Author + information
- Received July 31, 2000
- Revision received February 16, 2001
- Accepted March 1, 2001
- Published online June 15, 2001.
- ↵*Reprint requests and correspondence: Dr. Tim Kinnaird, Department of Cardiology, London Chest Hospital, Bonner Road, London, United Kingdom, E2 9JX
This study compared a prediction of mean left atrial pressure (Pla) ascertained by Doppler echocardiography of pulmonary venous flow (PVF), with predicted Plausing the pulmonary artery occlusion pressure (Ppao).
In select patient groups, PVF variables correlate with Ppao, an indirect measure of Pla.
In 93 patients undergoing cardiac surgery, we recorded with transesophageal echocardiography mitral valve early (E) and late (A) wave velocities, deceleration time (DT) of E (DTe), and pulmonary vein systolic (S) and diastolic (D) wave velocities, DT of D (DTd) and systolic fraction. The Ppaowas measured using a pulmonary artery catheter zeroed to midaxillary level. A further catheter was held at midatrial level to zero a transducer and was then inserted into the left atrium. A prediction rule for Plafrom DTdwas developed in 50 patients and applied prospectively to estimate Plain 43 patients.
A close correlation (r = −0.92) was found between Plaand DTd. Systolic fraction (r = −0.63), DTe(r = −0.61), D wave (r = 0.57), E wave (r = 0.52), and E/A ratio (r = 0.13) correlated less closely with Pla. The mean difference between predicted and measured Plawas 0.58 mm Hg for DTdmethod and 1.72 mm Hg for Ppao, with limits of agreement (mean ± 2 SE) of −2.94 to 4.10 mm Hg and −2.48 to 5.92 mm Hg, respectively. A DTdof <175 ms had 100% sensitivity and 94% specificity for a Plaof >17 mm Hg.
Deceleration time of pulmonary vein diastolic wave is more accurate than Ppaoin estimating left atrial pressure in cardiac surgical patients.
Pulmonary artery occlusion pressure (Ppao) is considered the clinical gold standard for estimation of mean left atrial pressure (Pla), an indirect indicator of left ventricular intracavity filling pressures (1,2). However, insertion of a pulmonary artery catheter is not a risk-free procedure, and a reliable, less-invasive alternative has been sought (3). Both pulsed-wave Doppler echocardiography of mitral inflow and, subsequently, pulmonary vein flow (PVF) have been extensively studied, and a clear relationship between selected variables and Ppaowas found (4–12). However, mitral inflow and PVF patterns are influenced by multiple factors including left atrial pressure, left ventricular relaxation (4,13), compliance and afterload (14,15), ventricular interaction (16,17), heart rate (18,19), cardiac output (20)and age (21). These confounding factors preclude routine clinical use of mitral inflow or PVF patterns to predict Pla.
Two recent studies have found a close relationship between the deceleration time of the diastolic wave (DTd) of PVF and Ppaoin selected patient groups (22,23). Therefore, this study set out to investigate the relationship between the DTdand directly measured Plain a more general group of cardiac surgical patients. We then attempted to predict Plain a test group using a regression equation developed from the correlation between DTdand Plain the study group. Finally, we compared the accuracy of this method of estimating Plawith Ppaoestimation of Pla.
Ninety-three patients scheduled for coronary artery bypass surgery and/or aortic valve replacement were studied in the operating room. Patients were divided into two groups: Patients in group 1 (50 patients)—the derivation group—were used to develop the prediction rule for Pla, and group 2 patients (43 patients)—the test group—were used to test the prediction rule. All patients had undergone cardiac catheterization ± transthoracic echocardiography prior to surgery. Patients with any degree of mitral stenosis, moderate or severe mitral regurgitation (24,25), or a history of prior cardiac surgery were excluded. The study protocol was approved by the St. Paul’s Hospital Research Ethics Board. All patients gave written, informed consent in a preadmission clinic or on the cardiac ward after full explanation of the study protocol.
Following induction of anesthesia, endotracheal intubation and placement of a pulmonary artery catheter (Model 131 F7, Baxter, Deerfield, Illinois), a multiplane 5-MHz transesophageal probe (Hewlett-Packard, Palo Alto, California) was placed in the esophagus. Data were obtained using a Hewlett-Packard Sonos 1500 ultrasound unit and recorded on videotape for later analysis. All measurements were obtained after pericardotomy, with the patient in a stable hemodynamic state, and ventilation briefly suspended at end expiration. Pulmonary venous flow was obtained by placing the pulsed-wave Doppler sample volume approximately 1 cm beyond the orifice of a superior pulmonary vein. Color flow Doppler was used when necessary to assist with optimal sample volume placement. Mitral flow was obtained in a four-chamber view with the pulsed-wave Doppler sample volume placed at the tips of the mitral leaflets. All Doppler tracings were recorded at 100 mm/s sweep speed.
