Heterogeneity in late-life depression: limited replication of resting-state EEG markers in a Korean community cohort

Background

Resting-state quantitative electroencephalography (qEEG) has been proposed as an objective biomarker of depression, supported by hypothesised links between left-lateralised frontal alpha asymmetry (FAA) and aberrant approach-withdrawal motivation, and between elevated beta power and altered fronto-limbic reward processing. A recent meta-analysis of 52 studies reported increased beta power and a trend for left-lateralised FAA as the most consistent findings in younger and middle-aged clinical samples [1]. Yet older adults are markedly under-represented in this evidence base: across the six studies in the beta power meta-analysis, only one cohort had a mean age above 65 years [1]. Whether qEEG markers established in younger samples generalise to community-dwelling older adults, who differ in neurobiology and clinical heterogeneity, remains untested at scale.

Methods

We analysed resting-state EEG from 280 community-dwelling adults aged ≥65 years (mean age 73.5 years; 61% female) from the BEODEUL (Behavioural and Emotional Observation in Digitally Enhanced Ubiquitous Living) cohort, a Gangwon-do living-lab subcohort within the Korean Longitudinal Study on Digitally Optimised Mental Healthcare (KLOSDOM). EEG was recorded under eyes-open (EO) and eyes-closed (EC) conditions using a 32-channel dry-electrode system (BIOS-S40, BioBrain Inc.). Across seven frequency bands (delta to gamma) we computed absolute and relative power, FAA, alpha-to-high-beta ratio, and spectral edge frequencies (SEF50/SEF90), yielding 864 EEG variables. Depressive symptoms were measured with the PHQ-9 (cutoff ≥10). Categorical comparisons used Welch t-tests with Hedges’ g; dimensional analyses used Pearson and Spearman correlations. Multiple comparisons were controlled with the Benjamini-Hochberg false discovery rate (FDR); sensitivity analyses adjusted for age and sex, and additionally for body mass index and daytime sleepiness given their potential influence on EEG.

Results

Zero of 864 EEG variables differed significantly between depressive-symptom (n = 90 to 100) and non-depressed (n = 180 to 188) groups in either EO or EC condition after FDR correction. Effect sizes were small overall (mean |g| = 0.11; maximum 0.46), substantially below the pooled effect for beta power reported in younger clinical samples [1]. FAA effects were negligible (|g| < 0.10; p > 0.50 in both EO and EC). Pearson correlations between PHQ-9 and EEG variables identified a small set of associations in EO (q < 0.05), strongest for posterior-temporal alpha-to-high-beta ratio; however, none survived Spearman correlation, indicating that these associations were not robust to non-parametric analysis. No association survived FDR with non-parametric testing. Findings were unchanged after age- and sex-adjusted sensitivity analyses. Signals previously observed in a preliminary analysis of the first 176 participants without multiple-comparison correction did not persist in this larger sample with FDR control.

Conclusions

In a large Korean community-dwelling older-adult cohort, established qEEG markers of depression, including beta power changes and FAA, did not replicate. Resting-state univariate qEEG markers derived from younger samples may not directly capture depressive symptoms in community-dwelling older populations. Advancing community-based detection of late-life depression will require biologically informed subtyping and the development of age-specific biomarkers suitable for real-world implementation.

References

1. Heitmann, H., Siani, J.-F., Zebhauser, P.T., Henningsen, P., Leucht, S., Priller, J., Ploner, M., 2026. Resting-state EEG activity in depression: a systematic review and meta-analysis. NeuroImage Clin., 103997.