Abstracts

Rationale: The NeuroPace RNS^TM System is an investigational device developed to treat intractable partial onset epilepsy. The implantable components include a neurostimulator and sensing/detection leads with cortical strip or deep brain electrodes. The neurostimulator detects seizure activity and responds with programmed bursts of electrical cortical stimulation. The purpose of this study was to determine the stimulation settings that would reduce seizure-occurrences. Methods: The neurostimulator and leads were implanted in 8-subjects enrolled at the Medical College of Georgia as part of a multicenter feasibility investigation. Seizure-occurrences and stimulus-variables were recorded. Eight stimulus variables were log-transformed: number of electrode contacts, charge density, stimulation current, pulse width, stimulus frequency, burst duration, burst number, and total delivered charge. The three leading singular vector components of these stimulus-variables were selected. Seizure-occurrences were modeled with a Markov chain. The probability of remaining seizure-free per unit time was a linear-logistic-regression (LLR) function of the stimulus-variables. Akaike’s Information Criteria was used to identify the optimum (LLR) function that best represented the relationship between the stimulus-variables and the probability of the subjects remaining seizure-free. Results: We recorded 28-262 seizures per subject (median 140-seizures) over intervals of 612-909 days (median 850-days). We evaluated regression models that depended on a range of zero to all three components of the stimulus-variables; the best model required all three components. The stimulus-variables that most influenced the probability of the subjects remaining seizure-free were total delivered charge, charge density, stimulus frequency and number of electrode contacts. Conclusions: These results may provide a quantitative method for selecting stimulus settings for the RNS^TM System neurostimulator. (Supported by NeuroPace, Inc.)