Valve thrombosis following transcatheter aortic valve replacement: significance of blood stasis on the leaflets

Abstract

OBJECTIVES: Leaflet thrombosis following transcatheter aortic valve replacement (TAVR) and valve-in-valve (ViV) procedures has been increasingly recognized. However, the factors affecting the post-TAVR/ViV thrombosis are not fully understood. This study aimed to investigate the effect of the geometric confinement of transcatheter aortic valve (TAV) on blood residence time (BRT) on the TAV leaflets and in turn on the post-TAVR valve thrombosis.

METHODS: Two computational models, representing a surgical bioprosthesis and a TAV, were developed to study the effect of the geometric confinement on BRT on the leaflets in ViV setting/TAVR Intra-annular positioning. 3D flow fields were obtained via a one-way fluid-solid interaction modelling approach validated by experimental testing. BRT was compared between the two models by quantification and statistical analysis of the residence time of randomly distributed particles in close proximity of the leaflets.

RESULTS: Significantly longer BRT on the leaflets was observed in the TAV compared to the surgical valve during different stages of the cardiac cycle. During forward flow, the mean value of BRT was found to be 39% higher in the TAV compared to the surgical bioprosthesis (P < 0.0001). During diastole, specifically from end-systole to mid-diastole and from mid-diastole to the beginning of systole, the amount by which the mean BRT was higher for TAV compared to the surgical valve was 150% and 40%, respectively (P < 0.0005).

CONCLUSIONS: The geometric confinement of TAV by the failed bioprosthesis or the calcified native valve increases the BRT on the TAV leaflets. This may act as a permissive factor in valve thrombosis.

Transcatheter heart valve replacement , Valve-in-valve , Thrombosis , Blood residence time

INTRODUCTION

Transcatheter aortic valve replacement (TAVR) is a well-established therapy for patients with severe symptomatic aortic stenosis at high risk for surgical aortic valve replacement (SAVR). Subclinical leaflet thrombosis associated with reduced leaflet motion is an increasingly recognized complication observed in post-TAVR patients [1–5]. Thrombus formation has also been observed in patients following valve-in-valve (ViV) procedures [2, 6–8]. Valvular thromboses were predominantly found on the aortic surface of the valve leaflets and to a lesser extent on the ventricular surface [1, 2]. Currently, post-TAVR antithrombotic or anticoagulation therapies are not well established and further investigation is required to determine the most favourable approach in the prevention and management of post-TAVR valve thrombosis. The present post-procedural management of TAVR is based on the history, outcomes, and treatment responses of SAVR [2]. The incidence of bioprosthetic thrombosis after SAVR is relatively low: 0.03–1.46% [4, 9, 10]. However, the frequency may be underestimated because computed tomographic or echocardiographic assessments of surgical bioprosthetic leaflet motion are not routinely done after SAVR. In stark contrast to SAVR patients, the occurrence of sub-clinical thrombosis and reduced leaflet motion in TAVR/ViV patients may be as high as 40% as shown in a recent clinical trial [1]. This has raised concerns about the antithrombotic therapies following TAVR. In addition, the underlying mechanism of post-TAVR leaflet thrombosis has remained unclear. Therefore, it is of the utmost importance to investigate the essential factors affecting the valvular thrombogenesis following TAVR/ViV procedures.

