Background
The acute and long-term effects of transcatheter aortic valve implantation (TAVI) in patients with aortic valve stenosis on left ventricular (LV) function are controversial. The aim of this study was to determine the effect of TAVI on LV function with two-dimensional (2D) and three-dimensional (3D) speckle-tracking analysis of LV deformation capability.
Methods
Patients underwent standardized 2D and 3D transthoracic echocardiography before TAVI and after 6 months of follow-up, including 3D and 2D LV deformation imaging.
Results
Forty-four patients (mean age, 81.7 ± 5.5 years; 21 men [47.7%]; mean body mass index, 26.3 ± 5.1 kg/m 2 ; mean logistic European System for Cardiac Operative Risk Evaluation score, 24.4 ± 13.7%) undergoing TAVI were prospectively included. After follow-up, mean 3D LV ejection fraction (LVEF) (35.4 ± 13.1% vs 40.6 ± 12.6%, P = .004), 3D LV volumes (end-systolic volume, 85.9 ± 41.8 vs 65.9 ± 33.7 mL, P < .001; end-diastolic volume, 127.6 ± 40.7 vs 106.4 ± 40.9 mL, P = .001), 3D global longitudinal strain (−9.9 ± 3.7% vs −12.6 ± 4.2%, P < .001), and 3D LV twist (6.1 ± 4.3° vs 8.5 ± 6.9°, P = .025) were relevantly improved. LV improvement was pronounced in patients with decreased baseline LV function (area under the curve, 0.78; P < .001), with a cutoff value for 3D LVEF of ≤37% to identify functional responders to TAVI. After follow-up, patients with 3D LVEFs ≤ 37% showed a significant improvements in 3D LVEF (26.0 ± 7.6% vs 35.9 ± 11.7%, P < .001), 3D LV volumes (end-diastolic volume, 147.4 ± 40.6 vs 117.1 ± 45.5 mL, P = .001; end-systolic volume, 110.9 ± 39.2 vs 77.5 ± 37.2 mL, P < .001), 3D global longitudinal strain (−7.8 ± 2.7% vs −11.3 ± 4.2%, P < .001), and 3D LV twist (5.6 ± 4.2° vs 8.0 ± 5.6°, P = .047), whereas in patients with 3D LVEFs > 37%, only 3D global longitudinal strain was relevantly altered (−12.5 ± 3.1% vs −14.2 ± 3.8%, P = .04). Compared with 2D transthoracic echocardiography, 3D LV functional imaging allowed significantly faster image acquisition and data analysis ( P < .0001). New York Heart Association functional class improved significantly in both groups (3D LVEF ≤ 37%, from 3.1 ± 0.5 to 2.0 ± 0.6, P < .001; 3D LVEF > 37%, from 2.7 ± 6.7 to 1.5 ± 0.7, P < .001), whereas a significant amelioration of N-terminal pro–brain natriuretic peptide was observed only in patients with baseline 3D LVEFs ≤ 37% (10,314.64 ± 11,682.2 vs 3,398.7 ± 3,598.9 pg/mL, P = .02; 3D LVEF > 37%, 10,306.4 ± 32,000.5 vs 2,868.0 ± 3,816.7 pg/mL, P = .12).
Conclusions
Our results indicate significant improvements of LV global and longitudinal function and clinical parameters 6 months after TAVI that are pronounced in patients with impaired baseline LV function. Compared with 2D LV functional imaging, 3D speckle-tracking imaging allowed significantly faster image acquisition and data analysis.
