Does sleep benefit memory consolidation in children with focal epilepsy?

Children with epilepsy have high rates of both cognitive impairment and sleep disruption. It is thus assumed that sleep-dependent memory consolidation is vulnerable to ongoing epileptic activity, but direct evidence of this is limited.

We performed a within-subject comparison of memory retention across intervals of wake or overnight sleep. Healthy children (n = 21, 6-16 years, 12 female) and children with focal epilepsy (n = 22, 6-16 years, 9 female) performed verbal and visuospatial memory tasks under each condition. Sleep was assessed with electroencephalography (EEG) polysomnography during the overnight interval. Interictal discharges were quantified manually.

Memory retention was greater in the sleep condition in both the verbal (F_1,39 = 10.8, p = 0.002, Cohen's d = 0.67) and the visuospatial (F_1,36 = 4.23, p = 0.05, Cohen's d = 0.40) tasks, with no significant interaction of group by condition in either task. Across the total sample, gain in memory retention with sleep in the verbal task correlated with duration of slow wave sleep (r = 0.4, p = 0.01). In patients, sleep-dependent memory consolidation was negatively correlated with interictal discharge rate in both the verbal (ρ = -0.49, p = 0.04) and visuospatial (ρ = -0.45, p = 0.08) tasks. On post hoc analysis, a longer history of epilepsy (r = 0.53, p = 0.01) and a temporal (t₁₀ = 1.8, p = 0.1, Cohen's d = 0.86) rather than an extratemporal seizure focus (t₁₀ = 0.8, p = 0.4, Cohen's d = 0.30) was associated with greater contribution of sleep to verbal memory retention.

We have demonstrated that memory consolidation in children with focal epilepsy benefits from sleep, showing the same correlation with slow wave sleep as in healthy children, but an inverse relationship with the interictal discharge load during sleep. This mechanism appears to be increasingly recruited with longer duration of illness, indicating a resilient homeostatic function which may be harnessed to aid learning.