After induction of anesthesia, a transducer for the pulmonary artery catheter was zeroed visually at the midaxillary level by the anesthesiologist and then fixed in relation to the chest. The Ppaomeasurements were taken by the anesthesiologist immediately after the echocardiographic data were acquired.
Following Ppaomeasurement, a fluid-filled catheter attached to a 21-gauge needle was held by the surgeon adjacent to the mid-right atrial wall to rezero the pressure transducer. The left atrium was then cannulated to directly record Pla. All measurements were obtained in a steady hemodynamic state with ventilation briefly suspended at end expiration. The maximum time to acquire all echocardiographic and hemodynamic data was 10 min.
Analysis of the echocardiographic data was performed offline by an interpreter (T.K.) blinded to the hemodynamic data. For all measurements, five consecutive beats were traced and the results averaged.
Pulmonary venous flow was analyzed for peak systolic (S) and diastolic (D) wave velocities, their ratio, and velocity time integrals. The DTd, and the peak velocity and duration of the atrial reversal wave were also measured. In the presence of a bimodal D wave deceleration slope, the initial, steeper part was extrapolated to zero to obtain the deceleration time (Fig. 1)(22). The systolic fraction of PVF was calculated as the ratio of the velocity-time integral of the S wave to that of the combined velocity-time integral of the S and D waves.
From mitral flow recordings, the velocities of peak early (E) and late (A) waves and their ratio, E wave deceleration time (DTE) and A wave duration were measured. The initial and steeper part of the E deceleration slope was also extrapolated to baseline where necessary to measure DTE. The difference between the duration of the A wave and the duration of the atrial reversal wave was also calculated.
Using Statistical Analysis Software (SAS Institute, Cary, North Carolina), quadratic regression analysis was performed to examine the correlation between Doppler variables and Plaand the correlation between Ppaoand Pla. A Plaprediction rule was developed based on the correlation between DTdand Plain the first 50 patients (group 1) and then applied prospectively to the subsequent 43 patients (group 2). To evaluate the agreement between predicted and actual Pla, and between Ppaoand Pla(in the same 43 patients), the data were processed by the Bland-Altman method, and the 95% confidence intervals (CI) expressed (26). Sensitivity and specificity were calculated with standard formulae.
The baseline clinical and hemodynamic characteristics of the study group are described in Table 1. Group 2 patients were slightly older and had a higher incidence of aortic stenosis than patients in group 1. Ejection fraction was measured during cardiac catheterization.
Correlation of mitral and PVF variables with Pla
A close correlation (r = −0.92) was found between DTdand Plafor the entire patient group (Fig. 2), whereas correlation of the other echocardiographic parameters was less close (Table 2). Among the PVF variables, DTd, systolic fraction and D wave peak velocity correlated most closely with Pla. Of the mitral inflow variables, DTEand E wave peak velocity correlated most closely with Pla. A DTdof <175 ms had 100% sensitivity and 94% specificity for a Plaof ≥17 mm Hg in the entire group, and 100% sensitivity and 90% specificity for a Plaof > 17 mm Hg in the test group. A DTd>275 ms predicted a Plaof ≤6 mm Hg , with 88% sensitivity and 95% specificity. There was no correlation between ejection fraction and DTd.
Estimation of Plafrom DTdin the test group
Using the DTdand Plaplot from group 1, the following regression equation was developed: This formula was then applied prospectively to group 2 to predict Pla. The correlation between the estimated Plausing DTdand actual Plais shown in Figure 3. Figure 4displays a Bland-Altman plot of the difference between estimated and actual Plaversus actual Pla. The mean difference between predicted and measured Plawas 0.58 mm Hg, with 95% CI (mean ± 2 SE) of −2.94 to 4.10 mm Hg.
Estimation of Plafrom Ppaoin the test group
There was also a close relationship between Ppaoand Pla(r = 0.93, Fig. 5) although a systematic error was introduced, in part, by the visual estimation of the midaxillary line. When the zero point from this level was referenced to the surgeon’s visual zero at midatrial level, the midaxillary estimation was, in general, consistently lower than midatrial level. Thus, the mean difference between predicted Plafrom Ppaoand measured Plawas 1.72 mm Hg, with 95% CI (mean ± 2 standard error) of −2.48 to 5.92 mm Hg.
