Multi-fidelity simulation of novel anticoagulation therapies in atrial fibrillation*

Direct Oral Anticoagulants (DOACs) are effective in patients with atrial fibrillation (AF) who need long-term therapy to reduce the ischemic stroke risk. Currently used DOACs target the activation of factor X and thrombin (IIa) in the common coagulation pathway, thus increasing bleeding risk. Recent efforts are targetting the intrinsic coagulation pathway (e.g., factor XI and factor XII inhibitors) to prevent thrombosis without increasing bleeding risk. However, these efforts are showing mixed outcomes in preventing stroke in AF patients clinical trials. Better understanding of how left atrial (LA) flow affects thrombosis and its inhibition in patient-specific anatomies is needed to select patients for trials and to select the optimal therapy for each patient. This study explores the influence of factors XI and XII and their inhibition on the intrinsic coagulation cascade described by a 32-equation model in 13 patient-specific LA anatomies under physiological flow conditions. Patient-specific anatomical models are created using 4D computed tomography (CT) images, which include the total flow rate through the pulmonary veins (PVs). A multi-fidelity (MuFi) coagulation model is verified and applied to accelerate the simulations. Leveraging the computational efficiency of the MuFi model, we study the coagulation dynamics from a procoagulant state and apply ten inhibition levels ranging from 50\% to 97.5\% for target factors XI and XII. This methodology allows us to quantitatively measure clotting time and clot volume, to identify the patient-specific inhibition levels necessary to impede thrombin generation.