Background
Three-dimensional (3D) echocardiographic (3DE) imaging is an alternative to multi–detector row computed tomography (MDCT) for aortic annular measurement before transcatheter aortic valve replacement (TAVR). A commonly used direct planimetry from a reconstructed short-axis view has not been compared with semiautomated 3DE methods. Typically accepted optimal cutoffs for percent prosthesis-area oversizing of the balloon-expandable SAPIEN or SAPIEN XT valve to native annular size are approximately 5% to 15%. The aim of this study was to compare semiautomated and direct planimetric 3DE methods for aortic annular sizing with a gold standard of MDCT to determine predictive value for paravalvular regurgitation (PVR) and balloon postdilatation.
Methods
In this retrospective analysis, aortic annular cross-sectional area was measured from pre-TAVR imaging using (1) MDCT (CT_Area), (2) a 3D transesophageal echocardiographic (TEE) semiautomated method (3DE_Area_SA), and (3) a 3D TEE direct planimetric method (3DE_Area_Direct). Annular area percent oversizing was calculated. PVR after TAVR was assessed from intraoperative TEE imaging. Need for balloon postdilatation was recorded.
Results
One hundred patients who underwent TAVR with either the SAPIEN or SAPIEN XT balloon-expandable prosthesis were analyzed. Twenty-three patients had mild or greater PVR after TAVR. CT_Area was 442 ± 79 mm 2 , 3DE_Area_SA was 435 ± 81 mm 2 , and 3DE_Area_Direct was 429 ± 82 mm 2 . Both 3DE_Area_SA and 3DE_Area_Direct underestimated MDCT ( P < .05). All methods were highly correlative ( R = 0.88–0.93, P < .0001). Percent oversizing obtained by the three methods significantly predicted mild or greater PVR and need for balloon postdilatation by receiver operating characteristic analysis, with optimal cutoffs for CT_Area (9%–10%) and 3DE_Area_SA (14%) within the recommended ranges for the studied transcatheter valves and for 3DE_Area_Direct higher than the recommended range (18%–19%). Inter- and intraobserver reproducibility were lowest for 3DE_Area_Direct.
Conclusions
Caution must be used when using 3D TEE direct planimetry of the aortic annulus, as optimal percent oversizing ranges approach the level associated with root injury, and measurements are less reproducible. Therefore, semiautomated 3DE planimetry is preferred to 3DE direct planimetry for aortic annulus sizing.
Cross-sectional measurement of the aortic annulus is required in the evaluation of patients for transcatheter aortic valve replacement (TAVR) for severe, symptomatic aortic stenosis. Although multi–detector row computed tomography (MDCT) has been a widely used modality for aortic annular measurements, its use is subject to important limitations. For example, the requirement for iodinated contrast may present a prohibitive risk in patients with chronic kidney disease, which is prevalent in the pre-TAVR population. Furthermore, arrhythmias and a multitude of artifacts (motion, misregistration, partial volume averaging, beam hardening) can significantly reduce multidetector computed tomographic image quality, necessitating repeat or alternative imaging. Three-dimensional (3D) transesophageal echocardiographic (TEE) measurements of the aortic annulus have been used as an alternative to measurements by MDCT in these cases. Previous studies comparing aortic annular cross-sectional measurements between 3D TEE imaging and MDCT have used a TEE direct planimetric method from a reconstructed cross-sectional view of the annulus and have shown a marked underestimation of MDCT-measured annular dimensions. A previously published method of 3D TEE aortic annulus measurement by our group showed very good agreement and correlation with MDCT-derived measurements. This semiautomated 3D echocardiographic (3DE) method has not been compared with the direct planimetric method typically used for 3DE annular measurement. The aim of the present study was to compare the previously published semiautomated 3DE method and a 3D TEE direct planimetric method with multidetector computed tomographic measurements of aortic annular cross-sectional area to determine potential differences, predictive value for paravalvular regurgitation (PVR) and need for balloon postdilatation and impact on chosen sizing for the balloon-expandable transcatheter heart valve (THV). A secondary aim was to consider the same 3DE measurements performed by a physician with less experience to better understand the limitations of these methods.
