INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The management of severe cardiovascular disorders is greatly facilitated when the patient's hemodynamic status can be quantitatively assessed. The cardiac output is one of the key parameters used by physicians to determine a therapeutic strategy and evaluate the effects of treatment. The thermodilution technique is extensively used for measurement of cardiac output in intensive care units (ICUs), regardless of the many pitfalls that may affect its precision. Despite concerns about its accuracy, thermodilution is widely accepted for clinical purposes and often regarded as the gold standard by many intensivists (1). Recently, the use of pulmonary artery catheters for invasive hemodynamic monitoring has been increasingly criticized on account of its unclear risk/benefit ratio (2, 3). Hence, a new technique that could provide a fair estimation of cardiac output in a less invasive and less expensive way is clearly desirable.

Among the proposed alternatives, ultrasound-based technologies are the most promising. By combining Doppler and echocardiographic measurements, instantaneous cardiac output can be obtained noninvasively by a transthoracic or transesophageal approach (4, 5). However, the procedure requires a trained operator, is often technically difficult in mechanically ventilated patients, and does not provide continuous monitoring. To overcome these limitations, transesophageal continuous-wave Doppler of the descending aorta has been proposed to monitor instantaneous cardiac output in anesthetized or critically ill patients (6). The technique, initially introduced by Daigle and colleagues (11), measures blood flow velocity in the descending thoracic aorta, using a transducer that can be inserted easily in sedated patients and left in place for several days. Stroke volume may then be derived using an algorithm based on: (1) the beat-to-beat maximum velocity-time integral (or stroke distance); (2) the cross-sectional area of the descending aorta; and (3) a correcting factor which transforms descending aortic blood flow into global cardiac output. The validity of this approach in mechanically ventilated patients is not fully established, and clinical use of transesophageal Doppler for continuous monitoring of cardiac output is not yet widely accepted.

The goal of the present study was to assess the validity of transesophageal Doppler-derived cardiac output measurement in mechanically ventilated, critically ill patients. A multicenter protocol, associating three different intensive care units, was designed to compare simultaneous cardiac output values obtained using transesophageal Doppler and thermodilution techniques. In one center, cardiac output was also obtained simultaneously using two other techniques: (1) suprasternal Doppler associated with two-dimensional echocardiography; and (2) indirect calorimetry using the Fick principle.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A total of forty-six consecutive critically ill patients were enrolled from one medical and two surgical intensive care units over a period of 6 mo. The protocol was approved by the local Ethics Committees and written informed consent was obtained from a patient's relative in each case. Patients were eligible if already mechanically ventilated, sedated, and monitored with a pulmonary artery catheter. Those with suspected or documented esophageal lesions (history of dysphagia or previous esophageal surgery) or aortic dissection were excluded. Measurements did not interfere with medical or nursing procedures, and the pattern of mechanical ventilation was not modified for the study.

Cardiac Output Measurements

Simultaneous measurements of cardiac output were obtained by thermodilution using the pulmonary artery catheter and transesophageal Doppler. Additional measurements of cardiac output were also attempted using suprasternal Doppler and indirect calorimetry in 24 patients from one center. In each case investigators were blinded to the values obtained using other techniques.

Transesophageal Doppler. The device consists of a continuous-wave Doppler transducer (4 MHz) mounted at 45° at the tip of a transesophageal probe (diameter = 8 mm). The probe is connected to a monitor (DOPTEK-ODM1; Deltex, Chichester, UK) displaying the blood flow velocity profile after spectral analysis of the reflected Doppler-shift signal (Fast Fourier Transformation). Following oral introduction, the probe was advanced gently until its tip was located in the mid-esophagus, approximately 35 cm from the incisors, and then rotated so that the transducer faced posteriorly and a characteristic aortic blood flow signal was obtained. Probe position was optimized to record peak velocity by slow rotation in the long axis and alteration of the depth of insertion to generate a clear signal. Gain setting was adjusted to obtain the best outline of the aortic velocity waveform and a 300-Hz high-pass filter eliminated the noise related to low-frequency vessel wall motion. Prior to each measurement, probe position was verified to ensure optimal acquisition of the maximal velocity signal. Stroke volume (SV, ml) was calculated as:

(1)

where V_Ao(t) represents instantaneous maximum aortic velocity, T is the cardiac ejection time (the integral of instantaneous maximum velocity during cardiac ejection representing the stroke distance), CSA_Ao is the cross-sectional area of the descending thoracic aorta (cm²), and K is a correcting factor (= 1.43) whose purpose is to transform the blood flow measured in the descending thoracic aorta into global cardiac output, assuming that a constant fraction (70%) of the total blood flow passes through the descending aorta. CSA_Ao is estimated from a nomogram based on the patient's age, weight, and height. The monitor was preset to calculate cardiac output (CO_TED, L/min) by averaging stroke volume over 10 beats and multiplying the value obtained by the heart rate.

