 16, 17 Briefly, the Doppler shift imparted
I
PEEP PICU PRISM
Positive end-expiratory pressure Pediatric intensive care unit Pediatric risk of mortality
by a moving stream of erythrocytes is used for sonographic determination o f aortic flow velocity. Imaging occurs through an acoustic window formed by the suprasternal notch.ll, 16 Cardiac output is calculated with the following formula: Cardiac output (ml/min) = Mean aortic blood
Volume 115 Number 4
flow velocity (cm/sec -1) X Aortic cross-sectional area (cm 2) X 60. Aortic cross-sectional area is determined by sonographic measurement of the aortic diameter (d) at the sampling site (area --- [~r . d2/4 cm2]). In most evaluations, aortic diameter is measured along the ascending aorta, s' 10, 13, 14 In children undergoing cardiac catheterization, pulsed Doppler measurements of cardiac output have corresponded well with measurements by the standard Fick method s or indocyanine green curve. 9 A comparison of pulsed Doppler with thcrmoditution cardiac output determinations in critically ill children has not been made. Pulsed Doppler measurement has also been employed in several studies in which cardiac output was noninvasively measured after administration of various vasoactive medications.13, 18 Analysis of this work is work is difficult without a broader understanding of the accuracy and precision of pulsed Doppler measurements. To evaluate the suitability of pulsed Doppler measurement for more general use in pediatric intensive care units, we compared paired pulsed Doppler and thermoditution cardiac output determinations in infants and children with critical illnesses. METHODS Children admitted to a PICU were eligible for study if their illness required insertion of a thermodilution pulmonary artery catheter (No. 5F, 7F, or 7.5F American Edwards Laboratories, Santa Ana, Calif.). Catheters were inserted percutaneously at the bedside, and their location was verified by radiograph before measurement of cardiac output. Aortic diameter was measured with standard instrumentation (ATL Mark-600 Cardioseries Ultrasound System, Advanced Technology Laboratories, Inc., Bothell, Wash.). A parasternal, long-axis image was obtained, and measurements were made in early systole (at the end of the QRS complex). With two-dimensional echocardi0graphy the internal diameter was measured at two sites: at the level of the annulus and in the ascending aorta, just above the sinuses of Valsalva. At each level measurements were made in three to five frames and the results were averaged. In two patients poor image quality secondary to respiratory interference precluded measurement in the ascending aorta; however, in none of the remMning patients was the difference between the two measurements greater than 1 mm, so aortic annulus measurements were used in determining all pulsed Doppler cardiac outputs. No patient had twodimensional or Doppler echocardiographic evidence of intracardiac shunt or aortic valve insufficiency, and every patient had laminar flow in the ascending aorta as observed with two-dimensional guided Doppler echocardiography at the time of aortic diameter measurements. Pulsed Doppler cardiac output was determined with a
Thermodilution and Doppler measurement
555
commercially available device (CardioFlo CF-1 Computer, Cardionics Inc., Houston, Tex.). For subjects less than approximately 10 kg a 3.5 mm transducer was employed; for larger subjects a 10 mm transducer was used: The transducer was manipulated in the suprasternal notch until reproducible triphasic waveforms, representing aortic flow velocity, were obtained as viewed on a strip chart recorder. These waveforms consist of an initial rapid upstroke during systole, negative flow during aortic valve closure, and positive flow during early diastole. 8' 14 Measurements not associated with these waveform characteristics were discarded. The CardioFlo CF-1 Computer does not provide a visual image of the ascending aorta. The operator achieves optimal placement and positioning of the transducer by examining the waveform display and monitoring an audio signal that varies with the amplitude of the measured flow velocity signal. The computer calculates the mean aortic flow velocity and, based on this calculation and the previously determined aortic root diameter, determines and displays the cardiac output in liters per minute. Several pulsed Doppler output determinations were made (between three and six) and the results averaged for each measurement. Immediately after pulsed Doppler cardiac output measurement, a paired thermodilution cardiac output was measured as the average of three determinations following chilled 5% dextrose solution injections. Standard procedures were followed, 19 and injection was synchronized with exhalation. Injectate temperature was less than 15 ~ C and was monitored by an in-line temperature probe. Injectate volumes were 5 ml in patients who had a No. 5F catheter and 10 ml in patients who had either a No. 7 or a No. 7.5F catheter. Clinically evident hemodynamic changes did not occur between paired pulsed Doppler and thermodilution measurements. Thermodilution determinations were completed within 5 minutes of the paired pulsed Doppler measurement. At least two and as many as six paired pulsed Doppler and thermodilution cardiac output measurements were performed on each subject. Values for thermodilution and pulsed Doppler measurement are reported as median and range. As described, each thermodilution and pulsed Doppler measurement was the average of several individual repeated determinations made with the respective technique. So that both technical and intrapatient variability could be examined, the coefficient of variation was calculated across the individual repeated determinations constituting each thermodilution and pulsed Doppler measurement. The difference between paired thermodilution and pulsed Doppler (Difference -- Thermodilution - Pulsed Doppler), and the absolute value of the percentage difference between paired thermodilution and pulsed Doppler (Percentage difference = [] (Thermodilution - Pulsed Doppler)t
556
N o t t e r m a n et al.
The Journal of Pediatrics October 1989
Table I. Effect of percentage difference on other patient characteristics Percentage difference <15% Age (yr) Weight (kg) Mechanical ventilation PEEP (cm H20) PRISM score Survived Thermodilution output (L/min) Pulsed Doppler output (L/min) Coefficient of variation Thermodilution? Pulsed Doppler~
Table II. Comparison of paired thermodilution and pulsed Doppler cardiac output measurements (40 paired measurements in 17 patients)
>15%
0.8-13 (3.25) 7-50 (12.5) 16/22 (73%) 0-20 (9) 5-39 (18)* 13/22 (59%) 1.10-5.60 (2.60)
0.7-15 (9) 5-75 (21.5) 14/18 (78%) 0-21 (5.5) 5-30 (15)* 9/18 (50%) 1.02-6.26(3.88)
1.13-4.98 (2.47)
1.22-6.35(3.29)
0.24%-7.6%(2.7%) 0.74-7.5 (1.7%) 0.34%-7.3%(2.4%) 1.4-10.6 (2.6%)
Values expressed as range (median) or proportion (%). *p = 0.02. "~Coefficient of variation across repeated individual thermodilution determinations. ~/Coefficient of variation across repeated individual pulsed Doppler determinations.
