J.-¥. Lefrant P. Bruelle A.G.M.Aya G. Sa'/ssi M. Dauzat J.-E. de La Coussaye J.-J. Eledj am
Training is required to improve the reliability of esophageal Doppler to measure cardiac output in critically ill patients
Received: 20 June 1997 Accepted: 17 December 1997
J. Y. Lefrant ([5~). R Bruelle • A. G. M. Aya. G. Sa~'ssi - J.-E. de La Coussaye. J.-J. Eledjam Ddpartement d'Anesth6sie-R6animation, Centre Hospitalier Universitaire de Ntmes, 5 rue Hoche, F-30 029 Nimes C6dex 4~ France Fax: + 33 (4) 66683624
M. Dauzat Laboratoire d'Anesth6sie et de Physiologie Cardio-Vasculaire, Centre Hospitalier Universitaire de Ntmes, 5 rue Hoche, F-30 029 Ntmes C6dex 4, France
Abstract Objectives: Assessment of and effect of training on reliability of esophageal Doppler (ED) versus thermodilution (TD) for cardiac output (CO) measurement. Design: Prospective study. Setting: Intensive care unit of a uni- versity hospital. Patients: 64 consecutive critically ill patients requiring a pulmonary ar- tery catheter, sedation, and mechan- ical ventilation. Interventions: Esophageal Doppler CO measurements were performed by the same operator, whereas TD CO measurements were carried out by other independent operators. A training period involving the first 12 patients made the operator self- confident. In the remaining patients, the reliability of ED was assessed (evaluation period), using correla- tion coefficients and the Bland and Altman diagram. Between training and evaluation periods, correlation coefficients, biases, and limits of agreement were compared.
Measurements and results: During training and evaluation periods, 107 and 320 CO measurements were performed in 11 out of 12 patients and in 49 out of 52 patients, respec- tively. Continuous CO monitoring was achieved in 6 out of 11 patients and in 38 out of 49 patients during training and evaluation periods, re- spectively. Between the two periods, correlation coefficients increased from 0.53 to 0.89 (p < 0.001), bias decreased from 1.2 to 0.1 1.min -1 (p < 0.001), and limits of agreement decreased from 3.2 to 2.2 1.min -1
< 0.001). Conclusion: A period of training in- volving no more than 12 patients is probably required to ensure reli- ability of CO measurement by ED.
With the routine use of pulmonary artery catheters, thermodilution (TD) has become the reference method of cardiac output (CO) measurement in intensive care units [1, 2]. However, pulmonary artery catheterization may cause rare but serious complications and/or misin- terpretations, especially for inexperienced operators [3-5]. Among noninvasive techniques used for CO mea- surement, ultrasonographic methods seem to be the
most suitable in intensive care units [6, 7]. Transthoracic and/or transesophageal Doppler echocardiography al- low CO measurement and assessments of left ventricu- lar preload, afterload, and contractility but have re- quired extensive training [6-9]. Doppler methods were first used to measure CO from the suprasternal notch [10]. However, suprasternal Doppler does not allow continuous monitoring of CO. In 1989, Singer et al., us- ing a nomogram to determine the aortic diameter, re- ported good agreement between esophageal Doppler
348
Triangle A
Triangle A
_/%
Pic velocity
i i .............................. I riangle B
Triangle B
I v
FTc
Fig. 1 Typical blood flow velocity waveform. Stroke volume is the product of the cross-sectional area and the area under the curve of blood velocity. The later area can be assessed by the method of triangles A and B. Peak velocity and corrected flow time FTc allow measurement of the stroke volume by the formula: SV = CSA. 1/2.(FT c . pic velocity) Cardiac output is the product of the latter with factor K and the heart rate: CO = CSA. 1/2.(FTc. pic velocity). HR. K
( E D ) and T D for C O changes in 38 cr i t ica l ly ill pa t i en t s [11]. H o w e v e r , this s tudy d id no t c o m p a r e E D a n d T D for C O m e a s u r e m e n t . M o r e o v e r , the i m p a c t of t r a in ing on the r e l i ab i l i t y of the t e c h n i q u e has no t b e e n assessed. T h e r e f o r e , the a ims of ou r s tudy were to c o m p a r e E D versus T D for C O m e a s u r e m e n t in c r i t ica l ly ill pa t i en t s and to assess the ef fec t of a p e r i o d of t r a in ing on the re- l i ab i l i ty of E D .