Thermodilution. Measurements of cardiac output using thermodilution (CO_TH) were performed using a 7-Fr balloon-tipped pulmonary artery catheter connected to a cardiac output monitor (Abbott, Mountain View, CA). Injections of iced 5% glucose (10 ml) were made throughout the respiratory cycle and CO_TH was obtained by averaging the results of at least five measurements (12). Variability of measurements was assumed to be mainly secondary to positive pressure ventilation (12, 13) and was not a criteria for rejection as long as the thermodilution curve was satisfactory (14, 15).

Suprasternal Doppler combined with 2D-echocardiography. Noninvasive estimates of instantaneous aortic flow were generated from pulsed-wave Doppler velocity recordings and echocardiographic measurements of aortic annular area (16). Ascending aortic blood velocity was recorded with a 2 MHz pulsed-wave Pedof Doppler transducer (Vingmed SD-50; Horten, Norway) positioned at the suprasternal notch. The aortic annular diameter (D_ao, cm) was measured from parasternal long-axis 2D echocardiographic recordings obtained with a 2.5-MHz transducer. Aortic annular area (AAA, cm²) was calculated, assuming a circular orifice, as:

(2)

Stroke volume was calculated by multiplying AAA by the Doppler velocity-time integral of the ascending aortic blood flow during systole. Five consecutive beats were averaged to allow for respiratory variation, and the value obtained was multiplied by heart rate to calculate cardiac output (CO_SSD) (4, 16, 17).

Indirect calorimetry using the Fick principle. Oxygen uptake (O₂) and carbon dioxide production were measured using a gas exchange monitor (Deltatrac, Datex, Helsinki, Finland). Inspiratory and expiratory oxygen concentration were determined using a fast response paramagnetic oxygen analyzer (OM-101; Datex). Before each study, the gas sensor was calibrated after a warm-up period of at least 30 min with air and a gas mixture containing 95% oxygen and 5% CO₂. This method has been validated in critically ill patients ventilated with a fraction of inspired oxygen (FI_O2) of less than 0.60 (18). Arteriovenous difference (AVD) was calculated as the difference between arterial and mixed venous oxygen contents and cardiac output (CO_IC) as (19):

(3)

Statistics

Correlation between techniques was determined using simple linear regression analysis with Bland and Altman representation (20) to explore the possibility of systematic errors. Significance of biases was tested using paired t test. In order to correct for the distribution of differences, data were logarithmically transformed to obtain new limits of agreement, as recommended by Bland and Altman (20). In all patients who underwent more than one measurement of cardiac output during the study, the variations between two time points in CO_TH and CO_TED were compared using two-way analysis of variance (ANOVA) for repeated measures. Correlation and agreement studies were also performed on variations. The reproducibility of each technique was assessed in six patients judged to be in stable hemodynamic condition. Intra-observer variability was calculated as the standard deviation of the differences between five measurements obtained over a short period of time divided by the mean cardiac output.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

One hundred thirty-six paired cardiac output measurements were made, using transesophageal Doppler and the thermodilution technique, in 46 mechanically ventilated, critically ill patients recruited from three different ICUs. Table 1 shows selected clinical characteristics, including information regarding inotropic support and positive end-expiratory pressure (PEEP). Esophageal intubation with the Doppler probe was easily accomplished in all patients, except three who required a direct laryngoscopy. In all cases an adequate signal was obtained within a few minutes, regardless of the presence of a gastric suction tube. The probe was left in place for up to 48 h (median: 36 h, quartile range: 27-42 h) without adverse effects.

