Combined use of pulsed and color M-mode doppler echocardiography for the estimation of pulmonary capillary wedge pressure: an empirical approach based on an analytical relation

Study population

We enrolled 66 consecutive patients who were admitted to the intensive care unit at our institution and had a balloon-tipped pulmonary artery catheter (Swan-Ganz). Exclusion criteria were: inadequate echocardiographic images to permit accurate measurements (eight patients), first-degree atrioventricular block that prevented a distinct separation between the E and A waves in the pulsed Doppler or color M-mode tracings (three patients) and severe mitral regurgitation (one patient). Three patients who had moderate mitral regurgitation as assessed by color Doppler flow imaging were included in the study. Thus, 54 patients (43 men and 11 women; mean age 64 years) formed our original study group. The total population was subsequently divided into two groups. The first 20 patients (14 men and 6 women; mean age 64 years) included in the study were used to generate an equation for estimation of Pw (training population). This equation was then assessed in the remaining 34 patients (29 men and 5 women; mean age 64 years) (test population). Most of the patients were in sinus rhythm (18 patients, 90% in the training group; 29 patients, 85% in the test group) and 48% (26 of 54 patients) were on mechanical ventilation (10 patients in the training group and 16 patients in the test group). Six patients (30%) in the training population and 15 (44.5%) in the test population had coronary artery disease. The primary diagnoses were: heart failure, 24 patients; aortic/peripheral vascular surgery, 7 patients; sepsis, 3 patients; trauma, 7 patients; respiratory distress syndrome, 4 patients, and other, 9 patients. To assess the usefulness of Doppler estimates for predicting changes in Pw, 13 patients were restudied after standard therapy, at least 24 h later than the first evaluation. The study was approved by the Human Subjects Review Committee of our institution.

Echocardiographic and Doppler examinations

Studies were performed with a Toshiba SSH-140 or SSH-160 (Tokyo, Japan) instrument equipped with a 2.5 MHz transducer. Left ventricular ejection fraction was determined by the area-length method from two-dimensional echocardiographic images obtained from the apical four-chamber view. Transmitral velocity was recorded from the same view with the pulsed sample volume placed at the tips of the mitral leaflets. The pulsed sample volume was then placed in the area of the anterior mitral valve leaflet to capture a LV outflow tract envelope and the mitral inflow profile simultaneously. Finally, the color Doppler sector map of the mitral inflow was displayed and the M-mode cursor was positioned within the mitral inflow stream, avoiding boundary regions and aligning the cursor as parallel to the filling flow as possible. Color M-mode recordings of the propagation of the early mitral inflow velocity into the LV were then obtained. Special effort was made to acquire a column of color flow from the mitral annulus to the distal third of the LV cavity. In all cases, color gain was set at subsaturation levels, using the same map and color processing filters. Pulsed wave and color M-mode Doppler tracings were recorded with a sweep speed of 100 mm/s and stored on video tape.

Pulsed and color Doppler measurements

Measurements were done off-line using the software package incorporated to echocardiographic instruments by a unique observer (FGV) who had no knowledge of the hemodynamic data. Five end-expiratory cardiac cycles were measured and the average used for analysis. In cases of atrial fibrillation, 10 cycles selected according to previously described criteria (12)were averaged. The following parameters were derived from the transmitral velocity: peak early mitral velocity (E wave) (cm/s), acceleration rate (cm/s²), acceleration time (ms), deceleration rate (cm/s)²and deceleration time (ms) of the E wave. In patients in sinus rhythm, the peak late mitral (A wave) velocity (cm/s) was also measured. The E peak velocity to A peak velocity ratio was calculated. Isovolumic relaxation time (ms) was measured from the end of the aortic flow to the onset of the mitral flow by use of pulsed wave Doppler (30)(Fig. 1A and C). As previously described (28), we measured the color M-mode Doppler FPV (cm/s) as the slope of the first aliasing velocity (55 cm/s) during early filling, from the mitral valve plane to 4 cm distally into the LV cavity (Fig. 1B). When peak E velocity was lower than aliasing velocity, we measured the slope of the transition no color/color (black/red) (Fig. 1D). We calculated the dimensionless parameter recently proposed by Garcia et al. (28)as the ratio of E peak velocity by pulsed Doppler to FPV by color M-mode Doppler.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Measurements of isovolumic relaxation time (IVRT) and flow propagation velocity (FPV) in a patient with normal wedge pressure (actual pressure = 10 mm Hg) (panels A and B)and in a patient with high wedge pressure (actual pressure = 31 mm Hg) (panels C and D). In the first case, FPV is determined by the slope of the first aliasing line during early filling, from the mitral valve plane distally into the left ventricular cavity (panel B). Estimated wedge pressure = 13.5 mm Hg. In the second case, FPV is determined by the slope of the first clearly demarcated isovelocity line during early filling (panel D). Estimated wedge pressure = 28.8 mm Hg.