+ Thermodilution] x 100), were comPUted for each pair of thermodilution and pulsed Doppler measurements. The percentage difference indicates the accuracy of the pulsed Doppler measurement relative to the standard method, thermodilution, by quantifying the discrepancy between paired determinations of cardiac output employing these two methods: The absolute value is employed because the magnitude (but not the direction) of any discrepancy is clinically relevant. Age, weight, need for mechanical ventilation, level of positive end-expiratory pressure, use of pressor support, and identity of the physician performing the cardiac output measurements were recorded during each observation. Severity of illness was examined by determining the patient's pediatric risk of mortality score 2~ and by noting whether the patient survived until discharge from the PICU. These patient characteristics are listed in Table I. Reproducibility studies have shown that a difference of at least 15% between successive measurements of cardiac output by commercial thermodilution equipment (when three determinations are averaged per measurement) is necessary to indicate a significant change in cardiac output. 21 For this reason we judged a percentage difference greater than 15% to represent a clinically meaningful difference between thermodilution and pulsed Doppler measurements. The proportion of paired measurements with a percentage difference greater than 15% was noted, as was the inflffence on this proportion of the patient variables described previously. Statistical analysis. Preliminary statistical examination indicated only a weak intraclass correlation between re-
Thermodilution Pulsed Doppler Difference~ Percentage difference:~
Range (L/min -I)
Median (L/min i)
1.02-6.26" 1.13-6.35* -3.13-+2.03 0.41%- 102.5%
2.77 2.57 +0.12 12.7%
*p = 0,25. ~Thermodilution- pulsed Doppler. $[[(Thermodilution - Pulsed Doppler) I + Thermodilution] X 100.
peated measurements of percentage difference made on the same patient (r = 0.31). We subsequently verified this assumption by averaging the percentage difference over all of the observations in each patient and again testing the asso~ ciation between the same patient variables and percentage difference. The results were similar to those obtained when each observation was treated independently. Therefore the data analysis treated all observations as independent measurements. The Wilcoxon test was employed to determine whether the ranges of thermodilution and pulsed Doppler measurements were significantly different. The linear regression equation and correlation coefficient for pulsed Doppler on thermodilution measurements were computer generated. The association, if any, between the previously described patient variables and the proportion of observations in which percentage difference was greater than 15% was tested by the Wilcoxon test or Fisher Exact Test, as appropriate. Statistical significance was p <0.05. RESULTS The study group consisted of 17 children (11 boys) with a median age of 4.3 years (range 0.7 to 15 years) and median weight of 15.5 kg (5 to 75 kg). Of the 17 patients, 10 (59%) survived; the median PRISM score was 17 (range 5 to 39). Five patients received pulmonary artery catheters for adult respiratory distress syndrome, nine for cardiogenic or septic shock, one for trauma, and five for miscellaneous problems. Some Patients had more than one indication. Forty paired measurements of cardiac output were obtained in the 17 patients. The median coefficient of variation across the individual repeated thermodilution determinations was 2.6% (range 0.24% to 7.6%), and across the individual repeated pulsed Doppler determinations it was 2.5% (0.34% to 10.6%). This demonstrates excellent reproducibility for both of these techniques. In addition, it indicates that the subjects did not display much physiologic variability at the time of cardiac output measurement.
Volume 115 Number 4
Thermodilution and Doppler measurement
557
Z Z
O
6
[]
F G_
F-
oo
5
/
O
o /
o
00~ 0
~3 H s CI2 CI2
O
4
0
3
co.~OO/
[24 LIJ
c~
~o
2
(23 [23
L~ O'3 C2
0 0
i
i
i
i
i
•
1
2
3
4
5
6
T H E R N O D I L U T I O N CARDIAC 0 U T P U T ( L / M I N )
Fig. t. Graph of pulsed Doppler versus thermodilution cardiac outputs (L/min), indicating line of regression. Individual paired measurements are indicated by open circles. Measurements associated with outlying patient are indicated by open squares (see text). Pulsed Doppler = 0.84 (thermodilution) + 0.39; r = 0.79.
For the group as a whole, thermodilution output ranged from 1.02 to 6.26 L / m i n (median 2.77 L / m i n ) and pulsed Doppler output from 1.13 to 6.35 L / m i n (median 2.57 L / rain) (p = 0.25; Table II). Pulsed Doppler measurements were significantly correlated with thermodilution (r = 0.79; p <0.01; Fig. 1). The regression line was defined by the following function: pulsed Doppler = 0.84 (thermodilution) + 0.39. The difference between thermodilution and pulsed Doppler outputs ranged from - 3 . 1 3 L / m i n to 2.03 L / m i n (median 0.12 L/rain). The percentage difference between thermodilution and pulsed Doppler output ranged from 0.41% to 102.5% (median 12.7%). In 23 pairs of observations, pulsed Doppler output underestimated thermodilution output; in 17 pairs, pulsed Doppler output overestimated thermodilution output. A discrepancy of 15% or more between pulsed Doppler and thermodilution cardiac outputs was noted in 18 (45%) of 40 paired measurements (95% confidence interval for this proportion: 29% to 61%). One fourth of the paired measurements differed by more than 25%. Need for ventilation, PEEP level, use of pressor agents, and survival until discharge from the P I C U did not affect the percentage difference between thermodilution and pulsed Doppler measurements. Three physicians performed cardiac output measurements. One was a pediatric cardiologist with extensive experience in performing Doppler echocardiography in the pediatric population, and the other two were special-
ists in pediatric critical care. The percentage difference was not affected by the identity of the physician performing the measurement (p = 0.43). The percentage difference was correlated with age (r = 0.40, p = 0.014) and weight (r = 0.53, p <0.01). However, the largest and the second largest percentage differences (102.6% and 56.6%, respectively) were both associated with observations in a single patient, who was also the heaviest (75 kg). When this patient's data were excluded, rieither age nor weight was significantly associated with percentage difference (for age, r = 0.22, p = 0.19; for weight, r = 0.06, p = 0.72). This suggests that the original observed correlations between percentage difference and age or weight were coincidental, the result of a single outlying patient. Inspection of Fig. 1 suggests that the discrepancy between thermodilution and pulsed Doppler measurements was larger at high cardiac outputs. However, as suggested by Fig. 2, there was no significant association between thermodilution cardiac output and percentage difference (r = 0.21, p = 0.19). When the outlying data from the previously described patient were excluded, a weak correlation between thermodilution and percentage difference was observed (r = 0.35, p = 0.03). Surprisingly, there was a significant negative correlation between percentage difference and severity of illness as measured by the P R I S M score (r = -0.48, p <0.01). Thus the discrepancy between methods of measuring cardiac
558
Notterman et al.