Materials and methods
After receiving approval from the local ethics committee and in- formed consent from the patient's family or guardian, consecutive patients requiring mechanical ventilation, pulmonary artery cathe- terization, and sedation were entered in to this prospective study. The exclusion criteria included the following: patients who were < 18 years old, patients with pharyngeal, laryngeal, or esophageal disease or surgery, and patients who were not mechanically venti- lated. Patients who had undergone recent gastric surgery or had significant bleeding diathesis and those with history of tricuspid re- gurgitation, intracardiac shunt, or aneurysm of the thoracic aorta were also excluded.
A pulmonary artery catheter (8 Fr, 5 lumen, Abbott Critical Care Systems, North Chicago, Ill., USA) was inserted and its cor- rect placement was confirmed by chest radiography. Measure- ments were performed using 10 ml cold saline asynchronously with ventilation. The CO value was the mean of the first four CO measurements without artifacts on the temperature curves.
We used a continuous wave Doppler device (DOPTEK ODM I, Deltex Medical, Chichester, Sussex, UK) with an esophageal probe consisting of a plastic tube 9 mm in diameter with a 5.1 MHz trans- ducer angled at 45 °. The probe was inserted orally into the esopha- gus to a depth of 35 to 40 cm from the teeth, until blood flow signals were heard. The probe was rotated as needed to obtain the best Doppler signal of blood flow in the midstream of the descending thoracic aorta. Correct probe placement was assumed when a sharp, well-defined waveform was seen on the monitor and crisp sound was heard through the loudspeaker [12]. ED measures the blood flow velocity in the descending thoracic aorta (Fig. 1). CO was automatically calculated as the product of the cross-sectional area, time averaged mean blood flow velocity in the descending thoracic aorta, heart rate, and a correcting factor (named factor K) that took into account the proportion of CO for the upper and low- er parts of the body, respectively. Cross-sectional area and factor K were taken from Singer's nomogram based on the patient's age, weight, and height [11]. Cardiac output was averaged over six beats to avoid variations induced by mechanical ventilation [13].
ED and TD CO measurements were performed simultaneously by two independent operators when the patient's hemodynamic status required CO measurement for either diagnostic or therapeu- tic purposes. TD was always performed by experienced physicians. ED was always performed by the same operator. However, the lat- ter was not used to CO measurement by ED at the beginning of the study. Another physician experienced in the Doppler technique taught the ED operator to place the probe and measure the CO value. Therefore, the study was separated in to a training and an evaluation period. When the operator was always able to get a loud and clear Doppler signal, with a well-defined sharp waveform, and when the ED value of CO was less than 15 % different from the TD value in four consecutive patients, the training period was ended and the evaluation period started. Nine to 11 and 6 to 8 CO measurements were performed per patient during the training and evaluation periods, respectively [14, 15]. There were more CO measurements per patient in the training period to train the operator quickly.
Results were expressed as means _+ SD. The number of pa- tients in whom CO was continuously monitored by ED was re- ported. The comparison between ED and TD was performed us- ing linear regression analysis, correlation coefficients, and the Bland and Altman method [16]. The ability to detect CO changes (ACO in percent) was also tested using a correlation coefficient and a linear regression analysis for CO changes with both meth- ods. Baseline patient characteristics in the training and evaluation groups were compared using Student's t test and the chi-square test. The correlation coefficients, biases, and limits of agreement were compared using Fisher's Z-transformation, two-sample T- test and the F-test, respectively. A p value < 0.05 was considered significant.
Results
Six ty - four pa t i en t s were e n r o l l e d in the study. T h e r e were no c o m p l i c a t i o n s a s soc i a t ed wi th e i the r E D or TD. A f t e r 12 pa t i en t s , the o p e r a t o r was ab le to ge t a l oud and c lear D o p p l e r s ignal , wi th a w e l l - d e f i n e d sha rp w a v e f o r m (end of the t r a in ing pe r iod ) . D u r i n g this pe - r iod , 1 p a t i e n t was e x c l u d e d b e c a u s e the T D t e c h n i q u e fa i led. T h e e v a l u a t i o n p e r i o d i n c l u d e d the fo l lowing 52 pa t ien ts . In 3 ou t of the 52 pa t i en t s , the D o p p l e r sig- na l cou ld no t be o b t a i n e d b e c a u s e of the indwel l ing na-
Septic shock 7 21 Pancreatitis 4 6 Major abdominal surgery 0 11 Acute respiratory distress syndrome 0 4 Cardiac surgery 0 3 Other 0 4
sogastric tube. In the remaining patients, the Doppler signal was obtained in less than 10 min. The patient's baseline characteristics (age, height, weight, sex ratio, and diseases) were not different between the training and the evaluation periods (Table 1).
