Three-Dimensional Echocardiography for Transcatheter Aortic Valve Replacement Sizing: A Systematic Review and Meta-Analysis.

Background Transcatheter aortic valve replacement (TAVR) is the standard of care for many patients with severe symptomatic aortic stenosis and relies on accurate sizing of the aortic annulus. It has been suggested that 3-dimensional transesophageal echocardiography (3D TEE) may be used instead of multidetector computed tomography (MDCT) for TAVR planning. This systematic review and meta-analysis compared 3D TEE and MDCT for pre-TAVR measurements. Methods and Results A systematic literature search was performed. The primary outcome was the correlation coefficient between 3D TEE- and MDCT-measured annular area. Secondary outcomes were correlation coefficients for mean annular diameter, annular perimeter, and left ventricular outflow tract area; interobserver and intraobserver agreements; mean differences between 3D TEE and MDCT measurements; and pooled sensitivities, specificities, and receiver operating characteristic area under curve values of 3D TEE and MDCT for discriminating post-TAVR paravalvular aortic regurgitation. A random effects model was used. Meta-regression and leave-one-out analysis for the primary outcome were performed. Nineteen studies with a total of 1599 patients were included. Correlations between 3D TEE and MDCT annular area, annular perimeter, annular diameter, and left ventricular outflow tract area measurements were strong (0.86 [95% CI, 0.80-0.90]; 0.89 [CI, 0.82-0.93]; 0.80 [CI, 0.70-0.87]; and 0.78 [CI, 0.61-0.88], respectively). Mean differences between 3D TEE and MDCT between measurements were small and nonsignificant. Interobserver and intraobserver agreement and discriminatory abilities for paravalvular aortic regurgitation were good for both 3D TEE and MDCT. Conclusions For pre-TAVR planning, 3D TEE is comparable to MDCT. In patients with renal dysfunction, 3D TEE may be potentially advantageous for TAVR measurements because of the lack of contrast exposure.

Clinical Perspective

What Is New?

The main findings of this study are that 3‐dimensional transesophageal echocardiography annular measurements (annular area, annular perimeter, and left ventricular outflow tract area) are strongly correlated to multidetector computed tomography measurements.

Intraobserver and interobserver concordances for annular cross‐sectional area, perimeter, and diameter were high for both techniques, with a correlation coefficient around 0.9.

In a sensitivity analysis limited to 3 studies, multidetector computed tomography and 3‐dimensional transesophageal echocardiography predicted post–transcatheter aortic valve replacement paravalular aortic regurgitation with similar accuracy.

What Are the Clinical Implications?

Three‐dimensional transesophageal echocardiography does not require contrast media and may potentially reduce the rate of kidney injury, which may be particularly beneficial in patients with impaired baseline renal function.

Introduction

Transcatheter aortic valve replacement (TAVR) is currently the standard of care for many patients with severe symptomatic aortic stenosis. TAVR implantation relies on accurate sizing of the aortic annulus to optimize aortic valve flow dynamics while minimizing paravalvular aortic regurgitation (PVAR). Historically, efforts to measure the aortic annulus used 2‐dimensional echocardiography; however, the current accepted gold standard has become multidetector computed tomography (MDCT) with the focus on area or perimeter measurements of the often eccentric annulus.1, 2 In recent years, several studies have shown accurate assessment of the aortic annulus with 3‐dimensional transesophageal echocardiography (3D TEE). As 3D TEE does not require contrast medium, this could clinically benefit patients with impaired renal function.

In this systematic review and meta‐analysis, we aim to summarize current evidence on the comparison between 3D TEE and MDCT for TAVR annular measurements.

Methods

The authors declare that all supporting data are available within the article and its online supplementary file.

Search Strategy

A medical librarian performed comprehensive searches to identify randomized trials and observational studies comparing aortic valve measurements by different imaging techniques. Searches were run on September 5, 2018, in the following databases: Ovid MEDLINE (ALL; 1946 to August 10, 2018); Ovid Embase (1974 to present); and the Cochrane Library (Wiley). The search strategy included all appropriate controlled vocabulary and keywords for the interventions: “two‐dimensional echocardiography,” “three‐dimensional echocardiography,” “multidetector row CT,” and “aortic valve.” The full search strategy for Ovid MEDLINE is available in Table S1. To limit publication bias, there were no publication date, language, or article type restrictions on the search strategy.

Study Selection and Quality Assessment

Searches across the chosen databases retrieved 4960 results. After results were de‐duplicated, 2 independent reviewers (I.H., M.R.) screened a total of 3835 citations. Discrepancies were resolved by the senior author (M.G.). Titles and abstracts were reviewed against predefined inclusion/exclusion criteria. Articles were considered for inclusion if they were written in English and were studies comparing 3D TEE with MDCT in patients with aortic stenosis. Animal studies, case reports, conference presentations, editorials, expert opinions, studies not clearly defining the imaging technique used, and studies not defining and/or reporting measurements of aortic valve were excluded.

