Abstracts

Rationale: In clinical practice, the choice of antiepileptic drug (AED) is often guided by the ease of use and side effect profile related to the specific patient. So far, there are no means to predict individual therapeutic response. This inevitably leads to patients having new seizures during the establishment of treatment as the epileptologist tries to set up the best therapy. This is not only a burden on the quality of life of patients but also holds a significant risk of injury from seizures or even sudden unexpected death in epilepsy. Therefore, a biomarker that allows assessing whether or not a patient will be a responder to a specific AED, without the need for waiting for the next seizure to happen, would be of great interest, as this would allow to decide more quickly how to continue therapy. In this study we investigate the development of an EEG biomarkers indicative of the efficacy of Brivaracetam. Methods: Fifty patients were included in this multi-centric study: 30 from the Epilepsy Center Hessen, University Hospital Marburg, Germany and 20 from Oslo University Hospital, Norway. EEG before and on Brivaracetam, patient details (gender, age, #AEDs tried) and outcome (19 responder, 31 non-responder) were collected retrospectively. A responder was identified as a patient who had at least a reduction of 50% in seizure frequency per month. Spike quantification, spectral analysis and functional connectivity analysis in source space was performed to extract features from both EEG recordings. For each type of analysis a multitude of features was extracted and ranked using a T-test. The 10 features with most discriminative power according to the T-test were used to train a classifier. One hundred balanced datasets containing 19 responders and 19-non responders were generated and classification was performed using leave-one-out cross-validation to assess the performance of the classifier. Area under the curve, sensitivity and specificity were computed as performance measures. Results: Spike quantification, spectral analysis and source space functional connectivity analysis led to an area under the curve of 0.35, 0.90 and 0.75, respectively. For spectral analysis delta and beta band characteristics showed most discriminative power resulting in an accuracy of 85%, sensitivity of 82% and specificity of 87%. Conclusions: This retrospective study shows the potential of EEG biomarker indicative of the efficacy of AEDs with high accuracy. Nevertheless, prospective studies are needed to assess the added value of the developed EEG biomarker in a clinical setting. Funding: No funding