The Journal of Pediatrics October 1989
100 90 80 LU tg Z LU OE LU LL II H D
70 60
O
50 40 O
o
O
0
80
O
22~22]2L
........2'~2~•
I0 0 0
O O
30
0 O
ggO ~0
,
OO
. . . . . . . &/•177177
~ 00 ~
1 2 THERMODILUTION
cO0 00 L
O
0 i
i
3 4 5 CRRDIRC OUTPUT (L/MIN)
6
Fig. 2, Relationship between cardiac output (L/min, determined by thermodilution method) and percentage difference between paired thermodilution and pulsed Doppler measurements. Open circles indicate paired measurements. Dashed line indicates median percentage difference. Dotted line indicates percentage difference of 15%.
output tended to be smaller in the more severely ill patients. In Table I the values of several patient variables are compared between observations having a percentage difference > 15% and those with a percentage difference < 15%. There were no significant differences between the two groups of observations, except for the association between percentage difference and PRISM. DISCUSSION In this group of 17 critically ill children, pulsed Doppler and thermodilution measurements of cardiac output were well correlated and median values of cardiac output determined by the two methods did not differ. Most previous evaluations of pulsed Doppler cardiac output measurement have examined only the strength of the correlation between pulsed Doppler and thermodilution measurements. However, this type of correlation analysis may obscure large discrepancies between methods of measurement. These discrepancies are made explicit and quantified by examination of the difference between paired pulsed Doppler and thermodilution measurements. Expressing this term as the absolute value of the percentage difference between pulsed Doppler and thermodilution measurements permits comparison of data derived from patients with different cardiac outputs. Although thermodilution measurements are also affected by well-described sources of error, which may be more pronounced in small children, 23 in our study the discrepancy between measurements made by the two different
methods was not greater at the low flow rates at which thermodilution is least reliable. By expressing the difference between Doppler and thermodilution measurements as a percentage of the thermodilution measurement, we acknowledge that at present the thermodilution method is the clinical standard by which physicians measure cardiac output in critically ill patients. An alternative approach, suggested by Bland and Altman,24 removes this potential source of bias by relating the difference between paired measurements to the average of the two measurements. We did not use this approach because our purpose was to compare the new method, pulsed Doppler measurement, with the clinical standard, thermodilution. However, when the data were reanalyzed by relating the difference between paired measurements to the average of the paired measurements (rather than to thermodilution measurement), the results of our study were unchanged and the magnitude of the discrepancy between thermodilution and pulsed Doppler methods remained as described. The data indicate that pulsed Doppler cardiac output did not provide a reliable estimate of thermodilution cardiac output. A discrepancy greater than 15% was noted in nearly half (45%) of measurement pairs, and one fourth of measurement pairs differed by more than 25%. Unfortunately, knowledge of such factors as age, weight, or need for ventilation or pressor therapy would not enable the physician to select patients in whom pulsed Doppler measurement would be valid. The relationship between the cardiac out-
Volume 115 Number 4
put and percentage difference was weak and would be too dilute to permit confident selection of patients. The percentage difference between thermodilution and pulsed Doppler measurements was not affected by the identity of the physician performing the measurement, and level of operator training was not a factor in the results of this study. Donovan et al., 1~ employing pulsed Doppler equipment similar to ours, performed 145 paired pulsed Doppler and thermodilution cardiac output measurements in 38 critically ill adult patients. They calculated the difference between paired pulsed Doppler and thermodilution measurements and found that the mean difference was large (0.51 L/min) with a wide standard deviation (+1.62 L/rain). Their study did not include children, and they did not express the difference between pulsed Doppler and thermodilution measurements as a percentage of the thermodilution measurement. They also concluded that pulsed Doppler measurements did not precisely estimate those performed with the thermodilution method. Several other studies have reported comparisons between pulsed Doppler and thermodilution measurements, in adult patients, some of whom were critically ill. 11, 12, 16, 17 In most cases these comparisons were based on evaluation of the strength of the correlation between pulsed Doppler and thermodilution measurements rather than the difference between paired measurements. We believe that this study is the first that systematically evaluates pulsed Doppler and thermodilution methods of measuring cardiac output in critically ill children. A related technique, continuous wave Doppler echocardiography, has been evaluated in critically ill children22 and found to provide good agreement with thermodilution determinations. However, this evaluation was also limited to evaluating the correlation between methods of determining cardiac output, rather than differences between paired outputs. Alverson et al. s compared Fick and pulsed Doppler cardiac outputs during cardiac catheterization in 33 children. They found excellent agreement between the two methods. Their patients were not critically ill, and paired differences were not examined. Sholler et al. 9 studied 21 children less than 2 years of age and found excellent correlation between cardiac output determined by pulsed Doppler measurement and that determined by indocyanine green curve. They employed two different sites for measuring aortic crosssectional area: at the aortic leaflets and in the ascending aorta. The correlation was slightly better with the aortic leaflet site (0.99 vs 0.93). Using data provided in the paper by Sholler et al., we calculated percentage difference between paired measurements for both aortic leaflet and ascending aortic sites. With the former, only 3 (14%) of 21 paired measurements differed by more than 15%; with the latter, 14 (67%) of 21 differed by more than 15%, a propor-
T h e r m o d i l u t i o n and Doppler m e a s u r e m e n t
559
tion similar to that observed in our study (45%). This analysis also emphasizes that a strong correlation (r = 0.93) does not imply that pulsed Doppler output accurately represents thermodilution output. The ideal site for measuring aortic diameter has not been established, and other authors have reported results different from those of Scholler et al. Rein et al. 22 compared cardiac output determined by thermodilution with that by a continuous wave Doppler technique. They used three areas to measure aortic diameter: aortic annulus, sinuses of Valsalva, and ascending aorta. The best correlation was achieved when the diameter was measured at the aortic annulus. Similarly, Ihlen et al., 12 when comparing pulsed Doppler with thermodilution measurements, measured aortic diameter at four locations: left ventricular outflow tract, aortic orifice (annulus), proximal aortic root, and distal aortic root. They achieved the closest correlation when aortic diameter was measured at the aortic annulus. In our study aortic diameter was measured at the aortic annulus. We employed a "blind" Doppler technique in which the angle of insonance of the transducer on the sternal notch was manipulated to produce the greatest mean flow velocity as indicated by the amplitude of an audio signal and the appearance of the associated waveform on a strip-chart recorder. In Doppler measurement the angle of insonance should be < 15 degrees.14 When this angle is > 15 degrees, cardiac output will be significantly underestimated by the pulsed Doppler method. In our study the proportion of observations in which pulsed Doppler measurement underestimated thermodilution measurement (23/40) was similar to the proportion in which pulsed Doppler measurement overestimated thermodilution measurement (17/40). This suggests that there was no systematic error in the angle at which the transducer was applied. Although use of two-dimensional, echocardiographically guided flow velocity determination might improve accuracy, this enhancement has not been examined in critically ill pediatric patients and is not available with the commercial equipment we employed. We conclude that, with the instrumentation and methods we used, the pulsed Doppler method of cardiac output determination is not sufficiently accurate to serve as a guide to therapy for critically ill children. Whether equipment providing visual images of the aorta during pulsed Doppler measurement would enhance accuracy remains to be evaluated. We gratefully acknowledge the assistance of Martin L. Lesser, PhD, North Shore University Hospital-Cornell Medical Center, in statistical analysis. REFERENCES
