Skip to main content
NCBI home page
JAMA Network logoLink to JAMA Network
. 2026 Jul 6;9(7):e2621568. doi: 10.1001/jamanetworkopen.2026.21568

Aerobic Exercise and Subthreshold Depressive Symptoms in Adolescents

Secondary Analysis of a Randomized Clinical Trial

Xiangbo Yan 1,2, Robin Shao 3, Jingwen Jin 3, Weicong Lu 4,5, Zhongqing Jiang 6, Guanghui Zhang 7,8, Qing Zhang 1,✉, Xiaoyue Li 5, Xinhe Tian 5, Brendon Stubbs 9, Davy Vancampfort 10,11, Marco Solmi 12,13,14,15, Michele Fornaro 16, Andre F Carvalho 17, Sonata Suk-yu Yau 18,19, Fengyu Cong 2,20, Kwok-fai So 21, Kangguang Lin 22,✉
PMCID: PMC13338804  PMID: 42406403

This secondary analysis of a randomized clinical trial evaluates associations between aerobic exercise and reductions in subthreshold depressive symptoms in adolescents and explores potential neural mechanisms underpinning symptom improvement.

Key Points

Question

Is aerobic exercise associated with reduced subthreshold depression in adolescents, and what are the potential underlying neural mechanisms?

Findings

In this secondary analysis of a randomized clinical trial including 206 adolescents, aerobic exercise was associated with reduced subthreshold depressive symptoms compared with the psychoeducation control. This outcome was mediated by 2 opposing electroencephalography (EEG)-based neural network pathways respectively representing processes ameliorating and exacerbating subthreshold depressive symptoms.

Meaning

The findings suggest aerobic exercise may be an effective, scalable, accessible intervention to alleviate adolescent subthreshold depression, with the identified EEG-based neural network pathways providing potential biomarkers for tailoring personalized treatment and monitoring responses across individuals.

Abstract

Importance

Adolescence is a high-risk period for subthreshold depression and a critical window for intervention. While aerobic exercise has shown efficacy in alleviating adult depressive symptoms, its efficacy and neural mechanisms in adolescents remain unclear.

Objective

To evaluate the association of aerobic exercise with reduced subthreshold depressive symptoms in adolescents and to identify potential neural mechanisms underpinning symptom improvement.

Design, Setting, and Participants

This was a prespecified secondary analysis of a 12-month, multicenter, cluster randomized clinical trial (RCT) conducted October 2021 to October 2022 among students aged 12 to 17 years in China. Data analysis took place between April and August 2024.

Interventions

The 12-month aerobic exercise intervention consisted of a 6-month supervised phase followed by a 6-month unsupervised phase. The control group received 6 psychoeducation sessions (once every 2 months) focusing on mood regulation, depression awareness, and stress management.

Main Outcomes and Measures

At baseline and postintervention, participants reported depressive symptoms (Patient Health Questionnaire-9 [PHQ-9]) and underwent electroencephalography (EEG) recording. Group differences in symptom change were assessed using linear mixed models. Resting-state EEG data were analyzed for functional connectivity and both global and nodal network topology. Mediation analyses were conducted to test the neural processes mediating the exercise effect.

Results

A total of 206 adolescents aged 12 to 17 years (110 males [53%]; median age, 13 years) with subthreshold depressive symptoms were included (111 in the exercise group and 95 in the psychoeducation group). The exercise group showed significantly greater reductions in subthreshold depressive symptoms than the control group (exercise: mean [SE] PHQ-9 score, 9.20 [0.05] points at baseline vs 7.33 [0.03] postintervention; t110, −4.49; P < .001; control: 7.74 [0.04] points at baseline vs 7.28 [0.04] postintervention; t94, −1.03; P = .30). The functional connectivity across 355 alpha-band, 127 beta-band, 30 theta-band, and 9 delta-band connections was reduced (all P < .05 after correction for false discovery rate), and global network efficiency, such as the clustering coefficient, was enhanced across all frequency bands after the exercise intervention (all P < .001 after correction for false discovery rate). Mediation analysis revealed 2 opposing neural pathways that represented depression-reduction (B, –2.28; P < .001) and depression-increase (B, 3.26; P < .001) processes.

Conclusions and Relevance

In this secondary analysis of an RCT, the findings suggest that exercise reduced subthreshold depressive symptoms in adolescents by engaging opposing EEG-based neural network processes. These findings support exercise as a potential accessible intervention for adolescent subthreshold depression, highlighting neurophysiologic signatures that may inform individualized intervention strategies and outcome monitoring that need to be replicated in adolescents with clinical depression.