In the present study, we seek to shed light on the precursor mechanism by which post-TAVR/ViV leaflet thrombosis is facilitated. In contrary to SAVR, in TAVR/ViV procedures, trans- catheter aortic valves (TAVs) are implanted within native or bioprosthetic valves. Therefore, the aortic portion of the stent frame is circumferentially surrounded by the calcified leaflets of the native valve in TAVR setting, or by the leaflets and frame of the degenerated bioprostheses in ViV setting (Fig. 1). This configuration is more pronounced in TAVs that tend to operate inside the annulus than the supra-annular TAVs. We hypothesize that the exceedingly long blood residence time (BRT) on the aortic surface of TAV leaflets due to the circumferentially surrounded TAV frame, could explain the relatively high rate of leaflet thrombosis following TAVR/ViV procedures. Although the increased BRT does not lead to thrombogenesis by itself, it has been suggested that once the clotting process is triggered, for instance by blood contact with the foreign surfaces of TAVs, thrombosis is more likely to occur in low flow regions characterized by large BRT [11, 12]. Hence, blood stasis on the aortic surface of leaflets following TAVR/ViV procedures should be considered as a permissive factor in thrombus formation and thromboembolic events. The hypothesis presented in this work is supported by the quantification of BRT and comparison of the calculated values via statistical analysis in two different model geometries representing a surgical bioprosthesis and a TAV. Such quantification was accomplished via an in-vitro assessment of a bioprosthesis heamodynamics along with computer simulations of the flow field that encompassed the complex 3D deformation of valve leaflets. Characterization of detailed flow field in close proximity of the leaflets is significantly difficult to achieve with the available imaging modalities in real-world invivo clinical cases. To the authors’ knowledge, this study is unique in terms of the novelty of both the hypothesis as well as the methods employed.

Figure 1:

(A) The aortic portion of TAV stent frame is circumferentially surrounded by the calcified leaflets in TAVR. Reprinted with permission from [23]. (B) Note the presence of a tan-white thrombus (red arrow) shown on the aortic surface of SAPIEN valve following TAVR. Reprinted with permission from [2]. (C) The aortic portion of TAV stent frame is circumferentially surrounded by leaflets and frame of the degenerated bioprosthesis in ViV setting. Reprinted with permission from [24]. (D) Thrombotic material adhered to the three leaflets of SAPIEN XT following ViV in CoreValve. Reprinted with permission from [6]. LCC: left coronary cusp; NCC: non-coronary cusp; RCC: right coronary cusp; TAVR: transcatheter aortic valve replacement; ViV: valve-in-valve.

MATERIALS AND METHODS

To study the effect of the confining geometry of ViV setting/TAVR intra-annular positioning on the BRT on TAV leaflets, and in turn on the likelihood of leaflet thrombosis, a fluid-solid interaction (FSI) modelling approach was incorporated. In this approach, computer simulations are employed to mimic the 3D motion and deformation of the valve leaflets along with the 3D time-dependent flow field of the bioprosthesis. The essential input parameter for the simulations, i.e. transvalvular pressure gradient, was obtained by a heart flow simulator experimental setup. Furthermore, experimental velocity measurements of the flow field was performed and used to confirm the validity of the modelling approach. It should be noted that the lack of optical access at the basal part of the TAV, where the leaflets are inserted to the metallic frame and surrounded by the calcified leaflets in TAVR, as well as the intrinsic complexity of the bioprosthetic 3D flow field prevent accurate experimental flow measurements in the vicinity of the leaflets. Therefore, computer simulations are motivated. In this section, a brief overview of the entire procedure is presented. The detailed description of each module employed in the method is reported in the Supplementary material.

A custom-built pulse duplicator system (Fig. 2) that simulates the heart physiological conditions was employed to obtain the transvalvular pressure waveform and the flow rate (Supplementary Material, Module S1). A 2D particle image velocimetry (PIV) system was used to perform flow velocity measurements downstream the bioprosthesis for validation of simulations (Supplementary Material, Module S2). The measured transvalvular pressure gradient was used in a finite element (FE) simulation to obtain the complex 3D large deformation of the valve leaflets during the entire cardiac cycle (Supplementary Material, Module S3) [13]. In this method, the complex geometry of the leaflets was subdivided into numerous simpler parts, to calculate the deformation response of their assembly to the transvalvular pressure. The obtained deformation of the valve leaflets was then implemented in a computational fluid dynamics (CFD) model to simulate the complex 3D flow field of the bioprosthesis (Supplementary Material, Module S4) in two different setups: (i) One representing the surgical bioprosthesis arrangement, hereafter being referred to as the surgical valve model (Fig. 3A) and (ii) one representing TAVR/ViV configuration, hereafter being referred to as the TAV model (Fig. 3B). The simulated flow fields of the two models were then used in a particle tracking procedure to calculate the BRT, in close proximity of valve leaflets (Supplementary Material, Module S5). In this method, the trajectory of imaginary particles that represented blood cell components were calculated from the obtained flow field and used to estimate blood stasis on the leaflets. Lastly, a statistical analysis was performed to compare the calculated values for BRT in the two models (Supplementary Material, Module S6).