1. Pollack M M, Reed TP, HolbrookPR, et al. Bedsidepulmonary artery catheterization in pediatrics. J PEDIATR1980;274-6.
560
N o t t e r m a n et al.
The Journal of Pediatrics October 1989
2. Swedlow DB. Invasive assessment of the failing circuiation. In: Swedlow DB, Raphaely RC, eds. Cardiovascular problems in pediatric critical care. New York: Churchill Livingston, 1986. 3. Tabata BK, Kirsch JR, Rogers MR. Diagnostic tests and technology for pediatric intensive care. In: Rogers MR, ed. Textbook of pediatric intensive care. Baltimore: Williams & Wilkins, 1987. 4. Katz RW, Pollack MM, Weibley RE. Pulmonary artery catheterization in pediatric intensive care. Adv Pediatr 1983; 30:169-90. 5. Shah KB, Rao T, Laughlin S, et al. A review of pulmonary artery catheterization in 6,245 patients. Anesthesiology 1984;61:271-5. 6. Boyd KD, Thomas S J, Gold J, et al. A prospective study of complications of pulmonary artery catheterizations in 500 consecutive patients. Chest 1983;84:245-8. 7. Raphaely R, Swedlow D, Kettric R, et al. Experience with pulmonary artery catheterizations in critically ill children (Abstract). Crit Care Med 1980;8:265. 8. Alverson DC, Eldridge M, Dillon T, et aI. Noninvasive pulsed Doppler determination of cardiac output in neonates and children. J PEDIATR 1982;101:46-50. 9. Sholler GF, Whight CM, Celermajer JM. Pulsed Doppler echocardiographic assessment, including use of aortic leaflet separation, of cardiac output in children with structural heart disease. Am J Cardiol 1986;57:1195-7. 10. Donovan KD, Dobb G J, Newman MA, et al. Comparison of pulsed Doppler and thermodilution methods for measuring cardiac output in critically ill patients. Crit Care Med 1987;15:853-7. 1 i. Loeppky JA, Hoekenga DE, Greene ER, et al. Comparison of noninvasive pulsed Doppler and Fick measurements of stroke volume in cardiac patients. Am Heart J 1984;107:339-46. 12. Ihlen H, Amlie JP, Dale J, et al. Determination of cardiac output by Doppler echocardiography. Br Heart J 1984;51:5460.
t 3. Walther F J, Sims ME, Siassi B, et al. Cardiac output changes secondary to theophylline therapy in preterm infants. J PEDIATR 1986;109:874-6. 14. Berman W Jr, ed. Pulsed Doppler ultrasound in clinical pediatrics. New York: Futura, 1983. 15. Walther F J, Siassi B, Ramadan NA, et al. Pulsed Doppler determinations of cardiac output in neonates: normal standards for clinical use. Pediatrics 1985;76:829-33. 16. Bernstein DP. Noninvasive cardiac output, Doppler flowmetry, and gold-plated assumptions. Crit Care Med 1987;15: 886-8. 17. Schuster AH, Nanda NC. Doppler echocardiographic measurement of cardiac output: comparison with a non-golden standard. Am J Cardiol 1984;53:257-9. 18. Padbury JF, Agata T, Baylen BG, et al. Dopamine pharmacokinetics in critically ill newborn infants. J PEDIATR 1987; 294:298. 19. Thermodilution cardiac output computer technique model 9520A: product information. Santa Ana, Calif.: American Edwards Laboratories, 1983. 20. Pollack MM, Ruttimann UE, Getson PR. Pediatric risk of mortality (PRISM) score. Crit Care Med 1988;16:1110-6. 21. Stetz CW, Miller RG, Kelly GE, et al. Reliability of the thermodilution method in the determination of cardiac output in clinical practice. Am Rev Respir Dis 1982;126:1001-4. 22. Rein A J, Hsieh KS, Elixson M, et al. Cardiac output estimates in the pediatric intensive care unit using a continuous-wave Doppler computer: validation and limitations of the technique. Am Heart J 1986;112:97-103. 23. Berman W Jr, Lister G Jr, Pitt BR, et al. Measurements of blood flow. Adv Pediatr 1988;35:427-8. 24. Bland M J, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307-10.
FELLOWSHIPS Available fellowships in pediatric subspecialties and those for general academic pediatric training are listed once a year, in May, in THE JOURNAL OF PEDIATRICS. Each October, forms for listing such fellowships are sent to the C h a i r m a n of the D e p a r t m e n t of Pediatrics at most m a j o r hospitals in the U n i t e d States and Canada. A sample application form also appears on page 19A of this issue; this form or photocopies thereof may be used instead of the forms mailed to pediatrics d e p a r t m e n t chairmen. Should you desire to list fellowships, a separate application must be made each year for each position. All-applications must be r e t u r n e d to T h e C.V. Mosby C o m p a n y by F e b r u a r y 15 of the listing year to ensure publication. Additional forms will be supplied on request from the Journal Editing D e p a r t m e n t , T h e C.V. Mosby C o m p a n y , 11830 Westline Industrial Drive, St. Louis, M O 63146-3318/314-872-8370.
Measurement of cardiac output by thermodilution is a standard method of evaluating hemodynamic function in critically ill children. 14 However, this technique requires insertion of a pulmonary artery catheter and is associated with measurable morbidity and mortality rates. 2-6 l~n addition, inserting a pulmonary artery catheter in a child may be difficult and time-consuming.7 A rapid, noninvasive method of accurately determining Cardiac output in critically ill children would be an important advance. Pulsed
Submitted for publication Dec. 9, I988; accepted May 23, 1989. Reprint requests: Daniel A. Notterman, MD, Divisionof Pediatric Critical Care Medicine,The New York Hospital 'Cornell Medical Center, 535 East 68th St., New York, NY 10021.