Trial Registration

ClinicalTrials.gov Identifier: NCT04816617

Introduction

Subthreshold depression is a prominent mental health concern, with approximately 34% of adolescents experiencing elevated depressive symptoms.1 If unaddressed, such symptoms are associated with a 2-to-3-fold increased risk of developing major depressive disorder (MDD) in adulthood,2 causing substantial academic and occupational impairments as well as poor clinical prognosis.3,4,5 Currently recommended treatments for adolescent depression include psychotherapy and antidepressants.6 However, psychotherapy yields remission in only 50% of cases,7 while antidepressants are more effective in adults than in adolescents and are associated with adverse effects, high relapse risks, and withdrawal symptoms.8,9,10 Thus, there is an urgent need for developing and testing a scalable, accessible, and cost-effective intervention method suitable for alleviating subthreshold depression among adolescents.

Existing evidence indicates therapeutic efficacy of exercise in treating MDD and depressive symptoms in adults,11,12,13 supporting its viability as a standalone intervention. However, the evidence in adolescents remains limited and inconsistent.12,13 Two reasons may explain the discrepancy in past findings. First, the intervention dosage for short-term exercise (≤2 months) may not have been sufficient to achieve the antidepressant effect. Second, there is a lack of high-quality randomized clinical trials (RCTs) specifically targeting adolescents with subthreshold depressive symptoms.

Furthermore, the neural correlates of the exercise effect in adolescents remain poorly understood. Past animal studies revealed putative molecular pathways that may underpin the antidepressant effects of exercise. For example, glutamatergic neurons in the anterior cingulate cortex (ACC) mediated the rapid antidepressant response elicited by a single exercise session in mice.14 Also, adiponectin-induced translocation of the APPL1 protein within the ACC in mice might be a key mechanism driving the rapid exercise effects.15 However, evidence on the neural processes linked to the antidepressant effect of exercise in humans is still lacking, partly due to the absence of large-scale RCTs in adolescents, which constitutes a major research gap.

Contemporary theories conceptualize depression as a system-level disorder involving neural network synchronizations.16,17 Emerging evidence suggests that alterations in brain connectivity may be critical treatment-responsive neural signatures.18,19 Therefore, investigating exercise-induced neural circuitry plasticity in adolescents holds significant potential for elucidating the neural markers and mechanisms underlying exercise’s antidepressant effects. Electroencephalography (EEG) serves as a pivotal tool for investigating the neurobiological mechanisms of depression.20 EEG measures the interregional synchronization of oscillatory brain activity21,22 that is critical for cross-network communication.23,24 Functional connection (FC) analysis in EEG, particularly in the alpha and beta bands, demonstrates high capability in predicting clinical phenotypes based on neural network connectivity profiles.25,26 Among patients with MDD, electroconvulsive and antidepressant therapy induced alterations in connectivity across the EEG alpha,27,28 beta, theta,29 and delta bands.29 However, it remained unclear whether EEG-based neural network changes mediated the antidepressant effect of exercise in adolescents.

Therefore, this secondary analysis of data from a large-scale, multicenter, 12-month RCT asked 2 questions: first, whether aerobic exercise was associated with reduced subthreshold depressive symptoms in adolescents and second, whether and how the exercise effects were mediated by EEG-based neural network features. We hypothesized that the exercise intervention would be associated with significantly reduced subthreshold depressive symptoms in adolescents compared with the psychoeducation control and that this reduction would be mediated by changes in EEG-derived functional connectivity and network topological features.

Methods

Study Design and Participants

This study was a prespecified secondary analysis of a 12-month, 2-arm cluster RCT (conducted October 2021 to October 2022) that evaluated the effects of moderate-intensity aerobic exercise on depression and cognition in adolescents (NCT04816617). Data analysis in the current study took place between April and August 2024. The protocol for the original trial received ethical approval from the institutional review board of the Affiliated Brain Hospital, Guangzhou Medical University, and complied with the Declaration of Helsinki30 (the research protocol is in Supplement 1). The intervention was implemented in a middle school in Huaiji, Guangdong Province, China, with written informed consent obtained from parents or guardians and assent from participants. This secondary analysis used existing deidentified data from the original RCT and was conducted under the original ethics approval granted by the institutional review board of the Affiliated Brain Hospital, Guangzhou Medical University. All participants provided written informed consent for the original trial; no additional consent was required for this secondary analysis. The current study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline.

The trial enrolled students aged 12 to 17 years from 14 classrooms. Participants were excluded if they had major past or present psychiatric illnesses or physical conditions that prevented them from engaging in regular aerobic exercises (eMethods in Supplement 2).

