A Comparison of Left Ventricular Performance Indices Measured by Transesophageal Echocardiography with Automated Border Detection 

Written, informed consent was obtained according to the guidelines of the institutional review board from 23 patients scheduled for elective coronary artery bypass surgery. In addition to routine noninvasive monitoring, a femoral artery catheter and a pulmonary artery catheter that could measure thermodilution cardiac output and mixed venous oxygen saturation were inserted in all patients. Anesthesia was induced with 15 to 20 [micro sign]g/kg fentanyl, 0.12 to 0.15 mg/kg pancuronium, and 3 or 4 mg midazolam administered intravenously for 5 - 8 min while the patients received 100% oxygen. Anesthesia was maintained with additional intravenous doses of fentanyl, diazepam, and pancuronium or vecuronium, as necessary. Isoflurane or halothane was used if necessary to enhance the depth of anesthesia in doses up to 1 end-tidal minimum alveolar concentration (MAC), as determined by continuous mass - spectrometry.

After anesthesia induction, transesophageal echocardiography (TEE) was performed with a 5-MHz Omniplane probe interfaced to a Sonos OR phased-array echoscanner (Hewlett-Packard, Andover, MA). Every patient underwent a complete TEE examination before and after cardiopulmonary bypass (CPB). To determine ventricular performance, LV cavity dimensions were obtained in the short-axis, mid-papillary plane using automated endocardial border detection by the Acoustic Quantification (AQ) method of the Sonos OR scanner. To achieve satisfactory endocardial border tracking throughout the cardiac cycle, image gain was adjusted manually in selected image sectors according to manufacturer recommendations. Scanner transmit gain was set at maximum, and time-to-gain and lateral gain controls were subsequently adjusted to track the endocardial border. Left ventricular cavity area was determined continuously and automatically within a region of interest designated manually on the video screen using a pointing cursor. The cavity area signal was sent continuously by the scanner to a dedicated personal computer work station (Bravo 3/25 s; AST, Irvine, CA), where the signals were digitized, displayed, and stored.

Near the end of CPB, a flexible 60-cm, 16-gauge catheter (I-Cath; CharterMed, Lakewood, NJ) was introduced in the LV cavity through the right superior pulmonary vein. The position of the catheter tip was confirmed by analyzing the pressure tracing and by echocardiographic imaging. Left ventricular pressure measurements were performed using a rigid 213-cm, 0.17-cm-diameter pressure tubing filled with a heparinized saline solution and connected to a Sorenson disposable, strain gauge transducer (Transpac IV; Abbott Critical Care Systems, North Chicago, IL). Pressure tracings were displayed on a SpaceLabs monitor (model 90305; Redmond, WA) and on a larger slave monitor. The analog LV and arterial pressure signals were sent to the personal computer at a gain of 1 V/100 mmHg of pressure along with the electrocardiograph signal.

The analog LV area, the electrocardiograph, and the pressure signals were digitized using commercial hardware and software (DI-430, AT-CODAS; DATAQ Instruments, Akron, OH) and input into the personal computer. Signals were acquired and stored at a rate of 225 Hz as they were displayed. The dynamic response characteristics of the pressure measurement system induced a time delay in the pressure signal that almost exactly matched the precessing time for the acoustic quantification signal ([tilde operator] 30 ms). [12]Therefore, no additional time delays were necessary to obtain adequate pressure-area loops, and attempts to introduce delays in either direction did not improve the quality of the loops.

All LV performance indices were based on short-axis, mid-papillary chamber area measurements. The measurement protocols were initiated after stable weaning from CPB. All pressure signals were calibrated and set at zero at the mid-axillary level. Thermodilution cardiac outputs were performed in triplicate, and the values were recorded manually. Patient ventilation was interrupted temporarily. While arterial and LV pressures, electric activity of the heart, and acoustic quantification - LV cavity area were recorded continuously, LV preload was reduced by partial occlusion of the inferior vena cava using gentle traction from a snare made of 0.5-cm cotton tape. Before the partial vena caval occlusion was released, 10 - 15 cardiac cycles were recorded during apnea (Figure 1). Ventilation was resumed within 30 - 45 s. To obtain satisfactory data, the measurements were acquired one to four times, with a 3-min interval between subsequent measurements to allow for restoration of ventilation and hemodynamics. A preload reduction run was considered satisfactory if at least six consecutive cardiac cycles were obtained in sinus rhythm with a gradual decrease in both EDA and end-systolic area. After the baseline LV performance measurements were completed, halothane or isoflurane was administered to 11 patients until a 1-MAC, stable end-tidal expired gas concentration was obtained. At that point, the data collection and inferior vena cava occlusion were repeated.