Rationale for a new estimate of Pw

During the LV pressure decay, instantaneous LV pressure (P_v), aortic closing pressure (P_o) and time from P_o(t) derived from the invasive LV pressure trace are related by the monoexponential equation: Pv=Po⋅e−t/τ+bwith a time constant of LV relaxation τ (29). For clinical purposes, Yellin et al. (32)have shown that the theoretical asymptote (b) of the pressure-decay curve can be neglected (b = 0). Furthermore, Thomas et al. (31)have demonstrated in a canine model that at mitral valve opening tequals IVRT. In the absence of severe mitral regurgitation, pressure at mitral valve opening relates closely to mean left atrial pressure (33)which instead can be approximated by Pw. In addition, P_ocan be considered as the end-systolic pressure and it can be estimated noninvasively by 0.9·systolic aortic blood pressure (SBP), as previously proposed (34). Finally, it has been recently reported that there is a strong inverse correlation between the FPV, as measured by color M-mode Doppler and tau (22,26,27). We hypothesized that τ can be substituted for 1/FPV. According to these assumptions, Equation 1 becomes: Pw=(Pest)=(0.9⋅SBP)⋅eIVRT×FPVwhere the second term can be entirely calculated by noninvasive methods. For calculations, IVRT is measured in seconds and FPV in cm/s.

Hemodynamic measurements

Mean Pw was determined automatically by the monitoring system from pressure tracings obtained with a pulmonary artery catheter (Swan-Ganz). Special care was taken to obtain reliable pressure tracings. Arterial blood pressure was measured noninvasively by a calibrated semiautomatic cuff connected to the monitor. All measurements were obtained within 5 min of the echocardiographic examination.

Reproducibility

Eight studies were randomly selected and interobserver and intraobserver variabilities were calculated for color M-mode FPV and IVRT measurements, and for derivation of Pw according to the multilinear regression equation. Reproducibility was assessed as the mean ± 1 SD difference between the two sets of observations. In addition, mean percent error was calculated as the absolute difference divided by the average of the two observations.

Statistical analysis

Continuous variables are given as mean ± SD. For the training group, the correlations between echocardiographic and Doppler variables and Pw were evaluated by univariate linear regression analysis and were expressed as a correlation coefficient. Subsequently, those variables that achieved a significance level <0.10 were introduced stepwise in a multilinear regression analysis to develop an equation for the estimation of Pw. We then applied the regression equation obtained to estimate Pw in the test population. Linear regression and Bland-Altman analyses were used to evaluate the agreement between the measured and the estimated Pw. These analyses were performed for the derived index 10³/([2·IVRT] + FPV) using the entire population. The ability of the selected cutoff values of different variables for the prediction of Pw ≤ 15 versus >15 mm Hg was assessed by the standard formulations of sensitivity, specificity, predictive values and accuracy. Comparisons of the linear regression between Doppler parameters and Pw for patients with normal and depressed LV systolic function were performed by analysis of covariance. Statistical calculations were carried out with the SPSS 6.1 (Cary, North Carolina) program. All tests were two-tailed and the level of significance was established at p < 0.05.

Relation of variables to Pw

The echocardiographic and hemodynamic data for the training and test populations are summarized in Table 1. Several echocardiographic Doppler parameters were significantly related to Pw in the training population (Table 2). Of all the variables, the best correlations were observed for P_est(r = 0.79, p < 0.0001) and IVRT (r = −0.78, p < 0.0001). Only marginal significance was observed in the correlation between FPV and Pw (r = −0.45, p = 0.046). In the stepwise multiple linear regression analysis, only IVRT (p = 0.0096) and P_est(p = 0.0043) remained significantly related to Pw, according to the derived equation Pw = 22.5–0.15 (IVRT) + 0.49 (P_est) (r = 0.88, r²= 0.77, [standard error of the estimate] SEE = 3.7 mm Hg). The equation obtained in the training group was then applied for the estimation of Pw in the test group. The resulting correlation had a value of r = 0.86 (r²= 0.73) and a SEE of 3.8 mm Hg. When all 54 patients were combined, the equation predicted Pw with a value of r = 0.86.