Methods
Patient Population and Procedure
This retrospective analysis included 100 patients who underwent TAVR with a balloon-expandable Edwards SAPIEN or SAPIEN XT THV (Edwards Lifesciences, Irvine, CA) from November 2011 to January 2013 who also underwent both preprocedural MDCT and intraprocedural TEE imaging. Patients were nonconsecutive because of the exclusion of patients who had not undergone both MDCT and intraprocedural TEE imaging. The procedural access route (transfemoral, transapical, or transaortic) and THV sizing were decided at the discretion of the treating physicians, with the use of all available imaging modalities. No patients were excluded from imaging analysis on the basis of image quality. The study was approved by the institutional review board for human research and a waiver of the requirement to obtain informed consent was obtained.
Image Acquisition
Echocardiography
Patients underwent intraprocedural TEE imaging using commercially available equipment (iE33; Philips Medical Imaging, Andover, MA), according to standard protocols. A full two-dimensional (2D) TEE imaging protocol was performed. User-defined 3D TEE volumes of the aortic valve complex were acquired (single-beat acquisition) by obtaining long-axis or short-axis 2D TEE views from imaging windows that minimized acoustic shadowing of the annular plane. The 3D volumes contained the left ventricular outflow tract, aortic annulus and valve, and aortic root to the sinotubular junction. Multibeat, spliced images were avoided.
MDCT
Before the TAVR procedure, patients underwent cardiac computed tomographic angiography using a 320-slice system (Toshiba Medical Systems, Otawara, Japan). During an inspiratory breath-hold, single-volume acquisition was performed with prospective electrocardiographic triggering. Data were acquired with collimation of 240 to 360 × 0.5 mm and a gantry rotation time of 350 msec. Intravenous injection of 39 to 60 mL of nonionic contrast agent (iodixanol) was performed at a rate of 3.5 mL/sec. The decision regarding the volume of contrast used was at the discretion of the physician conducting the scan. Tube current and potential were determined by the physician conducting the scan or by software automation according to the patient’s body habitus. Real-time bolus tracking with automated peak enhancement detection in the descending aorta was used for timing the scan. Data acquisition was initiated on the basis of a threshold of 180 Hounsfield units. The 3D data set from the contrast-enhanced scan was reconstructed at 5% increments throughout the cardiac cycle. Images were reconstructed with a slice thickness of 0.5 or 0.25 mm.
Aortic Annular Measurements and Calculations
The aortic annulus was defined as the plane of the virtual circumferential ring containing the basal attachment points of the three aortic valve leaflets. Aortic annular cross-sectional area was measured using MDCT and 3D TEE imaging (see the next section for details). Percent oversizing was calculated as {[(nominal THV area/aortic annular area) − 1] × 100%}. All measurements were performed in mid-systole, using the most optimal image at or near the time point of maximum aortic valve excursion.
Echocardiographic Measurements
Three-dimensional echocardiographic reconstruction for measurement of the aortic annulus was performed using two methods: (1) off-label use of commercially available QLAB MVQ software version 8.1 (Philips Medical Imaging) by a semiautomated, indirect planimetric technique (3DE_Area_SA), which has been previously described, whereby points are placed on long-axis plane and appear on the short-axis plane representing the aortic annulus ( Figure 1 A). The measurement is derived on the basis of primary placement of long-axis points. (2) Direct planimetry of the short-axis view of the aortic annulus (3DE_Area_Direct), whereby the aortic annular area is traced directly on the short-axis image ( Figure 1 B). Semiautomated echocardiographic measurements were performed intraoperatively at the time of THV implantation by a board-certified echocardiographer experienced in TAVR imaging (R.T.H.). With the semiautomated method, eight pairs of points (a total of 16 points) are placed by the software on the long-axis planes ( Figure 1 A, panels ii and iii) and are then visualized in the short-axis plane ( Figure 1 A, panel i). These points can then be adjusted on the long-axis plane by the user and will be automatically adjusted by the software on the short-axis plane. Echocardiographic direct planimetric measurements were performed retrospectively by a board-certified echocardiographer experienced in TAVR imaging (O.K.K.) using the commercially available QLAB 3DQ software version 8.1 (Philips Medical Imaging), blinded to the results of the original measurements. Both semiautomated and direct planimetric measurements were performed by a cardiologist novice in 3DE analysis (J.M.W.). The novice’s 3DE measurements were compared with expert multidetector computed tomographic measurements to assess differences in 3DE methodology. For those performing retrospective image analysis, no on-screen annotations were provided, and all available user-defined 3D volume data sets were provided for choice of analysis. For inter- and intraobserver reproducibility assessment, readers were blinded to the previous measurements.
