Authors: Karim Mithani; Simeon Wong; Mirriam Mikhail, MEng; George Ibrahim, MD, PhD (Toronto, Canada)
Although vagus nerve stimulation (VNS) is commonly applied to the treatment of epilepsy in children, response to stimulation is highly variable. Given overlapping thalamocortical circuitry between vagus and spinothalamic afferent pathways, we tested the hypothesis that somatosensory evoked fields (SEFs) could predict VNS response.
We analyzed pre-operative magnetoencephalography (MEG) data from 36 patients (26 responders, 10 non-responders) who underwent VNS implantation. All children underwent median nerve stimulation during MEG and features of the corresponding SEFs, including amplitudes, latencies, Euclidean distance of activation from S1, and sensorimotor network connectivity were analyzed. A support vector machine (SVM) classifier was used to predict VNS response.
VNS responders showed significantly more dispersed SEF localization compared to non-responders (p<0.001). Responders also demonstrated greater stimulation-locked connectivity within the sensorimotor network than non-responders (FWER-corrected p < 0.05). An SVM classifier demonstrated an accuracy of 88.9%, specificity of 100%, sensitivity of 50%, and AUC of 0.92 on 10-fold cross-validation.
We demonstrate that overlapping neuroanatomical circuitry can effectively be exploited to predict treatment response to VNS. These findings also highlight the importance of brainstem and thalamocortical afferent connectivity in the mechanism of action of VNS and its translational relevance to identify suitable candidates for surgery.