For the present analysis, participants with subthreshold depressive symptoms (Patient Health Questionnaire–9 [PHQ-9]31 scores ≥5 on a scale from 0-27, with higher scores indicating more severe depressive symptoms) and valid EEG data were included (Figure 1). Both subthreshold depressive symptoms and EEG features were prespecified secondary outcomes. The principal investigator (K.L.) generated the randomization sequence using IBM SPSS, version 20.0 (IBM Corp), with classrooms (clusters) assigned to either the aerobic exercise intervention or the psychoeducation control group (eMethods in Supplement 2).

Figure 1. CONSORT Diagram of Participant Flow Through Trial.

Flowchart of trial participant screening, randomization, follow-up, and analysis counts. Light blue and gray CONSORT-style flow diagram with rectangular nodes connected by gray arrows. Top center rectangle: 691 Individuals from 14 classes screened for eligibility. A rightward arrow leads to an upper right rectangle: 61 Excluded; 25 Declined to participate; 23 Had current major depressive disorder; 12 Had past major depressive disorder; 1 Had bipolar disorder. A downward arrow from the top node leads to a central oval: 630 Randomized. Two diagonal arrows branch downward. Left branch to a mid-left rectangle: 352 From 8 classes randomized to supervised aerobic exercise intervention. Downward arrow to a left rectangle listing assessments: 352 Underwent baseline P H Q 9 measurement; 155 Underwent baseline E E G measurement. A small left-side rectangle connected near this segment reads 1 Dropped out. Downward arrow to a left rectangle: 351 Received exercise intervention. Downward arrow to a left rectangle listing follow-up assessments: 351 Underwent P H Q 9 measurement at 12 mo postintervention; 137 Underwent E E G measurement at 12 mo postintervention. A left-side rectangle connected near this segment reads 26 Excluded; 20 P H Q 9 score less than 5 at baseline; 6 E E G impedance greater than 50 k ohms. Downward arrow to a lower-left rectangle: 111 Had baseline P H Q 9 score greater than or equal to 5 and completed both E E G assessments. Right branch to a mid-right rectangle: 278 From 6 classes randomized to psychoeducation intervention. Downward arrow to a right rectangle listing assessments: 278 Underwent baseline P H Q 9 measurement; 142 Underwent baseline E E G measurement. A small right-side rectangle connected near this segment reads 1 Dropped out. Downward arrow to a right rectangle: 277 Received psychoeducation intervention. Downward arrow to a right rectangle listing follow-up assessments: 277 Underwent P H Q 9 measurement at 12 mo postintervention; 120 Underwent E E G measurement at 12 mo postintervention. A right-side rectangle connected near this segment reads 25 Excluded; 21 P H Q 9 score less than 5 at baseline; 4 E E G impedance greater than 50 k ohms. Downward arrow to a lower-right rectangle: 95 Had baseline P H Q 9 score greater than or equal to 5 and completed both E E G assessments. Both lower rectangles connect by lines to a bottom center rectangle: 206 Included in mediation analysis; 111 Exercise intervention group; 95 Psychoeducation control group.

Open in a new tab

CONSORT indicates Consolidated Standards of Reporting Trials; EEG, electroencephalography; PHQ-9, Patient Health Questionnaire–9.

Intervention

Participants in the aerobic exercise group engaged in a 12-month program, comprising an initial 6-month supervised phase followed by a 6-month unsupervised maintenance phase. Supervised sessions led by trained school physical education teachers involved moderate-to-vigorous activities such as running and rope skipping, which lasted 30 minutes per session, for 3 or 4 sessions per week. During the unsupervised phase, participants were encouraged to maintain the frequency and intensity in their own preferred exercise, with weekly reminders to enhance adherence.

The control group received 6 psychoeducation sessions over 12 months (1 every 2 months) covering topics such as mood regulation, depression awareness, and stress management. The sessions were delivered by psychiatrists or psychologists in classrooms at the participating schools. Details about the intervention protocol, adherence, and monitoring procedures are included in the eMethods in Supplement 2.

Assessments

At baseline, all participants completed both self- and clinician-rated assessments prior to EEG recording. Preintervention physical activity was measured using the International Physical Activity Questionnaire–short form.32 The PHQ-9—which includes 9 core symptoms of depression, such as low mood and reduced energy—assessed participants’ depressive symptoms both before and after the intervention.

