Background: Childhood trauma (CT) significantly increases vulnerability to adolescent major depressive disorder (MDD), yet the underlying neurobiological mechanisms remain unclear.
Methods: This study applied graph-theoretical analysis to structural MRI data to examine how CT influences brain structural networks and antidepressant treatment outcomes in adolescent MDD. A total of 339 adolescents with MDD and 152 healthy controls (HC) were recruited. The MDD group was subdivided into those with (MDD-CT; n = 209) and without (MDD-nCT; n = 106) a history of CT.
Results: Compared with HC, both MDD subgroups exhibited a shift toward more randomized network properties. Notably, MDD-CT was characterized by more severe depressive and anxiety symptoms and showed focal disruptions within the default mode network (DMN), whereas MDD-nCT involved broader alterations across the central executive, salience, and DMN networks. Following antidepressant treatment, MDD-CT had a lower response rate than MDD-nCT (48.24% vs. 65.38%). Significant group-by-time interactions were observed in the right amygdala and caudate nucleus. Moreover, a machine learning model combining autoencoder-based feature extraction with support vector machine classification predicted treatment response with 89.8% accuracy.
Conclusions: This study revealed unique and common structural network abnormalities in MDD-CT and MDD-nCT, suggesting that childhood trauma leads to focused disruptions in the DMN. Treatment-induced changes in the amygdala and caudate nucleus may contribute to poorer outcomes in MDD-CT. These findings provide insights into the neurobiological underpinnings of trauma-related depression and highlight the potential of structural connectivity markers for guiding biomarker-informed treatment strategies in adolescent MDD.
Keywords: Adolescent; Childhood trauma; Machine learning; Major depressive disorder; Structural magnetic resonance imaging.
Copyright © 2026 Elsevier B.V. All rights reserved.