Fractional area change was calculated automatically by the echoscanner. For each measurement period, the values associated with the three to five cardiac cycles that immediately preceded inferior vena cava occlusion were recorded and averaged.

Circumferential fiber shortening, with heart rate correction, was calculated using the Equation describedby Kikura et al. [13]:

Vcf^c=(EDC - ESC)/(EDC [middle dot] ET)[middle dot](RR)(^1/2)

where EDC and ESC are the end-diastolic and end-systolic circumferences, ET is the ejection time, and RR is the R-to-R interval on the electrocardiogram. Ejection time was defined as the time between end-diastole and end-systole on the ABD area signal, whereas end-diastolic circumference and end-systolic circumference were calculated from EDA and end-systolic area with the equation:

circumference = 2 [middle dot](pi [middle dot] area)^1/2.

The same cardiac cycles were used to determine FAC and Vcf^c.

End-systolic elastance was determined from pressure-area loops using the personal computer (Figure 2). For each loop, the end-systolic point was identified by an automated mathematical procedure conducted in a Lotus 123 spreadsheet (Lotus Development Corp., Cambridge, MA). The first step consisted of determining the ratio of the largest area - smallest area to the largest pressure - smallest pressure for the loop (typically these measurements were not matched in time). Subsequently, all pressure values were normalized by multiplication with this ratio to avoid pressure bias in the subsequent equation. Finally, a “sort” operation was performed to identify the maximum value of the sum equation “normalized” LV pressure + LV area applied to each paired pressure and area data set. The sort identified the data point associated with the highest pressure and the smallest area. For each cardiac cycle, this point was consistently located in the upper left corner of the pressure-area loop and was considered to be the end-systolic point. End-systolic elastance was the slope of the regression line fitted from the pressures and areas of the end-systolic points of consecutive cardiac cycles obtained during inferior vena cava occlusion.

The PRSF was determined from the linear regression of stroke force on EDA for the cardiac cycles that were recorded during preload reduction. To identify the end-diastolic points, an automated mathematical process similar to the one already described was used. The end-diastolic point was defined as the point associated with the lowest pressure and the largest area. Stroke force was the area enclosed by a pressure-area loop and was obtained by the algebraic integration of pressure as a function of area.

The dP/dt^max[middle dot] EDA^-1was determined from the linear regression of the maximum first derivative of LV pressure on the preceding EDA. The same cardiac cycles were used for Ees and for PRSF.

The three pressure - dimension performance indices were quantified as the slopes (i.e., the coefficients) from the linear regressions of end-systolic pressure on end-systolic area (Ees), stroke force on end-diastolic area (PRSF), and dP/dt^maxon end-diastolic area (dP/dt^max[middle dot]- EDA^-1). To gauge the strength of the association between pressure-related data and area, Pearson's product-moment correlation coefficients (r) were calculated. The variability of the indices also was assessed. Variability of the indices would indicate whether one index was measured with more precision than another. However, variability could not be compared directly because each index was the result of regressed raw scores that had different units of measure. Therefore, the individual measurements of each patient were first converted into z scores. These standardized scores represent the positions of the original measurements in relation to the mean of the distribution (using the standard deviation as the unit of measurement), and they allow scores from different normal distributions to be reported on a single, comparable basis. [14]The regressions of the z scores resulted in indices (correlation coefficients) with unitless standard errors that could be compared.

Pearson's product-moment correlations also were used to evaluate the association between the various indices of ventricular performance. Changes in the indices were tested for significance by paired Student's t tests, with P < 0.05 considered significant. All data are expressed as mean +/- SD.

Ventricular performance indices were measured successfully in 23 patients. Mean age was 58 +/- 6 yr, and the preoperative LV ejection fraction was 48 +/- 14%(by ventriculography or transthoracic echocardiography).