Full text was pulled for the selected studies for a second round of eligibility screening. Reference lists for articles selected for inclusion in the study were also searched for relevant articles. The full Preferred Reporting Items for Systematic Reviews and Meta‐Analyses flow diagram outlining the study selection process is available in Figure S1.3 Two independent investigators (I.H., M.R.) reviewed all studies, and disagreements were resolved by the senior author (M.G.). For overlapping studies, the largest series were included.

Two investigators (I.H., F.K.) performed data extraction independently, and the extracted data were verified by a third investigator (M.R.) for accuracy. The following variables were included: study demographics (sample size, publication year, design, institution, and country) (Table 1),4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22 patient demographics (age, sex, body mass index, body surface area, and comorbidities [hypertension, previous myocardial infarction, diabetes mellitus], normal aortic valve or aortic stenosis, aortic valve morphology [bicuspid or tricuspid]) (Table S2), imaging procedure–related variables, and correlations between MDCT and 3D TEE in assessing distance to the left coronary ostia (Table 2).

Table 1

Summary of Included Studies

Study/Year	Study Period	Hospital/Country	Country	Type of Study	Total No. of Patients
Garcia‐Martin/20164	2012–2014	Ramon y Cajal University Hospital	Spain	Retrospective	31
Guez/20175	2014–2015	Thomas Jefferson University	United States	Retrospective	74
Hafiz/20176	2012–2015	University of Massachusetts Medical School	United States	Retrospective	111
Hammerstingl/20147	···	University Hospital Bonn	Germany	Retrospective	138
Husser/20138	2011–2011	University of Regensburg Medical Center	Germany	Retrospective	57
Jilaihawi/20139	···	Cedars‐Sinai Heart Institute	United States	Retrospective	256
Kato/201810	2016–2016	Tokyo Medical Dental University	Japan	Retrospective	43
Machida/201511	2011–2014	St. Marianna University School of Medicine	Japan	Prospective	126
Mediratta/201712	···	University of Chicago Medical Center	United States	Prospective	47
Otani/201013	2008–2009	University of Occupational and Environmental Health	Japan	Retrospective	35
Prihadi/201814	···	Leiden University Medical Centre	Netherlands	Retrospective	150
Stahli/201415	2008–2012	University Hospital Zurich	Switzerland	Retrospective	39
Tamborini/201216	2008–2011	Centro Cardiologico Monzino	Italy	Retrospective	119
Pinto Teixeira/201717	2014–2015	Hospital de Santa Marta	Portugal	Prospective	60
Vaquerizo/201618	2013–2014	McGill University Health Center	Canada	Prospective	50
Wu/201419	···	University of Occupational and Environmental Health	Japan	Retrospective	40
Ng/201020	···	Leiden University Medical Center	Netherlands	Retrospective	53
Khalique/201421	2011–2013	Columbia University Medical Center/New York Presbyterian Hospital	United States	Retrospective	100
Wiley/201622	2012–2014	Mount Sinai Medical Center	United States	Retrospective	70

