Aortic stenosis (AS), characterised by a pressure-overloaded left ventricle, is a major cause of heart failure among patients. Heart failure is the most important predictor of 1-year mortality and the primary reason for surgical rejection.1 2 Factors accounting for heart failure in AS are older age at diagnosis, left ventricular (LV) hypertrophy and depressed ejection fraction. Interestingly, most patients with AS presenting with heart failure have a normal ejection fraction; depressed LV function is seen in <20% of patients with AS.3 Hypertrophy is the cornerstone of heart failure. In a recent paper of isolated AS, increased LV mass predicted the presence of systolic dysfunction and heart failure independent of the severity of valvular obstruction. The authors concluded that LV hypertrophy may be maladaptive rather than beneficial in patients with AS.4 Furthermore, in a preoperative severely hypertrophied left ventricle, mass regression after aortic valve replacement is unsystematic.5 Persistent LV hypertrophy after surgery is correlated with an increase in death, reoccurrence of heart failure and decreased exercise capacity. Therefore, it would be attractive to differentiate independently between deleterious and beneficial preoperative LV hypertrophy before it becomes irreversible.

Conventional echocardiographic Doppler indices that evaluation of systolic ventricular function is unreliable in predicting clinical status, including cardiac symptoms and exercise capacity, among adults with AS but no LV hypertrophy. Recently, echocardiographic modalities such as tissue Doppler imaging (TDI) have demonstrated clinical relevance among patients with AS even in the presence of a normal ejection fraction.6 7 Regional myocardial deformation—that is, strain and strain rate, is abnormal in AS. This study aimed to determine whether TDI can predict the risk of death and development of cardiac symptoms in addition to exercise capacity in patients with AS after aortic valve surgery.

METHODS

Population and data collection

The work was approved by our institutional review board. Between 25 August 2003 and 21 November 2003, we prospectively enrolled consecutive patients with significant AS at Rouen University Hospital. Inclusion criteria consisted of patients with aortic valve area <1 cm² associated with heart failure, syncope or angina requiring aortic valve replacement. Thirty-two patients with AS were finally operated on and followed up during 1 year.

Demographic data included age, gender and weight. Patients were asked whether they had experienced angina, syncope or heart failure. Risk factors and past history of coronary artery disease were obtained along with medical treatment. Twenty-four hours before aortic valve replacement, New York Heart Association (NYHA) classification and 6-minute hall walk distance were recorded, and echocardiography was carried out for each patient (FM). The 6-minute hall walk distance was the longest walk distance covered by each patient over 6 minutes in a measured corridor. ECG, coronary angiogram and blood sample volume were collected within 1 month before surgery.

Study end point after aortic valve replacement

We used a composite end point of cardiovascular death, worsening of heart failure and limited exercise capacity. Worsening of heart failure was characterised by reoccurrence of congestive heart failure requiring admission to hospital or intensification of outpatient treatment at follow-up. Limitation of exercise capacity was defined as an improvement by <20% of 6-minute hall walk distance with no change or limited improvement of NYHA functional class as compared with baseline.8 The percentage of change of the 6-minute hall walk distance was measured as (distance at 3 months − distance at baseline)/distance at baseline. Non-responders were patients who died, or had worsening heart failure or limited exercise capacity. Information was collected by one “blinded” doctor (ND) at a 3-month follow-up during an inpatient visit and later by a telephone call at 6- and 12-month follow-up.

Echocardiographic analysis

Echocardiographic studies were performed with a Vivid 7 ultrasound system (GE Healthcare, General Electric, France). M-mode measurements included LV end-systolic and end-diastolic dimensions, posterior wall thickness, interventricular septal thickness and LV ejection fraction from parasternal long-axis view together with the aortic annulus diameter. Pulsed-wave Doppler was employed to record mitral valve inflow and LV outflow traces from four- and five-chamber views, respectively. The early to late transmitral velocity ratio E/A was calculated. The aortic valve area (cm²) was determined from the continuity equation. Colour M-mode velocity was used from the apical four chamber to record the velocity propagation Vp from which we calculated the ratio E/Vp. The modified Bernoulli’s equation served to calculate mean aortic pressure gradient (Ao − mean, mm Hg) from continuous wave (CW) Doppler interrogation through the aortic valve.

Tissue Doppler imaging

Tissue Doppler recordings with second harmonic imaging were collected with a frame rate above 110 frames per second during brief apnoea, if needed, from the short-axis view. Gains were adjusted to obtain tissue characterisation with minimal noise. Three (five in the case of atrial fibrillation) consecutive cardiac cycles were recorded as two-dimensional cine loops, and the acquired raw data were saved for offline analysis.

