Abstracts

Rationale: Continuous spike and wave of sleep with encephalopathy (CSWS) is a severe developmental epileptic encephalopathy characterized by abundant spikes during NREM sleep and declines in cerebral function. The mechanisms of the cognitive deficits in CSWS and related epileptic encephalopathies remain unknown, but high-dose diazepam may improve symptoms. Prior work has not been able to establish a relationship between spike activity and cognitive symptoms. Spikes can be generated by “hijacking” the same thalamocortical circuits used to generate sleep spindles. Sleep spindles are brief bursts of oscillatory sigma band (10-16 Hz) activity present only in non-rapid eye movement (NREM) sleep and are critical for sleep-dependent memory consolidation. We hypothesized that in CSWS: 1) spindle rate would be inversely related to spike rate, and 2) spindle density would increase after high-dose diazepam treatment.

Methods: All pediatric patients evaluated in the MGH EEG laboratory between 4/2010 and 3/2021 with documentation of an epileptic encephalopathy treated with high-dose diazepam (1mg/kg), abundant sleep-activated spikes, and readily available pre- and post-treatment raw EEG recordings of NREM sleep were included. 6 patients have been identified and analyzed to date. For each subject, EEG was reviewed and available NREM Stages 2 and 3 data at baseline and following 1 day (n=6) or 2 days (n=2) of diazepam treatment were included for analysis (average 199 minutes, range 17-453).

Sleep spindles from 1000 seconds of EEG (from both pre- and post-treatment recordings in 4 subjects) were manually marked by consensus between 2 reviewers, resulting in 10168 hand-marked spindles. These data were used to update a validated latent state spindle detector for use in EEGs with epileptiform spikes. The updated detector was validated on the CSWS dataset using leave-one-out cross-validation. To quantify spikes, we used Persyst13. To quantify spindle rate, the validated spindle detector was applied to the channel with the highest spike rate in each subject during NREM sleep pre- and post-diazepam. Spike and spindle rates (scaled to baseline for each subject) were compared using linear regression. Pre- and post-diazepam spindle rates (scaled to baseline) were compared using a mixed-effect regression model.

Results: The automated spindle detector worked well with an F1 score of 0.464 against expert markings. Spindle rate and log (spike rate) were anticorrelated (p=0.00062). There was an increase in spindle rate after diazepam treatment in 5/6 subjects. On group-level analysis, spindle rate trended to increase after diazepam treatment compared to pre- diazepam EEGs (p=0.065).

Conclusions: In this case series of children with CSWS, spindle density is inversely related to spike rate and increases after high-dose diazepam treatment. These data provide evidence that thalamocortical circuits are disrupted in CSWS and suggest that diazepam’s efficacy may be related to the positive impact on spindle rate.

Funding: Please list any funding that was received in support of this abstract.: CJC, DM, and MAK are funded by NIH NINDS R01NS115868.