Cross-sectional EEG study identifies two adolescent depression subgroups

Adolescents with major depressive disorder showed weaker connectivity between the default mode and limbic brain networks than healthy peers in a cross-sectional case-control study that combined resting-state EEG with childhood-trauma histories and depressive symptoms. The researchers also used multivariate analyses to identify two patient subgroups with distinct symptom and brain-network profiles, providing a preliminary description of heterogeneity within adolescent depression.

The study, published in BMC Psychology, included 67 medication-free adolescents with major depressive disorder and 35 matched healthy controls. The investigators estimated source-space functional connectivity from resting-state EEG, then applied sparse canonical correlation analysis to examine combinations of brain-network connections alongside clinical and childhood-trauma measures. This method identifies correlated sets of connectivity features and psychological characteristics. Because the study was cross-sectional, its results describe associations in the sampled adolescents but do not establish that trauma or any particular connectivity pattern caused depression.

Only one sparse canonical correlation analysis mode met the prespecified robustness criteria. It linked physical abuse, emotional neglect, and anxiety or somatization with distributed connectivity features involving the frontoparietal, limbic, visual, and salience networks. These findings concern coordinated patterns across measures rather than a single neural signature that independently diagnoses depression. The article’s abstract does not report the sizes or clinical characteristics of the resulting subgroups in enough detail to determine whether they represent clinically useful subtypes or how common they might be beyond this sample.

Connectivity within the visual network moderated the association between childhood trauma and depression severity. In statistical terms, the strength of that association varied with within-visual-network connectivity. The result does not show that altered visual-network connectivity increases the risk of depression following trauma, nor does it show that changing connectivity could change symptoms. The investigators then applied K-means clustering to canonical variates derived from the sparse canonical correlation analysis, separating the adolescents with major depressive disorder into two groups with different combinations of symptoms and network characteristics.

These findings remain preliminary because the sample was small and the design could not track changes over time. Including only medication-free adolescents with major depressive disorder also limits generalization to adolescents receiving treatment and to other clinical or demographic populations. Independent replication in larger samples will be required to assess whether the connectivity pattern, moderation result, and two-cluster solution persist. Even so, the study shows how EEG-derived network measures can be analyzed jointly with trauma exposure and symptom dimensions to characterize adolescent major depressive disorder, without delivering a diagnostic test or a treatment.

Subject of Research: EEG brain-network connectivity, childhood trauma, and symptom heterogeneity in adolescent major depressive disorder

Article Title: Data-driven subgroups of adolescent depression: integrating EEG-derived brain networks, childhood trauma, and depressive symptoms

Article References: Liu, X., Zeng, Y., Lan, X., Zhang, M., Li, W., Luo, Z., Liu, G., Xue, Y., Ning, Y., & Zhou, Y. (2026). Data-driven subgroups of adolescent depression: integrating EEG-derived brain networks, childhood trauma, and depressive symptoms. BMC Psychology. https://doi.org/10.1186/s40359-026-05753-2

Image Credits: AI Generated

DOI: 10.1186/s40359-026-05753-2

Keywords: adolescent depression, electroencephalography, brain networks, childhood trauma, functional connectivity, depressive symptoms, data-driven subgroups, cross-sectional study, default mode network, limbic network, Data-driven, subgroups