A single “blinded” sonographer performed the TDI echocardiographic analysis (FB). Tissue Doppler information from three consecutive cardiac cycles was analysed offline with dedicated software incorporated in the Vivid 7 System (Echopac, GE Healthcare). The sample volume was placed vertically at the posterior wall. For strain rate and strain measurement, an offset of 10 mm was used. The sample volume position was adjusted when three consecutive strain waveforms showed good reproducibility. We measured systolic peak strain rate (S-SR, 1/s) and systolic peak strain (S-S, %). Then, the processing was changed to velocity analysis and we measured systolic peak velocity (S-VEL, cm/s). The three tissue Doppler variables were measured from the same sample volume. Mitral annulus peak early diastolic velocity, Ea, was measured from the apical four-chamber view and E/Ea was calculated.

Statistical analysis

All enrolled patients were included in the statistical analysis. Statistical analysis was performed with StatView, version 5.0. Data are presented as mean (SD) for continuous variables. Correlation analysis was used to compare the relationship between demographic, clinical and echocardiographic variables, and the response after surgery, followed by multivariate analysis in a stepwise multiple regression model. Receiver operating characteristic (ROC) analysis was used to determine optimal cut-off points of TDI indices which could discriminate between the presence and absence of cardiac events at follow-up. ROC plots were obtained by plotting the sensitivity values against 1−specificity for all available thresholds of TDI indices. Event rates were plotted according to the Kaplan–Meier product limit method. Significance was taken as p<0.05.

RESULTS

Patient characteristics

Thirty-two patients with surgical AS (mean (SD) age 69 (9) years) were screened. Before surgery, angiotensin II converting enzyme inhibitors, diuretics, β blockers, calcium antagonists and digitalis were given to 38%, 31%, 19%, 13% and 9% of patients, respectively. The main reason for aortic valve replacement was heart failure and angina. Six patients were in atrial fibrillation. No patients presented with syncope. All patients amenable to the operating room were discharged 12 (4) days after surgery.

Primary end point

Fourteen patients were considered non-responders because they attained the primary end point at follow-up. One patient developed postoperative atrial fibrillation, persistent at discharge. Cardiovascular death was recorded in one patient who was had a cerebral haemorrhage while receiving anticoagulation for mechanical valve insertion. This event was not taken into account in the regression analysis. Table 1 lists the primary end points. Six patients had a reoccurrence of heart failure requiring oral treatment adjustment but no admissions to hospital. Seven patients had not improved their exercise capacity at the 3-month follow-up. In non-responders, mean (SD) NYHA functional class reduction was −0.8 (0.6) vs −0.7 (0.8) in responders (p = NS). The 6-minute hall walk distance improved by 45 (26)% in responders but remained unchanged in the group with poor outcome at a value of 8 (7)% (p<0.001).

There was no difference in baseline clinical and conventional echocardiographic parameters between responders and non-responders (table 2). In particular, both groups had similar ejection fraction and wall thickness. However, changes in Ea, Vp, E/Vp, peak strain rate and peak strain were seen in responders.

Comparison of demographic, clinical and echocardiographic parameters for the prediction of response after aortic valve surgery

As shown in table 3, LV filling parameters significantly predicted response after aortic valve replacement. Among them, Ea had the strongest predictive value (r = 0.55, p = 0.007). A significant correlation was also seen with peak strain rate (r = 0.43, p = 0.01) and peak strain (r = 0.49, p = 0.006). Peak systolic velocity was unable to predict the outcome.

On stepwise multivariate regression analysis in which the significant variables were entered into the statistical model, peak strain rate, Ea and E/Vp remained the independent predictors of response after aortic valve replacement. Other TDI parameters, including peak strain and peak systolic velocity, were unable to predict reverse remodelling. Since coronary disease was not correlated with response from linear regression analysis, we further tested ischaemic heart disease by multiple regression analysis together with stepwise multiple regression analysis and found no correlation with outcome (p = 0.15 and F = 2.3×10⁻⁵, F to enter >4, respectively). Importantly, patients with coronary disease did not have critical stenosis and thus severely impaired deformation (S-SR was 1.7 (0.5)/s in patients with pure AS and 1.5 (0.5)/s in patients with AS with coexisting coronary disease, p = 0.29).