Table 2

Summary of Imaging Variables of the Included Studies

Study/Year	MDCT Technique	3D TEE Technique (Axis)	Measurement Phase	Software for 3D Data Set Analysis	Imaging Modality Used for TAVR Sizing	LVEF (SD)	Correlation (r) Between MDCT and 3D TEE in Assessing Distance to Left Coronary Ostia	Baseline Echo‐Measured Mean TAG (SD), m/s	Baseline Echo‐Measured AVA of Patients Pre‐TAVR (SD), cm2
Garcia‐Martin/20164	Manual (coronal and sagittal)	Automatic (long‐axis)	Systole	Philips Q‐Lab	3D TEE and MDCT	58.2 (11)	NR	46.3 (16)	0.7 (0.2)
Guez/20175	Manual (3‐point measurement)	Manual (long‐axis)	Systole	Philips Q‐Lab	MDCT	···	NR	···	···
Hafiz/20176	Manual (multiplanar reformations)	Manual (long‐axis)	Systole	Philips Q‐Lab	MDCT	52.29 (13.57)	NR	43.33 (16.44)	0.71 (0.19)
Hammerstingl/20147	···	···	···	···	3D TEE and MDCT	50.5 (14.8)	NR	42.7 (16.8)	0.7 (0.2)
Husser/20138	Manual (coronal and sagittal)	Manual (long‐axis)	Systole	Philips Q‐Lab	3D TEE and MDCT	···	NR	50 (16)	0.67 (0.22)
Jilaihawi/20139	Manual (coronal and sagittal)	Manual (long‐axis)	Systole	Philips Q‐Lab	3D TEE and MDCT	···	NR	···	···
Kato/201810	Manual (coronal and sagittal)	Manual (long‐axis)	Systole	ACUSON SC2000 PRIME (Siemens Medical)	MDCT	58.3 (10.2)	NR	47.0 (16.8)	0.58 (0.12)
Machida/201511	Manual (coronal and sagittal)	Manual (long‐axis)	Systole	Philips Q‐Lab	NR	60 (13)	NR	34 (19)	0.79 (0.22)
Mediratta/201712	Manual (coronal and sagittal)	Manual (long‐axis)	Systole	Philips Q‐Lab	3D TEE and MDCT	57 (16)	NR	40 (13)	0.8 (0.2)
Otani/201013	Manual (coronal and sagittal)	Manual (long‐axis)	Systole	Philips Q‐Lab	MDCT	···	0.89	38 (20)	1.1 (0.4)
Prihadi/201814	Manual (coronal and sagittal)	Automatic (long‐axis)	Systole	iE33 and EPIQ7 (Philips Medical)	MDCT	50.0 (11.8)	NR	43.5 (19.6)	0.8 (0.3)
Stahli/201415	Manual (coronal and sagittal)	Manual (long‐axis)	Systole	Philips Q‐Lab	NR	56 (2)	NR	41 (2.3)
Tamborini/201216	Manual	Manual (long‐axis)	Systole	Philips Q‐Lab	3D TEE and MDCT	58 (12)	0.83	52 (15)	0.65 (0.16)
Pinto Teixeira/201717	Manual	Manual (long‐axis)	Systole	Philips Q‐Lab	NR	59.8 (13.9)	NR	49.4 (14.4)	0.62 (0.20)
Vaquerizo/201618	Manual (coronal and sagittal)	Manual (long‐axis)	Systole		MDCT	56.2 (13.1)	NR	···	···
Wu/201419	Manual (coronal and sagittal)	Manual (long‐axis)	Systole	X7‐2t (Philips)	3D TEE and MDCT	49 (11)	0.71	···	···
Ng AC/201020	Manual (coronal and sagittal)	Manual (long‐axis)	Systole	Philips Q‐Lab	3D TEE and MDCT	61.1 (27.7)	NR	···	···
Khalique/201421	Manual	Semiautomated (long‐axis)	Systole	Philips Q‐Lab	3D TEE and MDCT	···	NR	4.1 (0.76)	0.67 (0.17)
Wiley/201622	Manual	···	···	Philips Q‐Lab	3D TEE and MDCT	58.2 (11)	NR	46.3 (16)	0.7 (0.2)

3D indicates 3‐dimensional; AVA, aortic valve area; LVEF, left ventricular ejection fraction; MDCT, multidetector row computed tomography; NR, not reported; TAG, transaortic gradient; TAVR, transcatheter aortic valve replacement; TEE, transesophageal echocardiography.

For the pooled receiver operating characteristic analysis on PVAR, the incidence of PVAR, sensitivity, specificity, and receiver operating characteristic area under the curve (AUC) values of: (1) absolute differences between TAVR prosthesis size and 3D TEE and MDCT annulus diameter, area, perimeter; and (2) the covering indices of 3D TEE– and MDCT‐measured annular sizes in terms of nominal prosthesis sizes (diameter, area, and perimeter), as defined in the individual studies. Details of calculation of the pooled receiver operating characteristic are presented in Table S3.

The quality of the included studies was assessed using the Newcastle‐Ottawa Scale for observational studies (Table S4).23

Measurements

The primary outcome was the correlation coefficient between 3D TEE and MDCT measurements of annular area. Secondary outcomes were the correlation coefficients between the 2 techniques for mean annular diameter, annular perimeter, left ventricular outflow tract area (LVOT‐A), interobserver and intraobserver agreements, and mean differences between 3D TEE and MDCT measurements. In addition, sensitivity, specificity, and AUC value for prediction of post‐TAVR PVAR were calculated from the studies reporting this information.

Data and Statistical Analyses

The correlation coefficients between 3D TEE and MDCT measurements from each study were used as effect sizes and transformed using Fisher z. Summary estimates were then reconverted to correlations that were reported as correlation coefficient (r) and 95% CIs, while mean differences were pooled and reported as mean difference and 95% CIs using DerSimonian Laird (inverse variance) method and DerSimonian‐Laird estimator for tau2.24, 25 Random and fixed effect models were used. The Cochran Q statistic and the I 2 test were used to assess studies’ heterogeneity.26 Leave‐one‐out sensitivity analysis was performed for the primary outcome. Funnel plot and Egger's regression test were used to assess for potential publication bias (Figures S2 and S3).

Correlation coefficient (r) values between 0.7 and 1.0 (−0.7 and −1.0) were considered as indicating a strong positive (negative) correlation.