(Table 1) shows the hemodynamics obtained after CPB. The mean FAC was 31.1 +/- 7.9%(range, 17.4 - 45.3%), whereas Vcf^cwas 0.6 +/- 0.2 circ/s (range, 0.3 - 1.2 circ/s). Occlusion of the inferior vena cava resulted in significant changes in end-systolic LV pressure, end-systolic LV area, and end-diastolic LV area (Table 2). Mean ventricular contractility indices obtained during preload reduction were Ees: 25.8 +/- 11.6 mmHg [middle dot] cm^-2(range, 7.4 - 46.4 mmHg [middle dot] cm^-2); PRSF: 60.8 +/- 26.6 mmHg (range, 17.2 - 103.0 mmHg); and dP/dt^max[middle dot] EDA^-1: 245 +/- 123.4 mmHg [middle dot] s^-1[middle dot] cm^-2(range, 84.7 - 629.7 mmHg [middle dot] s^-1[middle dot] cm^-2). Correlation coefficients for the associations between end-systolic systolic pressure and end-systolic area, stroke force and end-diastolic area, and dP/dt^maxand end-diastolic area were 0.90 +/- 0.06, 0.93 +/- 0.05, and 0.91 +/- 0.07, respectively (all highly significant at P < 0.001). The mean standard errors (SE) of the regressions of the z scores were SE^Ees= 0.19, SE^PRSF= 0.17, and SE^dP/dtmax[middle dot] EDA^-1= 0.19, suggesting a similar precision for the three indices.

In two patients, continuous infusions of positive inotropes were begun at the end of CPB because of low arterial pressure. One patient received 0.01 [micro sign]g [middle dot] kg^-1[middle dot] min^-1norepinephrine, and the other received 5 [micro sign]g [middle dot] kg^-1[middle dot] min^-1dopamine. Both were retained in the analysis but did not receive a volatile anesthetic agent.

To compare the various ventricular performance indices, linear relations between FAC or Vcf^cand Ees, PRSF, or dP/dt^max[middle dot] EDA^-1were evaluated. No significant associations were observed between FAC and any of the pressure - dimension indices (FAC - Ees: r = 0.06, not significant [NS]; FAC - PRSF: r = 0.18, NS; FAC - dP/dt^max[middle dot] EDA^-1: r = 0.26, NS;Figure 3). Weak but significant associations were noted between Vcf^cand two of the pressure - dimension indices (Vcf (^c)- Ees: r = 0.65, P < 0.001; Vcf^c- PRSF: r = 0.71, P < 0.001; Vcf (^c)- dP/dt^max[middle dot]-EDA^-1: r = 0.16, NS). A highly significant association was found between Ees and PRSF (r = 0.88), but the associations of either of these variables with dP/dt^max[middle dot]-EDA^-1were not significant.

In the 11 patients to whom a volatile anesthetic agent was administered, systolic arterial pressure decreased significantly from 100 +/- 12 mmHg to 87 +/- 12 mmHg (P < 0.001;Table 3). There were no significant changes in heart rate, arterial diastolic pressure, or cardiac output. Fractional area change and dP/dt^max[middled dot] EDA^-1also remained unchanged (Table 4). A significant decrease in myocardial contractility was observed in Ees and PRSF. The decrease in PRSF was noted for each of 11 patients, whereas Ees decreased in 10 of 11 patients (Figure 4). After the volatile agent was administered, FAC increased in five patients, and Vcf^cand dP/dt^max[middle dot] EDA^-1increased in three patients.

This study found that conventional ventricular performance indices (FAC and Vcf^c) measured after CPB for myocardial revascularization showed weak or no association with pressure - dimension indices of contractility. In addition, reduction in myocardial contractility induced by the administration of volatile anesthetic agents could be detected by Ees and PRSF but not by FAC, Vcf^c, or dP/dt^max[middle dot] EDA^-1. Therefore, two of the three pressure - dimension indices of contractility, Ees and PRSF, appeared to be most sensitive for detecting changes in contractility. In the clinical environment, their measurement is, however, considerably more complex than the measurement of conventional indices of ventricular performance.

In previous human studies, the relation between ejection fraction by ABD and by other techniques has been shown to vary. Using transthoracic echocardiography and a long-axis view of the left ventricle, Chandra et al. [3]observed that acoustic quantification significantly overestimated radionuclide ejection fraction. Conversely, using TEE and a short-axis view of the left ventricle in critically ill patients, Liu et al. [2]found that acoustic quantification significantly underestimated the radionuclide ejection fraction.

