0630 Computational Modeling-Informed Design of Proximal Hypoglossal Nerve Stimulation Using a Multi-Contact Electrode for Targeted Activation in OSA

Introduction During sleep, tongue muscles coordinate protrusion, retraction, and shaping to maintain airway patency. Physiological studies suggest that antagonistic protrusor-retractor co-activation increases stiffness and lowers critical closing pressure compared with protrusor activation alone, whereas isolated retractor activation may worsen collapse. Our proximal hypoglossal nerve stimulation system (pHGNS) uses a six-contact cuff on the proximal nerve trunk; multiple contacts can be grouped and driven simultaneously at independent current levels. This enables activation patterns beyond traditional respiratory-gated airway rescue. In the OSPREY clinical trial, some patients showed limited response to single-contact dosing. We therefore used a biophysical model to design a PolySync™ contact-grouping algorithm that uses single-contact PSG responses and threshold-weighted dosing to explore additional tongue recruitment patterns in prior non-responders. Methods We used the ASCENT modeling pipeline (COMSOL finite-element model with NEURON axon models) to simulate activation from a six-contact cuff on the proximal hypoglossal trunk and validated the model against OSPREY sensory thresholds. Using this model, we evaluated two-contact groupings under different current-splitting schemes. Prior single-contact PSG responses were classified as therapeutic, neutral, or anti-therapeutic based on apnea-hypopnea index changes. Combinations of two anti-therapeutic contacts were excluded, while therapeutic or neutral contacts could pair with anti-therapeutic contacts to explore antagonistic co-activation. These contact-level PSG classifications, together with model-derived dosing rules, defined the PolySync™ titration strategy that was then employed in OSPREY non-responder re-titration. Results Model predictions reproduced OSPREY sensory threshold trends, with centrally located contacts (relative to cuff insulation) exhibiting lower thresholds than edge contacts. Single-contact stimulation created focal recruitment near individual contacts, whereas grouped contacts distributed current more broadly. Threshold-weighted grouping reduced dominance by low-threshold contacts and produced blended recruitment profiles that reflected both contacts, increasing the diversity of achievable recruitment patterns compared with single-contact dosing. In 10 OSPREY non-responders optimized with PolySync™, 71% (10/14) converted from non-responder to responder. Conclusion A computational model validated against clinical thresholds enabled design of contact-grouping strategies for pHGNS. Tailored therapy through electrode grouping broadened current injection, potentially enhancing tolerability and compliance. PolySync™-titrated results suggest that accessing additional recruitment patterns may support more personalized, tongue-stabilizing pHGNS. Support (if any) This work was supported by LivaNova.