Meta‐regression was used to explore the effects of age, sex, body mass index, body surface area, left ventricular ejection fraction, mean transaortic gradient, and baseline aortic valve area on the primary outcome using the DerSimonian‐Laird method.

Pooled sensitivity and specificity with corresponding 95% CIs and pooled receiver operating characteristic for PVAR were estimated using a random effects model (DerSimonian‐Laird method).

Statistical analyses were performed using “meta” and “metafor” packages27, 28 in R (version 3.3.3 R Project for Statistical Computing) within RStudio (0.99.489, http://www.rstudio.com) and Meta‐DiSc software (version 1.4).

Results

Search, Study Selection, and Quality Assessment

A total of 3835 studies were retrieved, of which 194, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22 met the inclusion criteria and were included in the final analysis.

Seven studies originated from the United States and Canada, 10 from Europe, and 2 from Japan. Overall, 1599 patients were included. The number of patients in the individual studies ranged from 31 to 256. The mean age ranged from 70 to 88 years. Men ranged from 22% to 63%, and mean body mass index and body surface area ranged from 25.4 to 27.0 kg/m2 and 1.45 to 1.84 m2, respectively. Implanted TAVR size was determined by MDCT in 6 studies, by both 3D TEE and MDCT in 10 studies, while 3 studies did not report this information. Correlations between 3D TEE and MDCT in assessing distance to the left main coronary ostia were reported by 3 studies and were generally high (Table 2). Description of the included studies and imaging modalities are presented in Tables 1and 2 and Table S3. The mean and SDs of 3D TEE– and MDCT‐derived measures for annular area, annular diameter, annular perimeter, and LVOT‐A are shown in Figure S4. The number of studies and patients analyzed for the primary and secondary outcomes are presented in Table 3. The assessment of the quality of the individual studies is reported in Table S4.

Table 3

Outcomes Summary

Outcome	No. of Patients	No. of Studies	Random Effects Model			Fixed Effect Model
			Effect Estimate (95% CI)	Heterogeneity, I 2 (P Value)	Tau2	Effect Estimate (95% CI)	Heterogeneity I 2 (P Value)	Tau2
Correlation (3D TEE and MDCT)
Annular area	1321	15	r=0.86 (0.80–0.90)	92.0% (P<0.001)	0.1399	r=0.86 (0.80–0.90)	92.0% (P<0.001)	0.1399
Annular perimeter	378	5	r=0.89 (0.82–0.93)	82.1% (P=0.001)	0.0676	r=0.90 (0.88–0.92)	82.1% (P=0.001)	0.0676
Annular diameter	1093	12	r=0.80 (0.70–0.87)	91.9% (P<0.001)	0.1347	r=0.78 (0.76–0.81)	91.9% (P<0.001)	0.1347
LVOT‐A	99	2	r=0.78 (0.61–0.88)	60.8% (P=0.11)	0.0352	r=0.78 (0.69–0.85)	60.8% (P=0.11)	0.0352
Mean difference between 3D TEE and MDCT
Annular area, cm2	681	9	MD=−0.12 (−0.24 to 0.00)	0.0% (P=0.64)	0	MD=−0.12 (−0.24 to 0.00)	0.0% (P=0.64)	0
Annular perimeter, cm	243	3	MD=−0.02 (−0.65 to 0.61)	49.4% (P=0.14)	0.1536	MD=0.04 (−0.40 to 0.47)	49.4% (P=0.14)	0.1536
Annular diameter, cm	357	4	MD=−0.03 (−0.15 to 0.10)	0.0% (P=0.95)	0	MD=−0.03 (−0.15 to 0.10)	0.0% (P=0.94)	0
LVOT‐A, cm2	92	2	MD=−0.40 (−1.05 to 0.26)	0.0% (P=0.88)	0	MD=−0.40 (−1.05 to 0.26)	0.0% (P=0.88)	0
Intraobserver agreement
Annular area (MDCT)	428	6	r=0.96 (0.91–0.98)	92.3% (P<0.001)	0.1846	r=0.96 (0.95–0.97)	92.3% (P<0.001)	0.1846
Annular area (3D TEE)	691	9	r=0.94 (0.90–0.97)	92.3% (P<0.001)	0.1682	r=0.95 (0.94–0.96)	92.3% (P<0.001)	0.1682
Annular perimeter (MDCT)	185	2	r=0.94 (0.58–0.99)	97.6% (P<0.001)	0.6003	r=0.96 (0.95–0.97)	97.6% (P<0.001)	0.6003
Annular perimeter (3D TEE)	288	2	r=0.95 (0.94–0.96)	0.0% (P=1.00)	0	r=0.95 (0.94–0.96)	0.0% (P=1.00)	0
Annular diameter (3D TEE)	288	2	r=0.94 (0.90–0.96)	75.9% (P=0.04)	0.0224	r=0.94 (0.92–0.95)	75.9% (P=0.04)	0.0224
Interobserver agreement
Annular area (3D TEE)	582	8	r=0.92 (0.88–0.95)	85.3% (P<0.001)	0.0875	r=0.92 (0.91–0.93)	85.3% (P<0.001)	0.0875
Annular perimeter (MDCT)	185	2	r=0.88 (0.52–0.97)	95.1% (P=0.001)	0.2951	r=0.91 (0.88–0.93)	95.1% (P=0.001)	0.2951
Annular perimeter (3D TEE)	288	2	r=0.94 (0.87–0.97)	91.9% (P=0.001)	0.0804	r=0.94 (0.93–0.95)	91.9% (P=0.001)	0.0804
Annular diameter (3D TEE)	419	4	r=0.92 (0.88–0.94)	72.4% (P=0.01)	0.0277	r=0.91 (0.90–0.93)	72.4% (P=0.01)	0.0277