9/20/14217 554
Doppler measurement of aortic flow velocity has recently been proposed as a noninvasive approach for determining cardiac output. This technique has been described in detail in both the adult and pediatric populations8-15 and has been critically reviewed. 16, 17 Briefly, the Doppler shift imparted
I
PEEP PICU PRISM
Positive end-expiratory pressure Pediatric intensive care unit Pediatric risk of mortality
by a moving stream of erythrocytes is used for sonographic determination o f aortic flow velocity. Imaging occurs through an acoustic window formed by the suprasternal notch.ll, 16 Cardiac output is calculated with the following formula: Cardiac output (ml/min) = Mean aortic blood
Volume 115 Number 4
flow velocity (cm/sec -1) X Aortic cross-sectional area (cm 2) X 60. Aortic cross-sectional area is determined by sonographic measurement of the aortic diameter (d) at the sampling site (area --- [~r . d2/4 cm2]). In most evaluations, aortic diameter is measured along the ascending aorta, s' 10, 13, 14 In children undergoing cardiac catheterization, pulsed Doppler measurements of cardiac output have corresponded well with measurements by the standard Fick method s or indocyanine green curve. 9 A comparison of pulsed Doppler with thcrmoditution cardiac output determinations in critically ill children has not been made. Pulsed Doppler measurement has also been employed in several studies in which cardiac output was noninvasively measured after administration of various vasoactive medications.13, 18 Analysis of this work is work is difficult without a broader understanding of the accuracy and precision of pulsed Doppler measurements. To evaluate the suitability of pulsed Doppler measurement for more general use in pediatric intensive care units, we compared paired pulsed Doppler and thermoditution cardiac output determinations in infants and children with critical illnesses. METHODS Children admitted to a PICU were eligible for study if their illness required insertion of a thermodilution pulmonary artery catheter (No. 5F, 7F, or 7.5F American Edwards Laboratories, Santa Ana, Calif.). Catheters were inserted percutaneously at the bedside, and their location was verified by radiograph before measurement of cardiac output. Aortic diameter was measured with standard instrumentation (ATL Mark-600 Cardioseries Ultrasound System, Advanced Technology Laboratories, Inc., Bothell, Wash.). A parasternal, long-axis image was obtained, and measurements were made in early systole (at the end of the QRS complex). With two-dimensional echocardi0graphy the internal diameter was measured at two sites: at the level of the annulus and in the ascending aorta, just above the sinuses of Valsalva. At each level measurements were made in three to five frames and the results were averaged. In two patients poor image quality secondary to respiratory interference precluded measurement in the ascending aorta; however, in none of the remMning patients was the difference between the two measurements greater than 1 mm, so aortic annulus measurements were used in determining all pulsed Doppler cardiac outputs. No patient had twodimensional or Doppler echocardiographic evidence of intracardiac shunt or aortic valve insufficiency, and every patient had laminar flow in the ascending aorta as observed with two-dimensional guided Doppler echocardiography at the time of aortic diameter measurements. Pulsed Doppler cardiac output was determined with a
Thermodilution and Doppler measurement
555
commercially available device (CardioFlo CF-1 Computer, Cardionics Inc., Houston, Tex.). For subjects less than approximately 10 kg a 3.5 mm transducer was employed; for larger subjects a 10 mm transducer was used: The transducer was manipulated in the suprasternal notch until reproducible triphasic waveforms, representing aortic flow velocity, were obtained as viewed on a strip chart recorder. These waveforms consist of an initial rapid upstroke during systole, negative flow during aortic valve closure, and positive flow during early diastole. 8' 14 Measurements not associated with these waveform characteristics were discarded. The CardioFlo CF-1 Computer does not provide a visual image of the ascending aorta. The operator achieves optimal placement and positioning of the transducer by examining the waveform display and monitoring an audio signal that varies with the amplitude of the measured flow velocity signal. The computer calculates the mean aortic flow velocity and, based on this calculation and the previously determined aortic root diameter, determines and displays the cardiac output in liters per minute. Several pulsed Doppler output determinations were made (between three and six) and the results averaged for each measurement. Immediately after pulsed Doppler cardiac output measurement, a paired thermodilution cardiac output was measured as the average of three determinations following chilled 5% dextrose solution injections. Standard procedures were followed, 19 and injection was synchronized with exhalation. Injectate temperature was less than 15 ~ C and was monitored by an in-line temperature probe. Injectate volumes were 5 ml in patients who had a No. 5F catheter and 10 ml in patients who had either a No. 7 or a No. 7.5F catheter. Clinically evident hemodynamic changes did not occur between paired pulsed Doppler and thermodilution measurements. Thermodilution determinations were completed within 5 minutes of the paired pulsed Doppler measurement. At least two and as many as six paired pulsed Doppler and thermodilution cardiac output measurements were performed on each subject. Values for thermodilution and pulsed Doppler measurement are reported as median and range. As described, each thermodilution and pulsed Doppler measurement was the average of several individual repeated determinations made with the respective technique. So that both technical and intrapatient variability could be examined, the coefficient of variation was calculated across the individual repeated determinations constituting each thermodilution and pulsed Doppler measurement. The difference between paired thermodilution and pulsed Doppler (Difference -- Thermodilution - Pulsed Doppler), and the absolute value of the percentage difference between paired thermodilution and pulsed Doppler (Percentage difference = [] (Thermodilution - Pulsed Doppler)t
556
N o t t e r m a n et al.
The Journal of Pediatrics October 1989
Table I. Effect of percentage difference on other patient characteristics Percentage difference <15% Age (yr) Weight (kg) Mechanical ventilation PEEP (cm H20) PRISM score Survived Thermodilution output (L/min) Pulsed Doppler output (L/min) Coefficient of variation Thermodilution? Pulsed Doppler~
Table II. Comparison of paired thermodilution and pulsed Doppler cardiac output measurements (40 paired measurements in 17 patients)
>15%
0.8-13 (3.25) 7-50 (12.5) 16/22 (73%) 0-20 (9) 5-39 (18)* 13/22 (59%) 1.10-5.60 (2.60)
0.7-15 (9) 5-75 (21.5) 14/18 (78%) 0-21 (5.5) 5-30 (15)* 9/18 (50%) 1.02-6.26(3.88)
1.13-4.98 (2.47)
1.22-6.35(3.29)
0.24%-7.6%(2.7%) 0.74-7.5 (1.7%) 0.34%-7.3%(2.4%) 1.4-10.6 (2.6%)
Values expressed as range (median) or proportion (%). *p = 0.02. "~Coefficient of variation across repeated individual thermodilution determinations. ~/Coefficient of variation across repeated individual pulsed Doppler determinations.