Cardiac output ranged from 1.7 to 14.8 L/min when measured using transesophageal Doppler, and from 1.4 to 15.3 L/ min using the thermodilution technique. Figure 1 demonstrates (1) the close correlation between CO_TED and CO_TH (correlation coefficients = 0.95; p < 0.0001), and (2) the agreement between the two techniques. The mean difference between the paired values (CO_TH  CO_TED), representing the bias of transesophageal Doppler with respect to thermodilution, was 0.24 L/min (p < 0.003), while the limits of agreement (bias ± 2 SD) were 2.04 and 1.56 L/min. The scatter of differences tended to increase with the increase in mean cardiac output (Figure 1B). We therefore performed a logarithmic transformation of these data and obtained a new bias of 0.03, and new agreement limits of 0.29 and 0.23. The antilogs of these new limits were 1.33 and 0.79, respectively, indicating that CO_TED may differ from CO_TH within a range of +33% to 21%. Twenty-three patients underwent more than one measurement of their cardiac output during the study, allowing us to compare the variations between two consecutive cardiac output values as measured with the two methods in 88 instances. The median delay between the two cardiac output determinations was 1.5 h (quartile range: 0.5-6.4 h). The two-way ANOVA for repeated measures showed no statistical difference between the variation measured by thermodilution and their paired counterparts measured using transesophageal Doppler. The absence of significant interaction confirmed a similar direction of change using both techniques. Variations in cardiac output measured using thermodilution (CO_TH) correlated very well with those measured using transesophageal Doppler (CO_TED), as shown in Figure 2A (r = 0.90, p < 0.0001). The agreement between these variations was characterized by the absence of systematic bias (0 L/min) and limits of agreement of 1.7 and 1.7 L/min (Figure 2B). In the subgroup of patients ventilated with PEEP levels higher than 10 cm of water (n = 10), the two methods remained closely correlated (n = 35; r = 0.95; p < 0.0001) with a bias of 0.2 L/min (p < 0.01), and limits of agreement of 2.1 and 1.7 L/min.

View larger version (20K): [in this window] [in a new window]   Figure 1.   (A) Scatterplot of paired cardiac output measurements obtained using thermodilution (COTH) and transesophageal Doppler (COTED), with the solid line representing linear regression. (B) Bland and Altman (20) representation of the agreement between the two techniques. The solid line represents the mean difference between COTH and COTED (systematic bias), and the dotted lines define the limits of agreement (95% confidence interval).

View larger version (18K): [in this window] [in a new window]   Figure 2.   (A) Scatterplot of paired variations in cardiac output obtained using thermodilution (COTH) and transesophageal Doppler (COTED), with the solid line representing linear regression. Dotted lines at zero intercept delineate four quadrants: upper left and lower right quadrants contain the points corresponding to opposite variations in COTH and COTED. (B) Bland and Altman (20) representation of the agreement between variations measured using the two techniques. The solid line represents the mean difference between COTH and COTED (systematic bias), and the dotted lines define the limits of agreement (95% confidence interval).

Suprasternal Doppler assessment was attempted in 24 patients but yielded a satisfactory signal in only 17 (71%). Fifty-three cardiac output measurements were obtained simultaneously using suprasternal Doppler, transesophageal Doppler, and the thermodilution technique. The correlation and agreement between CO_SSD and CO_TED are shown in Figure 3.

View larger version (18K): [in this window] [in a new window]   Figure 3.   (A) Scatterplot of paired cardiac output measurements obtained using suprasternal Doppler (COSSD) and transesophageal Doppler (COTED), with the solid line representing linear regression. (B) Bland and Altman (20) representation of the agreement between the two techniques. The solid line represents the mean difference between COSSD and COTED (systematic bias), and the dotted lines define the limits of agreement (95% confidence interval).

The requirements for reliable oxygen consumption measurement using indirect calorimetry were met in 13 of the same 24 patients (54%). The 46 cardiac output measurements obtained using this technique also correlated well with those obtained using transesophageal Doppler (Figure 4A), although the agreement between the two techniques was not as close as with the two previous methods (Figure 4B).

View larger version (18K): [in this window] [in a new window]   Figure 4.   (A) Scatterplot of paired cardiac output measurements obtained using indirect calorimetry (Fick principle, COIC) and transesophageal Doppler (COTED), with the solid line representing linear regression. (B) Bland and Altman (20) representation of the agreement between the two techniques. The solid line represents the mean difference between COIC and COTED (systematic bias), and the dotted lines define the limits of agreement (95% confidence interval).

The results of linear regression and agreement analyses between the four techniques are presented in Table 2.

Variability in cardiac output measurement was 8% for CO_TED, 12% for CO_TH, and 9% for CO_SSD. No reproducibility study could be performed using indirect calorimetry due to the prolonged equilibration time required for each cardiac output determination.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Transesophageal Doppler is a new minimally invasive method for the continuous monitoring of cardiac output in critically ill patients. In this multicenter trial, the cardiac output values obtained using this technique correlated well with those obtained using thermodilution. The systematic underestimation of CO_TED with respect to CO_TH was small (< 0.25 L/min), and the limits of agreement were 0.25 ± 1.8 L/min. More importantly, variations in cardiac output with time were comparable using both techniques. Other measures of cardiac output using suprasternal Doppler or the Fick principle also demonstrated excellent correlation and good agreement with transesophageal Doppler. We conclude that transesophageal Doppler can provide a clinically useful estimate of cardiac output and reliably detect its variations in mechanically ventilated, critically ill patients.

