Abstracts

Rationale: If seizures were predictable they might become more preventable. In a paper diary study, we have previously demonstrated that a subset of patients with epilepsy successfully predicted their seizures. Herein, we repeated this evaluation using electronic diaries with time stamping. Methods: Eligible subjects were 18 or older, had localization-related epilepsy; with 3 or more seizures per month. Subjects maintained an electronic seizure diary using a handheld device, providing data upon awakening and before bed on a daily basis (fixed interval AM and PM), ad lib in relation to seizure and at quasi-random control periods. Though we collected data on triggers and premonitory features this abstract focuses only on self-prediction assessed by the following question: How likely are you to experience a seizure [today (AM diary)/in the next 24 hrs (PM diary)? Reponses included: Almost certain (>95% chance); Very likely (75-94% chance); Fairly likely (50-74% chance); Quite unlikely (25-49% chance); Very unlikely (<25% chance). Relative incidence of seizure within 6,12,18 and 24 hours was examined using Poisson models with log normal random effects to adjust for multiple observations. Results: Nineteen subjects returned 3274 interval contingent diary entries (1680 AM entries; 1594 PM entries) and reported 257 seizures. Total number of responses for each prediction choice were: almost certain, n=15; very likely, n=77; fairly likely, n=346; quite unlikely, n=985; very unlikely, n=1851. A positive response to self-prediction (almost certain, very likely, fairly likely) was significantly associated with seizure occurrence within the next 12 hours, while negative responses (quite unlikely, very unlikely) were not. Odds ratios associated with endorsing specific prediction choices are presented in Table 1. Combining all positive responses yielded an OR of 5.6, 95% CI 3.4-9.4, p<0.001 for any positive prediction. When examined by time of entry, morning diary self prediction ratings were more significant than those collected in the evening diaries. The overall pattern of results was similar for 6 - 24 hour prediction windows. Conclusions: Conclusions: Subjects with epilepsy can successfully self-predict seizures using time stamped electronic diaries. Morning self-prediction ratings significantly predicted seizures for the subsequent 12 hours, while evening ratings did not. Subjects are likely integrating objective and subjective cues such as seizure precipitants and premonitory features to self-predict; further prediction-modeling work will incorporate these biological variables and consider individual as well as group models.