Table 3 and fig 1 show the ROC for variables with significant correlation with response in the univariate model. The area under the curve was largest for peak strain rate (0.85, p = 0.002) and was significant for peak strain (0.77, p = 0.03). On the basis of the data on the ROC curve, we concluded that peak strain rate value of 2.0/s has a sensitivity of 83% and a specificity of 76% to predict significant reverse remodelling. When this cut-off value was applied, all but two patients with S-SR >2/s were correctly identified as good responders after aortic valve replacement. The areas under the ROC curve for Ea, Vp and E/Vp were 0.45 (p = NS), 0.71 (p = NS) and 0.34 (p = NS), respectively

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 (A) Receiver operating characteristic curve for identification of response after aortic valve replacement in all patients for peak strain and peak strain rate. (B) Kaplan–Meier estimate shows a significantly higher incidence of worsening heart failure and limited exercise capacity in patients with a lower peak strain rate.

Intra- and interobserver variability for S-VEL were 0.1 (0.3) cm/s (2 (8)%) and 0.1 (0.1) cm/s (3 (5%), respectively. Intraobserver variability in the analysis of S-SR was 0.12 (0.25)/s (5 (10)%) of the absolute measured values and interobserver variability was 0.17 (0.27)/s (9 (13)%). Intra- and interobserver variability for S–S were 1.9 (2.5)% (8 (10)%) and 3.4 (2.3) cm/s (10 (6)%), respectively.

Event-free survival curve

In the event-free survival curve (fig 1B), there were significantly (p<0.02) more heart cardiac incidents—that is, therapeutic intensification and worsening of exercise capacity, in patients whose S-SR declined before aortic valve replacement than in patients with normal S-SR. Figure 2 shows an example of one responder and one non-responder.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Analyses of tissue Doppler imaging in (A) one responder (peak strain rate 2.61/s¹); (B) one non-responder (peak strain rate 1.50/s).

DISCUSSION

This study demonstrates that the presence of abnormal preoperative systolic peak strain rate in patients with severe AS identifies patients at higher risk of cardiac morbidity after valve replacement.

Determinant of poor outcome after aortic valve replacement

At an early stage of LV remodelling, it was suggested that two-dimensional echocardiographic preoperative LV hypertrophy is associated with an increase in morbidity and mortality.9 LV hypertrophy is promoted by neurohormonal activation/tissue growth factor β1 triggered by prestenotic raised LV pressure and poststenotic low cardiac output.10 11 Histologically, pressure-overloaded remodelling is characterised by fibrosis and myocyte degeneration.12 This remodelling must be deleterious rather than compensatory in AS despite partial reversibility after stenosis relief.4 13 14 Therefore, regression of LV hypertrophy after aortic valve replacement is variable and correlates with poor outcome if persistent.15 16 A recent study in patients with AS showed that moderate to severe diastolic dysfunction assessed on mitral and pulmonary venous flow was also an independent predictor of late mortality after valve replacement.17

At an advanced stage of LV remodelling, low gradient, low cardiac output is the most important echocardiographic predictor of poor outcome in significant AS. Outcome is even worse when a preoperative low-dose dobutamine stress test fails to improve stroke volume, cardiac output or ejection fraction, identifying patients without contractile reserve from those with. There is a continuum in this remodelling, starting from subtle and undetectable myocardial structural changes, progressing to obvious myocardial abnormalities with diastolic dysfunction followed by LV hypertrophy and at late stage, depressed LV systolic function preceding death.

In patients with AS, ejection fraction deterioration is considered as an absolute signal for early surgery. However, LV hypertrophy is less taken into account, particularly when symptoms are unnoticeable. From previous reports and our results, it is certainly not correct to consider LV hypertrophy as acceptable. We here differentiate between two types of LV hypertrophy. LV hypertrophy with low deformation carries a high risk of morbidity and can be considered to be detrimental. Conversely, LV hypertrophy with normal deformation by TDI and no serious events after valve replacement can be considered to be compensatory.

Early detection by TDI of LV remodelling in patients with AS

In large cohort of patients with AS, LV hypertrophy is variable and significant deterioration of the ejection fraction is seen in <20%.18 Therefore, most patients present with an apparently normal left ventricle. Fortunately, TDI is a powerful tool to characterise LV abnormalities when the heart is apparently normal.

