Vagus nerve stimulation-induced laryngeal motor evoked potentials for response prediction and intensity titration in drug-resistant epilepsy

Abstract

Objective: The objective of the study was to record Laryngeal Motor Evoked Potentials (LMEPs) in Vagus Nerve Stimulation (VNS)-implanted patients suffering from Drug-Resistant Epilepsy (DRE). Based on these recordings, LMEPs characteristics were evaluated and compared between responders (R) and non-responders (NR). Finally, possible under- or over-stimulation was assessed based on a physiological indicator of fiber engagement.

Methods: Mean dose-response curves were compared between R and NR. A Support Vector Machine (SVM) model was built based on both LMEP and dose-response curves features, to discriminate R from NR. For the exploration of possible under- or over-stimulation, a ratio between the clinically applied stimulation intensity and the intensity yielding to LMEP saturation was computed for each patient.

Results: A trend towards a greater excitability of the nerve was observed in R compared to NR. The SVM classifier discriminated R and NR with an accuracy of 80%. An ineffective attempt to overstimulate at current levels above what is usually necessary to obtain clinical benefits was suggested in NR.

Conclusions: The SVM model built emphasizes a possible link between vagus nerve recruitment characteristics and treatment effectiveness. Most of the clinically responding patients receive VNS at a stimulation intensity 1-fold and 2-fold the intensity inducing LMEP saturation.

Significance: LMEP saturation could be a practical help in guiding the titration of the stimulation parameters using a physiological indicator of fiber engagement.

Keywords: Biomarker; Epilepsy; Larynx; Motor-evoked potentials; SVM; Vagus nerve stimulation.