Resting-State EEG Measurements

EEG data were recorded using a 128-channel high-density EEG system with the HydroCel Geodesic Sensor Net (Electrical Geodesics, Inc).33 Electrode placement followed the international 10-10 system for spatial localization. The FC analysis was performed using the Brainstorm toolbox.34 The weighted phase lag index (wPLI) was used to estimate FCs in the brain network analyses.35 The FC matrix is output in 4 frequency bands of interest: alpha (8-13 Hz), beta (>13 to 30 Hz), theta (4 to <8 Hz), and delta (2 to <4 Hz). Graph theory analysis was performed using the GRETNA toolbox36 in MATLAB (MathWorks). The 68 × 68 wPLI matrices used a binary threshold for calculating global and nodal network properties. Global metrics included global efficiency, local efficiency, clustering coefficients, characteristic path length, and small-worldness.37 Nodal metrics included betweenness centrality, degree centrality, nodal clustering coefficients, nodal efficiency, and nodal local efficiency. These EEG features are fundamental to characterize brain network connectivity and topological features, which have been widely used to study psychiatric conditions.38,39,40 Further details about EEG data processing and analysis are included in the eMethods in Supplement 2.

Statistical Analysis

Intervention effects were evaluated using linear mixed-effects models (LMMs) on participants’ PHQ-9 scores and EEG metrics (both were secondary outcomes of the original trial), which were entered as dependent variables. Fixed effects included group (exercise = 1, control = 0), time (baseline = 0, postintervention = 1), and their interaction, while individual participants and class were specified as random effects on intercept. For the analysis of the association of exercise with subthreshold depressive symptoms, a random effect of time on slope was also added to allow individual difference in PHQ-9 change pattern, and the analysis was repeated after controlling for participants’ baseline PHQ-9 scores and sex. Other variables, including age and baseline activity, were not included as covariates, since they showed identical mean values or group proportions for individuals receiving the exercise or psychoeducation intervention. For outcomes showing significant group × time interactions, post hoc paired-sample t tests were performed. The analyses were performed using the lme4 package in R, version 1.1.37, and lmerTest package in R, version 3.1.3 (R Project for Statistical Computing). EEG features comprised FCs (wPLI) and global and nodal network topological variables (eTable 1 in Supplement 2).

To test whether neural variables mediated the association between exercise and subthreshold depressive symptoms, we implemented a 3-step analytical framework. The first step was mediation analysis with intervention group (0 = control, 1 = exercise) as the binary independent variable, changes in EEG metrics (functional connectivity and network properties) as mediators, and PHQ-9 score change as the dependent variable (changes in EEG metrics and PHQ-9 scores were calculated as postintervention minus baseline values), using false discovery rate (FDR) methods to conduct multiple-testing correction. In the second step, among the significant mediating features from step 1, we selected those that mediated reduction of subthreshold depressive symptoms following the exercise vs psychoeducation intervention, which were simultaneously entered into a composite depression-reduction mediation model, controlling for baseline PHQ-9 scores. As the third step, we performed a similar procedure to select neural features that mediated increases in subthreshold depressive symptoms and included them together in a composite depression-increase mediation feature set, which was added to the depression-reduction mediation model from step 2 to form a combined total model including both opposing pathways. The neural features tested in the mediation analysis were those that showed a significant exercise (vs psychoeducation) intervention effect, since we primarily aimed to identify neural features that statistically accounted for the exercise effect on subthreshold depressive symptoms.41 As such, the mediation analysis results should be considered as being post hoc in nature. The same models were repeated after controlling for sex as an additional covariate. The statistical significance threshold was FDR-corrected, 2-sided P < .05, with 95% CIs generated by bootstrapping (1000 times). The analysis was conducted using the lavaan and semTools packages in R, version 4.1.3.

Results

Demographic and Clinical Data

A total of 206 participants aged 12 to 17 years with subthreshold depression (PHQ-9 score ≥5) who completed EEG assessments were included in this secondary analysis (Table). Two individuals participated in the baseline measurement (1 allocated to the exercise group, 1 allocated to the psychoeducation group) but dropped out of the original trial before starting the intervention; their data were excluded from all analyses. The final sample for this study comprised 96 females (47%) and 110 males (53%) with median age of 13 years; the exercise group had 111 participants (46 females [41%] and 65 males [59%]; median age, 13 years) and the control group had 95 participants (50 females [53%] and 45 males [47%]; median age, 13 years). There was no significant difference between the exercise and control groups in age, sex, baseline depressive symptoms, or baseline physical activity.

Table. Baseline Participant Characteristics.