Methodologic Considerations

The estimation of cardiac output using transesophageal Doppler is based on four assumptions: (1) an accurate measurement of the velocity of descending aortic blood flow; (2) a "flat" velocity profile throughout the aorta; (3) an estimated aortic cross-sectional area close to the mean value during systole; and (4) a constant division of blood flow between the descending aorta (70%) and the brachiocephalic and coronary arteries (30%). The accuracy of velocity measurement requires a good alignment between the Doppler beam and blood flow and knowledge of the angle at which the blood flow is insonated. Alignment is best assessed subjectively by optimizing the quality of the obtained signal with the aid of the visual display of instantaneous velocity waveform and the Doppler sound. The angle between the Doppler beam and blood flow is roughly the same as that between the transducer and the probe (45°), since the esophagus and aorta are parallel in the thorax. However, even a 10% change in the alignment of the Doppler beam with respect to blood flow would result in less than 5% error in the final velocity measurement since the principal determinant is the cosine of the angle (6). A "flat" velocity profile implies that all red blood cells move at the same speed through the vessel. In fact, the flow velocity profile in the descending thoracic aorta is rather parabolic than flat (21), i.e., the red blood cells at the center of the vessel move faster than those at the periphery. Hence, use of the maximum velocity envelope to compute stroke distance may result in overestimation of stroke volume. Bedside measurement of the cross-sectional area of the descending aorta can only be performed by using transesophageal echocardiography. This technique is not available everywhere, and the manufacturers of the transesophageal Doppler probe have incorporated a nomogram to estimate the cross-sectional area of the descending aorta based on the patient's age, weight, and height. Systematic errors due to a discrepancy between the actual area and the nomogram value would not affect the trend of cardiac output variation with time (7). Finally, the division of blood flow between cephalic and caudal territories may also vary according to hemodynamic conditions, reflex activation, or metabolic activity within different organs. Therefore, the assigned constant ratio of 70:30% may become inaccurate under a variety of pathophysiologic conditions (7, 22). To overcome the uncertainty surrounding this arbitrary ratio, previous studies have incorporated calibration of the descending aortic velocity with respect to simultaneous suprasternal Doppler measurements of cardiac output (7, 8, 10, 23). In fact, this calibration process was time-consuming, required a trained operator for suprasternal Doppler assessment, was a source of error, and could not account for possible secondary variations in the flow division between supra-aortic vessels and the descending aorta.

All techniques for cardiac output measurement currently available in clinical practice have well-documented limitations. The thermodilution technique has been extensively scrutinized and shown to be affected by a variety of external factors (15, 19, 24) as well as patient-related conditions (i.e., tricuspid regurgitation, intracardiac shunts, and low cardiac output states) (27, 28). Indirect calorimetry is unreliable in patients with an (FI_O2) greater than 0.6 (18), and cardiac output estimation using the Fick principle may be adversely affected by the lack of hemodynamic stability frequently encountered in mechanically ventilated, critically ill patients (19, 24). A significant proportion of patients was unsuitable for cardiac output measurement using suprasternal Doppler, due to a poor ultrasound window. In addition, although its accuracy is excellent, this approach is time-consuming and can only be performed by a trained operator.

Present Findings

Despite the abovementioned methodologic limitations, the correlation between cardiac output obtained using the thermodilution technique and transesophageal Doppler was good. Single-center studies comparing thermodilution and transesophageal Doppler in anesthetized patients have reported a wide range of correlation coefficients (r) between the two techniques. Poor correlation (r = 0.52) was observed in a study where repeated measurements were performed without readjusting the probe position after initial placement (23). However, the majority of authors reposition their probes, usually by a small rotation, to recover the optimal signal and obtain a better correlation (r = 0.90-0.98) (6, 7, 10). Our bias analysis demonstrated that the mean difference between the two techniques was small (0.24 L/min), indicating a minimal systematic underestimation of cardiac output by transesophageal Doppler with respect to thermodilution. The limits of agreement between the two techniques, 1.8 L/min above and below the bias, were in the range of previous observations (10, 22). The relatively large scatter of differences between the two techniques reflects the lack of precision of both methods of measurement. Indeed, the 95% confidence interval is highly dependent on the repeatability and accuracy of the techniques compared. According to Bland and Altman, a new method that is perfect will not agree with a "reference" method, which is the most variable (20). We noted a greater variability of CO_TH than CO_TED (12% versus 8%), consistent with previous reports (7, 9). A meta-analysis evaluating the reproducibility of the thermodilution technique suggested that differences between measurements greater than 15% were necessary to consider a change significant (1). Therefore, although isolated CO_TED values should be interpreted with caution, our findings indicate that transesophageal Doppler is adequate for differentiating clinically important ranges of cardiac output. More importantly, we found that variations in cardiac output with time were similar regardless of the measurement technique employed. Variations were in the same direction and the agreement demonstrated the absence of systematic bias (0 L/ min) and a spread of ±1.7 L/min in CO_TH  CO_TED, similar to that observed in CO_TH  CO_TED. This confirms that the trend in cardiac output variation evaluated using transesophageal Doppler is comparable to that measured with thermodilution, as already suggested by other investigators (7).