3D indicates 3‐dimensional; LVOT‐A, left ventricular outflow tract area; MD, mean difference; MDCT, multidetector row computed tomography contrast angiography; r, correlation coefficient; TEE transesophageal echocardiography.

Meta‐Analysis

The overall heterogeneity and results for pooled correlations, mean differences, and intraobserver and interobserver agreements are summarized in Table 3.

The correlations between 3D TEE and MDCT annular area, annular perimeter, annular diameter, and LVOT‐A measurements were strong (0.86 [CI, 0.80–0.90]; 0.89 [CI, 0.82–0.93]; 0.80 [CI, 0.70–0.87]; and 0.78 [CI, 0.61–0.88], respectively) (Figure S5).

The mean differences between 3D TEE and MDCT for annular area, annular perimeter, and mean annular diameter measurements were small and not statistically different (−0.12 cm2 [CI, −0.24 to 0.00], −0.02 cm [CI, −0.65 to 0.61], and −0.03 cm [CI, −0.15 to 0.10]). The mean difference between 3D TEE and MDCT for LVOT‐A was −0.40 cm2 [CI, −1.05 to 0.26] (Figure S6).

The pooled intraobserver agreements for:

Annular area in 3D TEE and MDCT were 0.94 (CI, 0.90–0.97) and 0.96 (CI, 0.91–0.98), respectively.

Annular perimeter in 3D TEE and MDCT were 0.95 (CI, 0.94–0.96) and 0.94 (CI, 0.58–0.99), respectively.

Annular diameter in 3D TEE was 0.94 (CI, 0.90–0.96) (Figure S7).

The pooled interobserver agreements for:

Annular area in 3D TEE was 0.92 (CI, 0.88–0.95).

Annular perimeter in 3D TEE was 0.94 (CI, 0.87–0.97).

Annular diameter in 3D TEE was 0.92 (CI, 0.88–0.94).

Annular perimeter in MDCT was 0.88 (CI, 0.52–0.97) (Figure S8).

On meta‐regression, mean body mass index, body surface area, and transaortic gradient were associated with lower correlations between 3D TEE and MDCT annular area measurement (Table 4).

Table 4

Meta‐Regression of Patient and Imaging Variables on the Primary Outcome of Correlation Between 3D TEE and MDCT Annular Area Measurement

Variable	No. of Studies	β±SD* (P Value)
Age	13	0.00090±0.022 (P=0.98)
Male sex	13	−0.0075±0.0065 (P=0.24)
Body mass index	4a	−2.78±0.82 (P<0.001)a
Body surface area	3a	−2.55±0.68 (P<0.001)a
Left ventricular ejection fraction	9	−0.040±0.029 (P=0.16)
Transaortic gradient	10a	−0.036±0.013 (P<0.01)a
Aortic valve area	10	0.61±0.94 (P=0.52)

3D indicates 3‐dimensional; MDCT, multidetector row computed tomography contrast angiography; TEE, transesophageal echocardiography.

Positive β implies stronger correlation with increase in the explored variable while negative β implies weaker correlation with increase in the explored variable.

P‐value significant.

The pooled sensitivities, specificities, and receiver operating characteristic of 3D TEE and MDCT included 3 studies and are presented in Table 5. Discriminatory abilities for PVAR were good for both 3D TEE (annular area cover index AUC 0.83, standard error [SE] 0.04 and annular perimeter cover index AUC 0.93, SE 0.06; mean difference between prosthetic valve and 3D TEE–measured diameter AUC 0.63, SE 0.10) and MDCT (annular area cover index AUC 0.89, SE 0.03 and annular perimeter cover index AUC 0.93, SE 0.04; mean difference between prosthetic valve and MDCT‐measured diameter AUC 0.75, SE 0.13, respectively) (Figure and Figures S9 and S10).