Characteristic Participants (N = 206)
Exercise group (n = 111) Control group (n = 95)
Sex, No. (%)
Female 46 (41) 50 (53)
Male 65 (59) 45 (47)
Age, median (range), y 13 (12-17) 13 (12-17)
PHQ-9 score, mean (SE)a 9.20 (0.05) 7.74 (0.04)
Activity level, No. (%)b
Medium-low 51 (46) 44 (46)
High 60 (54) 51 (54)
Open in a new tab

Abbreviation: PHQ-9, Patient Health Questionnaire–9.

a

Total score range is 0 to 27, with higher scores indicating more depressive symptoms.

b

High activity was considered vigorous-intensity activity at least 3 days a week, achieving minimum total physical activity of at least 1500 metabolic equivalent of task (MET) min/wk, or 7 days of walking, moderate-intensity, or vigorous-intensity activities, achieving at least 3000 MET min/wk. Moderate-low activity was considered not meeting the criteria for high activity level.

Interventions and Subthreshold Depressive Symptom Outcomes

Among the 206 participants, LMM analysis revealed a significant group × time interaction for PHQ-9 scores (β, −1.38; SE, 0.62; t204.00, −2.21; P = .03). The between-class variance component and the intraclass correlation were 0.48 and 0.05, respectively. In the exercise group, mean (SE) PHQ-9 scores were significantly lower at postintervention (7.33 [0.03] points) compared with baseline (9.20 [0.05] points) (t110, −4.49; P < .001). In contrast, the control group showed no significant mean (SE) score change between baseline (7.74 [0.04] points) and postintervention (7.28 [0.04] points) (t94, −1.03; P = .30) (Figure 2). The same results were obtained after additionally controlling for participants’ baseline PHQ-9 scores and sex (β, −1.38; SE, 0.44; t403.20, −3.12; P = .002).

Figure 2. Associations of Exercise vs Psychoeducation With Depressive Symptom Scores.

Two-panel paired dot plots of P H Q 9 at baseline and postintervention. Two side-by-side panels labeled A and B. Panel A title at upper left: Control group. Panel B title at upper left: Exercise group. Both panels share the same vertical axis label, P H Q 9 score, with tick marks from 0 to 30 in steps of 5 and light horizontal gridlines. In each panel, the horizontal axis has two categories: Baseline at left and Postintervention at right. Dark teal circular markers form two vertical columns per panel, one column at Baseline and one at Postintervention. Numerous light blue line segments connect each participant’s Baseline dot to the corresponding Postintervention dot, creating a paired before-after display. In panel A, the Baseline dots cluster mostly between about 5 and 17, with one dot near about 22. Postintervention dots cluster mostly between about 2 and 16, with two high dots near about 25 and 27. A black dashed horizontal mean line with a short solid black segment and small vertical error bars appears near about 7 to 8 at Baseline and near about 7 to 8 at Postintervention. Above the two time points, a thin gray bracket spans Baseline to Postintervention with centered text P greater than or equal to .05. In panel B, Baseline dots span roughly 5 to 27 with several dots between about 12 and 18 and two high dots near about 26 to 27. Postintervention dots span roughly 0 to 26, with many dots between about 5 and 14, several low dots near 0 to 3, and one high dot near about 26. A black dashed mean line with a short solid black segment and small vertical error bars appears near about 9 to 10 at Baseline and near about 7 to 8 at Postintervention. Above the two time points, a thin gray bracket spans Baseline to Postintervention with centered text P less than .001.

Open in a new tab

PHQ-9 indicates Patient Health Questionnaire–9 (score range, 0-27, with higher scores indicating more severe depressive symptoms).

Interventions and EEG Data Outcomes

FCs

FC analyses were conducted using LMMs with FDR correction. A total of 2278 FCs spanning 68 brain regions were examined. Significant group × time interactions were predominantly observed in the alpha band (355 FCs), with fewer interactions in the beta (127 FCs), theta (30 FCs), and delta (9 FCs) bands (eFigure 1 in Supplement 2).

In the alpha band, 322 FCs (91%) showed a significant reduction in wPLI values in the exercise group at postintervention compared with baseline (all FDR-corrected P < .05), whereas the control group exhibited no significant changes. Among the 33 remaining alpha band FCs with significant interaction effects (9%), 4 (1%) decreased in the exercise group but increased in the control group, 3 (1%) increased in the control group only (all FDR-corrected P < .05), and 26 (7%) showed no significant change in either group (eFigure 1 in Supplement 2).

In the beta band, 60 FCs (47%) significantly increased in the control group at postintervention compared with baseline (all FDR-corrected P < .05), while 67 FCs (53%) increased in both groups (all FDR-corrected P < .05) (eFigure 1 in Supplement 2). In the delta and theta bands, all FCs in the exercise group decreased significantly after intervention, whereas no changes were observed in controls (all FDR-corrected P < .05) (eFigure 1 in Supplement 2).