Figure 2:

Custom-built pulse duplicator system and the particle image velocimetry setup.

Figure 3:

Computational models for the surgical valve (A) and the transcatheter aortic valve (B).

RESULTS

A 23mm Carpentier-Edwards Perimount bioprosthesis was tested in the pulse duplicator to obtain valvular hemodynamic data. The transvalvular pressure gradient waveform obtained from the in-vitro test was then used as the dynamic loading in the FE simulations [13]. Figure 4A demonstrates the results of the computer simulation of the deformation of the leaflets during one cardiac cycle. A well-matched leaflet deformation was observed throughout the cardiac cycle between the simulation and the experimental data of valve opening and closing in the pulse duplicator captured by the high speed camera.

Figure 4:

(A) Comparison of the in vitro leaflet motion and deformation with FE simulation throughout a complete cardiac cycle. (B) Comparison of the in vitro PIV measurement of the detailed flow structure (left) with the CFD simulation results for the flow structure (right) at a time instant during opening. The flow configuration is represented by contours of the flow velocity magnitude and the streamlines in the mid-plane longitudinally cutting through the computational domain. The simulations presented in this figure correspond to the surgical valve model. CFD: computational fluid dynamics; FE: finite element; PIV: particle image velocimetry.

Once the complex 3D deformation of the valve leaflets was obtained, it was implemented in the CFD models to simulate the bioprosthesis flow field. To confirm the validity of the flow simulations, a comparison was made between the 3D simulation and the 2D PIV flow measurements in terms of the detailed flow structure. The results are presented in Fig. 4B. The contours of the flow velocity magnitude as well as the stream traces at a time instance during the valve opening, which are obtained from the PIV flow measurements and the computer simulations, are presented in Figs 4B (left) and 4B (right), respectively. Note that the plane at which the flow parameters are demonstrated in Fig. 4B (right) is the mid-plane longitudinally cutting through the bioprosthetic valve. Again, good agreement between the experiment and the simulation was observed. Overall, the results presented in Fig. 4 confirmed the validity of the approach employed in this study.

Once validated, the modelling approach was used to simulate the 3D flow field of the two models representing the surgical valve and the TAV. The simulated flow fields were then used in a particle tracking scheme to visualize the effect of the geometric confinement of the TAV on the BRT on the leaflets. This module was performed as follows: First, a random distribution of 3000 particles in the region of interest (ROI) was generated. Note that the ROI was defined as the volume between the inner surface of the confining cylinder, which modelled the degenerated bioprosthesis or the calcified native valve that circumferentially surround the TAV stent, and the outer surface of the leaflets. In fact, the ROI represents the region in close proximity to the aortic side of the leaflets. The cardiac cycle was divided into four different stages, namely: I-forward flow, II-closing, III-end of closing to mid-diastole, and IV-mid-diastole to beginning of systole. Thereafter, in the beginning of each stage, the generated initial distribution of particles were released and tracked throughout the time period. Note that having the time-dependent 3D flow field of the bioprosthesis available, the trajectory of a particle can be generated using the flow velocities and their variations in time.