Training period
A total of 107 paired CO measurements were perform- ed in 11 patients. CO was continuously monitored by ED in 6patients (54%). Mean CO values were 9.2+1.61.min -1 (range 4.5 to 13.31.min -1) and 8.0 + 1.7 1.min < (range 3.4 to 12.9 1.min -1) by TD and ED, respectively. The linear regression equation was ED CO = 0.56 TD CO + 2.86. The correlation coeffi- cient was 0.53 (Fig. 2). The bias and limit of agreement were 1.2 and 3.2 1.min -1, respectively (Fig.3). There were 96 CO changes. The linear regression equation was AED CO = 0.62 ATD CO + 0.82 (r = 0.57).
Evaluation period
Altogether, 320 paired CO measurements were per- formed in 49 patients. CO was continuously monitored by ED in 38 patients (77%). Mean CO values were 7.5+2.41.rain -1 (range 2.5 to 14.71.rain 1) and 7.4 + 2.2 1.min q (range 1.5 to 14.9 1.min -1) by TD and ED, respectively. These mean CO values were signifi- cantly lower than mean values obtained during the training period (p < 0.001 for TD comparison, p < 0.01 for ED comparison). The linear regression equation was ED CO = 0.85 TD CO + 1.06. The correlation coef- ficient was 0.89 (Fig. 4). The bias and limit of agreement were 0.1 and 2.21.min q, respectively (Fig.5). There were 271 CO changes. The linear regression equation was AED CO = 0.99 ATD CO + 0.66 (r = 0.85).
~, 14
"E" 12 N
10
8 ~
6-
o o 4 - (3 14,1
2
o
. . , ' .
$ ee
~ o .itltlltliltttltll trill
D i •
i ; i r ~ •
=| e, e
0'JJ 0 2 4 6 8 10 12 14
TD CO I.min -1
ED CO = .56 TD CO + 2.86 r = .53 n = 107 measurements N = 11 patients
Fig.2 Regression line and individual data during the training peri- od. CO cardiac output, TD thermodilution, ED esophageal Dop- pler, dotted line identity line
5 ¸ '7, e-
T= 4 m
3
2
O O
0 '
UJ n "1
O o -2 a I - -3
+ 2 S D
- 2 S D
ee • •
• • " B
• t e e •
• • e
I o
, . , . , . , . , . , . , . ,
0 2 4 6 8 10 12 14
1/2(TD CO + ED CO) I .min "1
bias = 1.2 I .min "1 2 SD = 3.2 I .min -1
n = 107 measurements N = 11 patients
Fig.3 Bland and Altman diagram for the training period. CO car- diac output, TD thermodilution, ED esophageal Doppler
Effect of training
The correlation coefficient increased to 0.89 in the eval- uation period compared with the training period (r = 0.53,p < 0.001). The bias and limit of agreement de- creased from 1.2 to 0.1 1.min -1 (p < 0.001) and from 3.2 to 2.2 1.rain -1 (p < 0.001), respectively. For CO changes, the correlation coefficient increased from 0.57 to 0.85 (p < 0.01).
350
"7. e-
1 6 /
12 ° • • # l l l l l f
• • • • =e ji l l Ill = l ~ • • •
10 , , ~ I • • •
u _ o ° h ~ o ° l b • ~ • •
.
"hmkl lR~ - -
CI ILl
4
ijJJ jjl 0
T D C O I .min ~
E D C O = .85 T D C O + 1 , 0 6
r = .89
n = 320 measurements N = 4 9 patients
Fig.4 Regression line and individual data during the evaluation period. CO cardiac output, TD thermodilution, ED esophageal Doppler, dotted line identity line
5 - • r, .
'~ 4 -
3 -
2
0 1 0
w 0 - - n
0 -1 0
-2
+ 2 S D • "
j: : • . . . . i u ~ • ! i • • •
• • in • • in • • .=k . . . V ~ . . . . . r i~m''
i ~ l l [ l I ~ i i
• " - ~ . " ' . ~ v . . " " .