Global Network Topological Properties

LMMs revealed significant group × time interactions for clustering coefficients, global efficiency, local efficiency, and characteristic path length across 4 frequency bands (all FDR-corrected P < .001) (eFigure 2 in Supplement 2). A significant group × time interaction for small-worldness was observed only in the theta band. In the exercise group, clustering coefficients, global efficiency, and local efficiency significantly increased, whereas characteristic path length significantly decreased, at postintervention compared with baseline in all frequency bands. By contrast, no significant changes were observed in the control group (eFigure 2 in Supplement 2).

Local Network Topological Properties

Across all 4 frequency bands, the LMM analysis revealed significant group × time interactions; these were predominantly in the alpha band (eFigure 3 in Supplement 2), with fewer interactions in the delta, theta, and beta bands (eFigure 4 in Supplement 2). After the intervention, the exercise group showed both increases and decreases in alpha band nodal attributes. The increases were observed in degree centrality (6 nodal regions), nodal clustering coefficients (26 regions), nodal efficiency (17 regions), and nodal local efficiency (27 regions). The decreases occurred in betweenness centrality (2 nodal regions), degree centrality (1 region), nodal efficiency (2 regions), nodal clustering coefficients (11 regions), and nodal local efficiency (12 regions) (all FDR-corrected P < .05). By contrast, no significant nodal changes were observed in the control group (eFigure 3 in Supplement 2).

EEG Features Correlated With Reduction of Subthreshold Depressive Symptoms

The eResults and eFigures 5 and 6 in Supplement 2 show the correlations between changes in PHQ-9 scores and in EEG features following the exercise intervention. For example, following exercise, the changes in FC between the right insula (region 19) and left pericalcarine sulcus (region 44) in the beta frequency band was significantly and negatively correlated with the changes in participants’ PHQ-9 scores (r = −0.35; P < .001).

Post Hoc Mediation Analysis

Mediation analysis was performed with group (0 = control, 1 = exercise) as the independent variable, changes in FCs and network topology as mediators, and PHQ-9 change as the outcome. In the first depression-reduction model, 18 neural features (neural factor 1) significantly mediated the indirect association between the exercise intervention and reduced subthreshold depressive symptoms (total effect size: B, −1.36 [P = .02]; indirect effect size: B, −2.28 [P < .001]; direct effect size: B, 0.92 [P = .07]) (Figure 3; full names of the brain regions are presented in eTable 2 in Supplement 2).

Figure 3. Mediation Model of Exercise-Induced Reductions in Depressive Symptoms via Electroencephalography (EEG)-Based Neurophysiologic Mechanisms.

Flow diagram of exercise, neural factor one, and change in P H Q 9 score. Title at top: Mediation of exercise-induced reductions in depressive symptoms. Left side: a light blue rectangle labeled Exercise. From Exercise, one arrow extends downward to a light blue rectangle labeled delta P H Q 9 score, with text beside the arrow reading beta, zero point one five with superscript a. A second connector from Exercise extends rightward and then down into a light blue circle labeled Neural factor one; text near this connector reads beta, minus zero point six three with superscript c. From the Neural factor one circle, an arrow extends down and then left into the delta P H Q 9 score rectangle; text near this arrow reads beta, zero point five six with superscript c. Right side: a header reading Brain regions showing changes in E E G features, with a smaller header beta aligned to the left of the list. A vertical bracket-like line connects Neural factor one to multiple right-pointing arrows, each arrow aligned with a beta value on the left and a light blue labeled box on the right. From top to bottom, the beta values and corresponding boxes read: zero point three seven superscript b to F C Alpha 51 and 8; zero point four five superscript c to F C Alpha 52 and 27; zero point three nine superscript c to F C Alpha 60 and 27; zero point four zero superscript c to F C Alpha 52 and 43; zero point three one superscript c to F C Theta 42 and 35; zero point two two superscript a to F C Delta 57 and 8; zero point four nine superscript c to F C Beta 62 and 8; zero point five six superscript c to F C Beta 68 and 8; zero point five six superscript c to F C Beta 59 and 14; zero point five zero superscript c to F C Beta 59 and 23; zero point one nine superscript a to F C Beta 32 and 28; zero point two eight superscript b to F C Beta 43 and 36; zero point six zero superscript c to F C Beta 68 and 59; minus zero point three four superscript c to N C C, alpha 30; minus zero point five three superscript c to N C C, alpha 40; zero point one nine superscript a to N C C, alpha 49; minus zero point three six superscript c to N L E, alpha 30; minus zero point five five superscript c to N L E, alpha 40.