Aortic valve stenosis (AVS) is the most common native valve disease, and its incidence increases in the elderly. In the early stages of almost asymptomatic AVS, left ventricular (LV) ejection fraction (LVEF) and cardiac output are maintained. However, when increasing wall stress exceeds the compensating mechanisms of the left ventricle, LV systolic function declines because of afterload mismatch, and patients experience typical symptoms of chronic heart failure, such as angina pectoris and exertional dyspnea. Surgical aortic valve replacement (AVR) is known to partly decrease the negative impact of AVS on LV function, and clinical outcomes after surgical AVR are mainly good in patients with preserved LV function. However, the optimum treatment of patients with severely reduced LV function is still the subject of debate, because of relevantly elevated perioperative mortality rates in such patients. In recent years, transcatheter aortic valve implantation (TAVI) has been established for the treatment of patients at high risk for cardiovascular surgery, and numerous studies have shown the beneficial effects of TAVI on LV hemodynamics and patients’ prognoses.
However, findings concerning LV functional remodeling after TAVI are not congruent throughout the literature and in part contradict experiences seen in patients after surgical AVR. Speckle-tracking (ST)–derived strain imaging has been proven to be an appropriate method to examine global and regional LV functional properties and may enable better characterization of changes in LV performance after TAVI compared with conventional echocardiographic parameters. Recently, three-dimensional (3D) functional imaging has been introduced into clinical practice, potentially enabling better understanding of complex LV mechanics and pathologies.
The aims of this study were (1) to determine the prevalence of echocardiographically detectable pathologic myocardial LV functional parameters in patients with severe AVS before undergoing TAVI using 3D ST, (2) to evaluate changes in measurable LV functional parameters 6 months after TAVI, and (3) to evaluate the correlation of baseline echocardiographic parameters with functional response to TAVI (defined as an increase in 3D LVEF of ≥10%).
Methods
Patients
From January 2010 to March 2011, consecutive patients undergoing TAVI with the Medtronic CoreValve prosthesis (Medtronic, Inc., Minneapolis, MN) were prospectively included. The aortic valve prosthesis was inserted retrograde in all subjects via transfemoral access.
All patients provided written informed consent before study inclusion. The study was approved by local ethics committee and in was accordance with the Declaration of Helsinki.
Follow-Up Procedure and Definitions of End Points
We performed standardized two-dimensional (2D) and 3D transthoracic echocardiography on all patients before the procedure and 6 months ± 14 days after TAVI. Functional response to TAVI after follow-up FU was defined as an improvement of 3D LVEF of ≥10% according to the available data on patients undergoing cardiac resynchronization therapy ; clinical response was defined by a decrease in New York Heart Association functional class after therapy of ≥2.
Transthoracic Echocardiography
Each patient underwent standardized 2D transthoracic echocardiography for the determination of LV and right ventricular functional parameters and dimensions according to the recommendations of the American Society of Echocardiography using commercially available ultrasound scanners (Vivid 7, GE Healthcare, Waukesha, WI; iE33, Philips Medical Systems, Best, The Netherlands) with a 2.5-MHz phased-array transducer. Echocardiographic views, including apical four-chamber and two-chamber views and parasternal long-axis and short-axis views, with the patient in the left lateral decubitus position, were obtained in 2D, 3D, and color Doppler tissue imaging modes. The endocardial border was manually traced in the apical four-chamber and two-chamber views in end-diastole and end-systole. Two-dimensional LVEF was calculated by the computer software from volumes obtained by the summation of a stack of elliptical disks at end-diastole and end-systole. To assess triplanar 2D LVEF, a modified Simpson’s method was carried out in two-chamber, three-chamber, and four-chamber views, and the results were averaged for each patient. Mitral inflow velocities were recorded using standard pulsed-wave Doppler at the tips of the mitral valve leaflets in an apical four-chamber view. Doppler tissue imaging–derived systolic and diastolic velocities were obtained from the septal mitral and lateral tricuspid valve annuli according to national echocardiographic guidelines.
During echocardiography, systolic pulmonary artery pressure was estimated by measuring the peak systolic tricuspid regurgitant velocity flow on continuous-wave Doppler if applicable. Aortic valve area was calculated using the continuity equation. The maximum pressure gradient across the restrictive orifice was evaluated using the modified Bernoulli equation (4 v 2 ).