- 2 SD . • •
B B
2 4 6 8 t0 12 14 t6 1/2 (TD CO + ED CO) I . rn in "1
bias = 0.1 l . m i n "1
2 S D = 2 .2 I .rnin "1
n = 320 measurements N = 4 9 patients
Fig.5 Bland and Altman diagram for the evaluation period. CO cardiac output, TD thermodilution, ED esophageal Doppler
D i s c u s s i o n
In the present study, a training period including 12 pa- tients allowed an increase in the correlation coefficients (between CO measurements and CO changes) and a de- crease in bias and limits of agreement. Moreover, ED al- lowed rapid CO measurement in more than 90 % of pa- tients and continuous CO monitoring in more than 75 % of patients during the evaluation period. Finally, no complications related to ED occurred.
ED has been used since the 1980s to measure CO in surgical patients [17-22]. Correlation coefficients, bias, and limits of agreement have ranged from 0.52 to 0.94, from 0.16 to 0.4 1.•in -1 and from 1.4 to 3.6 1.min -~, re- spectively [17-22]. In these studies, the determination of the aortic diameter with A-mode sonography or with a nomogram did not seem to alter the reliability of the different devices used [17-22]. In 21 critically ill pa- tients, Lavandier et aI. obtained a correlation coefficient of 0.97 [23]. However, these authors did not plot a Bland and Altman diagram. In the present study, correlation coefficients, bias, and limits of agreement during the evaluation period are similar to those reported in the previous studies. Using the DOPTEK ODM I in 38 crit- ically ill patients, Singer et al. reported that this device was able to assess relative changes in CO (ability to de- tect a CO change > 10%) [11]. This was confirmed by Klotz et al. in 6 patients during aortic surgery [24]. The present study showed that the correlation between CO changes increased from 0.57 to 0.85 with training. This confirms that ED is able to monitor CO changes.
There are several reasons for the discrepancies be- tween ED and TD. First, the use of TD as a reference method can be debated. Errors may alter CO measure- ment by TD [25]. Recently, an experimental study and a clinical trial demonstrated that the error in CO mea- surement can increase in the higher range of CO values [26, 27]. This phenomenon, called heteroscedasticity, makes the standard deviation higher when the CO value increases. This could explain the scattering of CO values > 8 1.•in -1 on Fig.3. Therefore, the 13 % error in CO measurement reported by Stetz et al. would probably be underestimated in high CO values [28].
ED also leads to errors in CO measurement [12]. It measures blood flow velocity in the descending thoracic aorta. CO is calculated by the product of this velocity and the aortic diameter. A correcting factor (the so- called factor K) is applied to take into account the dis- tribution of CO between the vessels originating from the aortic arch and the remainder which constitutes flow in the descending aorta [12]. ODM I uses a nomo- gram based on the patient's age, height, and weight to determine aortic diameter and factor K [12]. These as- sumptions could lead to pitfalls in CO measurement [12]. It was shown that ED is not reliable during aortic surgery because the CO distribution between supra-aor-
351
tic vessels and descending aorta changes during the clamping period [21, 24]. Alterations in CO distribution could partly explain discrepancies between ED and TD in our study, especially in patients with atherosclerosis and during pharmacological intervention. Moreover, the accurate measurement of blood flow by ED requires that the esophageal probe is strictly parallel to the aorta. We did not check the probe position with radiography. However, in 4 patients, abdominal tomodensitometry was required. It was performed with the esophageal probe that was always seen near and parallel to the de- scending thoracic aorta. However, the misplacement of the esophageal probe can explain why the continuous monitoring of CO could not be always achieved. In this case, probe repositioning was justified. The need for re- positioning was recently confirmed by Krishnamurthy et al. [29]. In this study, the availability of a trained oper- ator to reposition the probe improved the accuracy of ED to measure CO in patients undergoing coronary ar- tery revascularization. Finally, aortic diameter assess- ment using a nomogram can be debated. However, pre- vious studies using echography to measure aortic diame- ter did not report better results than those using the no- mogram [17-22]. Moreover, echography can under- and overestimate aortic diameter when the ultrasound beam is not strictly perpendicular to the aorta or does not pass in the center of the aorta.
Few studies have assessed the influence of training on CO measurement methods. Rather they have con- cerned pulmonary artery catheterization and echocardi- ography [4, 8, 9]. Iberti et al. showed that knowledge, understanding of the use of the pulmonary artery cathe- ter, and interpretation of its data increased with train- ing, frequency of use, frequency of insertion, and the op- erator's qualification [4]. These authors suggested that more than ten pulmonary artery catheter insertions are required to attain competence and more than five inser- tions per year to maintain mastery in pulmonary artery catheter use and data interpretation. These recommen- dations were always complied with by the operators who measured CO by TD in the present study.