+ Thermodilution] x 100), were comPUted for each pair of thermodilution and pulsed Doppler measurements. The percentage difference indicates the accuracy of the pulsed Doppler measurement relative to the standard method, thermodilution, by quantifying the discrepancy between paired determinations of cardiac output employing these two methods: The absolute value is employed because the magnitude (but not the direction) of any discrepancy is clinically relevant. Age, weight, need for mechanical ventilation, level of positive end-expiratory pressure, use of pressor support, and identity of the physician performing the cardiac output measurements were recorded during each observation. Severity of illness was examined by determining the patient's pediatric risk of mortality score 2~ and by noting whether the patient survived until discharge from the PICU. These patient characteristics are listed in Table I. Reproducibility studies have shown that a difference of at least 15% between successive measurements of cardiac output by commercial thermodilution equipment (when three determinations are averaged per measurement) is necessary to indicate a significant change in cardiac output. 21 For this reason we judged a percentage difference greater than 15% to represent a clinically meaningful difference between thermodilution and pulsed Doppler measurements. The proportion of paired measurements with a percentage difference greater than 15% was noted, as was the inflffence on this proportion of the patient variables described previously. Statistical analysis. Preliminary statistical examination indicated only a weak intraclass correlation between re-
Thermodilution Pulsed Doppler Difference~ Percentage difference:~
Range (L/min -I)
Median (L/min i)
1.02-6.26" 1.13-6.35* -3.13-+2.03 0.41%- 102.5%
2.77 2.57 +0.12 12.7%
*p = 0,25. ~Thermodilution- pulsed Doppler. $[[(Thermodilution - Pulsed Doppler) I + Thermodilution] X 100.
peated measurements of percentage difference made on the same patient (r = 0.31). We subsequently verified this assumption by averaging the percentage difference over all of the observations in each patient and again testing the asso~ ciation between the same patient variables and percentage difference. The results were similar to those obtained when each observation was treated independently. Therefore the data analysis treated all observations as independent measurements. The Wilcoxon test was employed to determine whether the ranges of thermodilution and pulsed Doppler measurements were significantly different. The linear regression equation and correlation coefficient for pulsed Doppler on thermodilution measurements were computer generated. The association, if any, between the previously described patient variables and the proportion of observations in which percentage difference was greater than 15% was tested by the Wilcoxon test or Fisher Exact Test, as appropriate. Statistical significance was p <0.05. RESULTS The study group consisted of 17 children (11 boys) with a median age of 4.3 years (range 0.7 to 15 years) and median weight of 15.5 kg (5 to 75 kg). Of the 17 patients, 10 (59%) survived; the median PRISM score was 17 (range 5 to 39). Five patients received pulmonary artery catheters for adult respiratory distress syndrome, nine for cardiogenic or septic shock, one for trauma, and five for miscellaneous problems. Some Patients had more than one indication. Forty paired measurements of cardiac output were obtained in the 17 patients. The median coefficient of variation across the individual repeated thermodilution determinations was 2.6% (range 0.24% to 7.6%), and across the individual repeated pulsed Doppler determinations it was 2.5% (0.34% to 10.6%). This demonstrates excellent reproducibility for both of these techniques. In addition, it indicates that the subjects did not display much physiologic variability at the time of cardiac output measurement.
Volume 115 Number 4
Thermodilution and Doppler measurement
557
Z Z
O
6
[]
F G_
F-
oo
5
/
O
o /
o
00~ 0
~3 H s CI2 CI2
O
4
0
3
co.~OO/
[24 LIJ
c~
~o
2
(23 [23
L~ O'3 C2
0 0
i
i
i
i
i
•
1
2
3
4
5
6
T H E R N O D I L U T I O N CARDIAC 0 U T P U T ( L / M I N )
Fig. t. Graph of pulsed Doppler versus thermodilution cardiac outputs (L/min), indicating line of regression. Individual paired measurements are indicated by open circles. Measurements associated with outlying patient are indicated by open squares (see text). Pulsed Doppler = 0.84 (thermodilution) + 0.39; r = 0.79.
For the group as a whole, thermodilution output ranged from 1.02 to 6.26 L / m i n (median 2.77 L / m i n ) and pulsed Doppler output from 1.13 to 6.35 L / m i n (median 2.57 L / rain) (p = 0.25; Table II). Pulsed Doppler measurements were significantly correlated with thermodilution (r = 0.79; p <0.01; Fig. 1). The regression line was defined by the following function: pulsed Doppler = 0.84 (thermodilution) + 0.39. The difference between thermodilution and pulsed Doppler outputs ranged from - 3 . 1 3 L / m i n to 2.03 L / m i n (median 0.12 L/rain). The percentage difference between thermodilution and pulsed Doppler output ranged from 0.41% to 102.5% (median 12.7%). In 23 pairs of observations, pulsed Doppler output underestimated thermodilution output; in 17 pairs, pulsed Doppler output overestimated thermodilution output. A discrepancy of 15% or more between pulsed Doppler and thermodilution cardiac outputs was noted in 18 (45%) of 40 paired measurements (95% confidence interval for this proportion: 29% to 61%). One fourth of the paired measurements differed by more than 25%. Need for ventilation, PEEP level, use of pressor agents, and survival until discharge from the P I C U did not affect the percentage difference between thermodilution and pulsed Doppler measurements. Three physicians performed cardiac output measurements. One was a pediatric cardiologist with extensive experience in performing Doppler echocardiography in the pediatric population, and the other two were special-
ists in pediatric critical care. The percentage difference was not affected by the identity of the physician performing the measurement (p = 0.43). The percentage difference was correlated with age (r = 0.40, p = 0.014) and weight (r = 0.53, p <0.01). However, the largest and the second largest percentage differences (102.6% and 56.6%, respectively) were both associated with observations in a single patient, who was also the heaviest (75 kg). When this patient's data were excluded, rieither age nor weight was significantly associated with percentage difference (for age, r = 0.22, p = 0.19; for weight, r = 0.06, p = 0.72). This suggests that the original observed correlations between percentage difference and age or weight were coincidental, the result of a single outlying patient. Inspection of Fig. 1 suggests that the discrepancy between thermodilution and pulsed Doppler measurements was larger at high cardiac outputs. However, as suggested by Fig. 2, there was no significant association between thermodilution cardiac output and percentage difference (r = 0.21, p = 0.19). When the outlying data from the previously described patient were excluded, a weak correlation between thermodilution and percentage difference was observed (r = 0.35, p = 0.03). Surprisingly, there was a significant negative correlation between percentage difference and severity of illness as measured by the P R I S M score (r = -0.48, p <0.01). Thus the discrepancy between methods of measuring cardiac
558
Notterman et al.
