Abstracts

Rationale: SUDEP occurs more commonly in sleep. Current pathophysiological models suggest a link with arousal impairment. In animal studies, when mice are forced into a seizure during REM sleep to override its anti-epileptic properties, seizures are associated with mortality. Therefore, we sought to assess the arousal threshold during REM sleep in persons with epilepsy (PWE). We also addressed the possibility of local sleep by analyzing arousability during sequential REM sleep cycles over brain space.

Methods: Arousability was represented by the odds ratio product (ORP), which is a measure of wake propensity on a scale from 0-2.5, with higher values indicating higher probabilities of awakening. We calculated a modified ORP (mORP) based on high density 10-10 EEG from 13 PWE over 1-3 nights of recording (MY, GB) using a referential average of the TP9/TP10 electrodes. We only used mORP values from REM sleep segments in each person’s night(s) with varying number of data points per REM sleep segment(s) in each night(s). Artifactual electrodes were discarded from analysis. To account for the varying number of nights and segments per patient, a Generalized Estimating Equation (GEE) was modeled using the GEEQBOX MATLAB toolbox. The data was assumed to be equicorrelated due to each night coming after one another, as well the REM sleep segments being sequentially one after another from the beginning to end of sleep. GEE was run by inputting a column to indicate patient, mORP values, and night; with a correlation matrix that contained a column of nights, REM sleep segments, and ones (as per GEEQBOX). Significance was set at p≤0.000833 (Bonferroni correction for 60 electrodes).

Results: Of 13 adult PWE, 2 had genetic generalized epilepsy, 5 had left hemispheric foci, 3 had right hemispheric foci, and 2 had bilateral independent foci. Over 60 channels, there were 3 million mORP data points, Fig.1. The 3D boxplot is a representation of all the mORP values (z-axis) regarding each of the 30 nights (x-axis) by each channel (y-axis). With this dataset, we were able to run a GEE model which resulted in 57/60 channels achieving statistical significance. In Fig.2, the beta (b) values represent the regression coefficients of the ‘segments’ variable from the significant channels (if not significant, b=0). These values signify that mORP, and by extension arousability, increases by b in each channel into the next sequential REM sleep cycle. Fig.2 also shows that for the 13 PWE, REM sleep over the left cerebral hemisphere has a higher wake propensity (0.017≤b≤0.059; disregarding CP5 due to artifact) than the right hemisphere (0≤b≤0.036).

Conclusions: Our findings are proof-of-concept that arousability can be spatially assessed over sequential REM sleep cycles across nights and PWE. In our cohort, arousability over the left hemisphere selectively increased over sequential REM sleep cycles despite a variety of generalized and focal epileptogenic localizations. Future studies can focus on normative data sets of REM sleep arousability over brain space across nights in healthy controls, and examine greater numbers of PWE with particularly high SUDEP risk.

Funding: Please list any funding that was received in support of this abstract.: Mathematics of Information Technology and Complex Systems.