Abstracts

Rationale: The RNS System is an FDA-approved direct-brain responsive system for adults with medically intractable focal onset seizures. A recent multicenter retrospective study evaluated the real-world experience with direct brain-responsive neurostimulation in a group of 150 patients treated at eight comprehensive epilepsy centers¹. The median percent seizure reduction was 67%, 75%, and 82% in years 1, 2, and 3 respectively, which exceeds the outcomes in patients in the clinical trials leading to FDA approval². The authors speculated that this improved response could be because physicians were following the guidance of a standard therapy protocol that emerged from the clinical trial experience and was made available as suggestions to potentially consider as they developed patient-specific treatment protocols. However, approximately 15% of the retrospective study patients had a 25% or less reduction in seizures², suggesting that some patients require a non-standard approach to stimulation programming. In this case study, we explore the consideration and use of distinct treatment options generated by NeuroPace using an artificial intelligence model.

Methods: Both patients were identified as non-responders because they had less than a 25% reduction in clinical seizure frequency after 2 years of RNS System treatment, and after physicians had reviewed and decided to implement the standard, recommended therapy protocol. The deep learning algorithm compares the ECoG features in all of the electrographic seizure ECoG records for a patient of interest, in this, case the non-responding patient, to all electrographic seizure ECoGs from a pool of 256 RNS System IDE clinical trial patients. The algorithm then identifies a group of patients that have the closest ECoG signal features that match to the non-responder and identifies the stimulation settings at which these patients were programmed. The physician can then elect whether to adjust programming in the non-responding patient to settings at which these similar patients had the best outcome.

Results: Both of our patients reported significant reductions in seizure frequency within three months of using the algorithm-derived, stimulation settings. The clinical outcomes and programmed settings before and after the algorithm suggested changes are in Table 1.

Conclusions: The algorithm is still in the research and development phase and the impact of applying this tool on safety and effectiveness of RNS System treatment will need to be evaluated further. However, this small case series supports the exploration of this novel algorithm to suggest alternative stimulation settings for non-responder patients after exhausting the standard suggested therapy settings.

References:
1.Razavi B, Rao VR, Lin C, Bujarski KA, Patra SE, Burdette DE, et al. Real-world experience with direct brain-responsive neurostimulation for focal onset seizures Epilepsia. 2020 Jul 13.

2.Nair DR, Laxer KD, Weber PB, Murro AM, Park YD, Barkley GL, et al. Nine-year prospective efficacy and safety of brain-responsive neurostimulation for focal epilepsy Neurology. 2020 Sep 1;95:e1244-e1256.

Funding: Please list any funding that was received in support of this abstract.: None.