Thus, although both Ppaoand DTdpredict the Plawith a similar SE, the DTdmethod is not influenced by the systematic error introduced by visual estimation of the midaxillary line.
Intraobserver and interobserver variability
This was assessed from 20 random Doppler recordings. In measuring the DTd, the mean percentage of variation between observers was 6% and for repeated measurement was 4%.
We have shown that the DTdcorrelates strongly with the Plain a group of general cardiac surgical patients. The other mitral inflow and PVF variables measured correlate less well with Pla.
Comparison with previous studies
Two previous studies have directly examined the DTdand Plarelationship. Chirillo et al. (22)studied the correlation between the two variables in patients with atrial fibrillation in whom more traditional measures of diastolic function such as the E/A ratio or systolic fraction cannot be used. They found a very close correlation between DTdand Ppao, (r = −0.91), and they concluded that in patients with atrial fibrillation, DTdcould be used to estimate Ppao. More recently, Yamamuro et al. (23)studied the relationship in patients within one week of an acute myocardial infraction and also found a close relationship between Ppaoand DTd(r = −0.89). The correlation between DTdand Ppaoor Plais remarkably similar in all three studies, as are the regression lines. To our knowledge, our study is the first to show the strong correlation between mean Plaand DTdin a more general group of cardiac patients and to compare directly the prediction of Plafrom Ppaoand DTd.
Mechanism of relationship between DTdand Pla
Controversy exists as to whether the left atrium is a passive structure through early diastole and ventricular systole. Little et al. (14), in an experimental model, found that DTEdepended strictly on left ventricular chamber stiffness and assumed that, in early diastole, the left atrium and left ventricle act as a common conduit. However, Henein et al. (27)believe that the left atrium is active throughout most of the cardiac cycle. The discrepancy between DTdand DTEfound in the Henein et al. (27)study, and by other investigators, suggests that the left atrium in the patient group studied behaves as a receiving chamber in its own right (22,23). Differing left ventricular and left atrial compliances may have an important role in modulating both PVF and mitral inflow patterns. If this is the case, then the driving pressure between the pulmonary veins and the left atrium and the compliance of the left atrium itself might be the most important determinants of the deceleration time of the DTD. This would explain the much closer correlation between DTdand Plathan between DTEand Plafound in the present study.
Thus, in early diastole, blood flowing into the left ventricle will cause a rapid pressure drop in a poorly compliant left atrium (with volume loss), resulting in blood accelerating in from the pulmonary veins (28). Rapid pulmonary vein inflow associated with low left atrial compliance will result in a rapid rise in left atrial pressure, an early abolition of the driving pressure gradient and a short deceleration time of early diastolic pulmonary flow. We did not examine the relationship between DTdand left atrial volumes in this study because of the inherent difficulties in accurate measurement of left atrial diameters from the transesophageal route, and because of time constraints. Future studies investigating left atrial compliance and PVF are needed.
Comparison between echocardiography and Ppaoestimation of Pla
To our knowledge, this is the first study in which left atrial pressure was measured directly rather than estimated using Ppao. Kuecherer et al. (9)used direct left atrial pressure measurement in a third of the periods studied in his series and measured Ppaoin the remainder. Cannulation of the left atrium enables a comparison between echocardiographic estimation and Ppaoestimation of Plato be made.
We validated the regression equation developed from the initial patient data in the test group and were able to predict Plawithin limits that would make it clinically useful. The 95% CIs for the estimate are narrower than in a previous study, which may reflect comparison with direct measurement, rather than estimation of Pla(22). The SE of the estimate of Plausing DTdwas similar to the SE of the estimate using Ppao. However, there was a tendency for the Ppaoto consistently overestimate the Plaas reflected by a mean difference of 1.72 mm Hg. The explanations for this consistent error are threefold: 1) the tendency for the estimate of midatrial level (as referenced to the midaxillary level) to be too low; 2) as found in the original study relating Ppaoto left atrial pressure, the Ppaodoes overestimate the Plabecause of the contribution of pulmonary venous resistance (29); and (3)the contribution of the right ventricular systolic pressure wave to Ppao(30). Our findings suggest that the prediction of Plafrom DTdis more accurate than the prediction from Ppao—the current clinical practice.
Risks of pulmonary artery catheters
Recent controversy has centered on whether pulmonary artery catheters improve or worsen survival in critically ill patients (31). Irrespective of this controversy, there are well-recognized risks of pulmonary artery catheter placement including pneumothorax, pulmonary artery rupture and sepsis (3,32). Alternative and less invasive techniques to obtain hemodynamic data such as esophageal Doppler echocardiography and thoracic bioimpedance are emerging technologies (33,34).