In studies using short-axis views by TEE in patients having cardiac surgery, FAC by ABD underestimated FAC obtained by manual tracing of the echocardiographic images. [4,5]End-systolic area was systematically underestimated in one of these studies [5]and slightly overestimated in the other. [4]Both studies indicated that ABD slightly underestimated EDA.

In the current study, the values for Ees and PRSF were lower than those obtained by Gorcsan et al. [11]before CPB but higher than those measured after CPB by the same investigators. Because many factors may influence myocardial contractility immediately after CPB, it is not clear why our patients appeared to have better contractility after CPB. It is worth noting, however, that 60% of the patients studied by Gorcsan et al. required inotropic support after CPB compared with only 9% of patients in the current study.

When pressure-dimension indices of contractility were compared in conscious dogs, Little et al. [9]observed that the slopes of three indices, similar to those studied here, increased in response to an augmented contractile state. The slope of dP/dt^max[middle dot] EDV^-1(where EDV is end-diastolic volume) was the most sensitive but also displayed the greatest variability between determinations. The slope of preload recruitable stroke work was the most stable but least sensitive to changes in inotropic states. The authors concluded, however, that because of its stability, preload recruitable stroke work was the preferred index to evaluate serial performance.

In humans, the three indices were compared during cardiac catheterization in a study by Feneley et al. [10]They concluded that preload recruitable stroke work had the greater clinical utility, because it remained highly linear and was least sensitive to changes in afterload. An additional advantage in humans was that preload recruitable stroke work decreased more during preload reduction by inferior vena cava occlusion because stroke work is the product of stroke volume and ejection pressure. These same authors found that dP/dt^max[middle dot] EDV^-1could not be used as an index of contractility under the circumstances of their study. They attributed this to the narrow range over which dP/dt^maxis studied during inferior vena cava occlusion.

Our findings correspond with those of both studies. We found a close association between Ees and PRSF, but none between either of these indices and dP/dt^max[middle dot]-EDTA^-1. A significant serial change in contractility induced by volatile agents was measured by PRSF. This change was also detected by Ees but was not observed in the measurements of dP/dt^max[middle dot] EDA^-1, FAC, or Vcf^c. However, in the current study, the precision of the three pressure-dimension indices was similar (similar standard errors for the z-scores).

Both FAC and Vcf^cmeasure ventricular performance during the ejection phase of cardiac contraction. In a detailed review of ejection fraction, Robotham et al. [15]concluded that, because ejection fraction depends on preload, afterload, and contractility, it describes the entire cardiovascular system rather than the intrinsic properties of the myocardium. The absence of association between FAC and pressure-dimension indices is, therefore, not surprising. The weak association between Vcf^cand the pressure-dimension indices confirms that Vcf^cis a better measure of contractility than FAC. [13,16] 

Several animal and human investigations have used pressure-dimension indices to measure changes in contractility induced by volatile agents. [17–19]Using pressure-diameter relations in anesthetized dogs and in patients undergoing coronary artery bypass surgery with fentanyl anesthesia, Van Trigt et al. [17]observed a decrease of 20% in Ees at 1 MAC of halothane. The magnitude of that decrease is similar to the one we observed in this study. In dogs fitted with measuring instruments and receiving isoflurane, Hettrick et al. [18]found a comparable decrease in preload recruitable stroke work (17% at 0.9 MAC and 28% at 1.2 MAC vs. 23% at 1 MAC in the current study), but a larger decrease in Ees (29% at 0.9 MAC and 32% at 1.2 MAC vs. 17% at 1 MAC in the current study). Given significant differences in methods and experimental conditions, we must be cautious, however, in comparing our findings with those of the previously cited studies.

In conclusion, pressure-dimension indices of LV performance can be measured with TEE and ABD after CPB in patients undergoing myocardial revascularization. These indices show weak or no association with conventional measures of ventricular performance, but they can detect the changes in myocardial contractility associated with the administration of inhalation anesthetics. Such changes are not detected by the measurement of either FAC or Vcf^c. These observations suggest that LV performance should be assessed by pressure - dimension indices rather than by FAC or Vcf (^c), when the effects of new agents or procedures on LV performance are studied after CPB.