Abstracts

Rationale: Long-term video EEG monitoring (LTM) is a resource intensive procedure for diagnosing and characterizing behavioral episodes and seizures, and for presurgical evaluation of medically refractory epilepsy. How long should we monitor patients to establish a diagnosis of epilepsy or psychogenic nonepileptic seizures? Methods: We prospectively collected video EEG data from consecutive patients admitted for noninvasive LTM at the UW Regional Epilepsy Center between July 1, 1999 and June 1, 2006. We included only the initial LTM admissions, and classified any captured event as diagnostic, confirmatory, indeterminate or irrelevant, and recorded their time of occurrence relative to admission. We also coded diagnostic events as psychogenic nonepileptic, generalized, focal, or epileptic-undefined, the latter when the classification as focal or generalized could not be made. We estimated the probability of capturing a diagnostic event at different monitoring times using the product-limit (Kaplan Meier) estimator. Results: 1453 LTM studies (59.0% female) were selected. The first captured event was diagnostic in 51.4%, indeterminate in 30.3%, and irrelevant in 5.0% of patients. 13.3% of patients did not experience any events. 69.2% of patients had at least one diagnostic event. In these studies, the first diagnostic event was preceded by indeterminate or irrelevant events (range 1-26) in 25.8%. The first diagnostic event type was psychogenic nonepileptic in 40.8%, focal in 46.0%, generalized in 13.0%, and epileptic undefined in 0.2% of patients with diagnostic events. The median length of stay was 3.92 days (range 0.17-11.92) for all patients, and 3.92 days (range 0.21-10.92) and 4.92 days (range 0.17-11.92) for patients with and without captured diagnostic events, respectively. The median time to first diagnostic event was 0.7 days (range 0.01-7.03) for psychogenic nonepileptic, 0.99 days (range 0-7.35) for generalized, and 1.27 days (range 0.04-8.21) for focal events. The probability of capturing a diagnostic event in the average patient referred for LTM was 0.36 after 1 day, 0.5 after 47 hours, 0.6 after 3 days, 0.72 after 5 days, and 0.84 after 9 days of monitoring. Conclusions: LTM is a useful but resource intensive tool for event diagnosis, epilepsy characterization, and presurgical evaluation. This study was the first step in providing clinicians with a probability estimate for capturing a diagnostic event during LTM. This exceeds 50% in 2 days of recording.