SPARC: Biophysical Modeling of Vagus Nerve Stimulation for Translational Scaling of Stimulation Parameters Across Species

Vagus nerve stimulation (VNS) is used clinically to treat epilepsy, depression, and obesity, and is under investigation for other applications. In many cases, the stimulation parameters used in preclinical VNS studies did not translate to successful clinical outcomes. We propose that differences in nerve morphology across preclinical models (e.g., rats and pigs) and humans, as well as variance within those populations, contribute to variance in therapeutic response due to variable levels of neural activation or block. Therefore, we quantified vagal morphology across species and simulated excitation and block thresholds using computational models. We quantified the morphology of the cervical and subdiaphragmatic vagus nerves (VN) in rats, pigs, and humans. The human and pig VNs had similar effective diameters (~3.7 mm at the cervical level and ~2.4 mm at the subdiaphragmatic level), although the pig nerves had almost 10x more fascicles (~41 fascicles for pig VN vs. ~5.8 fascicles for human VN). Conversely, the rat nerves were ~10x smaller (0.265 mm diameter at the cervical level and 0.147 mm at the subdiaphragmatic level) and had only one or a couple of fascicles. For our rat samples, the ratio of perineurium thickness to fascicle diameter corresponded well to the previous estimate of 3% of the fascicle diameter (Grinberg et al., 2008). Conversely, for our human samples, the perineurium was substantially thicker (~8.0% and ~21% at the cervical and subdiaphragmatic levels, respectively), and the ratio of perineurium thickness to fascicle diameter decreased with increased fascicle diameter. We designed and implemented a computational pipeline to construct volume conductor finite element models (FEMs) in COMSOL for multiple vagus nerve samples from each species, instrumented with a cuff electrode as used preclinically or clinically. We coupled the FEMs to models of mammalian myelinated and unmyelinated fibers in NEURON to quantify excitation and block thresholds. The intensity of VNS stimulation required to achieve activation of 50% of the 5.7 to 10 μm model myelinated fibers in human models were ~5–10x greater than in rat nerve samples, and intensities to activate 0.8 μm model unmyelinated fibers were ~20x greater. Further, thresholds of excitation and block varied across individuals within a species and were governed by species‐ and individual‐specific factors such as fascicle diameter and fascicle organization within the nerve. Human and pig cervical models show comparable within‐species variability (the highest thresholds to activate 50% of model nerve fibers were ~2x larger than the lowest) and greater variability than across the rat cervical models (the highest thresholds to activate 50% of model nerve fibers were ~1.25x larger than lowest). Morphologically‐based computational models of the cervical and subdiaphragmatic VN in rats, pigs, and humans indicate that cross‐ and within‐species differences in nerve morphology are important considerations for understanding differences in VNS responses between individuals and that substantial translation scaling of stimulation intensity is required to achieve comparable responses in different species. Support or Funding Information Funding: This work was supported by the NIH SPARC Program (OT2‐OD025340).