Night-to-night sleep EEG variability over one year

This study aimed to explore night-to-night variability of multiscale sleep patterns by analyzing subcutaneous electroencephalography (sqEEG) from 20 healthy participants over one year (205–388 nights per participant, 6,429 nights in total). We utilized the time series of aperiodic slopes, sigma and slow-wave power as a new whole-night unit of sleep macrostructure. Using dynamic time warping, we calculated the distances (differences) between those time series to assess night-to-night sleep macrostructure dissimilarity. We found that the overall sleep macrostructural patterns were relatively similar across nights (20% dissimilarity), while their temporal alignment was quite variable (time series warped by ∼60% for the best alignment). Lower variation in macrostructure dissimilarity was associated with better subjective sleep quality (r=-0.25).

Then, we qualitatively compared yearlong variation in macroscale, microscale (sleep stage proportions, mean spectral power) and mesoscale (sleep cycle duration) metrics. We found that intra-individual night-to-night variation was “low” (coefficients of variation < 20%) for spectral power, sleep duration, N2 and REM sleep; “medium” (20–40%) – for N3 and macrostructure dissimilarity; and “high” (>40%) – for sleep cycle duration, wake and N1. In summary, different sleep metrics showed differential night-to-night variability, which was more metric-specific than scale-dependent. This might reflect a distinction between more trait-like versus more dynamically varying features of sleep, although this assumption needs further clarification.