Open in a new tab

Δ indicates change; FC, functional connection; NCC, nodal clustering coefficient; NLE, nodal local efficiency; PHQ-9, Patient Health Questionnaire–9. eTable 2 in Supplement 2 includes full names of the brain regions.

aP < .05.

bP < .01.

cP < .001.

We then added 26 EEG features (neural factor 2 [depression-increase feature set]), which mediated the indirect association between exercise and increased depressive symptoms in the first model, resulting in a combined total model (total effect size: B, −1.35 [P = .02]; direct effect size: B, −2.20 [P = .004]). Within this model, the 18 depression-reduction features significantly mediated the indirect association between exercise and decreased subthreshold depressive symptoms (B, −2.41 [P < .001]), while the 26 depression-increase features significantly mediated the indirect association between exercise and exacerbated depressive symptoms (B, 3.26 [P < .001]) (Figure 4). Pearson correlations between changes in the significant alpha band local topological measures and changes in PHQ-9 score after intervention compared with baseline are depicted in eFigure 7 in Supplement 2. These results highlight 2 opposing EEG pathways that mediated the association between exercise and subthreshold depressive symptoms. The mediation results were fully replicated after controlling for sex (eResults in Supplement 2).

Figure 4. Mediation Model of Exercise-Induced Bidirectional Modulations of Depressive Symptoms via Opposing Electroencephalography (EEG)-Based Neurophysiologic Mechanisms.

Multi-panel mediation schematic with brain maps, E E G links, and beta coefficients. Title at top: Mediation of exercise-induced bidirectional modulations of depressive symptoms. Upper row contains four translucent dorsal brain renderings labeled, left to right, Alpha N C C, Alpha N L E, Theta, and Delta. In Alpha N C C, blue circular nodes labeled 30, 40, and 49 and one red node labeled 38 appear on the cortical surface. In Alpha N L E, blue nodes labeled 30 and 40 appear. In Theta, two blue nodes labeled 35 and 42 are connected by a single green line. In Delta, two blue nodes labeled 57 and 8 are connected by a single green line. Center of the figure contains a mediation diagram: a rectangle labeled Exercise at upper center; two circles labeled Neural factor 1 on the left and Neural factor 2 on the right; and a rectangle labeled delta P H Q hyphen 9 score at lower center. A vertical arrow from Exercise to delta P H Q hyphen 9 score is labeled beta, minus zero point two seven with superscript b. A leftward arrow from Exercise to Neural factor 1 is labeled beta, minus zero point six four with superscript c; an arrow from Neural factor 1 to delta P H Q hyphen 9 score is labeled beta, zero point five four with superscript c. A rightward arrow from Exercise to Neural factor 2 is labeled beta, minus zero point seven five with superscript b; an arrow from Neural factor 2 to delta P H Q hyphen 9 score is labeled beta, minus zero point six nine with superscript c. Left side includes a column header Brain regions showing changes in E E G features and a beta column, followed by stacked boxes listing connections: F C Alpha 51 and 8; 52 and 27; 60 and 27; 52 and 43; F C Beta 62 and 8; 68 and 8; 59 and 14; 59 and 23; 32 and 28; 43 and 36; 68 and 59; F C Theta 42 and 35; F C Delta 57 and 8; N C C, alpha 30; N C C, alpha 40; N C C, alpha 49; N L E, alpha 30; N L E, alpha 40, each with a rightward arrow and a beta value. Right side contains a taller list under beta and Brain regions showing changes in E E G features, with many boxes labeled F C Alpha followed by paired numbers and a few F C Beta entries, each aligned to a beta value. Lower row contains two lateral brain network renderings labeled Alpha and Beta. The Alpha panel contains many numbered blue nodes connected by numerous red and green lines; the Beta panel contains fewer nodes with several green lines and a few red lines.

Open in a new tab

Red nodes and functional connections (FCs) indicate prodepressive effects and blue nodes and green FCs indicate effects linked to symptom reduction. Δ indicates change; NCC, nodal clustering coefficient; NLE, nodal local efficiency; PHQ-9, Patient Health Questionnaire–9. eTable 2 in Supplement 2 includes full names of the brain regions.

aP < .05.

bP < .01.

cP < .001.

Discussion

This secondary analysis of an RCT revealed that a 12-month exercise intervention was associated with significantly reduced subthreshold depressive symptoms among adolescents. Compared with psychoeducation, exercise was associated with decreased EEG-based network FCs across 4 frequency bands and changes in both global and local network topology, particularly in the alpha band. Mediation analysis revealed EEG features associated with both depression-reduction and depression-increase effects, which uncovered novel neurophysiologic pathways and candidate intermediate markers of intervention response across individuals.