Figure 5 demonstrates sample snapshots of the particle tracking during the forward flow (Stage I) for the two models. The simulation of the flow field of the surgical valve is presented in Fig. 5A, and that of the TAV is demonstrated in Fig. 5B. The fluid flow structure is shown by 3D stream traces as well as the contours of the velocity magnitude in the mid-plane of the computational domain. The configuration of particles is shown by black dots, each dot representing one particle. As Fig. 5 demonstrates, it was observed that the general characteristics of the flow field, including the formation of a jet flow and recirculating regions, were quite similar between the two models. It was also observed that the difference in the distribution of particles was negligible during the opening stages of the bioprosthetic valve (first three columns in Fig. 5). However, a significant difference in the washout of particles was observed between the two models during the final stages of forward flow (the last two columns in Fig. 5). While the particles were almost completely washed out of the ROI in the surgical valve model (clearly demonstrated in the Video 1), a significant number of them still resided in ROI, specifically on the leaflets in the TAV model (clearly demonstrated in the Video 2). This demonstrates the prominent effect of the confining geometry around the valve on BRT on the leaflets.

Figure 5:

Snapshots showing the configuration of the flow field and the particles released in the region of interest at different time instances during Stage I (forward flow) in a cardiac cycle for the surgical valve model (A) and the transcatheter aortic valve model (B). The contours of the velocity magnitude are shown in the mid-plane longitudinally cutting through the computational domain.

Video 1:

Simulation of the flow in the model for surgical valve during forward flow.

Video 2:

Simulation of the flow in the model for transcatheter aortic valve during forward flow.

Figure 6 shows the results of the particle tracking for the duration from the end of closing to mid-diastole (Stage-III), where the valve is completely closed. Figure 6A corresponds to the surgical valve and Fig. 6B corresponds to the TAV. Again, a considerable difference in the distribution of particles between the two models at the end of this stage was observed. Similar to the forward flow (Stage-I), a considerable number of particles were observed to still reside on the leaflets in the TAV model, while most of them had been washed out in the surgical valve model.

Figure 6:

Snapshots showing the configuration of the flow field and the particles released in the region of interest at different time instances during Stage III (end of closing until mid-diastole) in a cardiac cycle for the surgical valve model (A) and the transcatheter aortic valve model (B). The contours of velocity magnitude are shown in the mid-plane longitudinally cutting through the computational domain.

To quantify the observed difference between the two models in terms of BRT on the leaflets, the residence time of the particles in the ROI, $T_R$⁠, was calculated based on the velocity and the displacement of the particles residing in the ROI. Once the values were obtained for all of the particles in all of the four stages of a complete cardiac cycle in the two models, a statistical analysis was performed (Supplementary Material, Module S6 - paired-sample sign test). The analysis revealed that $T_R$ was significantly larger for the TAV versus the surgical valve, in all of the four stages: During forward flow (Stage I) P < 0.0001, during closing (Stage II) P < 0.0005, from end of closing to mid-diastole (Stage III) P < 0.0005, and from mid-diastole to the beginning of systole (Stage IV) P < 0.0001. The calculated values of $T_R$ during each stage were also averaged over the number of particles and compared in the two models. The results are presented in Fig. 7. The values of the standard deviation are also shown as error bars. As it is observed, in all of the stages throughout a complete cardiac cycle, average $T_R$ was significantly larger for the TAV than the surgical valve. In fact, the value of the average $T_R$ during forward flow (Stage I) was 39% larger for TAV compared to the surgical valve. During closing (Stage II), TAV exhibited a value of $T_R$ which was 132% larger than that of surgical valve. Average $T_R$ was also 150% and 40% larger for TAV with respect to surgical valve during closing to mid-diastole (Stage III) and mid-diastole to systole (Stage IV), respectively. It should also be noted that a significant difference was obtained between $T_R$ in Stages I through IV for both models (P < 0.0005) (Supplementary Material, Module S6 – Friedman test). Moreover, Fig. 7 demonstrates that in both models, the particles were washed out of the ROI significantly faster during forward flow (Stage I) compared to diastole (Stages III and IV). The results of the statistical analysis also confirmed that $T_R$ was smaller in Stage I than Stage IV (P < 0.0005) for surgical valve. Similarly, $T_R$ was found to be significantly smaller in Stage I than Stage III (P < 0.0005) and Stage IV (P < 0.0005) for TAV.