Thus, Kowalski et al analysed velocities together with both radial and longitudinal deformation in patients with AS and normal ejection fraction.6 The maximal systolic velocity and deformation in the patients with AS were significantly reduced compared with normal subjects. Furthermore, they were able to differentiate between patients with pure AS and those with coexisting coronary disease. Indices of radial and longitudinal deformation correlated with aortic valve area. In a similar designed study, Moreno et al found a negative correlation between isovolumic contraction time and the severity of aortic valve stenosis.19 They established a cut-off point between 73 and 85 ms to predict critical AS. Bruch et al confirmed these findings, additionally showing regional diastolic dysfunction and raised end-diastolic pressure in patients with AS but normal ejection fraction.7

After percutaneous valve implantation, TDI is informative, demonstrating immediate improvement in regional systolic deformation.20 Such a fast recovery is afterload-dependent according to the Frank–Starling law with better actin–myosin molecular configuration and cross bridge. Longlasting reverse remodelling by TDI has been evaluated by Giorgi et al. One year after aortic valve surgery, they found enhancement in global, regional systolic and diastolic function together with decrease LV preload.21 However, their results are debated. Experimentally, Derumeaux et al demonstrated abnormal deformation in chronically pressure-overloaded left ventricle.22 Recovery of regional function was obvious at early, but not late, debanding. Improvement in contractile function is the combination of quantitative and qualitative myocyte function recovery together with matrix stabilisation.13

TDI as prognostic determinant of morbidity and mortality in various diseases and AS

The relation between TDI and adverse outcomes has demonstrated varying prevalence and predictive values in population-based studies.23 In systolic heart failure, Wang et al found that early diastolic peak velocity measured by TDI provides an incremental predictive power for cardiac mortality compared with clinical data and standard echocardiographic measurements.24 After myocardial infarction, tissue Doppler velocity and tissue tracking predict death or reinfarction at follow-up in similar manner to a pharmacological stress test.25 26 In patients with significant mitral regurgitation but normal LV function, left lateral wall velocity <10.5 cm/s predicted a postoperative ejection fraction reduction >10%.27 TDI has been extensively used to predict functional status recovery or reduction in morbidity and mortality in patients receiving biventricular pacing. The magnitude of dyssynchrony correlates with improvement of NYHA functional class, 6-minute hall walk distance, quality of life and mortality.28^–30

As far as we know, the role of TDI in AS outcome after aortic valve replacement has never been reported. We concluded that radial deformation <2/s was a marker of persistent heart failure but not mortality. It is established that strain rate is a marker of myocardial remodelling but also depends on elongation/shortening of myocytes—that is, contraction and relaxation coupling.22 Therefore, we can assume that deterioration of the strain rate in patients with a poor outcome was associated with more severe preoperative diastolic function than in responders. This pathophysiological approach is reasonable since we found higher E/Ea and E/Vp ratios used as surrogate for end-diastolic pressure in non-responders than in responders (table 2). In addition, the relaxation had deteriorated more in patients with poor outcome, as suggested by the low Ea in non-responders as compared with responders in our study. Thus, the severity of the preoperative diastolic dysfunction may have a crucial role in the functional recovery after aortic valve surgery. We may reasonably conclude that chronic heart failure or exercise limitation relates to preoperative diastolic function. The reason why measurement of strain rate is better than measurement of velocity is quite clear. Strain rate describes regional deformation more precisely than does myocardial velocity, which is always influenced by overall heart motion, tethering effects and external parameters. The superiority of strain rate over strain is less understandable. Strain is presumably more load dependent but less remodelling dependent and less influenced by diastolic dysfunction.

Clinical application and limitations

Guidelines for the right timing of an operation in patients with severe AS are unambiguous. A prompt aortic valve replacement should be performed in symptomatic patients with AS. In patients with AS with low gradient/low cardiac output, operative risk stratification requires a dobutamine stress test. On the other hand, the right timing of an operation is less clear for asymptomatic patients. Although our work was not designed for this clinical situation, radial deformation represents a useful predictor of the eventual re-occurrence of symptoms or limited functional improvement after aortic valve replacement and, thus, in defining a new risk.

A small number of patients and subjective outcome variables are the two major limitations of our work. Furthermore, functional capacity within weeks after surgery is not only dependent on haemodynamic enhancement but also on recovery from surgery, and must be interpreted with caution.

CONCLUSION

A peak strain rate <2/s at the LV posterior wall is a useful and powerful predictor of clinical outcome after aortic valve replacement in patients with AS. Regional remodelling by TDI in combination with other echo parameters could be taken into account to improve the decision for surgery in patients with AS.