Abstracts

Rationale: It has long been recognized that some people with epilepsy experience seizures that occur with shorter than usual inter-seizure intervals. For these people, the occurrence of seizure clusters increases morbidity and mortality, and negatively impacts quality of life due to the unpredictability of the next seizure and fear of evolution into status epilepticus. There are currently a variety of definitions of seizure clusters in clinical usage, but the prognostic significance of these definitions is not clear. The ability to predict the onset of a seizure cluster on an individualized patient basis would allow for more aggressive intervention to prevent seizure clusters or evolution into status epilepticus. There is an urgent need for a more uniform definition of seizure clusters and clarification of the clinical significance of current definitions. We propose a new method for defining seizure clusters to achieve this purpose. Methods: Patient-reported seizure diaries from SeizureTracker.com were collected from over 13,000 people with epilepsy from December 2007 to February 2016. Invalid and duplicate seizure diary entries were excluded, and seizure diaries recorded for at least 24 hours with either aura, simple partial, complex partial, or secondarily generalized seizures were included. A total of 377,282 seizures reported by 6,369 people with epilepsy were analyzed for inter-seizure interval (ISI) and number of seizures in the 24 hours following each preceding seizure pair. The mathematical relationship between ISI and seizure clustering was assessed based on the Akaike information criterion. Survival analysis was used to estimate the probability of evolving into a seizure cluster within the next 24 hours for increasing ISI lengths. Results: As expected, shorter ISI between any preceding pair of seizures was significantly associated with more seizures in the 24 hours following the preceding pair. This relationship empirically followed an inverse polynomial function of the form Y=aX^b-1 (Y=number of seizures, X = ISI length). The prognostic significance of the inter-seizure interval for predicting the probability that a shorter-than-usual ISI length would evolve into a seizure cluster was estimated for various ISI lengths and baseline seizure frequencies. The conventional definitions of seizure clusters based on any two or more consecutive seizures with ISI less than 4 or 8 hours were found to correspond to an elbow in the survival probability curve and the ISI at which there was an approximately 50% chance of evolving into a seizure cluster, respectively, providing new empiric evidence for the clinical significance of these thresholds. Conclusions: We propose a new method for defining seizure clusters based on the prognostic significance of the preceding inter-seizure interval for predicting the probability of evolving into a seizure cluster. We analyzed probabilities of having 3 or more seizures in 24 hours, but this approach is generalizable to other time units of interest (shorter or longer than 24 hours). Rather than proposing a single predetermined definition in the form of ISI < N units of time or number of seizures / N units of time, we recommend a new probabilistic approach which considers the probability of evolving into a seizure cluster. This new approach to defining seizure clusters allows clinicians to define clusters on an individual patient basis, based on what probability they are willing to tolerate considering the risks and benefits of aggressive intervention. Our results provide clinicians with a new prognostic calculator to estimate the risk of evolving into a seizure cluster on an individual patient basis, based on the patient’s individual clinical risk factors. Funding: Not applicable