Feasibility of large-scale in silico transcatheter aortic valve implantation trials using a fast-to-evaluate model.

Although transcatheter aortic valve implantation (TAVI) has been demonstrated to be a successful treatment for aortic stenosis, it remains associated with complications, such as paravalvular leakage (PVL). To address these, TAVI devices continue to undergo iterative development. Integration of in silico trials into the regulatory validation pathway offers a promising approach to accelerate the development and clinical implementation of novel TAVI devices. This study addresses the feasibility of conducting large-scale in silico TAVI trials using a virtual cohort generator (VCG) combined with a fast-to-evaluate model. The objective is to investigate anatomical and procedural predictors of PVL in silico, as was done in an earlier clinical study. A virtual cohort of 500 synthetic aortic stenosis patients was generated, that matched anatomical and demographic characteristics of the clinical population. Using a novel fast-to-evaluate TAVI deployment model, nearly 29,000 simulations were performed across multiple model parameter combinations per patient. Shape and demographic distributions in the in silico trial, remained within the bounds of the clinical study. Among the investigated anatomical parameters, a higher angle between left ventricular outflow tract and ascending aorta was found in patients with significant PLV, in both clinical and virtual cohorts. Additionally, the relationship between PVL and implantation depth appeared highly patient-specific, which is in line with findings in clinical studies. The ability to systematically test multiple TAVI deployments scenarios per patient, which is unfeasible in clinical practice, provides valuable insights for procedure design and optimisation. Overall, the results support the feasibility of implementing large-scale in silico TAVI trials, using the VCG and a fast-to-evaluate model, into the regulatory validation chain.