Virtual Reality to Predict Paravalvular Leak in Severe Bicuspid Aortic Valve Stenosis in Transcatheter Aortic Valve Implants

© 2024 HMP Global. All Rights Reserved.
Any views and opinions expressed are those of the author(s) and/or participants and do not necessarily reflect the views, policy, or position of the Journal of Invasive Cardiology or HMP Global, their employees, and affiliates.

J INVASIVE CARDIOL 2024. doi:10.25270/jic/24.00019. Epub March 11, 2024.

Watch the accompanying author interview here.

Objectives. Severe aortic stenosis (AS) in bicuspid aortic valves (BAV) is associated with an increased risk of paravalvular leak (PVL) after a transcatheter aortic valve replacement (TAVR). Virtual reality (VR) has been shown to be an effective tool in surgical training, but its utility in clinical practice has not been studied. Here we present the first study to evaluate the use of VR simulation in pre-procedure planning and prediction of PVL in TAVR in patients with severe BAV AS. 

Methods. Twenty-two patients with severe BAV AS undergoing TAVR between 2014 and 2018 at the University of Minnesota were included in the study. VR simulation of TAVR implants was performed and implants were analyzed for PVL. The primary endpoint was the percent circumference of valve malapposition in VR as compared to the severity of PVL on post-procedure echocardiography.

Results. The median age was 78.26 years (IQR 63.77-86.79) and 40.9% (n = 9) were female. Our VR model accurately predicted the presence and absence of PVL in all patients (17/17 and 5/5, respectively). The mean circumferential PVL was 3.73 % ± 7.71. The receiver operator characteristic curve showed an area under the curve of 0.83 (0.59-1.00, P = .03) for malapposition in the VR-TAVR simulated model.

Conclusions. VR-TAVR implantation may predict PVL in severe BAV AS patients undergoing TAVR. 

Transcatheter aortic valve replacement (TAVR) is the mainstay therapy for severe tricuspid aortic stenosis. Severe bicuspid aortic valve (BAV) aortic stenosis (AS) accounts for at least 50% of severe aortic valve stenosis requiring aortic valve replacement (AVR).¹ It is the most common congenital cardiac defect, with a prevalence of 0.8% to 2%.^2-3 BAV AS has been traditionally treated with surgical aortic valve replacement (SAVR) even after the advent of TAVR.⁴ Severe BAV AS patients have also been excluded from many landmark trials.^5-9 Moreover, patients with severe BAV AS are often young, mechanical prostheses candidates, are at low surgical risk, and are at higher risk of TAVR-related complications.^2,10,11 Severe BAV AS treated with TAVR increased from 0.39% in 2011 to 4.16% in 2014, ¹² and the 2019 FDA approval of TAVR as a therapeutic option in low-risk populations and the BAV will accelerate this trend.¹⁰ As a result, a significant increase in TAVR in severe BAV AS is expected, and improved periprocedural planning is paramount, given the higher rate of periprocedural TAVR complications such as paravalvular leak (PVL) in these patients.^2,10,11

The development of PVL in the BAV is often due to asymmetric distribution of calcium, commissural calcification, and eccentricity of the aortic annulus, which can affect the sizing and increase malapposition of the prosthesis.¹³ PVL that is mild or greater is associated with increased short- and long-term mortality; therefore, it is imperative to identify those at risk, especially as TAVR is expanded to low-risk populations and the BAV.^14,15 Presently, correct identification of aortic root anatomy and valve size relies on noninvasive imaging, with multi-detector computed tomography (MDCT) as the gold standard.^13,16

Several factors have been put forth to mitigate the periprocedural complications of TAVR including predictive algorithms, improved pre-procedure planning with MDCT, newer generation TAVR valves designed to attenuate PVL, and procedural techniques such as post-dilation, valve-in-valve TAVR, and a variety of PVL closure devices.¹⁷ The BAV poses a particular challenge in almost all strategies designed to combat PVL associated with severe tricuspid AS. Increased annular size, calcium, and eccentricity are risk factors that may occur in severe tricuspid AS but are universal in the BAV and impact pre-procedural planning, sizing, and procedural techniques.

