Abstracts

Rationale: Sleep has a complex, bidirectional relationship with seizures. Sleep deprivation may be a seizure trigger, and seizures often have a profound effect on sleep duration and quality. While some data is available in controlled, clinical environments or patient surveys, the relationship between sleep and seizures in real-world settings is poorly understood, owing partly to the challenges of measuring sleep unobtrusively and longitudinally in the home environment. Here, we aim to investigate the relationship between sleep and seizures using wearable biometric sensors in a real-world setting. 

Methods: A wearable photoplethysmographic sensor and inertial motion sensor (Neureka Sleep Monitor) recorded 5309 sleep sessions longer than 4 hours from 98 individuals with epilepsy over a 18 month period. During this time, each individual maintained an electronic seizure diary. The inertial motion sensor data was processed using an actigraphy-based approach to identify periods of restlessness during sleep, which was then used to calculate total sleep duration and sleep efficiency. By comparing these metrics to expected values for healthy individuals of a similar age, a sleep quality score was generated for each night. We performed a linear regression for each individual with the sleep quality score as the independent variable and the normalized number of reported seizures in the subsequent 48 hours as the dependent variable. To compare the relationship between the sleep quality score and the seizure occurrences, the analysis was limited to patients with at least 10 reported seizures and 10 recorded sleep nights. In addition to this, we also examined the sleep efficiency (defined by total time asleep divided by total time in bed) of sleep sessions immediately preceding days with seizures and compared these with sleep sessions preceding seizure-free days. 

Results: The analysis included 29 individuals with 2435 sleep sessions (mean duration = 6.8±1.6hours, mean sleep quality score = 91.8±11.5%) and 3199 reported seizures (mean seizure frequency = 5.7±7.0 seizures/week). 23 (79.3%) individuals reported seizures with significant circadian patterns (p < 0.05 with the Omnibus test for circular uniformity), and none had exclusively nocturnal seizures. Among the 29 individuals, 5 (17.2%) had a significant negative correlation (p < 0.05) between the sleep quality score and seizures in the subsequent 48 hours. This proportion was found to be significant (p = 0.0136, Binomial Test). Moreover, nights immediately preceding seizures had a significantly lower sleep efficiency when compared to nights preceding seizure free days (p = 0.0437, Paired T-test, Normality verified with a Lilliefors Test).