Age-Adjusted Exploratory Analysis of Spectral, Aperiodic, and Microstate EEG Features in Multiple Sclerosis

Electroencephalography (EEG) can characterize functional brain alterations in multiple sclerosis (MS), but many EEG features vary with age. We therefore examined resting-state EEG associations with MS using age-adjusted models while explicitly accounting for the severe age imbalance in the available dataset.

Resting-state 128-channel EEG data from the BrainLat dataset were analyzed. Demographic files included 32 patients with MS and 32 healthy controls (HC); global-feature models included 29 patients with MS and 32 HC, and channel-level sample sizes varied by feature. Diagnosis and age were entered in linear models. Benjamini-Hochberg false discovery rate (FDR) correction was applied separately to five global spectral/aperiodic tests, 16 global microstate tests, and the 128 channel-wise tests within each feature. Within-MS continuous age regression and exploratory age-stratified ANOVA were evaluated with the same family definitions.

After correction, diagnosis coefficients were significant for global bandwidth (β = 1.928, q = 0.0287), global aperiodic offset (β = 1.816, q = 3.29 × 10⁻⁷), and offset at all 128 channels (β range = 1.52-2.41; q range = 3.56 × 10⁻⁸ to 7.43 × 10⁻⁵). Global central frequency, exponent, PSD, and all 16 microstate measures did not survive FDR correction. No within-MS continuous-age or age-stratified result survived FDR correction.

Aperiodic offset showed the strongest diagnosis association under the specified age-adjusted model. However, the MS and HC age ranges did not overlap (21-55 vs 56-83 years), so the diagnosis coefficient depends on extrapolation and cannot establish an MS-specific effect. The findings are exploratory and require age-matched replication with detailed clinical, vigilance, site, sex, and MRI characterization.