Emerging medical technologies such as 3-dimensional (3D) printing and digital modeling have shown promise in pre-procedure planning for TAVR in severe tricuspid AS.¹⁸ Virtual reality (VR) is another emerging technology that has been primarily utilized as a teaching tool in surgical subspecialties with various successes.^19,20 However, VR has not been utilized in TAVR planning and has never been used in the pre-procedural planning of patients with severe BAV AS who are at high risk of TAVR complications. VR is particularly well-suited for this approach due to rapid deployment, low cost, and intuitive operation. In this study, we demonstrate that VR simulations are also effective at accurately predicting the presence, location, and severity of PVL in severe BAV AS patients undergoing TAVR.

This retrospective study was performed at the University of Minnesota Medical Center (UMMC). Its structural cardiovascular division has performed over 1000 TAVRs since 2012. Twenty-two patients with severe BAV AS who underwent TAVR at the UMMC between 2014 and 2018 were included in the study. All patients underwent a preprocedural MDCT with at least a transthoracic echocardiogram performed before, during, and after the TAVR. Seven patients were excluded due to prior SAVR, valve-in-valve TAVR, and non-aortic TAVR.

VR simulations were performed using a commercially available Valve Index VR headset (Valve) and Elucis VR medical image processing software (Realize Medical). MDCT digital imaging and communications in medicine (DICOM) datasets were imported into Elucis. DICOM data was used to create interactive VR models and respective TAVR prostheses. The selection of TAVR type and size was true to the TAVR implant at the time of the procedure. An Elucis VR environment was used to simulate TAVR deployment. The combined VR model (VR-TAVR) was assessed for the presence, location, and severity of malapposition, and findings were compared with post-procedural PVL on echocardiography. The severity of the PVL was determined according to the Valve Academic Research Consortium 3: Updated Endpoint Definitions for Aortic Valve Clinical Research.²¹

The primary endpoint was the presence or absence of malapposition and the percent circumference of valve malapposition in VR-TAVR as compared to the severity of PVL on post-procedure echocardiography. The secondary endpoint was the specificity and sensitivity of the VR-TAVR to predict PVL as compared to calcium and annular eccentricity index (AEI) (calculated as 1 – minimal diameter/maximal diameter).¹⁸ Baseline characteristic variables were collected from electronic medical records and are listed in Table 1.

Categorical variables were described as percentages, and continuous variables as medians and interquartile ranges (IQR) or means and standard deviations. Chi-square and Fisher’s exact tests were used for categorical variables, and the k-sample median test and t-test to compare medians and means, respectively. A receiver operator characteristic (ROC) curve was used to determine the accuracy of malapposition on virtual simulation to predict PVL. A P-value of less than .05 was considered significant. SPSS version 27.0 (IBM) was used.

The study was approved by the Institutional Review Board at the University of Minnesota. Informed consent was waived due to the retrospective nature of the study.

Twenty-two patients who had TAVR for severely stenotic BAVs were included. The median age was 78.26 years (IQR 63.77-86.79) and 40.9% (n = 9) were female. Five patients (22.7%) were Sievers type 0 and 17 (77.3%) were Sievers type 1. Twelve patients (54.5%) had a balloon-expandable valve implanted, and 10 (45.5%) received a self-expanding valve (Table 1). The median valvular calcium score was 2237 (IQR 1825-4781). Patients who developed PVL after TAVR were more likely to receive a self-expanding valve (n = 7 [77.8%] vs n = 3 [23.1%], P = .03), and had a higher AEI (median of 0.21 [IQR 0.13-0.25] vs 0.12 [0.02-0.19], P = .03) (Table 1). Videos 1 and 2 illustrate examples of a combined VR model (VR-TAVR) of the aortic root and TAVR valve.

Figure 1 features 3 patient scenarios. Patient 1 had a 25% malapposition in VR-TAVR and was found to have a moderate PVL. His cardiac CT prior to TAVR showed moderate annular calcification. Patient 8 had an approximately 10% malapposition and was found to have mild PVL. He had no annular calcification. Patient 12 had no malapposition, no PVL, and mild annular calcification.