2D and 3D ST Analysis of LV Deformation Capabilities
Figure 1 shows a screen shot of 3D LV ST analysis after TAVI.
Two cine loops from the apical four-chamber view were digitized and stored in an echocardiographic imaging server (Xcelera; Philips Medical Systems). Offline 2D and 3D LV ST analyses of the grayscale images obtained by 2D and 3D echocardiography were performed using a commercially available software package (TomTec Imaging Systems GmbH, Unterschleissheim, Germany). For 2D LV ST analysis, the endocardium of the free LV wall was manually traced starting from the lateral mitral annulus to the LV apex and was tracked by the 2D strain software along the border throughout two cardiac cycles. Accuracy of border tracking was manually verified and adjusted if needed.
For 3D LV ST analysis, cine loops of three apical views (two-chamber, three-chamber, and four-chamber views) and two short-axis views (basal and apical) derived from the volume data set were automatically displayed on a screen, allowing manual border alignment if necessary. Automatic endocardial border delineation of the whole LV volume was processed at end-diastole and end-systole after positioning two landmarks on the mitral annulus and one on the LV apex at each apical view. Manual correction was performed if necessary to ensure an optimal LV delineation.
Two-dimensional untwist was calculated using the formula (end-systolic twist − end-diastolic twist/end-systolic twist) × 100, as described previously.
To determine the reproducibility of 2D and 3D echocardiography for the assessment of LV functional imaging, 20 randomly selected patients were analyzed again by the same observer for the determination of intraobserver agreement and by a second experienced echocardiographer to evaluate interobserver agreement.
Statistical Analysis
Exploratory data analysis was performed, and no adjustment was made for multiple tests.
Normal distribution of continuous variables was examined using the Kolmogorov-Smirnov test. Continuous data are expressed as mean ± SD. Two-tailed P values were calculated and considered to be significant if <.05. Comparisons between two groups were performed using Student’s t tests for paired samples or pairwise comparisons with Wilcoxon’s signed-rank tests for paired continuous variables. For categorical data, Fisher’s exact tests were performed. The method of Bland and Altman was used for the assessment of interobserver agreement. For the assessment of intraobserver variability, 20 randomly chosen patients were analyzed by the same investigator twice. Intraobserver variability was evaluated using intraclass correlation coefficients for total agreement, with good agreement defined as >0.80. Mean values and standard deviations between the measurements were obtained, and total agreement among the observation was calculated using intraclass correlation analysis.
Statistics were performed using SPSS for Windows (PASW version 17.0.2; SPSS Inc., Chicago, IL) and MedCalc version 11.4.1.0 (MedCalc Software, Mariakerke, Belgium).
Results
Baseline Data
A total of 51 patients with severe AVS, defined as aortic valve area < 1 cm 2 , completed clinical and echocardiographic FU. Images of nine patients were excluded for the determination of LV twist and torsion because of bad image quality. Therefore, 44 patients (mean age, 80.5 ± 5.6 years; 16 men [57.1%]; mean body mass index, 26.1 ± 5.3 kg/m 2 ) were included in the final analysis. Concomitant coronary artery disease was present in 15 patients (53.6%). According to local standard operating procedure for TAVI procedures, all patients underwent coronary angiography 3 months before the intervention, with coronary revascularization if necessary. Logistic European System for Cardiac Operative Risk Evaluation scores ranged from 10.3% to 64.5% (mean, 24.4 ± 13.7%), identifying patients at high surgical risk. Further baseline characteristics are presented in Table 1 .
