Intrinsic electrophysiological activity maps a latent dimension of poor sleep quality and reduced cognitive performance: a magnetoencephalography study using Cam-CAN data.

Sleep quality and cognition vary as functions of lifestyle, genetics, and health. However, poor sleep quality is prevalent, reportedly affecting approximately 38% of the adult population. Poor sleep quality is a significant risk factor for mood disorders, a predictive factor in cognitive decline in later life, and sleep disturbance is a putative precursor to severe cognitive impairment and is predictive of dementia onset. The direct relationship between these factors and intrinsic neural function is poorly understood. Integrating neurophysiology, cognition, and sleep quality to reveal latent factors would help understand the neurophysiology of sleep disturbances and their relation to cognitive performance. Here, we used data from the Cam-CAN dataset with a partial least squares (PLS) approach, producing a multivariate cross-decomposition model to map resting-state magnetoencephalography (MEG) data and cognitive/sleep scores of healthy controls (n = 490, age 18-86). Normative modelling was applied to MEG data, correcting for effects of age, sex, and handedness. We identify a significant relationship between poorer self-reported sleep quality and lower cognitive performance across multiple domains, characterised by excess low-frequency neural activity and reduced high-frequency activity, particularly in the alpha band. Moreover, globally increased low frequency and decreased alpha-beta functional connectivity - the dominant frequency channel in neural connectivity at rest across the brain - contribute significantly to this relationship. Our multivariate mapping approach enabled us to parse the electrophysiological signatures directly related to sleep quality and cognitive performance, independent of the confounding effects of age, and show that neural slowing and dysconnectivity are intrinsically linked to self-reported sleep quality and cognitive performance.