Our VR-TAVR simulation accurately predicted the presence and absence of PVL in all patients (17/17 and 5/5, respectively). The mean circumferential PVL was 3.73% ± 7.71% (Table 2). Based on our VR TAVR simulation, the median oversizing with balloon-expandable valves was 16.3% (IQR 4.81-30.37) and only 7.95% (IQR 5.45-17.1) in the self-expanding valves (Table 2).

The ROC curve showed an area under the curve (AUC) of 0.83 (0.59-1, P = .03) for malapposition in VR-TAVR, an AUC of 0.65 (0.36-0.94, P = .33) for annular calcium, an AUC of 0.58 (0.29-0.88, P = .59) for calcified raphe, and an AUC of 0.48 (0.18-0.78, P = .91) for the AEI (Figure 2).

There were other adverse events noted in our patient population directly related to pre-procedure planning and sizing. In patient 1 (Sievers type 0), there was no significant valve oversizing based on VR-TAVR, and moderate annular calcification was present (Table 3). After valve implantation, a moderate PVL was noticed on transesophageal echocardiography. Post-dilation, PVL did not resolve with a 23-mm balloon, and another 26-mm Edwards Sapien valve was implanted 3 mm lower in the left ventricular outflow tract (LVOT). This was complicated by annular rupture, cardiac arrest, and death in the immediate post-procedural period. In patient 16 (Sievers type 0), 3 days after successful placement of a 23-mm Edwards Sapien valve, the valve dislodged with a significant degree of stenosis and required urgent SAVR with a 21-mm St Jude Trifecta tissue valve. Patient 12 (Sievers type 1) had significant annular calcification at the right coronary cusp level and developed a ventricular septal defect post-procedure. The TAVR valve in patient 9 (Sievers type 0) did not expand completely and had 7% oversizing based on VR-TAVR, although only mild annular calcification was present. Patient 15 (Sievers type 1) had a low implant (-1.2% oversizing based on VR-TAVR). Patient 8 (Sievers type 0) developed mild PVL. Patient 2 (Sievers type 1) had an annular area of 7.5 cm², annular perimeter of 95.6 mm, 10% malapposition, and a mild PVL even with a 34-mm Evolut valve.

In our small cohort of all comers with severe BAV AS, VR simulation and implantation of both balloon-expandable and self-expanding TAVR prostheses identified common periprocedural complications such as PVL with high fidelity. Both traditional risk factors and 3D-printed modeling have been shown to help predict or identify patients at high risk for PVL.^22,23

In tricuspid severe AS, the incidence of greater than mild PVL has been largely attenuated in both self-expanding and balloon-expandable prostheses.²⁴ These advances have been driven by newer generation valves with sealing ‘skirts’ placed at the base of the prostheses, improved procedural planning, and procedural modifications such as post-dilation and oversizing. However, all are associated with potential complications such as high degree atrioventricular block, complete heart block, pacemaker requirement, and annular rupture.^25,26 These complications occur with higher frequency in patients with bicuspid severe AS where traditional risk factors such as calcium, annular size, and annular eccentricity are almost universal.^22,27 Ex-vivo simulations can provide insight into procedural complications and guide valve selection (self- vs balloon-expandable) and size without exposing the patient to the additional risk of post-dilation and multiple valve deployments.¹⁸

We have previously reported that ex vivo-simulated TAVR implantation in 3D-printed models and visual analysis of implanted models can predict PVL in tricuspid severe AS with a sensitivity and specificity of 80% and 90%, respectively.¹⁸ Not only can 3D printing and modeling identify anatomy at risk of PVL, but it can predict the location of post-procedural PVL.^18,28 Its accuracy and peri-procedural characterization exceed the predictive capacity of conventional risk factors such as annular calcium, annular size, and AEI.¹⁸ In a retrospective study of 18 TAVR patients with severe tricuspid AS, TAVR implantation was simulated ex vivo in 3D-printed aortic root models and allowed for quantification of the annular bulge index, a novel marker of annular strain unevenness. Compared to more common PVL predictors, such as annular calcium and ellipticity, the annular bulge index was the only significant predictor of moderate-to-severe PVL (AUC of 0.95, P < .001).²⁸ However, 3D printing and implantation can be a cumbersome and expensive process, requiring a 3D printer capable of composite prints, 3D printing filaments, laboratory space, 3D modeling software, and the availability of TAVR prostheses or high-fidelity replicas. This requires not only resources and expense, but critical knowledge in multiple disciplines such as 3D printing techniques, software proficiency, and implant availability, none of which are commonly available in the community setting.