Table 5

Diagnostic Performance of 3D TEE and MDCT in Discriminating Post–TAVR PVAR

Parameter	Imaging Modality	Pooled Sensitivity (95% CI)	Pooled Specificity (95% CI)	Pooled ROC (AUC, SE)
∆ Mean annular diameter	MDCT	0.83 (0.73–0.90)	0.63 (0.58–0.68)	0.75 (0.13)
	3D TEE	0.80 (0.70–0.87)	0.54 (0.49–0.59)	0.63 (0.10)
Cover index area	MDCT	0.81 (0.71–0.88)	0.78 (0.74–0.82)	0.89 (0.03)
	3D TEE	0.73 (0.62–0.82)	0.67 (0.62–0.71)	0.83 (0.04)
Cover index perimeter	MDCT	0.81 (0.71–0.88)	0.75 (0.70–0.79)	0.93 (0.04)
	3D TEE	0.73 (0.62–0.82)	0.64 (0.59–0.69)	0.93 (0.06)

∆ Indicates the difference between prosthetic valve and measured size; 3D, 3‐dimensional; AUC, area under curve; MDCT, multidetector row computed tomography; PVAR, paravalvular aortic regurgitation; ROC, receiver operating characteristic curve; SE, standard error; TAVR, transcatheter aortic valve replacement; TEE, transesophageal echocardiography.

Figure 1

Pooled receiver operating characteristic curves of (A) 3‐dimensional transesophageal echocardiography (3D TEE) and (B) multidetector row computer tomography (MDCT) annular area covering index for predicting paravalvular aortic regurgitation (PVAR). The red circles of different diameters represent different studies. Their true positive rates (sensitivity) and false positive rates (1‐specificity) for determining PVAR can be traced to the y‐ and x‐axes, respectively. Both 3D TEE and MDCT annular area cover indices are good in predicting PVAR (area under curve [AUC] 0.8268, standard error [SE] 0.0371; and AUC 0.8914, SE 0.337, respectively).

Discussion

Evaluation of aortic root anatomy has undergone significant evolution since the early days of TAVR. Minor axis assessment via 2‐dimensional echocardiography was the sole modality used to measure aortic annulus for sizing in early trials. These measurements often correlated loosely with true annular dimensions.29 Not surprisingly, PVAR rates were high and many of these were more than mild in nature.30

The introduction of MDCT imaging developed the ability to accurately assess annular dimensions. Significant data were also gained on root anatomy including coronary heights and sinus of Valsalva and sinotubular junction diameters.31 With this additional data in hand, it became possible to plan for and prevent many potential root complications such as coronary obstruction and root rupture. Guidelines were developed for valve selection based on annular perimeters and areas.32 As MDCT was included as a mandatory part of the evaluation of patients for entry into TAVR clinical trials, the contrast requirements were not of major concern as patients with significant renal insufficiency were excluded from participation by trial guidelines.33

As the technology gained commercial approval, the real‐world experience began to include patients with significantly impaired renal function. In this setting, 3D TEE was introduced as an alternative to standard MDCT evaluation. As all manufacturer‐provided sizing tables are based on MDCT measurements, an assumption was made that 3D TEE measurements could correlate closely with those obtained by MDCT. This study is the most comprehensive attempt to validate this assumption.

The main findings of this study are that: (1) 3D TEE annular measurements (annular area, annular perimeter, and LVOT‐A are strongly correlated with MDCT measurements; (2) intraobserver and interobserver concordance for annular cross‐sectional area, perimeter, and diameter was high for both techniques, with a correlation coefficient around 0.9; (3) in a sensitivity analysis limited to 3 studies,13, 15, 27 MDCT and 3D TEE predicted post‐TAVR PVAR with similar accuracy.

There has been only one previous attempt to systematically compare 3D TEE and MDCT for TAVR sizing.34 In a meta‐analysis of 13 studies and 1228 patients, Elkaryoni and colleagues34 described a high correlation between the 2 techniques for annular area, the only outcome that was investigated. However, the authors focused on a single measurement and did not evaluate correlation in terms of other important variables usually considered when planning TAVR.

Our study expands on the previous finding to include aortic annular perimeter and LVOT‐A measurements, which are important for TAVR preparation. Additionally, we compared both modalities in terms of prediction of post–valve deployment PVAR and evaluated interobserver and intraobserver agreement for both techniques.

There are modality‐specific reasons for differences in annular measurements between 3D TEE and MDCT. Due to different imaging techniques, the cross‐sectional plane chosen for annular measurements may be different. This may also explain the relatively large difference of 0.40 cm2 in LVOT‐A between 3D TEE and MDCT, where the location of the LVOT‐A cross‐sectional area measured by both modalities may not be consistent. Also, because of the differences in temporal resolution and appearance of calcifications, the image analysis may have been performed at different points in the cardiac cycle, as well as have different errors introduced by calcification.35

There are also modality‐specific strengths specific to 3D TEE and MDCT. Three‐dimensional TEE can provide real‐time intraoperative guidance, assess and aid in the decision‐making process for PVAR, and diagnose intraoperative complications such as annular rupture or coronary occlusion. If MDCT measurements are imprecise or fall in between 2 valve sizes (a common occurrence), 3D TEE may be used to add key information for sizing. MDCT instead is key in providing information on vascular access, coronary height, aortic valve, and LVOT‐A calcifications. Although 3D TEE is not primarily used for coronary height, there was good correlation for left coronary height with MDCT (Table 2).

