Prediction of epileptic seizures for on-demand vagus nerve stimulation

Abstract

Vagus nerve stimulation (VNS) is a therapy for refractory epilepsy. Typically, VNS is performed for 30 s every 5 min, and the therapeutic effect increases over a time scale of several months. In addition, VNS exerts an immediate seizure suppressing effect. To use this effect, VNS should be done ondemand, before or at the early start of seizures. This requires that seizures can be predicted or detected at an early stage. This study was performed to develop an epilepsy model suitable for identifying seizure predictors for use in an on-demand VNS system. The present work was performed in 9 rats. Animals were anesthetized by continuous intravenous (I.V.) administration of a mixture of ketamine and xylazine. Seizures were induced by I.V. infusion of 50 mg/kg pentylenetetrazole (PTZ) over a 5 min period. During the experiments EEG, ECG and EMG (one animal) were recorded. Seizures were defined from the EEG activity and RR interval (RRInt) changes were investigated as a seizure predictor. Four different stages of EEG activity were identified. Stage 2 was associated with the onset of the tonic phase of the seizure. Therefore the start of this EEG stage was defined as the onset of tonic seizure. In 78% of cases, it was possible to predict seizures based on significant changes in the RRInt. Predictions were done 114 ± 29 s (mean ± SD) prior to seizure onset. In 89% of cases tonic seizure was associated with bradycardia. The results show that the model is suitable for identifying precursors for epileptic seizures. RRInt changes could be used as a predictor in the model, however, further work is needed to clarify if RRInt changes can be used for seizure prediction in other animal models or in human patients. © 2008 Springer-Verlag.