Abstracts

Rationale: Memory impairment is common in people with epilepsy (PWE). Although sleep is known to facilitate memory consolidation processes, the extent to which nocturnal seizures disrupt memory in PWE is unknown. Investigating the potential effect of nocturnal seizures on memory requires a task designed to assess memory for material learned the evening before a period of sleep. We created a psychometrically sound, computerized word paired-associates memory task (WPA) with 5 alternate forms to assess overnight memory longitudinally in adults ages 18-55. We administered the task over multiple days to a group of healthy controls (HC) and PWE admitted for seizure monitoring. We hypothesized HC would perform better than PWE, and that for PWE, memory would be poorer in the morning following nights with nocturnal seizures compared to nights without seizures in a within-subjects design.

Methods: Stimuli included 200 concrete nouns with low semantic relatedness (latent semantic analysis < 0.16 cosine similarity score) and low emotional valence. The five WPA versions were matched for word frequency in spoken language and number of syllables. A pilot study with 26 healthy adults determined feasibility of remote, computerized task administration. Learning trials were completed at night before sleep and word-pair memory was assessed the following morning. Performance variables included learning and delayed trial response times and accuracy, and percent retention.


We collected normative data from 37 HC over five consecutive days (Age=24.0±7.0, Education =15.0±2.1, IQ=98.3±11.8). Bayesian analyses compared WPA performance across test versions in HC to determine comparability of the alternate forms. PWE completed three days of the WPA task (N=25, 17 TLE, Age=29.6±7.5, Education =14.9±2.3, IQ=98.8±16.2). One-way ANCOVAs (age as covariate) with bootstrapping compared WPA performance between HC and PWE. We present only descriptive data for comparison of WPA memory performance after nights with and without seizures due to the small sample size (n=4).






Results: Bayesian repeated measures ANOVAs confirmed comparability of the 5 versions (Bayes factors for alternative model of differences < 0.13). PWE performed poorer than HC on all WPA performance variables (bootstrapped p values for mean differences < 0.05, partial eta^2 >0.08). Descriptive data trended towards poorer memory performance after nights with nocturnal seizures (delayed recall 5.63±5.22, % retention 45.91±31.62) compared to nights without seizures (delayed recall 9.13±6.19, % retention 67.81±29.35).

Conclusions: We created a multi-form remote instrument to assess memory after multiple nights of sleep that, ultimately, will be available for clinical and for research use in multiple disciplines and populations. Preliminary normative data for the WPA task confirms comparability across task forms, and suggest feasibility and validity for use with PWE. Preliminary patient results suggest that nocturnal seizures may disrupt memory consolidation for material learned before sleep, although definitive conclusions require a larger sample.

Funding: American Epilepsy Society Predoctoral Research Fellowship