Recent studies have shown that 3D TEE may underestimate annulus cross‐sectional area by ≈10% compared with MDCT.9, 20, 35 Our data confirm that 3D TEE measurements are on average smaller than corresponding MDCT measurements for annular area, diameter, and perimeter measurements. However, the mean differences (−0.12 cm2 [CI, −0.24 to 0.00] and −0.02 cm [CI, −0.65 to 0.61], respectively) may be irrelevant for TAVR planning purposes and are not statistically significant.

The clinical relevance may be inferred from whether this difference resulted in a different valve size, and possible undersizing, of the implanted valve, which may lead to increased risk of PVAR. A subset of 4 studies (Husser, Kato, Prihadi, Vaquerizo) analyzed transcatheter valve sizing agreement between 3D TEE and MDCT with considerable heterogeneity and varying results. Husser et al8 found that 3D TEE predicted final valve size in 84% of patients, while MDCT predicted 79%, with a 77% agreement between both modalities. Kato et al10 found that semiautomated 3D TEE measurements showed 77% agreement with final prosthesis implanted. Prihadi et al14 did not compare the 2 modalities to size of prosthesis implanted but found that there was 93.3% agreement in valve size between 3D TEE and MDCT. Finally, Vaquerizo et al18 found that there was only 38% agreement between MDCT and 3D TEE for valve size with up to 50% of patients measured by 3D TEE as having a hypothetical inappropriate valve size according to manufacturer‐recommended sizing algorithms (based on MDCT). Future studies including the incidence of PVAR and agreement in valve sizing may offer clinical information to the correlations found in this study.

The incidence of MDCT contrast–related kidney injury in patients evaluated for TAVR range from 7% to 10.5%.36 Three‐dimensional TEE does not require contrast media and may potentially reduce the rate of kidney injury, which may be particularly beneficial in patients with impaired baseline renal function. Analysis of the Society of Thoracic Surgery/American College of Cardiology Transcatheter Valve Therapy Registry shows that among the overall population of 44 778 TAVR candidates, there were 19 266 (43.03%) patients with stage 3 chronic kidney disease, 2413 (5.39%) with stage 4 chronic kidney disease, and 206 (0.46%) with stage 5 chronic kidney disease, suggesting that 3D TEE may be more appropriate than MDCT for TAVR measurements in a significant proportion of the current TAVR population.

While 3D TEE can substantially reduce contrast load requirements with MDCT, it does not address the issue of access and evaluation therein. With the elimination of MDCT evaluation of the aortic root, patients may have isolated iliofemoral computed tomography angiography with a substantially reduced contrast load. Alternatively, a noncontrast study can provide adequate measurements as well.

Study Limitations

Our study has several important limitations that must be acknowledged. Statistical heterogeneity (I 2) was high for most outcomes. However, this is to be expected in meta‐analyses evaluating continuous outcomes where traditional interpretations of heterogeneity may not be appropriate.37 A few studies used automated software for calculation of annular area, but most used manual tracing techniques, which are more likely operator dependent. However, we found little interobserver variability between studies, so this effect was likely minimal. Additionally, some heterogeneity between studies may be attributable to varying numbers of bicuspid and tricuspid valves included, although only one study mentioned bicuspid valves and we assumed that the other studies were performed in patients with tricuspid valves.

Conclusions

The present data suggest that 3D TEE is comparable to MDCT for TAVR planning. As 3D TEE does not require the use of contrast media, it may be advantageous in patients with preexisting renal dysfunction. Since the first TAVR performed in 2002, TAVR has transitioned from a novel procedure to standard of care for many patients with aortic stenosis. Early on, procedural safeguards such as general anesthesia and intraprocedural TEE were recommended.38 As TAVR has become increasingly safer with fewer complications, many programs are taking a minimalist approach including percutaneous transfemoral vascular access, sedation without general anesthesia, lack of intraprocedural TEE guidance, and early discharge protocols. This has not been shown to have improved outcomes, however, and some studies show that there is an increased incidence of PVAR from this approach. In the future, there may be a role for 3D TEE in patients presenting for TAVR with renal dysfunction, patients with indeterminate valve sizing with MDCT, and patients with complex anatomy at high risk for PVAR.