MDCT Measurements
Commercially available 3mensio software (Pie Medical Imaging, Maastricht, The Netherlands) was used for multidetector computed tomographic annular measurements ( Figure 1 C). The appearance of partial volume averaging artifacts (“blooming”) due to calcification was reduced by adjusting window and level settings. Images with suboptimal contrast opacification were enhanced by adjusting window and level settings to better delineate the boundaries of the annular lumen. In cases of both suboptimal contrast opacification and calcified hinge points, window and level settings were adjusted to alternately decrease partial volume averaging or increase lumen/tissue contrast, to optimize visualization of the annular boundaries. Computed tomographic annular measurements were performed retrospectively by a board-certified MDCT reader experienced in TAVR imaging (O.K.K.), blinded to 3DE measurements.
Postprocedural PVR Assessment
Final assessment of PVR was performed by a modified Valve Academic Research Consortium–2 scheme, with direct planimetry of 3D TEE color Doppler reconstruction of effective regurgitant orifice area as the primary method within the multiparametric approach. When 3D color Doppler reconstruction was not possible, assessment was performed by a combination of visual estimation of 2D color Doppler imaging and quantitative Doppler assessment of relative stroke volumes across the left ventricular outflow tract and right ventricular outflow tract. In cases in which 3D color Doppler was performed, grading of PVR was performed using the following cutoffs for effective regurgitant orifice area: trace, >0 to 4 mm 2 ; mild, 5 to 9 mm 2 ; moderate, 10 to 19 mm 2 ; moderate to severe, 20 to 29 mm 2 ; and severe, ≥30 mm 2 . The need for postdilatation to reduce PVR was decided by the treating physicians and was typically based on the immediate postdeployment TEE imaging of more than mild PVR, relying primarily on the short-axis view just apical to the THV stent.
Statistical Analysis
Analyses were performed MedCalc version 12.4.0.0 (MedCalc Software, Mariakerke, Belgium). Statistical significance was defined as P < .05. Continuous variables are reported as mean ± SD. Normality of distributions for continuous variables was tested using the Shapiro-Wilk test before performing t tests. Given that all compared continuous variables were normally distributed, comparisons between two measurements were performed using a paired 2-sided Student t test, and comparisons between groups were performed using an independent t test. Pearson correlation coefficients were used to assess the correlations between measurements from echocardiography and MDCT. Intraclass correlation coefficients were used to assess interobserver (R.T.H. and O.K.K. for 3DE_Area_SA, O.K.K. and N.B.H. for 3DE_Area_Direct, and O.K.K. and J.M.P. for aortic annular area from MDCT [CT_Area]) and intraobserver reliability (R.T.H. for 3DE_Area_SA, O.K.K. for 3DE_Area_Direct, and O.K.K. for CT_Area) between expert readers in 20 randomly selected patients from the cohort, using a two-way mixed-model absolute-agreement method. Test-retest variability analysis was performed using Pearson’s correlation coefficient. Logistic regression analysis was performed to determine univariate predictors of mild or greater PVR and balloon postdilatation. Receiver operating characteristic (ROC) curves were generated using mild or greater PVR and the occurrence of balloon postdilatation as classification variables, using the method of DeLong et al. Agreement between measurement methods was displayed with plots using the Bland-Altman method.
