Author + information
- Peter de Jaegere, MD, PhD∗ (, )
- Gianluca De Santis, PhD,
- Ramon Rodriguez-Olivares, MD,
- Johan Bosmans, MD, PhD,
- Nico Bruining, PhD,
- Tim Dezutter, MSc,
- Zouhair Rahhab, BSc,
- Nahid El Faquir, BSc,
- Valérie Collas, MSc,
- Bart Bosmans, MSc,
- Benedict Verhegghe, PhD,
- Claire Ren, MD, PhD,
- Marcel Geleinse, MD, PhD,
- Carl Schultz, MD, PhD,
- Nicolas van Mieghem, MD, PhD,
- Matthieu De Beule, PhD and
- Peter Mortier, PhD
- ↵∗Thoraxcenter, ‘s-Gravendijkwal 230, 3015 CE, Rotterdam, the Netherlands
Outcome of transcatheter aortic valve replacement (TAVR) depends on a combination of patient-, procedure-, and operator-related variables. Specific device–host-related interactions may also be involved and may result in, for instance, incomplete and/or nonuniform frame expansion that in turn may lead to aortic regurgitation (AR) (1). Due to the large variability of the aortic root anatomy, the occurrence and severity of AR is hard to predict, indicating the need of tools that help the physician to select the type and size of valve that best fits the individual patient in addition to the optimal landing zone. Computer simulation of a TAVR procedure that is based upon the integration of the patient-specific anatomy, the physical and (bio)mechanical properties of the valve, and recipient anatomy may serve this goal (2). We herein describe such a model for AR prediction that was validated in a series of 60 patients who underwent TAVR with the Medtronic CoreValve Revalving System (MCS) (Medtronic, Dublin, Ireland).
For that purpose, pre-operative multislice computed tomography (MSCT) was used to generate patient-specific 3-dimensional models of the native aortic root using image segmentation techniques (Mimics v17.0, Materialise, Leuven, Belgium). Subsequently, implantation of virtual CoreValve models in these aortic root models was retrospectively simulated using finite-element computer modelling (Abaqus v6.12, Dassault Systèmes, Paris, France), resulting in a prediction of frame deformation and native leaflet displacement. Details of this method, as well as the validation of the predicted frame deformation, have been described before (3). In each computer-simulated implantation, all steps of the clinical implantation were respected, consisting of pre-dilation, valve size selection, depth of implantation, and post-dilation if applied. The depth of implantation was matched with the actual depth of implantation derived from contrast angiography performed immediately after TAVR.
The blood flow domain including the paravalvular leakage channels (if any) was then derived from the predicted frame and aortic root deformation, and computational fluid dynamics (OpenFOAM v2.1.1, OpenCFD, Bracknell, United Kingdom) was used to model blood flow during diastole with the aim of assessing the severity of aortic regurgitation after TAVR. For this purpose, a fixed pressure difference of 32 mm Hg was imposed from the ascending aorta to the left ventricle. The actual pressure difference post-TAVR was intentionally not used as the aim is to validate a model predicting AR based on pre-operative MSCT only (i.e., when the pressure post-TAVR is unknown). The value of 32 mm Hg is an average obtained from a large group of patients. The resulting flow, expressed in ml/s, was compared with the clinically assessed AR. The modelling of AR is illustrated in Figure 1 showing 2 patients with different severities of AR.
Contrast angiography and Doppler echocardiography were used for the assessment of AR. Analogous to the CHOICE (A Comparison of Transcatheter Heart Valves in High Risk Patients With Severe Aortic Stenosis) study, AR severity by contrast angiography was defined by visual estimation of the contrast density in the left ventricle using the Sellers classification (0 = none/trace, 1 = mild, 2 = moderate, 3 = severe; the latter comprised grades 3 and 4 according to Sellers) (4). Two observers independently from one another scored the angiograms. In case of discrepancies, consensus was reached by consulting a senior cardiologist. The intraobserver and interobserver variability for the assessment of AR post-TAVR according to the Sellers classification were κ 0.70 and 0.78, respectively. Doppler echocardiography was performed before discharge. AR severity was defined by the circumferential extent of the Doppler signal at the inflow of the MCS frame in the parasternal short-axis view (VARC-2 [Valve Academic Research Consortium-2]) (5). Echocardiography was available in 56 of the 60 patients. Distinction was made between none (grade 0), mild (<10%, grade 1), moderate (10% to 29%, grade 2), and severe (≥30%, grade 3) AR. Physicians performing TAVR and engineers performing the simulations were blinded to one another’s results.