Sources of Funding

Rong is supported in part by the Foundation for Anesthesia Research and Education Training Grant (FAER MTRG‐CT‐08‐15‐2018‐Rong)––significant. Wijeysundera reports research funding from Edwards Lifesciences and Medtronic—significant. Fremes is supported in part by the Bernard S. Goldman Chair in Cardiovascular Research—significant.

Disclosures

Angiolillo reports receiving payments as an individual for: consulting fee or honorarium from Amgen, Aralez, AstraZeneca, Bayer, Biosensors, Boehringer Ingelheim, Bristol‐Myers Squibb, Chiesi, Daiichi‐Sankyo, Eli Lilly, Haemonetics, Janssen, Merck, PLx Pharma, Pfizer, Sanofi, and The Medicines Company; participation in review activities from CeloNova and St. Jude Medical; institutional payments for grants from Amgen, AstraZeneca, Bayer, Biosensors, CeloNova, CSL Behring, Daiichi‐Sankyo, Eisai, Eli‐Lilly, Gilead, Janssen, Matsutani Chemical Industry Co., Merck, Novartis, Osprey Medical, and Renal Guard Solutions—modest. Salemi reports being a clinical proctor for Medtronic and Edwards Lifesciences. The remaining authors have no disclosures to report.

Table S1. Ovid MEDLINE Search Strategy

Table S2. Demographics of Included Patients

Table S3. Details of Calculation of PROC on Paravalvular Aortic Regurgitation

Table S4. Summary of Critical Appraisal of Included Observational Studies Using the Newcastle‐Ottawa Quality Assessment Scale for Cohort Studies

Figure S1. Preferred reporting items for systematic reviews and meta‐analyses (PRISMA) flowchart of our analysis.

Figure S2. (A) Leave‐one‐out analysis and (B) funnel plot of included studies reporting correlation between 3‐dimensional transesophageal echocardiography (3D TEE) and multidetector row computed tomography (MDCT) for primary outcome (annular area).

Figure S3. Funnel plots (with trim‐and‐fill method) of included studies reporting correlation or mean difference between 3‐dimensional transesophageal echocardiography (3D TEE) and multidetector row computed tomography (MDCT) for (A) annular area (correlation; bias=1.62, P=0.60), (B) annular perimeter (correlation; bias=5.14, P=0.28), (C) annular diameter (correlation; bias=1.92, P=0.59), (D) annular area (mean difference; bias=−0.70, P=0.04), (E) annular perimeter (mean difference; bias=−8.09, P=0.23), and (F) annular diameter (mean difference; bias=−1.37, P=0.17). Only 2 studies reported left ventricular outflow tract area and could not be analyzed by funnel plot.

Figure S4. The mean and SDs of 3‐dimensional transesophageal echocardiography (orange) vs multidetector computed tomography (MDCT) (blue)–derived measures, (A) annular area (cm2), (B) annular diameter (cm), (C) annular perimeter (cm), and (D) left ventricular outflow tract area (LVOT‐A) (cm2).

Figure S5. Correlation between 3‐dimensional transesophageal echocardiography (3D TEE) and multidetector row computed tomography (MDCT) in measuring (A) annular area, (B) annular perimeter, (C) annular diameter, and (D) left ventricular outflow tract area (LVOT‐A).

Figure S6. Mean difference between 3‐dimensional transesophageal echocardiography (3D TEE) and multidetector row computed tomography (MDCT) in measuring (A) annular area (cm2) (P=0.05), (B) annular perimeter (cm) (P=0.95), (C) annular diameter (cm) (P=0.68), and (D) left ventricular outflow tract area (LVOT‐A) (cm2) (P=0.23). Negative numbers indicate 3D TEE underestimation compared with MDCT.

Figure S7. Intraobserver agreement in (A) multidetector row computed tomography (MDCT) annular area, (B) 3‐dimensional transesophageal echocardiography (3D TEE) annular area, (C) MDCT annular perimeter, (D) 3D TEE annular perimeter, and (E) 3D TEE annular diameter.

Figure S8. Interobserver agreement in (A) 3‐dimensional transesophageal echocardiography (3D TEE) annular area, (B) 3D TEE annular perimeter, (C) 3D TEE annular diameter, (D) multidetector row computed tomography (MDCT) annular perimeter, (E) MDCT annular diameter, and (F) MDCT annular area.

Figure S9. Pooled receiver operating characteristic curves of (A) 3‐dimensional transesophageal echocardiography (3D TEE) and (B) multidetector row computer tomography (MDCT) annular perimeter covering index for predicting paravalvular aortic regurgitation (PVAR).

Figure S10. Pooled receiver operating characteristic curves of (A) 3‐dimensional transesophageal echocardiography (3D TEE) and (B) multidetector row computer tomography (MDCT) difference between measured mean annular diameter and diameter of prosthetic valve for predicting paravalvular aortic regurgitation (PVAR).

Click here for additional data file.