Patients were studied after pericardotomy to allow left atrial cannulation to take place immediately after echocardiography and Ppaomeasurements. It is possible that the relationship found between DTdand Plawould be different with a closed pericardium. However, in nine patients we measured DTdimmediately before and after pericardotomy and found no significant difference in the predicted Pla.
A further study limitation is that only three patients were in atrial fibrillation and, therefore, it is not possible to conclude from this study alone that the DTdcan be used to estimate Plain patients in atrial fibrillation. However, in a previous study examining only patients in atrial fibrillation (22), there was a similar correlation between DTdand Ppaoas found in the present study. Therefore, the combined evidence suggests that the DTdcan be routinely applied to predict Plain patients with atrial fibrillation as well as to patients in sinus rhythm. The present study considered only the relationship between DTdand Plain a steady hemodynamic state. If echocardiography is to replace the pulmonary artery catheter in certain situations, further work is needed to investigate whether changes in hemodynamic parameters and Plaare reflected by appropriate changes in the DTd.
Measurements of PVF were made using transesophageal ultrasound because the study was conducted during cardiac surgery. Routine clinical application would be facilitated if transthoracic measurements were feasible. Pulmonary vein flow can be recorded in over 80% of patients from the transthoracic approach, and measurements taken correlate closely with simultaneous transesophageal recordings (35,36). Previous studies (22,23)showing a similar correlation between Ppaoand DTdas found in our study were conducted using transthoracic ultrasound. Therefore, the use of the transesophageal rather than the transthoracic approach should not prevent extrapolation of the study results to wider clinical practice.
Finally, we conclude that, in cardiac surgical patients, measurement of the DTdusing echocardiography can reliably estimate Pla, and it may obviate the need for invasive hemodynamic measurement with its attendant risks.
- late mitral inflow
- confidence interval
- diastolic pulmonary vein flow
- deceleration time of diastolic pulmonary vein flow
- deceleration time of early mitral inflow
- early mitral inflow
- mean left atrial pressure
- pulmonary artery occlusion pressure
- pulmonary venous flow
- systolic pulmonary vein flow
- standard error
- Received July 31, 2000.
- Revision received February 16, 2001.
- Accepted March 1, 2001.
- American College of Cardiology
- Civetta J.M.
- Appleton C.P.,
- Hatle L.K.,
- Popp R.L.
- Mulvagh S.,
- Quinones M.A.,
- Kleiman N.S.,
- Jorge C.B.,
- Zoghbi W.A.
- Nishimura R.A.,
- Abel M.D.,
- Hatle L.K.,
- Tajik A.J.
- Kuecherer H.F.,
- Muhiudeen I.A.,
- Kusumoto F.M.,
- et al.
- Rossvoll O.,
- Hatle L.K.
- Appleton C.P.,
- Galloway J.M.,
- Gonzalez M.S.,
- Gaballa M.,
- Basnight M.A.
- Ishida Y.,
- Meisner J.S.,
- Tsujioka K.,
- et al.
- Little W.C.,
- Ohno M.,
- Kitzman D.W.,
- Thomas J.D.,
- Cheng C.P.
- Brutsaert D.L.,
- Rademakers F.E.,
- Sys S.U.
- Baker A.E.,
- Dani R.,
- Smith E.R.,
- Tyberg J.V.,
- Belenkie I.
- Atherton J.J.,
- Moore T.D.,
- Thomson H.L.,
- Frenneaux M.P.
- Appleton C.P.,
- Carucci M.J.,
- Henry C.P.,
- Olajos M.
- Van Dam I.,
- Fast J.,
- de Boo T.,
- et al.
- Chirillo F.,
- Brunazzi M.C.,
- Barbiero M.,
- et al.
- Yamamuro A.,
- Yoshida K.,
- Hozumi T.,
- et al.
- Spain M.G.,
- Smith M.D.,
- Grayburn P.A.,
- Harlamert P.A.,
- DeMaria A.N.
- Hatle L.,
- Angelsen B.,
- Tromsdal A.
- Henein M.Y.,
- Gibson D.G.
- Smiseth O.A.,
- Thompson C.R.,
- Lohavanichbutr K.,
- et al.
- Luschinger P.C.,
- Seipp H.W.,
- Patel D.J.
- Nishimura R.A.,
- Tajik J.A.,
- Holmes D.R. Jr.