Development and Validation of a Digital Twin for Simulating the Hemodynamic Response to Antithrombotic Drugs: Predicting Impact on the Prevention of Vascular Complications (Dialysis and Blindness)

Abstract

Antithrombotic therapy is crucial for preventing thrombotic events in cardiovascular and metabolic diseases. However, its effect on microvascular hemodynamics in target organs such as the kidney and retina is heterogeneous and difficult to predict, influencing the risk of complications such as chronic kidney disease requiring dialysis and retinopathy leading to blindness. Objective: To develop and validate a personalized multiscale digital twin (DT) capable of accurately simulating the systemic and microvascular hemodynamic response to new antithrombotic drugs and quantifying its predictive impact on the long-term preservation of renal and visual function. Methods: A DT architecture was developed that integrates: 1) a macrocirculatory finite element model of the patient's vascular geometry derived from CTA, 2) a renal (glomerulus) and retinal (capillary plexuses) microcirculation model based on pore networks and fractal capillary beds, and 3) a coupled pharmacokinetic-pharmacodynamic (PK/PD) model for factor Xa inhibitors, direct thrombin inhibitors, and antiplatelet agents. It was calibrated with longitudinal data from 150 patients (50 healthy, 50 with type 2 diabetes, 50 with atrial fibrillation) including central blood pressure, serum biomarkers, retinal optical coherence tomography (OCT), and glomerular filtration rate (GFR) measurements. Prospective validation was performed in 30 patients by simulating the administration of a new antithrombotic drug (candidate ABC123) and comparing the DT's predictions with clinical outcomes and renal Doppler and OCT-angiography parameters at 6 and 12 months. Results: The DT achieved high fidelity in simulating hemodynamic profiles (correlation >0.95 for aortic pressure and flow). The prediction of changes in glomerular barrier permeability and retinal capillary density showed a sensitivity of 89% and specificity of 94% for identifying patients who subsequently showed preserved GFR (>60 ml/min/1.73m²) and no progression to proliferative retinopathy at 12 months. The model predicted a relative risk reduction of 34% (95% CI: 25-42%) for dialysis initiation and 28% (95% CI: 19-36%) for blindness from vascular complications. Conclusion: The personalized digital twin represents an innovative paradigm for the *in silico* evaluation of new antithrombotic drugs, allowing quantification of their beneficial hemodynamic impact at the microvascular level and predicting their efficacy in preventing devastating complications. This can optimize clinical trial design and guide personalized therapies.