| Variable | Value |
|---|---|
| Age (y) | 81.7 ± 5.5 |
| Men | 21 (47.7%) |
| Body mass index (kg/m 2 ) | 26.3 ± 5.1 |
| Society of Thoracic Surgeons mortality score (%) | 10.0 ± 8.8 |
| Logistic European System for Cardiac Operative Risk Evaluation score (%) | 24.4 ± 13.7 |
| Cardiovascular risk factors | |
| Arterial hypertension | 38 (86.4%) |
| Systolic blood pressure (mm Hg) | 121.6 ± 24.7 |
| Diastolic blood pressure (mm Hg) | 72.5 ± 14.6 |
| Chronic heart failure | 13 (29.5%) |
| Diabetes mellitus | 17 (38.6%) |
| History of stroke | 10 (22.7%) |
| Coronary artery disease | 29 (65.9%) |
| Smoking | 10 (22.7%) |
| Dyslipidemia | 31 (70.5%) |
| New York Heart Association functional class | 2.9 ± 0.6 |
| NT-proBNP (pg/mL) | 10,311.1 ± 22,468.5 |
2D Echocardiographic Data at Study Inclusion and after FU
The baseline echocardiographic characteristics of the study population are summarized in Table 2 . Only diastolic interventricular septal thickness significantly decreased 6 months after TAVI (1.2 ± 0.5 vs 0.9 ± 0.6 cm, P = .01), whereas other systolic 2D functional parameters were not relevantly altered. Regarding LV diastolic functional parameters, only E′ was changed significantly after FU (0.05 ± 0.02 vs 1.0 ± 0.03 m/sec, P < .001), whereas other variables showed no amelioration (2D untwist, 45.1 ± 32.2% vs 49.6 ± 31.5 %, P = .67; E/A ratio, 1.8 ± 5.3 vs 1.2 ± 0.7, P = .24; Table 2 ).
| Variable | Baseline ( n = 44) | FU ( n = 44) | P ∗ |
|---|---|---|---|
| Systolic pulmonary artery pressure (mm Hg) | 39.6 ± 17.9 | 32.5 ± 18.6 | .58 |
| Diastolic interventricular septal thickness (cm) | 1.2 ± 0.5 | 0.9 ± 0.6 | .01 |
| Aortic valve area (cm 2 ) | 0.7 ± 0.2 | NA | |
| LV end-systolic diameter (cm) | 3.5 ± 1.3 | 3.6 ± 1.3 | .05 |
| LV end-diastolic diameter (cm) | 5.3 ± 1.1 | 5.6 ± 1.2 | .08 |
| SVI (mL/m 2 ) | 23.3 ± 7.4 | 22.6 ± 8.2 | .62 |
| LVEF (%) | 55.7 ± 20.6 | 56.2 ± 16.8 | .64 |
| E′ (m/sec) | 0.05 ± 0.02 | 1.0 ± 0.03 | <.001 |
| E/A ratio | 1.8 ± 5.3 | 1.2 ± 0.7 | .24 |
| 4-CV GLS (%) | −11.5 ± 5.2 | −11.6 ± 5.1 | .86 |
| 2-CV GLS (%) | −4.5 ± 6.1 | −4.2 ± 5.4 | .34 |
| 3-CV GLS (%) | −5.7 ± 7.8 | −5.7 ± 7.1 | .98 |
| 2D twist (°) | 7.4 ± 4.7 | 7.5 ± 5.7 | .95 |
| 2D torsion (°/cm) | 0.8 ± 0.4 | 0.7 ± 0.3 | .21 |
| 2D untwist (%) | 45.1 ± 32.2 | 49.6 ± 31.5 | .67 |
| 3D end-diastolic volume (mL) | 127.6 ± 40.7 | 106.4 ± 40.9 | .001 |
| 3D end-systolic volume (mL) | 85.9 ± 41.8 | 65.9 ± 33.7 | <.001 |
| 3D stroke volume (mL) | 41.7 ± 12.2 | 40.4 ± 14.3 | .59 |
| 3D LVEF (%) | 35.4 ± 13.1 | 40.6 ± 12.6 | .004 |
| 3D twist (°) | 6.1 ± 4.3 | 8.5 ± 6.9 | .025 |
| 3D torsion (°/cm) | 0.8 ± 0.6 | 1.3 ± 2.4 | .17 |
| 3D GLS (%) | −9.9 ± 3.7 | −12.6 ± 4.2 | <.001 |
3D Functional Analysis
After 6 months of FU, 3D functional analysis showed significant improvement in measured values for 3D LVEF (35.4 ± 13.1% vs 40.6 ± 12.6%, P = .004), end-systolic volume (85.9 ± 41.8 vs 65.9 ± 33.7 mL, P < .001), end-diastolic volume (127.6 ± 40.7 vs 106.4 ± 40.9 mL, P = .001), 3D LV twist (6.1 ± 4.3° vs 8.5 ± 6.9°, P = .025), and 3D global longitudinal strain (GLS) (−9.9 ± 3.7% vs −12.6 ± 4.2%, P < .001), whereas 3D stroke volume (41.7 ± 12.2 vs 40.4 ± 14.3 mL, P = .59) and 3D torsion (0.8 ± 0.6°/cm vs 1.3 ± 2.4°/cm, P = .17) remained almost unchanged ( Table 2 ).