Simplification for clinical use

Although the equation obtained by multilinear regression analysis demonstrated a relatively high accuracy for predicting individual Pws, its formulation is rather complex for the daily practice. In analyzing this equation, it becomes apparent that the main variables related to Pw are IVRT and FPV. The exclusion of 0.9 × SBP in P_estdid not change the results of the multivariate analysis. For this reason, we investigated the usefulness of several empirical parameters based on a combination of IVRT and FPV in the entire population (Table 3). The parameter 10³/([2·IVRT] + FPV) (in which the numerator 10³is used to avoid extremely low and impractical values) achieved a strong correlation with Pw (r = 0.89, p < 0.0001; SEE = 3.3 mm Hg) according to the regression equation Pw = 4.5 (10³/[2·IVRT] + FPV) − 9 (Fig. 2A). The mean difference between measured and predicted pressures was −0.24 ± 3.3 mm Hg (range −6.1 to 7.6 mm Hg) (Fig. 2B). The sensitivity and specificity for Pw > 15 mm Hg (n = 21) were 90% and 100%, respectively. The positive and negative predictive values were 100% and 94%, respectively. Accuracy was 96%. In 47 (87%) of 54 patients the estimated Pw was within 5 mm Hg of the observed Pw and in 50 patients (93%) was within 6 mm Hg. The correlation between observed and predicted Pw was the same in patients with and without mechanical ventilation (both; r = 0.88).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

(A)Linear regression between Doppler estimates and catheter measurements of pulmonary capillary wedge pressure (Pw). (B)Bland-Altman analysis of agreement between the estimated and measured Pw. (The middle solid lineindicates the average difference between the two methods, whereas the outer dashed lines represent 2 SD or the 95% limits of agreement.) Circles= test population; Squares= training population.

Influence of LV systolic function (Table 4)

The correlations between transmitral pulsed-Doppler derived parameters and Pw were significantly stronger in patients with depressed systolic function than in those with normal systolic function. This was also true for the parameter proposed by Garcia et al. (28), E peak velocity/FPV. Conversely, the correlations between the parameters obtained by the combined use of IVRT and FPV and Pw were not significantly influenced by the LV systolic function (Fig. 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Influence of left ventricular systolic performance on the correlation between mean pulmonary capillary wedge pressure (Pw) and: (A)E peak velocity/Flow propagation velocity (FPV) and (B)the parameter 10³/([2·IVRT] + FPV). IVRT = isovolumic relaxation time. Dashed linesand open squares= patients with left ventricular ejection fraction >50%; Solid linesand solid squares= patients with left ventricular ejection fraction ≤50%.

Detection of serial changes after treatment

To test if the Doppler index 10³/([2·IVRT] + FPV) could accurately track directional changes in Pw, 13 patients underwent repeat hemodynamic and Doppler measurements after standard therapy. The changes induced in Pw were reflected by those in the Doppler index (Fig. 4; r = 0.82; SEE = 3.6 mm Hg).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Comparison of predicted and observed changes in pulmonary arterial wedge pressure (Pw) in response to standard therapy.

Reproducibility

The interobserver and intraobserver reproducibilities for FPV and IVRT measurements are reported in Table 5. The variabilities of estimated Pw, in absolute values, were 1.32 ± 0.96 mm Hg (range: 0.17 to 2.83 mm Hg) for the same observer and 1.98 ± 1.4 mm Hg (range: 0.05 to 4.56 mm Hg) between observers.

Theoretical framework

The analytical relationship among left atrial pressure, aortic closing pressure, tau and IVRT described by the Weiss’ equation (29)has been previously validated in an animal model (31)and in humans (30)and constitutes a fundamental theoretical support of our approach. This can explain the strong correlation found between Pw and IVRT in our study. Similar results have been previously reported (8,11,12). Nevertheless, other authors have found a weaker relation (3,9). Isovolumic relaxation time is determined not only by Pw but also by the rate of LV relaxation. The relative influence of each of these parameters could depend on the study population. Thus, in patients with a more stable condition than those included in our study (3,9), the range of Pw could be narrower and changes in IVRT could probably be more dependent on LV relaxation.

According to the theoretical framework (1,4,30,31), it becomes apparent that the addition of an index of LV relaxation similar to tau may improve the value of IVRT for the prediction of Pw. Our results show that FPV, as assessed by color M-mode Doppler, can be used as a surrogate for tau and, along with IVRT, is a key parameter for the prediction of Pw. Both parameters can be easily obtained, even when patients are ventilated mechanically (12,28), and their use in clinical practice does not require complex regression equations. This approach is not entirely new. Recently, Garcia et al. (28), based on the relation among peak E wave velocity, tau and left atrial pressure (4), have reported a strong correlation between the dimensionless parameter E peak velocity/FPV and Pw, and their results have been confirmed in patients with atrial fibrillation by Nagueh et al. (12). In our patients, the combined use of IVRT and FPV allowed us a better prediction of Pw than the use of E peak velocity/FPV. We hypothesize that these findings could be in relation to theoretical and practical reasons. On the one hand, E peak velocity depends on more complex influences (left atrial pressure, minimal LV diastolic pressure, compliance of the left atrium, the rate of ventricular relaxation, LV suction and mitral inertance [1,3,4,18,19,21]) than IVRT. On the other hand, we have observed that the correlation between E peak velocity/FPV and Pw depends significantly on the LV systolic performance, reaching higher values in patients with depressed LV ejection fraction. Overall, our patients had ejection fractions (mean 52%) higher than those reported by Garcia et al. (mean 40%) (28).

