A 3D Non-human Primate Digital Model for Pharmacokinetic Prediction of Intra-Cerebrospinal Fluid Drug Neuraxial Dispersion

Background Intra-cerebrospinal fluid (CSF) drug delivery bypasses the blood-brain barrier, making it a promising route of delivery to treat central nervous system (CNS) diseases. Optimizing this delivery route is challenging because of complex interactions among drug kinetics, CSF flow dynamics and anatomical variations. Non-human primate (NHP) models provide an approximation to human physiology, making a suitable surrogate for studying intra-CSF drug dispersion. We present a NHP digital model for pharmacokinetic prediction of intra-CSF solute neuraxial dispersion that incorporates craniospinal compliance and other key physiological features. Methods A 3D subject-specific digital model of the NHP CSF system was formulated using a 3D multi-phase computational fluid dynamics (CFD) approach with flow and geometric boundary conditions using animal-specific in vivo MRI data. Initial digital model drug dispersion predictions were carried out assuming rigid dura and pial surfaces and verified by comparison to a 3D-printed NHP bench-top model replicating the in vivo measurements utilizing fluorescein as a surrogate drug tracer. Once verified, the digital model was extended to mimic craniospinal compliance by incorporating a dynamic mesh to allow dura surface motion that replicated the non-uniform CSF flow along the neuroaxis. Results were quantified over a one-hour period after a 1 mL drug injection via lumbar puncture needle in terms of spatial-temporal drug dispersion along the neuroaxis for the rigid, compliant and bench-top models. Regional percent of injected dose was assessed across the lumbar, thoracic, cervical and cranial regions, while total exposure at each 1 mm section was calculated as the area-under-the-curve (AUC) along the neuroaxis. Results The rigid digital model tracer dispersion predictions were verified through comparison with the NHP bench-top model, showing high spatial-temporal agreement (R² = 0.88). The introduction of dynamic mesh motion in the compliant digital model resulted in ~ 10X reduction in peak lumbar CSF flowrate compared with the rigid model (0.065 versus 0.65 mL/min). This decrease in peak CSF flowrate contributed to a reduction in the average Reynolds number along the neuroaxis, dropping from 250 in the rigid model to under 100 in the compliant model leading to decreased tracer dispersion in the lumbar region. At 1 hour following injection, tracer distribution to the lumbar, thoracic, cervical and intracranial CSF was 91.9, 8.1, 0 and 0% of injected dose for the compliant model, while a model not including these physiological factors predicted 72.9, 20.4, 5.6 and 1.1%. Conclusion The developed NHP-specific digital model, verified with NHP bench-top model simulations, provides a platform to understand and potentially improve intrathecal drug delivery protocols and devices. This study highlights the potentially important role of craniospinal compliance in CSF solute dispersion along the neuroaxis. Incorporating physiological factors such as compliance and varying flowrates into digital models of CSF transport can enhance the predictive capability of drug distribution within the CNS, aiding the design of more effective therapeutic strategies for CNS diseases.