Figure 7:

Comparison of the average BRT between the surgical valve and the TAV models at different stages of the cardiac cycle; I-forward flow, II-closing, III-the end of closing to mid-diastole, and IV-mid-diastole to the beginning of systole. TAV: transcatheter aortic valve; BRT: blood residence time.

DISCUSSION

This study performed an analysis of the effect of the confining geometry of the TAV on the likelihood of the leaflet thrombosis, via quantification of BRT on the TAV leaflets. Significantly longer BRT on the leaflets was observed in the TAV compared to the surgical valve during all stages of the cardiac cycle. During forward flow, the mean value of BRT was found to be 39% higher in the TAV compared to the surgical bioprosthesis (P < 0.0001). During diastole, specifically from end-systole to mid-diastole and from mid-diastole to the beginning of systole, the amount by which the mean BRT was higher for TAV compared to the surgical valve was 150% and 40%, respectively (P < 0.0005). The results confirmed that the geometric confinement of TAV by the failed bioprostheses or the calcified native valves increases the BRT on the TAV leaflets. The confinement may act as a permissive factor in valve thrombosis following TAVR and ViV procedures.

The occurrence of valve thrombosis after bioprosthetic SAVR has been considered to be very low, which in part may be due to a general lack of awareness of its existence. Valve thrombosis and leaflet immobility are detectable only by 4D computed tomographic or transoesophageal echocardiographic assessments, which are not routinely performed in the follow-up period for patients. Recent clinical studies have raised concerns about sub-clinical thrombosis and reduced leaflet motion following bioprosthetic SAVR and TAVR. For instance, Makkar et al. [1] reported reduced leaflet motion in 13% of the patients who had undergone either SAVR or TAVR, and in 40% of the ones who had TAVR. In general, patients with bioprosthetic heart valves are at a higher risk of ischaemic stroke or peripheral embolism than the normal population. Therefore, for aortic bioprosthetic valves, vitamin K antagosists (VKAs) are recommended for 3–6 months according to the American Heart Association/American College of Cardiology (AHA/ACC) guideline, while aspirin is preferred over VKA in the European Society of Cardiology (ESC) guideline [14, 15]. Following TAVR, dual antiplatelet therapy with clopidogrel and aspirin is currently recommended and used in most centres worldwide, but the duration of clopidogrel varies among studies ranging from one to six months [16]. To date, there are no clinically tested guidelines to recommend appropriate therapy for post-TAVR thrombus formation. Most TAVR patients with valve thrombosis have been successfully managed with oral anticoagulation therapy with significant hemodynamic improvement and resolution of thrombosis [1]. Nevertheless, the duration of oral anticoagulation therapy is not known and should be determined on a case-by-case basis considering bleeding risks.

The majority of post-TAVR valve thrombosis cases occurred following the implantation of TAVs with intra-annular design, such as SAPIEN valve (Edwards Lifesciences, Irvine, CA, USA) and Portico valve (St. Jude Medical, Saint Paul, MN, USA) [1, 3, 17]. It is difficult to draw a firm conclusion on the likelihood of valve thrombosis in TAVs with intra-annular design as the available data are based on the results of isolated case reports and small case series. Nonetheless, considering the results of the presented study, it can be inferred that valvular thrombosis may be more pronounced in TAVs that tend to operate inside the annulus than supra-annular TAVs. Supra-annular positioning of TAVs may reduce the BRT on the aortic surface of TAV leaflets due to the partial covering of the TAV stent frame [18]. Further investigations are required to confirm such notions in this regard. In addition, the risk of thrombus formation may be high in patients who have had ViV procedure as well as in patients with small aortic roots due to possible low flow regions on the TAV leaflets [8]. Considering the nature of ViV configuration and presence of greater foreign surface contact in the ViV setting, leaflet thrombosis may be more likely to occur in ViV patients than TAVR patients.

