Feasibility of Patient-detectable Neurostimulation as a Seizure Warning System
Abstract number :
3.086
Submission category :
2. Translational Research / 2A. Human Studies
Year :
2022
Submission ID :
2204234
Source :
www.aesnet.org
Presentation date :
12/5/2022 12:00:00 PM
Published date :
Nov 22, 2022, 05:24 AM
Authors :
Erin Yeagle, MD – UCSF; Tyler Gray, B.S. – Neurology – Yale University School of Medicine; Hitten Zaveri, Ph.D. – Neurology – Yale University School of Medicine; Lawrence Hirsch, M.D. – Neurology – Yale University School of Medicine; Imran Quraishi, M.D., Ph.D. – Neurology – Yale University School of Medicine
Rationale: There is a pressing need for technology that can warn patients of seizures before they occur or while they are occurring (if they are unaware of their seizures), and to notify patients when they are at high or low risk of seizures. The RNS SystemTM (NeuroPace, Inc.) delivers brain-responsive neurostimulation in response to epileptiform patterns, which can be modified to be detectable by the patient. This modification has been used successfully in one patient at our center, at the 5th and last therapy in a train, as a warning system for seizure clusters. Patient-detectable neurostimulation can be programmed in other patients with RNS devices (Quraishi and Hirsch 2021), but the accuracy of such a modification to predict or identify seizures in additional patients has yet to be formally assessed.
Methods: We annotated a single-center database of RNS patients with markers of clinical seizures in order to evaluate utility of each therapy delivered (1st-5th) as a seizure warning for the upcoming hour. Patients included in analysis were those for whom a clear correlate of clinical seizures could be identified, at least 90% of data were available over a period of at least 6 months, and stimulation on the device was enabled. Data were included from 11 patients (6 female) for a total of 13 data epochs. We evaluated sensitivity and time in warning of each therapy in a train of up to 5 therapies to warn of a seizure within the upcoming hour, and calculated area under the receiver operating characteristic (ROC) curve of sensitivity vs. time in warning for each epoch of patient data.
Results: Sensitivity and time in warning of patient-detectable neurostimulation to warn of seizures in the upcoming hour varied across patients. The greatest positive predictive value was observed at a threshold of 5 therapies. When evaluating the ability of a certain number of therapies to warn of a seizure in that event or over the coming hour, the mean area under the ROC curve of sensitivity vs. time in warning (AUC) was above chance (AUC .75, 95% confidence interval .56-.93). When evaluating only the ability of a certain number of therapies to warn of a seizure in the coming hour (excluding the event in which the threshold was reached), the mean AUC was at chance (.47), although AUC remained greater than .8 in 18% of patients.
Conclusions: Our findings suggest that patient-detectable neurostimulation is a feasible seizure warning system. Such a system would be most effective in alerting patients of seizures as they are happening. In a minority of patients, detectable neurostimulation also showed potential to function as a seizure prediction system (at least on the time horizon of one hour). We propose that patient-detectable neurostimulation may be most useful for patients who are not aware of their seizures, and those who are prone to seizure clusters. Further research will be needed to optimize thresholds and time windows to use as a seizure warning system, to identify patients most likely to benefit from this modification, and to see if this can lead to effective intervention and improved quality of life.
Funding: Swebilius Foundation (IHQ)
Translational Research