Interestingly, there was a significant correlation between baseline 3D LVEF (area under the curve, 0.78; P < .001), baseline stroke volume index (SVI) (area under the curve, 0.65; P = .07), and functional response to TAVI. Receiver operating characteristic curve analysis defined optimal cutoff values of ≤37% for 3D LVEF and ≤24 mL/m 2 for SVI to separate functional responders from nonresponders to this therapy ( Figures 2 and 3 ). We included SVI as parameter to discriminate between functional responders and nonresponders because it was proven to be very important in the setting of AVS, as it has been shown that patients with AVS with normal LVEFs but low gradients on the basis of low forward stroke volumes have a poor prognosis.
However, compared with discrimination by LVEF, discrimination of functional responders according to SVI showed less good results ( Figure 2 , Supplementary Figure 1 , Supplementary Table 1 ).
When separating groups according to cutoff values for 3D LVEF, patients with baseline 3D LVEF ≤ 37% experienced a significant improvement in 3D LVEF (26.0 ± 7.6% vs 35.9 ± 11.7%, P < .001), 3D GLS (−7.8 ± 2.7% vs −11.3 ± 4.2%, P < .001), 3D LV twist (5.6 ± 4.2° vs 8.0 ± 5.6°, P = .047), and 3D LV end-diastolic volume (147.4 ± 40.6 vs 117.1 ± 45.5 mL, P = .001), whereas in patients with baseline 3D LVEFs > 37%, only 3D GLS was significantly improved after FU (−12.5 ± 3.1% vs −14.2 ± 3.8%, P = .04) ( Table 3 , Figure 3 ).
| Variable | 3D LVEF ≤ 37% ( n = 24) | 3D LVEF > 37% ( n = 20) | ||||
|---|---|---|---|---|---|---|
| Baseline | 6-month FU | P ∗ | Baseline | 6-month FU | P ∗ | |
| 3D end-diastolic volume (mL) | 147.4 ± 40.6 | 117.1 ± 45.5 | .001 | 103.9 ± 25.9 | 93.6 ± 31.2 | .17 |
| 3D end-systolic volume (mL) | 110.9 ± 39.2 | 77.5 ± 37.2 | <.001 | 55.9 ± 19.0 | 52.1 ± 22.8 | .47 |
| 3D stroke volume (mL) | 36.5 ± 9.4 | 39.6 ± 15.4 | .31 | 48.0 ± 12.3 | 41.5 ± 13.1 | .07 |
| 3D LVEF (%) | 26.0 ± 7.6 | 35.9 ± 11.7 | <.001 | 46.6 ± 8.6 | 46.1 ± 11.5 | .86 |
| 3D twist (°) | 5.6 ± 4.2 | 8.0 ± 5.6 | .047 | 6.8 ± 4.6 | 9.7 ± 8.1 | .13 |
| 3D torsion (°/cm) | 0.6 ± 0.5 | 0.7 ± 0.8 | .62 | 1.1 ± 0.5 | 2.1 ± 3.3 | .20 |
| 3D GLS (%) | −7.8 ± 2.7 | −11.3 ± 4.2 | <.001 | −12.5 ± 3.1 | −14.2 ± 3.8 | .04 |
| 2D twist (°) | 5.9 ± 4.1 | 9.1 ± 7.4 | .87 | 7.8 ± 5.3 | 8.9 ± 1.7 | .11 |
| 2D torsion (°/cm) | 0.5 ± 0.2 | 0.6 ± 0.4 | .56 | 1.3 ± 0.5 | 1.9 ± 2.0 | .43 |
