Longitudinal resting‐state EEG along the Alzheimer’s disease continuum: Sample size considerations for clinical trials.

Abstract

AbstractBackgroundTo enable successful inclusion of resting‐state electroencephalography (EEG) outcome measures in Alzheimer’s disease (AD) clinical trials, we (retrospectively) mapped the progression of EEG markers over time in amyloid‐positive subjects.MethodsResting‐state 21‐channel EEG was recorded in 173 subjects positive for amyloid deposition (subjective cognitive decline, n = 16; mild cognitive impairment, n = 94; dementia due to AD, n = 63). Two or more EEG recordings were available for all subjects. We computed whole‐brain and regional relative power (i.e., theta (4‐8Hz), alpha1 (8‐10Hz), alpha2 (10‐13Hz), beta (13‐30Hz)), peak frequency, signal variability (i.e., theta Permutation Entropy) and functional connectivity values (i.e., alpha and beta corrected Amplitude Envelope Correlation, theta Phase Lag Index, weighted Symbolic Mutual Information, inverted Joint Permutation Entropy). Whole‐group linear mixed effects models were used to model the development of EEG markers over time. Group‐wise analysis was performed to investigate potential differences in change trajectories between diagnostic groups. The longitudinal relation between EEG and cognition was evaluated. Finally, we estimated the minimum sample size required to detect a stabilizing treatment effect using EEG markers, in hypothetical clinical trials of 1‐ or 2‐year duration.ResultsWhole‐group analysis revealed significant regional and global oscillatory slowing over time (i.e., increased relative theta power, decreased alpha2 and beta power), with strongest effects for temporal and parieto‐occipital regions. Signal variability showed stronger decline on a global scale. We report no significant longitudinal changes in functional connectivity strength. Disease severity at baseline influenced EEG markers’ rates of change, with fastest deterioration in MCI subjects. A decrease of the parieto‐occipital peak frequency, temporal relative alpha2 and beta power, as well as an increase in temporal relative theta power, were significantly associated with a decline in Mini Mental State Examination score. We estimate that 1‐year trials, focusing on amyloid‐positive MCI subjects, require 148 or 438 subjects per arm to detect a stabilizing treatment effect on temporal relative theta‐ or beta power, respectively.ConclusionResting‐state EEG markers can facilitate early detection of target engagement in an objective, non‐invasive, low‐cost manner. Spectral measures, particularly recorded from temporal regions, present robust secondary endpoints for AD interventions.