The use of FPV as a noninvasive surrogate for tau has both internal and external foundation:

• 1. In our study, the substitution of τ for the inverse of FPV into the Weiss’ equation resulted in a acceptable correlation with Pw;

• 2. some authors have recently reported strong correlations between tau and FPV (22,26,27)or the time difference of peak velocity in the apex and at the mitral tips (23–25). Moreover, both color M-mode indexes share very important properties of τ, like their relative independence on preload (23,26)or their behavior with myocardial ischemia (24,25,27).

Influence of LV systolic function

Recently, Yamamoto et al. (35)have clearly shown that the correlations between LV filling pressure and indexes derived from pulsed wave Doppler transmitral flow velocity curves depend strongly on the LV systolic function. One of the most promising findings in our study is the clear independence on LV systolic function of indexes based on the combined use of IVRT and FPV.

Detection of serial changes

Our patients were not specifically manipulated to assess directional changes in Pw. Thus, the range of the variations in filling pressures was relatively narrow. As expected, small changes were poorly reflected by the noninvasive estimation. Overall the reliability of the Doppler index 10³/([2·IVRT] + FPV) to track such changes was fairly accurate.

Study limitations

The number of patients in the present study was relatively small. Despite this limitation, the range of baseline hemodynamic data and ejection fraction was wide, in agreement with the variety of underlying cardiac conditions. Nevertheless, patient characteristics could be responsible for both similarities (12)and differences (28)with previous studies. Only a few patients had atrial fibrillation, and results must be taken with caution for this subset of patients. One previous study (12)has demonstrated that both IVRT and FPV can be measured reliably in patients with atrial fibrillation and found results very similar to ours. Likewise, patients with severe mitral regurgitation were excluded. Although color M-mode Doppler indexes seem to be unrelated to preload (23,26), additional studies are necessary to elucidate their applicability in patients with significant mitral regurgitation or other reasons for significant left atrial v wave.

We did not explore pulmonary venous flow. Though most of the parameters derived from pulmonary venous flow recordings have the same limitations as transmitral early diastolic Doppler indexes for the prediction of LV filling pressure (12,35), the difference in the duration at atrial contraction between pulmonary venous and transmitral flow velocity curves has proved to be a useful method for this purpose, regardless of systolic function (35). Adequate tracings, however, can be difficult to obtain, particularly in critically ill patients (11,12,35). Therefore, further studies must elucidate the value of the indexes based on M-mode color as compared with the standard evaluation of pulmonary venous flow. Like other authors (12,28), we used Pw as an estimate of left atrial pressure. It must be remembered that there is also variability in catheter measurements, particularly in patients who are critically ill.

The color M-mode Doppler exploration of early LV filling is feasible, even in the setting of an intensive care unit and in patients on mechanical ventilation. In the present study, 12 out of 66 patients (18%) had inadequate images for analysis. Similar results (17%) has been reported by Nagueh et al. (12)in a population of the same characteristics. These authors also found 13% of patients with inadequate pulmonary venous recordings. It is likely that feasibility of color M-mode Doppler exploration could improve in a more favorable environment. The FPV of early LV filling can be measured by different methods (22,26,28). We used the method proposed by Garcia et al. (28). This method does not require special computer implementations but is prone to subjectivity. Interobserver and intraobserver reproducibilities were within reasonable limits and compare favorably with those reported by others (12,28). We believe that standardization of methods and measurements of color M-mode Doppler would be warranted for future investigations and clinical application.

Clinical implications

Our findings suggest that simple algorithms could be useful in a wide variety of cardiac conditions, in which the standard assessment of the diastolic function by pulsed Doppler is more difficult (1). This applies, for instance, to patients in atrial fibrillation (12). Likewise, our results showed that such methods are useful for the estimation of LV filling pressure both in patients with depressed and preserved systolic function. It remains to be proved that this approach is also valid in patients with significant LV hypertrophy. The correlation of FPV with a key parameter of diastolic physiology (tau) and its relative independence on preload would allow us to separate the contribution of an abnormal relaxation from other parameters (compliance, volume overload) in individual patients. Accordingly, this information could guide a more specific therapeutic approach. The simplicity of the method and its sound physiological basis could place it as an attractive tool for the noninvasive serial assessment of patients with heart failure.

Conclusions

We have shown that FPV as assessed by color M-mode Doppler can work as a surrogate for tau. Therefore, it can serve to correct the effect of altered relaxation on transmitral pulsed Doppler parameters, particularly IVRT. The combined use of FPV and IVRT provides an empirical tool for the estimation of Pw that is easy to obtain in daily practice and could be reliable in many patients.