Concerning echocardiography, this technique can help to diagnose valvular and aortic diseases and allows the assessment of left ventricular preload, afterload, and cardiac function in the intensive care unit [7, 30-32]. However, echocardiography requires a long pe- riod of training (about 6 months), first with the trans- thoracic approach and then with the transesophageal approach [8, 9]. Moreover, the equipment needed is ex- pensive.
The influence of training on the reliability of ED was considered by Freund [18]. Comparing ED with TD dur- ing general anesthesia in 23 patients, Freund showed that the correlation coefficient increased in the last 13 patients. In the present study, a training period in- cluding 12 patients was required for the same operator
to obtain a good Doppler audio signal. This period is much shorter than the training period required for echo- cardiography, and significantly improved the reliability and accuracy of ED. However, the level of CO values during the training period was higher than those mea- sured during the evaluation period. This difference could partly explain the improvement in ED accuracy because of the heteroscedasticity. Nevertheless, the pre- sent study prospectively consecutive patients and pa- tients could not be chosen according to their CO level. If the duration of training could be shown to be the same for several operators, the present study suggests that a training period of about 12 patients improves the reliability of ED to measure CO.
ED allowed rapid CO measurement and continuous CO monitoring in 77 % of patients. The impact of ED on decision making could be questioned. For instance, Steingrub et al. [5] demonstrated that the knowledge of pulmonary artery catheter data leads to alteration in the therapeutic decision in about 50 % of patients. As echocardiography can assess left ventricular preload, af- terload, and cardiac function, and diagnose valvular and aortic diseases [7, 30, 31], it probably influences decision making. In contrast, no study concerning the influence of ED on decision making has been reported. Moreover, ED does not measure loading pressure like pulmonary capillary wedge pressure. ED can measure parameters related to the shape of the blood flow velocity wave- form, such as mean acceleration, peak velocity, and cor- rected flow time [11]. These parameters allow the evalu- ation of hyper- or hypocontractile and hypo- or hyper- dynamic states [11, 12]. Therefore, ED may also have therapeutic implications. However, no study concerning the influences of ED on decision making has been re- ported and further clinical trials are needed in this re- gard. Finally, because ED requires an esophageal probe, it cannot easily be used in awake patients.
In conclusion, ED is reliable in critically ill patients to measure CO. A short period of training improves its reliability. However, because no study has been per- formed to assess its impact on decision making about management of patients, its place among the methods of CO measurement remains to be specified.
352
References
1. American Society of Anesthesiologists Task Force on Pulmonary Artery Cath- eter (1993) Practice guidelines for pul- monary artery catheterization. Anes- thesiology 78:380-394
2. Forrester JS, Ganz W, Diamond G, MacHigh T, Chonette DW, Swan HJC (1972) Thermodilution cardiac output determination with a single flow direct- ed catheter. Am Heart J 83:306-311
3. Shah KB, Rao TLK, Laughlin S, E1-Etr AA (1984) A review of pulmonary ar- tery catheterization in 6,245 patients. Anesthesiology 61:271-275
4. Iberti TJ, Fisher ER Leibowitz AB, Pa- nacek EA, Silverstein JH, Alberston TE and The Pulmonary Artery Cathe- ter Group (1992) A multicenter study of physicians' knowledge of the pulmo- nary artery catheter JAMA 264: 2928-2932
5. Steingrub JS, Celoria G, Vickers-Lahti M, Teres O, Bria W (1991) Therapeutic impact of pulmonary artery catheteriza- tion in a medical/surgical ICU. Chest 99:1451-1455
6. Darmon PL, Hillel A, Mogtader A, Mindich D, Thys D (1994) Cardiac out- put by transesophageal echocardiogra- phy using continuous-wave Doppler across the aortic valve. Anesthesiology 80:796-805
