Early detection of antiseizure medication inefficacy using an implantable continuous EEG system and a personalized model: a case study

Evaluating anti-seizure medication (ASM) efficacy can be limited by inaccurate seizure diaries and periodic fluctuations in seizure frequency, known as seizure cycles. These limitations may prolong monitoring of ineffective treatments. This study explores implantable continuous EEG monitoring (iCEM™) with Epiminder’s Minder® system and timeseries modelling to improve efficacy assessments.

This retrospective case study examines a 49-year-old female with drug-resistant focal epilepsy with iCEM (Nov. 2019). The participant maintained a seizure diary and was followed-up for 3 years. ASMs were changed in Aug.-Oct. 2020. A personalised autoregressive model incorporating interictal epileptiform discharge cycles to project 3-monthly seizure rates was trained and validated on post-drug data then tested on four held-out datasets (two pre-drug and two post-drug). The Kruskal-Wallis test assessed model performance between drug periods (α = 0.05).

Only 37 % of seizures were reported. Post-drug, diary-reported seizures increased while detected seizures decreased, but both remained within normal seizure rate variability. ASM inefficacy was addressed after 3 years. The autoregressive model mean squared error post-drug was 0.17 and 0.13 seizures per day over 3-months², which were significantly different from pre-drug (0.49 and 0.58 seizures per day over 3-months², $H$=336.82, p = 2.44exp.^-89), suggesting the model could recognise when an altered drug regimen affected seizure rate and interictal epileptiform discharges.

Seizures identified using iCEM combined with an individualised model may be able to distinguish drug-induced changes in seizure rate from normal variability. This proof-of-concept study offers useful information towards the development of methods that can support early treatment assessments, potentially shortening the time to find an optimal therapy.