Monitoring Anti-seizure Medication Efficacy with Sub-scalp EEG
Abstract number :
3.205
Submission category :
2. Translational Research / 2B. Devices, Technologies, Stem Cells
Year :
2024
Submission ID :
211
Source :
www.aesnet.org
Presentation date :
12/9/2024 12:00:00 AM
Published date :
Authors :
Presenting Author: Ashley Reynolds, MD – The University of Melbourne
Alan Lai, PhD – St Vincent's Hospital Melbourne, The University of Melbourne
Amy Halliday, MD – St Vincent's Hospital Melbourne, The University of Melbourne
David Grayden, PhD – The University of Melbourne
Andre Peterson, PhD – St Vincent's Hospital Melbourne, The University of Melbourne
Mark Cook, MD – The University of Melbourne
Rationale: A reliable method to assess anti-seizure medication (ASM) efficacy is desperately needed. Current methods are hampered by inaccurate seizure diaries and cross-sectional assessments of complex seasonal data, resulting in patients undergoing years of trial and error, failed therapies, with a third never achieving seizure freedom.
Ultra-long sub-scalp electroencephalography (EEG) has the potential to address both inaccurate reporting and seizure cycles. We aim to model long-term trends in seizure rate, using sub-scalp EEG detected seizures and interictal epileptiform discharge (IED) cycles. We further aim to demonstrate that large changes in the model’s residual values signal ASM efficacy.
Methods: This is a retrospective longitudinal case study of a 49-year-old female with bilateral periventricular nodular heterotopia and focal seizures with impaired awareness who participated in the sub-scalp EEG system Minder® trial (ACTRN12619001587190). The subject provided written and informed consent and was implanted in Nov. 2019. Study protocol was approved by St Vincent’s Hospital Melbourne Human Research Ethics Committee (HREC158/19).
Seizures and IEDs were detected from sub-scalp EEG using a convolutional neural network (Clarke et al. 2021 Epilepsy Behav, 121, 106556). A 90-day seizure rate was used in a moving average (MA) model. The continuous wavelet transform of IEDs estimated the instantaneous phase of seizure cycles and was included with seizure rate in an autoregressive integrated moving average model with exogenous variables (ARIMAX). Pre-sample data was 52 days long for a 14-day forecast. Models were trained on 250 days when ASMs were stable and hyperparameters were optimised using repeated hold one out validation and Bayesian information criterion. Final models were tested on four sets of data (pre-drug 2019, pre-drug 2020, post-drug 2020, post-drug 2022). The Kruskal-Wallis test (p< 0.05) compared the root mean squared error from test data.
Results: In Nov. 2019 the subject was taking lacosamide, pregabalin and valproate. Between Aug.-Oct. 2020, valproate was titrated off and brivaracetam was initiated. Her 90-day seizure rate reduced from 0.26 to 0.11 but 2 years later increased to 0.33. The root mean squared error from the ARIMAX(20,1,1) were significantly different between all test periods except pre-drug 2019 compared to post-drug 2022 (χ2=510.4914, p=4.3406e-108). The MA(50) model also found significant differences pre- and post-drug (χ2=807.0751, p=3.4094e-172).
Conclusions: Ultra-long sub-scalp EEG can prognosticate long-term trends in seizure rate using a MA or ARIMAX model. Large deviations in expected residual values indicate a change in seizure rate due ASM. A non-sustained change in residuals is an early indicator of ineffective ASM, flagging the need to try another therapy. Sub-scalp EEG could be used to make early and reliable anti-seizure medication efficacy assessments, significantly improving the management of epilepsy.
Funding: EpiminderPTY LTD and the Australian Government Research Training Program Scholarship from the University of Melbourne.
Translational Research