Results
Study Population
The population consisted of 55 women and 45 men with a mean age of 87.8 ± 8.3 years. Mean pre-TAVR calculated aortic valve area and peak transaortic velocity were 0.67 ± 0.17 cm 2 and 4.1 ± 0.76 m/sec, respectively. TAVR was performed in 85 patients via transfemoral access, in nine via transapical access, and in six via transaortic access. Sixty patients received SAPIEN THVs, and 40 received SAPIEN XT THVs. Ten patients received 29-mm THVs, 57 patients received 26-mm THVs, and 33 patients received 23-mm THVs. Balloon postdilatation was performed in 27 patients. No annular or root perforations occurred in the study population.
PVR
Immediate postprocedural echocardiographic assessment revealed no PVR in 49 of 100 patients. In the 51 patients with PVR, assessment was graded primarily by 3D color Doppler reconstruction in 44 of 51 patients. In seven of 51 patients with PVR, assessment was performed by a combination of visual estimation by 2D color Doppler and quantitative Doppler assessment of relative stroke volumes across the left ventricular outflow tract and right ventricular outflow tract, with six classified as trace and one classified as mild. At the conclusion of the procedure, 49 patients had no PVR, 28 had trace PVR, 16 had mild PVR, and seven had moderate PVR. No patient had more than moderate PVR.
Comparison of 3D TEE and Multidetector Computed Tomographic Measurements
Area Measurements Performed by Experts
Table 1 compares 3D TEE and computed tomographic annular measurements. Area measurements showed excellent correlation among the modalities (CT_Area, 3DE_Area_SA, and 3DE_Area_Direct; R = 0.88–0.93 for all comparisons). Although 3DE_Area_Direct was the lowest of the three methods, it was not statistically different from 3DE_Area_SA ( P = .14). Bland-Altman plots comparing computed tomographic and 3DE methods are shown in Figure 2 . The limits of agreement for CT_Area versus 3DE_Area_Direct were generally broader (−56.9 to +84.8 mm 2 ) than for comparison with 3D_Area_SA (−49.1 to +65 mm 2 ) Direct planimetric measurements exhibited the lowest interobserver and intraobserver reliability of all measurements ( Table 2 ). Test-retest variability was calculated as follows: CT_Area, R = 0.97; 3DE_Area_SA, R = 0.97; and 3DE_Area_Direct, R = 0.92.
| Area (mm 2 ) | Difference from MDCT (mm 2 ) | P | R vs MDCT ∗ | Percent oversizing | |
|---|---|---|---|---|---|
| CT_Area | 443 ± 79 | — | — | — | 15.5 ± 13.3 |
| Expert | |||||
| 3D_Area_SA | 435 ± 81 | 8 ± 29 | .007 | 0.93 | 17.7 ± 12.4 |
| 3D_Area_Direct | 429 ± 82 | 14 ± 36 | .0002 | 0.90 | 19.7 ± 15.6 |
| Novice | |||||
| 3D_Area_SA | 406 ± 90 | 37 ± 45 | <.0001 | 0.87 | 28.0 ± 20.7 |
| 3D_Area_Direct | 443 ± 93 | 0.3 ± 57 | .96 | 0.79 | 17.0 ± 19.3 |
| Interobserver (95% CI) | Intraobserver (95% CI) | |
|---|---|---|
| CT_Area | 0.90 (0.51–0.97) | 0.97 (0.93–0.99) |
| 3D_Area_SA | 0.91 (0.79–0.96) | 0.96 (0.91–0.99) |
| 3D_Area_Direct | 0.59 (0.21–0.82) | 0.90 (0.72–0.96) |
Percent Oversizing in Relation to PVR and Balloon Postdilatation
Percent oversizing was significantly lower in cases of mild or greater PVR and occurrence of balloon postdilatation, respectively ( Table 3 ). Multidetector computed tomographic area, direct planimetered area, and semiautomated planimetered area were all univariate predictors of PVR ( Table 4 ). Table 5 summarizes ROC analyses with areas under the curve for percent oversizing using the three measurement methods, using mild or greater PVR and occurrence of balloon postdilatation as classification variables. The upper cutoff values for oversizing with the highest combination of sensitivity and specificity are also listed. Discriminatory ability for mild or greater PVR was good for all methods. However, the highest cutoff for percent oversizing occurred with direct planimetry. A similar trend was present when dichotomizing percent oversizing values by occurrence of balloon postdilatation.