7. Gorcsan J (1995) Quantification of left ventricular function using transesopha- geal echocardiography. In: Vincent JL (ed) Yearbook of intensive care and emergency medicine. Springer, Brus- sels, pp 575-592
8. Cahalan MK, Foster E (1995) Training in transesophageal echocardiography: in the lab or on the job? Anesth Analg 81:217-218
9. Savage RM, Licina MG, Koch CG, Hearn CJ, Thomas JD, Start N J, Stew- art WJ (1995) Educational program for intraoperative transesophageal echo- cardiography. Anesth Analg 81: 399-403
10. Hunstman LL, Stewart DK, Barnes SR, Franklin SB, Colocousis JS, Hessel EA (1983) Non invasive Doppler determi- nation of cardiac output in man. Clini- cal validation. Circulation 67:593-602
11. Singer M, Clarke J, Bennett D (1989) Continuous hemodynamic monitoring by esophageal Doppler. Crit Care Med 17:447-452
12. Singer M (1993) Esophageal Doppler monitoring of aortic blood flow: beat- by-beat CO monitoring. Int Anesthesi- ol Clin 31:99-125
13. Kristensen BD, Goldbert SJ (1987) Number of CO required to accurately determine mean velocity of blood flow in the ascending aorta and pulmonary trunk. Am J Cardio160:746-747
14. Lamantia K, O'Connor T, Barash PG (1990) Comparing methods of measure- ment: an alternative approach. Anes- thesiology 72:781-783
15. Siegel LC, Pearl RG (1992) Noninva- sive cardiac output measurement: trou- bled technologies and troubled studies. Anesth Analg 74:790-792
16. Bland JM, Altman DG (1986) Statisti- cal methods for assessing agreement be- tween two methods of clinical measure- ment. Lancet I: 307-310
17. Mark JB, Steinbrook RA, Gugino LD, Maddi R, Hartwell B, Shemin R, Disesa V, Rida WN (1986) Continuous nonin- vasive monitoring of cardiac output with esophageal Doppler ultrasound during cardiac surgery. Anesth Analg 65:1013-1020
18. Freund PR (1987) Transesophageal Doppler scanning versus thermodilu- tion during general anesthesia. An ini- tial comparison of cardiac output tech- niques. Am J Surg 153:490-494
20. Spahn DR, Schmid ER, Tornic M, Jenni R, yon Segesser L, Turina M, Baetscher A (1990) Noninvasive versus invasive assessment of cardiac out put after car- diac surgery: clinical validation. J Car- diothorac Vasc Anesth 4:46-59
22. Schmid ER, Spahn DR, Tornic M (1993) Reliability of a new generation transesophageat Doppler device for car- diac output monitoring. Anesth Analg 77:971-979
23. Lavandier B, Cathignol D, Muchada R, Bui Xuan B, Motin J (1985) Noninva- sire aortic blood flow measurement us- ing an intraesophageal probe. Ultra- sound Med Biol 11:451-460
24. Klotz KE Klingsiek S, Singer M, Wenk H, Eleftheriadis S, Kuppe H, Schmucker P (1995) Continuous mea- surement of cardiac output during aor- tic cross-clamping by the esophageal Doppler monitor ODM 1. Br J Anaesth 74:655-660
25. Nishikawa T, Dohi S (1993) Errors in the measurement of cardiac output by thermodilution. Can J Anaesth 40: 142-153
26. Renner LE, Morton MJ, Sakuma GY (1993) Indicator amount, temperature, and intrinsic cardiac output affect ther- modilution cardiac output accuracy and reproducibility. Crit Care Med 21: 586-597
27. Lefrant JY, Bruelle R Ripart J, Ibanez F, Aya G, Peray R Saissi G, de La Cous- saye JE, Eledjam JJ (1995) Cardiac out- put measurement in critically ill pa- tients: comparison of continuous and conventional thermodilution tech- niques. Can J Anaesth 42:972-976
28. Stetz CW, Miller RG, Kelly GE, Raffin TA (1982) Reliability of the thermodi- lution method in the determination of cardiac output in clinical practice. Am J Respir Crit Care Med 126:1001-1004
29. Krishnamurthy B, McMurray TJ, McClean E (1997) The peri-operative use of the esophageal Doppler monitor in patients undergoing coronary artery revascularisation. Anaesthesia 52: 624-629
30. Smith MD, Cassidy MJ. Souther S, Morris EJ, Sapin PM, Johnson SB, Kearney PA (1995) Transesophageal echocardiography in the diagnosis of traumatic rupture of the aorta. N Engl J Med 332:356-362
31. Daniel WG, Mtigge A (1995) Trans- esophageal echocardiography. N Engl J Med 332:1268-1279
32. Practice guidelines for perioperative transesophageal echocardiography (1996) A report by the American Soci- ety of Anesthesiologists and the Society of Cardiovascular Anesthesiologists task force on transesophageal echocar- diography. Anesthesiology 84:986-1006