Moderate–severe AR (Sellers AR ≥2) post-TAVR was seen in 15 patients (25%) by angiography. The agreement between the observed (i.e., Sellers, angiography) and predicted AR (i.e., ml/s, model) is shown in Table 1. Receiver-operating characteristic curve analysis revealed that 16.25 ml/s is the cutoff value that best differentiated patients with none-to-mild and moderate-to-severe AR. Sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 0.80, 0.80, 0.57, 0.92, and 0.80, respectively. By echocardiography, moderate–severe AR was seen in 9 patients (15%). The agreement between the observed and predicted AR is shown in Table 1. Receiver-operating characteristic curve analysis revealed that 16.0 ml/s is the cutoff value that best differentiated patients with none-to-mild and moderate-to-severe AR. Sensitivity, specificity, positive predictive value, negative predictive value, accuracy were 0.72, 0.78, 0.35, 0.94, and 0.73, respectively. Besides simulating the actual TAVR procedure (i.e., device size and positioning), a few alternative scenarios were investigated in a subset of cases in which the impact of device sizing (Figure 2) and implantation depth were investigated (Figure 3).
We, thus, found that computer simulation using dedicated software integrating the MSCT-derived patient-specific anatomy and the geometric, and mechanical properties of the valve accurately predicts the severity of AR that will occur after the implantation of the self-expanding MCS valve when measured by angiography or echocardiography.
These findings indicate both the feasibility and clinical utility of computer simulation of a TAVR procedure with the objective to improve outcome by helping the physician to select the size of valve that best fits the individual patient. AR, which was the outcome of interest in this study, also depends on the depth of implantation that in particular depends on physician’s performance. As illustrated in the case study (Figure 3), the simulation can inform the physician which size of valve at which optimal landing zone is associated with the least amount of AR. Although the choice of valve size is easy to follow in clinical practice, this is less so for the depth of implantation. The use of repositionable valve technologies, however, may overcome this technical issue, thereby enforcing the clinical power and utility of the herein proposed simulation workflow.
Also, in order to meet the goal of tailored or patient-specific treatment planning, all clinically available valves should be incorporated into the simulation program that also should have the capacity to predict all clinically relevant device–host-related interactions or complications. The most recent valve technologies are reported to be associated with substantially less AR, but possibly with a higher than expected or accepted incidence of new conduction abnormalities. The simulation program should, therefore, follow suit and provide a comprehensive output containing all outcomes that are relevant to the patient and physician.
The current validation is not without limitations, in particular because we used contrast angiography and echocardiography for the assessment AR. These widely used clinical tools have clear limitations and are inferior to magnetic resonance imaging for the assessment of AR. Magnetic resonance imaging should, therefore, have been used but is logistically demanding and difficult to perform in patients who underwent TAVR. Nevertheless, patient-specific computer simulation using dedicated software accurately predicts the severity of AR and may improve outcome of TAVR by helping the physician to select the size of valve that best fits the individual patient.
Please note: Dr. Bosmans is supported by a PhD grant, partially funded by Materialise N.V. Dr. de Jaegere is proctor for Medtronic. Dr. De Santis and Mr. Dezutter are employees of FEops. Drs. Verhegghe, De Beule, and Mortier are cofounders of and shareholders in FEops. Dr. van Mieghem has received research grant support from Edwards Lifesciences, St. Jude Medical, Abbott Vascular, Boston Scientific, and Medtronic. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. de Jaegere and De Santis contributed equally to this work.
- American College of Cardiology Foundation
- Schultz C.J.,
- Weustink A.,
- Piazza N.,
- et al.
- Kappetein A.P.,
- Head S.J.,
- Généreux P.,
- et al.