| 2D untwist (%) | 39.1 ± 37.4 | 53.1 ± 39.7 | .49 | 71.1 ± 17.2 | 46.1 ± 25.3 | .14 |
| NT-proBNP (pg/mL) | 10.314.64 ± 11.682.2 | 3.398.7 ± 3.598.9 | .02 | 10.306.4 ± 32.000.5 | 2.868.0 ± 3.816.7 | .12 |
| New York Heart Association functional class | 3.1 ± 0.5 | 2.0 ± 0.6 | <.001 | 2.7 ± 6.7 | 1.5 ± 0.7 | <.001 |
| Systolic blood pressure (mm Hg) | 118.0 ± 28.5 | 123.8 ± 29.5 | .10 | 126.3 ± 18.1 | 127.4 ± 9.5 | .85 |
| Diastolic blood pressure (mm Hg) | 72.0 ± 17.5 | 74.8 ± 17.6 | .27 | 73.2 ± 10.2 | 74.5 ± 10.4 | .69 |
Comparison of 3D LV Imaging with 2D Functional Parameters
Pearson’s correlation coefficient for the correlation of 3D LVEF and two-dimensionally measured LVEF from the four-chamber view was r = 0.66 ( P < .001) at baseline and r = 0.56 ( P < .001) for FU measurements, which was increased when using a triplanar approach to estimate 2D LVEF ( r = 0.85, P < .001).
Association with 3D GLS of 2D GLS values from the two-chamber or three-chamber view alone was rather weak (two-chamber view, r = −0.07, P = .66; three-chamber view, r = 0.15, P = .32). Interestingly, 2D GLS derived from the four-chamber view ( r = 0.46, P = .002) and the combination of mean values from 2D GLS from the two-chamber, three-chamber, and four-chamber views (the triplanar approach) were more strongly associated with 3D GLS ( r = 0.58, P = .005).
Regarding diastolic functional parameters, mitral inflow velocity was not significantly altered 6 months after TAVI (LVEF ≤ 37%, 1.1 ± 0.8 vs 1.1 ± 0.5, P = .94; LVEF > 37%, 2.7 ± 0.1 vs 1.3 ± 0.9, P = .17), whereas measured e′ values experienced relevant changes in both groups (LVEF ≤ 37%, 0.04 ± 0.02 vs 1.0 ± 1.3, P = .001; LVEF > 37%, 0.04 ± 0.02 vs 0.97 ± 0.3, P < .001).
When analyzing the correlation of untwist with values of E/A and e′, a significant negative correlation was observed between values of untwist and E/A only in group 1 ( r = −0.98, P = .02), whereas other parameters were not relevantly correlated with untwist (e′: LVEF ≤ 37%, r = 0.28, P = .65; LVEF > 37%, r = −0.07, P = .91; E/A: LVEF > 37%, r = −0.53, P = .46).
Intraobserver and Interobserver Variability of 3D ST Measurements
Bland-Altman analysis showed no evidence of any systematic difference regarding interobserver variability. Intraobserver agreement for the average measures of 3D GLS, expressed by the intraclass correlation coefficient, was 0.96, and for 3D LVEF was 0.94 ( Table 4 ).
