# Childhood chronic stress associated with abnormal brain white matter networks in first-episode, drug-naïve MDD adolescents: a machine learning study

## TL;DR

This study finds that chronic childhood stress is linked to abnormal brain white matter networks in adolescents with depression, which could help identify subtypes of the disorder.

## Contribution

The study is the first to link chronic childhood stress with specific white matter network abnormalities in drug-naïve adolescents with MDD using machine learning.

## Key findings

- MDD adolescents showed larger network radius and diameter, and lower small-worldness and global efficiency in white matter networks.
- Local abnormalities were found in brain regions like the prefrontal cortex and cingulate cortex.
- Chronic childhood stress correlated with altered topological properties in specific white matter regions.

## Abstract

This study aims to identify the topological abnormalities of the brain white matter structural network in adolescents with first episode, drug-naïve major depressive disorder (MDD), and to explore the relationship between these abnormalities and chronic childhood stress.

T1-weighted images and diffusion tensor imaging (DTI) data were collected from 77 first-episode, drug-naïve MDD adolescents and 31 healthy controls (HCs). A whole-brain white matter structural network was constructed for each participant. Graph-theoretical analyses were employed to investigate the topological properties of white matter structural networks. Subsequently, we applied four machine learning methods—univariate linear regression, stepwise regression, LASSO regression, and random forest—to screen for relevant variables. The selected variables were then integrated into a final multivariate linear regression model to examine their correlations with Childhood Chronic Stress Questionnaire (CCSQ) scores.

Compared with HCs, adolescents with MDD showed significantly larger network radius (t = -2.647, p = 0.009) and network diameter (t = -2.619, p = 0.010), as well as significantly lower small-worldness (t = 2.066, p = 0.041) and global efficiency (t = 2.083, p = 0.040) in brain white matter structural network. Local topological abnormalities were observed in multiple brain regions of adolescents with MDD, primarily involving paracentral lobular and mid cingulate cortex, MT+ complex, early auditory cortex, dorsolateral prefrontal cortex, orbital and polar frontal cortex, as well as anterior cingulate and medial prefrontal cortex. The final regression model showed that the weighted betweenness centrality in left area 1, left area 6mp, left frontal opercular area 4, left posterior OFC complex and left lateral belt complex, as well as the weighted local efficiency of the right ventral intraparietal complex and right orbital frontal complex (adjusted R2 = 0.270, Cohen’s f2 = 0.630).

We found that chronic childhood stress is associated with local topological abnormalities in multiple regions of the white matter structural networks of adolescents with MDD. These altered white matter network topological properties may be utilized in the future to aid in identifying depressive subtypes with high childhood stress exposure.