The Journal of Pediatrics October 1989
100 90 80 LU tg Z LU OE LU LL II H D
70 60
O
50 40 O
o
O
0
80
O
22~22]2L
........2'~2~•
I0 0 0
O O
30
0 O
ggO ~0
,
OO
. . . . . . . &/•177177
~ 00 ~
1 2 THERMODILUTION
cO0 00 L
O
0 i
i
3 4 5 CRRDIRC OUTPUT (L/MIN)
6
Fig. 2, Relationship between cardiac output (L/min, determined by thermodilution method) and percentage difference between paired thermodilution and pulsed Doppler measurements. Open circles indicate paired measurements. Dashed line indicates median percentage difference. Dotted line indicates percentage difference of 15%.
output tended to be smaller in the more severely ill patients. In Table I the values of several patient variables are compared between observations having a percentage difference > 15% and those with a percentage difference < 15%. There were no significant differences between the two groups of observations, except for the association between percentage difference and PRISM. DISCUSSION In this group of 17 critically ill children, pulsed Doppler and thermodilution measurements of cardiac output were well correlated and median values of cardiac output determined by the two methods did not differ. Most previous evaluations of pulsed Doppler cardiac output measurement have examined only the strength of the correlation between pulsed Doppler and thermodilution measurements. However, this type of correlation analysis may obscure large discrepancies between methods of measurement. These discrepancies are made explicit and quantified by examination of the difference between paired pulsed Doppler and thermodilution measurements. Expressing this term as the absolute value of the percentage difference between pulsed Doppler and thermodilution measurements permits comparison of data derived from patients with different cardiac outputs. Although thermodilution measurements are also affected by well-described sources of error, which may be more pronounced in small children, 23 in our study the discrepancy between measurements made by the two different
methods was not greater at the low flow rates at which thermodilution is least reliable. By expressing the difference between Doppler and thermodilution measurements as a percentage of the thermodilution measurement, we acknowledge that at present the thermodilution method is the clinical standard by which physicians measure cardiac output in critically ill patients. An alternative approach, suggested by Bland and Altman,24 removes this potential source of bias by relating the difference between paired measurements to the average of the two measurements. We did not use this approach because our purpose was to compare the new method, pulsed Doppler measurement, with the clinical standard, thermodilution. However, when the data were reanalyzed by relating the difference between paired measurements to the average of the paired measurements (rather than to thermodilution measurement), the results of our study were unchanged and the magnitude of the discrepancy between thermodilution and pulsed Doppler methods remained as described. The data indicate that pulsed Doppler cardiac output did not provide a reliable estimate of thermodilution cardiac output. A discrepancy greater than 15% was noted in nearly half (45%) of measurement pairs, and one fourth of measurement pairs differed by more than 25%. Unfortunately, knowledge of such factors as age, weight, or need for ventilation or pressor therapy would not enable the physician to select patients in whom pulsed Doppler measurement would be valid. The relationship between the cardiac out-
Volume 115 Number 4
put and percentage difference was weak and would be too dilute to permit confident selection of patients. The percentage difference between thermodilution and pulsed Doppler measurements was not affected by the identity of the physician performing the measurement, and level of operator training was not a factor in the results of this study. Donovan et al., 1~ employing pulsed Doppler equipment similar to ours, performed 145 paired pulsed Doppler and thermodilution cardiac output measurements in 38 critically ill adult patients. They calculated the difference between paired pulsed Doppler and thermodilution measurements and found that the mean difference was large (0.51 L/min) with a wide standard deviation (+1.62 L/rain). Their study did not include children, and they did not express the difference between pulsed Doppler and thermodilution measurements as a percentage of the thermodilution measurement. They also concluded that pulsed Doppler measurements did not precisely estimate those performed with the thermodilution method. Several other studies have reported comparisons between pulsed Doppler and thermodilution measurements, in adult patients, some of whom were critically ill. 11, 12, 16, 17 In most cases these comparisons were based on evaluation of the strength of the correlation between pulsed Doppler and thermodilution measurements rather than the difference between paired measurements. We believe that this study is the first that systematically evaluates pulsed Doppler and thermodilution methods of measuring cardiac output in critically ill children. A related technique, continuous wave Doppler echocardiography, has been evaluated in critically ill children22 and found to provide good agreement with thermodilution determinations. However, this evaluation was also limited to evaluating the correlation between methods of determining cardiac output, rather than differences between paired outputs. Alverson et al. s compared Fick and pulsed Doppler cardiac outputs during cardiac catheterization in 33 children. They found excellent agreement between the two methods. Their patients were not critically ill, and paired differences were not examined. Sholler et al. 9 studied 21 children less than 2 years of age and found excellent correlation between cardiac output determined by pulsed Doppler measurement and that determined by indocyanine green curve. They employed two different sites for measuring aortic crosssectional area: at the aortic leaflets and in the ascending aorta. The correlation was slightly better with the aortic leaflet site (0.99 vs 0.93). Using data provided in the paper by Sholler et al., we calculated percentage difference between paired measurements for both aortic leaflet and ascending aortic sites. With the former, only 3 (14%) of 21 paired measurements differed by more than 15%; with the latter, 14 (67%) of 21 differed by more than 15%, a propor-
T h e r m o d i l u t i o n and Doppler m e a s u r e m e n t
559
tion similar to that observed in our study (45%). This analysis also emphasizes that a strong correlation (r = 0.93) does not imply that pulsed Doppler output accurately represents thermodilution output. The ideal site for measuring aortic diameter has not been established, and other authors have reported results different from those of Scholler et al. Rein et al. 22 compared cardiac output determined by thermodilution with that by a continuous wave Doppler technique. They used three areas to measure aortic diameter: aortic annulus, sinuses of Valsalva, and ascending aorta. The best correlation was achieved when the diameter was measured at the aortic annulus. Similarly, Ihlen et al., 12 when comparing pulsed Doppler with thermodilution measurements, measured aortic diameter at four locations: left ventricular outflow tract, aortic orifice (annulus), proximal aortic root, and distal aortic root. They achieved the closest correlation when aortic diameter was measured at the aortic annulus. In our study aortic diameter was measured at the aortic annulus. We employed a "blind" Doppler technique in which the angle of insonance of the transducer on the sternal notch was manipulated to produce the greatest mean flow velocity as indicated by the amplitude of an audio signal and the appearance of the associated waveform on a strip-chart recorder. In Doppler measurement the angle of insonance should be < 15 degrees.14 When this angle is > 15 degrees, cardiac output will be significantly underestimated by the pulsed Doppler method. In our study the proportion of observations in which pulsed Doppler measurement underestimated thermodilution measurement (23/40) was similar to the proportion in which pulsed Doppler measurement overestimated thermodilution measurement (17/40). This suggests that there was no systematic error in the angle at which the transducer was applied. Although use of two-dimensional, echocardiographically guided flow velocity determination might improve accuracy, this enhancement has not been examined in critically ill pediatric patients and is not available with the commercial equipment we employed. We conclude that, with the instrumentation and methods we used, the pulsed Doppler method of cardiac output determination is not sufficiently accurate to serve as a guide to therapy for critically ill children. Whether equipment providing visual images of the aorta during pulsed Doppler measurement would enhance accuracy remains to be evaluated. We gratefully acknowledge the assistance of Martin L. Lesser, PhD, North Shore University Hospital-Cornell Medical Center, in statistical analysis. REFERENCES