Blood stasis due to flow stagnation and formation of low-flow regions in the sinuses of Valsalva has been investigated in previous studies. For instance, Ducci et al. [19] performed flow measurements on a model aortic root and showed considerable reduced flow in the bases of the sinuses. Groves et al. [20] studied the effect of the positioning of TAV on the flow residence time and wall shear stress in the aortic root. Midha et al. [21] investigated the different axial positioning of the ViV setting on the thrombotic potential by focusing on the sinus wall shear stress. As observed, the focus of these studies has been mainly the flow characteristics in the aortic root. However, the connection between the low-flow regions in the sinuses and the leaflet thrombosis has remained unclear. The novelty of the current work was the definition of ROI for the calculation of BRT in the region between the aortic side of the TAV leaflets and the confining geometry of the TAV surrounding, i.e. the failed bioprosthesis or the calcified native valve that circumferentially surrounds the TAV stent. The choice of the ROI is supported by the observation that thrombus has been mostly observed to be located on the aortic side of the leaflets in the patients [1, 2].

Besides the effects of the geometric confinement that was focused in the present work, additional hemodynamic factors may affect TAV thrombosis. Recently, it has been shown that patients with depressed cardiac function, i.e. left ventricular ejection fraction <35%, may carry a higher risk of early TAV thrombosis [3]. In addition, differences may exist between thrombus formation on coronary and noncoronary leaflets [2]. However, due to the limited valve thrombosis data that is currently available, further research should be conducted to characterize early thrombus formation on coronary and noncoronary leaflets. In addition to the hemodynamic causes, other predisposing factors including haemostatic factors such as inadequate antithrombotic treatment and concurrent prothrombotic conditions, and endothelial factors such as delayed endothelialization of the metallic TAV stent frame, may impact TAV thrombosis [3, 22]. It is also important to note that valve thrombosis is unlikely to be caused by leaflet injury as shown by pathological findings at autopsy [2]. Therefore, structural changes due to crimping of pericardial leaflets of balloon-expandable or self-expandable TAVs are not likely to affect valvular thrombosis.

The present study employed an invitro setup to create a well-defined experimental testing environment that can be used for the validation of the computer simulations. Considering the technical characteristics of the currently available invivo imaging modalities, the utilization of an invitro setup was inevitable as very high temporal and special resolution of the valve motion was required for validation of the computational modelling. Such a setup along with the validations presented in this study, also provides a basis for future simulations of more realistic geometries. Although the absence of the aortic root geometry in the model was considered as a limitation of the present work, considering the relevance of the defined ROI to the realistic conditions, the proposed hypothesis is claimed to be valid and not affected by the geometric simplification. Another limitation of the present work was the absence of geometric details of the bioprosthesis frame, TAV stent, and the calcified native valve or failed bioprosthesis around the TAV stent. Since the proposed hypothesis is mainly based on the comparison of the BRT between the surgical and the TAV models rather than the actual values of the calculated particles’ residence time, it is believed to remain valid despite the aforementioned simplifications. All in all, despite the above limitations, the most striking finding of the present study, which is the defective washout of blood from TAV leaflets, is unlikely to vary significantly by introducing more details and complexities to the model.

In summary, the results of the present study indicate that in TAVR, as opposed to SAVR, the flow field around the leaflets is disrupted due to the geometric confinement introduced by the degenerated bioprosthesis or the calcified native valve that circumferentially surround the TAV stent. The geometric confinement increases the BRT on the leaflets and consequently increases the likelihood of leaflet thrombosis. Based on the results of the present study, further investigations regarding geometric variability associated with TAVR setting such as TAV type, size, and positioning, as well as anatomic considerations like coronary versus noncoronary leaflets in patient-specific models are motivated. Lastly, the findings of our study have important implications on the potential requirement for a more evidence based approach to post-TAVR antithrombotic treatment regimens for patients considering their bleeding risks.

SUPPLEMENTARY MATERIAL

Supplementary material is available at EJCTS online.

Funding

This work was supported by the Knoebel Institute for Healthy Aging (Grant number 89546-142235), Professional Research Opportunity Funds administered by University of Denver (Grant number 89610-142235), and by University of Denver Postdoctoral Fellowship Award.

Conflict of interest: none declared.