VR simulations are low cost, require few resources, and necessitate a rapid learning curve due to intuitive interaction in the VR space. Our study presents several important and novel points. First, this is the first study that utilizes VR-simulated implants of MDCT-generated aortic annuli and both balloon- and self-expanding TAVR prostheses. These digital models can be analyzed in the VR space for any potential problems that may arise during the planned procedure. Second, this is the first study using VR-simulated implants in severe BAV AS. Third, our data support the initial hypothesis that VR-simulated implants can predict PVL with very high specificity and sensitivity, including accurate PVL localization. Fourth, we show that the use of traditional risk factors such as annular calcium, AEI, and annular size in severe BAV AS does not correlate with the incidence of PVL like it does in cases of severe tricuspid AS.

The predictive failure of traditional risk factors for PVL in severe BAV AS is related to their ubiquity in the BAV. Prior studies have observed higher procedural mortality rates in type 0 compared with type 1 BAV.^29-31 Type 0 BAV predisposes to the under-expansion of prosthetic heart valves due to anatomical variation— 2 commissures opening in an elliptical fashion— which impose greater restrictive forces on the prostheses.³²  This is associated with higher post-TAVR peak and mean aortic gradients, an increased risk of PVL, and lower aortic valve areas.^33,34 Ultimately, TAVR in type 0 BAV has been observed to have a lower Valve Academic Research Consortium-2 device success rate and may explain our observations.³⁵

VR is changing the medical landscape and may be the solution to future cost-effective periprocedural planning to reduce complications, guide valve size and selection, and streamline procedures. Future routine implementation of additional algorithms to simulate tissue elasticity as well as computational fluid dynamics can augment the VR simulation with realistic interactions and hemodynamic assessment of valves including post-procedural gradients, PVL quantification, and early identification of leaflet thrombosis.³⁶

Limitations. Our study has several limitations. First, this is a retrospective single-center observational study with a small sample size, therefore, further larger studies are needed to validate the current model. Second, there was no echocardiography core laboratory to evaluate PVL severity, so these interpretations were not standardized. Third, multiple generations of TAVR valves were used, which may impact the observed occurrence of PVL. Fourth, CT quality remains the rate-limiting step in generating high fidelity models, and significant artifact cannot be entirely overcome and may limit prognostication (though all artifact was manageable in the study). Fifth, the VR model may not fully simulate the behavior of the surrounding tissues and further larger studies are necessary to validate the current predictive model. Sixth, while we considered the depth of the TAVR valve when reconstructing the LVOT, we did not consider the balloons’ volumes and valvular angles, so this could limit accurate reconstruction. Seventh, the TAVR outcomes were not blinded to the VR operators, which may introduce bias.

This is the first study to demonstrate that preoperative VR simulation and planning of both balloon-expandable and self-expanding TAVR prostheses have the potential to identify patients at high risk of common peri-procedural complications such as PVL in patients with severe BAV AS.

From the ¹Department of Cardiovascular Disease, University of Minnesota Medical School, Minneapolis, Minnesota, USA; ²Department of Internal Medicine, University of Minnesota Medical School, Minneapolis, Minnesota, USA.

Disclosures: The authors report no financial relationships or conflicts of interest regarding the content herein.

Address for correspondence: Sergey Gurevich, MD, Department of Cardiovascular Disease, University of Minnesota Medical School, 420 Delaware Street, SE, MMC 508, Minneapolis, MN 55455, USA. Email: gure0011@umn.edu; X: @ChahineJohnny