1. Pollack M M, Reed TP, HolbrookPR, et al. Bedsidepulmonary artery catheterization in pediatrics. J PEDIATR1980;274-6.
560
N o t t e r m a n et al.
The Journal of Pediatrics October 1989
2. Swedlow DB. Invasive assessment of the failing circuiation. In: Swedlow DB, Raphaely RC, eds. Cardiovascular problems in pediatric critical care. New York: Churchill Livingston, 1986. 3. Tabata BK, Kirsch JR, Rogers MR. Diagnostic tests and technology for pediatric intensive care. In: Rogers MR, ed. Textbook of pediatric intensive care. Baltimore: Williams & Wilkins, 1987. 4. Katz RW, Pollack MM, Weibley RE. Pulmonary artery catheterization in pediatric intensive care. Adv Pediatr 1983; 30:169-90. 5. Shah KB, Rao T, Laughlin S, et al. A review of pulmonary artery catheterization in 6,245 patients. Anesthesiology 1984;61:271-5. 6. Boyd KD, Thomas S J, Gold J, et al. A prospective study of complications of pulmonary artery catheterizations in 500 consecutive patients. Chest 1983;84:245-8. 7. Raphaely R, Swedlow D, Kettric R, et al. Experience with pulmonary artery catheterizations in critically ill children (Abstract). Crit Care Med 1980;8:265. 8. Alverson DC, Eldridge M, Dillon T, et aI. Noninvasive pulsed Doppler determination of cardiac output in neonates and children. J PEDIATR 1982;101:46-50. 9. Sholler GF, Whight CM, Celermajer JM. Pulsed Doppler echocardiographic assessment, including use of aortic leaflet separation, of cardiac output in children with structural heart disease. Am J Cardiol 1986;57:1195-7. 10. Donovan KD, Dobb G J, Newman MA, et al. Comparison of pulsed Doppler and thermodilution methods for measuring cardiac output in critically ill patients. Crit Care Med 1987;15:853-7. 1 i. Loeppky JA, Hoekenga DE, Greene ER, et al. Comparison of noninvasive pulsed Doppler and Fick measurements of stroke volume in cardiac patients. Am Heart J 1984;107:339-46. 12. Ihlen H, Amlie JP, Dale J, et al. Determination of cardiac output by Doppler echocardiography. Br Heart J 1984;51:5460.
t 3. Walther F J, Sims ME, Siassi B, et al. Cardiac output changes secondary to theophylline therapy in preterm infants. J PEDIATR 1986;109:874-6. 14. Berman W Jr, ed. Pulsed Doppler ultrasound in clinical pediatrics. New York: Futura, 1983. 15. Walther F J, Siassi B, Ramadan NA, et al. Pulsed Doppler determinations of cardiac output in neonates: normal standards for clinical use. Pediatrics 1985;76:829-33. 16. Bernstein DP. Noninvasive cardiac output, Doppler flowmetry, and gold-plated assumptions. Crit Care Med 1987;15: 886-8. 17. Schuster AH, Nanda NC. Doppler echocardiographic measurement of cardiac output: comparison with a non-golden standard. Am J Cardiol 1984;53:257-9. 18. Padbury JF, Agata T, Baylen BG, et al. Dopamine pharmacokinetics in critically ill newborn infants. J PEDIATR 1987; 294:298. 19. Thermodilution cardiac output computer technique model 9520A: product information. Santa Ana, Calif.: American Edwards Laboratories, 1983. 20. Pollack MM, Ruttimann UE, Getson PR. Pediatric risk of mortality (PRISM) score. Crit Care Med 1988;16:1110-6. 21. Stetz CW, Miller RG, Kelly GE, et al. Reliability of the thermodilution method in the determination of cardiac output in clinical practice. Am Rev Respir Dis 1982;126:1001-4. 22. Rein A J, Hsieh KS, Elixson M, et al. Cardiac output estimates in the pediatric intensive care unit using a continuous-wave Doppler computer: validation and limitations of the technique. Am Heart J 1986;112:97-103. 23. Berman W Jr, Lister G Jr, Pitt BR, et al. Measurements of blood flow. Adv Pediatr 1988;35:427-8. 24. Bland M J, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307-10.
FELLOWSHIPS Available fellowships in pediatric subspecialties and those for general academic pediatric training are listed once a year, in May, in THE JOURNAL OF PEDIATRICS. Each October, forms for listing such fellowships are sent to the C h a i r m a n of the D e p a r t m e n t of Pediatrics at most m a j o r hospitals in the U n i t e d States and Canada. A sample application form also appears on page 19A of this issue; this form or photocopies thereof may be used instead of the forms mailed to pediatrics d e p a r t m e n t chairmen. Should you desire to list fellowships, a separate application must be made each year for each position. All-applications must be r e t u r n e d to T h e C.V. Mosby C o m p a n y by F e b r u a r y 15 of the listing year to ensure publication. Additional forms will be supplied on request from the Journal Editing D e p a r t m e n t , T h e C.V. Mosby C o m p a n y , 11830 Westline Industrial Drive, St. Louis, M O 63146-3318/314-872-8370.