Abstracts

Seizures often occur in clusters or flurries, but a specific definition of clustering and robust estimates of its frequency are lacking. Conceptually, seizure clustering implies that the probability of a seizure is increased by the occurrence of recent prior seizures. In a prospective seizure diary study, we applied a widely used clinical definition of clustering, and compared the results to a statistical definition derived from testing the empirical distribution of seizures for homogeneity (constant rate over time), to provide a foundation for better detection and management of seizure clustering. Subjects were recruited from the Epilepsy Management Center and neurology clinics at MMC. Inclusion criteria: Age [gt]= 18 years; localization-related epilepsy; [gt]=1 seizure within the prior 12 months; ability to maintain a daily seizure diary. Seizure clustering was defined using a standard clinical definition (3 or more seizures in 24 hours) and a statistical definition based on deviation from a Poisson distribution (variance of observed numbers of seizures that was significantly greater than the mean) utilizing a formula described by Boots and Getis (Point Pattern Analysis, 1988). Data were analyzed for the first 55 subjects completing the study. Mean follow up was 227 diary days. Nine subjects (16%) had no seizures, 19 (35%) were non-clusterers by either criterion, 26 (47%) met the definition of clustering by the clinical definition while 12 (22%) had a seizure distribution which met the statistical definition. All subjects identified by the statistical method also met the clinical definition. Review of the 14 discrepant cases (meeting the clinical but not the statistical definition) revealed that these subjects usually had a high baseline seizure frequency with occasional episodes of 3 sz. in 24 hrs by chance, or had a typical pattern of single seizures with a rare occurrence of 3 or more sz. in 24 hrs. All subjects identified as clusterers had at least one isolated single seizure. The relative frequency of seizure clustering among our patients was 22% using a statistical definition and 47% using a clinical definition. Analysis of discrepant cases suggests that the clinical definition generates false positives. While the clinical definition of clustering may have utility (i.e. assessing seizure independence for presurgical evaluation), the statistical approach appears to be a more robust method of identifying a subset of patients who truly cluster. The results also suggest that even in individuals who meet a statistical definition of clustering, isolated seizures do occur. Optimal management will require clinical methods for the identification of individuals at high risk for seizure clustering. (Supported by K23NS02192 (PI: Dr. Haut))