To further establish the validity of transesophageal Doppler, we obtained simultaneous cardiac output measurements using suprasternal Doppler and indirect calorimetry using the Fick principle. These techniques were only available at one center and could not be applied successfully in every patient due to the technical limitations mentioned earlier. Cardiac output derived using both techniques correlated well with the CO_TED values. Interestingly, the best correlation and agreement involving transesophageal Doppler was observed with suprasternal Doppler (Figure 3), indicating that the variability related to factors other than Doppler velocimetry itself was relatively small. Among those factors, the algorithm used to compute stroke volume appears to generate very few inaccuracies. Hence, the assumptions concerning the precision of the velocity measurement, the flow velocity profile, and the cross-sectional area of the descending aorta as well as the division of flow between supra-aortic vessels and the descending aorta appear valid.

The best correlation and agreement was observed between CO_TH and CO_SSD (Table 2). All other techniques agreed similarly, indicating that all methods were roughly of equal merit.

Transesophageal Doppler for Cardiac Output Monitoring

Transesophageal Doppler is a simple technique, and most users acknowledge that it is fairly easy to achieve adequate probe positioning and obtain reproducible results (9, 29). Freund (8) noted a dramatic improvement in the skills of untrained operators after performing only 10 measurements. However, in the present study only trained operators performed each type of measurement. Interobserver variability has previously been shown to be less than 10% using transesophageal Doppler (6). We found intra-observer variability to be slightly less with transesophageal Doppler (8%) than with thermodilution (12%), confirming previous observations (7, 9).

Potential side effects of transesophageal cardiac output monitoring include esophageal damage. Some authors have left the probe in place for over 2 wk without adverse effects (29), but there has been no systematic fibroscopic evaluation of esophageal mucosal injury at the time of probe removal.

A major advantage of transesophageal Doppler over other currently available techniques is its ability to provide continuous, real-time monitoring. In contrast to "continuous" cardiac output monitoring using a pulmonary artery catheter, which only provides repeated measurements of mean cardiac output, transesophageal Doppler displays the instantaneous aortic velocity spectrum. Probe displacement can occur during prolonged monitoring as a result of various causes (e.g., nursing procedures, deglutition, gravity) and results in a poorly defined velocity envelope. It is mandatory to ensure an adequate signal prior to interpreting Doppler-derived data. Failure to reposition the probe prior to each measurement may lead to grossly erroneous cardiac output values and poor agreement with other techniques (23). The monitor can be set to calculate cardiac output either on a beat-to-beat basis or after averaging stroke volume over a number of consecutive beats (maximum 20), according to the operator's choice. In addition, the device derives parameters from the shape of the velocity envelope, such as peak velocity, mean acceleration of blood flow, and the rate-corrected flow time during systole. Some insight into volemic status, peripheral resistance, and left ventricular function can be obtained from these parameters in the ascending aorta (30, 31). Although additional studies are needed to establish the validity of these indices when obtained in the descending thoracic aorta using transesophageal Doppler, several reports support their usefulness for optimizing therapeutic strategies in critically ill patients (32). In a recent randomized trial, Sinclair and colleagues (35) showed that patients with femoral neck fracture undergoing intravascular colloid resuscitation guided by transesophageal Doppler during surgical repair had a significantly reduced hospital stay (39%) compared with control subjects (35).

In conclusion, although transesophageal Doppler cannot replace the pulmonary artery catheter, it offers an instantaneous, reasonably accurate, and noninvasive estimate of cardiac output. In addition, the variations in cardiac output occurring spontaneously or in response to treatment are tracked similarly by the two techniques. Transesophageal Doppler may prove clinically useful in a large number of critically ill patients in whom hemodynamic monitoring is desirable without the risks associated with right heart catheterization.