Abstracts

Rationale: Identifying patients as drug-resistant or drug-responsive is a central aspect of epilepsy clinical care. Since 2009, the International League Against Epilepsy (ILAE) Commission has defined seizure freedom as a minimum of three times the longest preintervention interseizure interval or 12 months, whichever is longer, providing a precise definition to facilitate classification of treatment response. However, recent evidence suggests that the natural history of epilepsy fluctuates over time, and that apparent responsiveness to treatment may merely represent changes in seizure frequency unrelated to treatment. Therefore, proper classification of patients as drug-resistant or drug-responsive sometimes is confounded by natural history. In this study, we investigate the natural variability of epilepsy as well as differences for various patient subgroups. In contrast to traditional approaches to epilepsy natural history, which have investigated temporal patterns in seizure frequency based on observed seizure counts, we propose and investigate a new concept of epilepsy natural history as the pattern of fluctuations in the underlying seizure threshold. Methods: We developed a statistical model for modeling the natural history of epilepsy based on the pattern of fluctuations in an underlying latent seizure threshold. Seizure occurrence and clinical history data was obtained for the period of 12/1/2007 to 2/25/2016 from the seizure log application developed by Seizure Tracker, LLC. The dataset included 925 patients who collectively had 384,399 seizures, and diary durations ranging 365 to 3005 days. To capture the underlying seizure threshold, seizures from each individual subject are modeled through a Hidden Markov model as the observed manifestation of each subject's underlying seizure threshold. Differences in the probability of transitioning between various seizure threshold ranges, the probability of remaining in a state of low predisposition for seizures, and long-run proportion of time for remaining in a state with low predisposition for seizures were evaluated for nine clinical covariates, including seizure etiology, seasonal timing, and seizure type. Results: Analysis is currently under way. Results will be shared at the conference. Conclusions: Understanding the natural history of epilepsy is a crucial component to evaluating treatment responsiveness. A patient's observed response to treatment may be easily confounded by natural fluctuations in seizure frequency. Improved understanding of the influence of various clinical factors on natural history may lead to improved clinical evaluation of drug response and therapeutic decision-making. It also may impact seizure prediction through identification of periods of greater seizure risk. Funding: Funding in support of this abstract was provided by the National Library of Medicine Training Fellowship in Biomedical Informatics, Gulf Coast Consortia for Quantitative Biomedical Sciences (Grant #2T15-LM007093-21) (SC); the National Institute of Health (Grant #5T32-CA096520-07) (SC); NINDS Division of Intramural Research (DMG, WT); NIH-NINDS K23 Grant NS044936 (JMS); and The Leff Family Foundation (JMS).