At the neurophysiologic level, exercise was associated with decreases in FCs within the alpha and beta bands, a pattern similar to the outcomes of electroconvulsive therapy in patients with MDD.29,42 Alpha band FC reductions, particularly in the frontal regions,43 may mitigate maladaptive rumination44 and reduce depressive symptoms, while theta and delta band changes may support exercise as an effective intervention for subthreshold depressive symptoms that induces beneficial neural synchronization changes across wider frequency bands. Exercise was also associated with enhanced global network properties such as clustering coefficients, global efficiency, and local efficiency while reducing characteristic path length, suggesting improved information integration and preserved local specialization.45,46 The increased local and global network efficiency we found after the exercise intervention may normalize the reduced brain network efficiency previously observed in patients with MDD.47,48 The consistent pattern of changes found in the global network attributes across the 4 frequency bands suggests that exercise-induced neural network remodeling may alleviate depressive symptoms through remedying distributed brain network deficiencies rather than affecting individual brain regions.49,50 On the other hand, nodal-level changes, particularly in clustering coefficients and local efficiency, emerged primarily in the alpha band, indicating candidate neural markers for intervention response.39

Clinical and psychobiological models of depression consider dysregulated emotions as their core feature, which is accompanied by other sensorimotor, physiological (eg, sleep), and cognitive abnormalities.51 The cognitive control model of depression views emotion dysregulation as being consequent to deficient top-down cognitive regulation of emotions,52 which is a popular target for antidepressant interventions.53 The dysregulated emotion system, in turn, has overarching associations with perceptual, cognitive, and motor functions that are biased toward oversensitivity to negative processes but insensitivity to hedonic processes.54

Following these conceptual frameworks, amelioration of subthreshold depressive symptoms after the exercise intervention is likely to have been associated with normalizing the core brain emotion-processing and regulation systems. Consistent with this, the antidepressive mediating pathway involved exercise-induced reductions in functional synchronization within the default mode network including the precuneus, parahippocampal gyrus, and middle temporal cortex,55,56,57 which is the central network implicated in self-related affective processing that has consistently shown hyperactivity in depression.58 In addition, the exercise intervention was associated with increased network connectivity and efficiency of the medial orbitofrontal cortex, which is a key structure involved in reward processing and positive emotions.53 Collectively, these neural network changes following the exercise intervention may exert therapeutic effects through dampening the negatively biased, self-related emotion processes while enhancing the reward-related positive emotion functions, thereby ameliorating the core emotion symptoms of depression. Other exercise-induced neural changes included decrease in functional synchronization in the visual and auditory sensory systems, which might indicate reduced hypersensitivity to negative sensory information in concordance with the general reduction of negative emotion function.59 In addition, exercise was associated with decreased functional connectivity between the superior frontal cortex and cuneus in the delta band and between the inferior frontal gyrus and parahippocampal gyrus in the theta band, which might result in improvement of sleep symptoms of depression.60,61 Sleep disturbances are both potent risk factors for and consequences of depressive symptoms, the amelioration of which is pertinent for recovery from depression.62

On the other hand, the prodepressive pathway of the exercise intervention was expected to involve neural changes that are opposite to the normalized patterns of emotion processing and cognitive regulation. Specifically, exercise was associated with decreased functional connectivity in an extended central executive network important for top-down cognitive control of emotions (eg, reappraisal), including the dorsal and medial prefrontal cortex, left ventrolateral prefrontal cortex, caudal middle frontal cortex, and superior parietal cortex.63,64 Reduced connectivity of this network with the default mode network (eg, the inferior parietal cortex) and with the salience network (eg, insula) involved in selective orienting and vigilance to emotional stimuli could be associated with the heightened and dysregulated processing of both internal and external negative stimuli.65 Besides the core emotion functions, the prodepressive pathway also included reduced functional synchronization among the multimodal sensory networks encompassing the postcentral somatosensory, visual, and auditory sensory cortices, which might indicate the reduced sensory seeking and sensory sensitivity that are associated with anhedonia and decreased motivation in depression.66 Collectively, aerobic exercise may result in decreased functions of the central executive network and top-down emotion regulation, along with dampened perceptual processing across sensory modalities, which are associated with depressive symptom increase.67,68

Limitations

Several limitations of the current study need to be considered with regard to future directions in validating and extending the opposing-pathway model of exercise. First, the current data were obtained from a single school in a Chinese city. The findings need to be externally validated using data collected from an independent site. Second, the present study design did not allow for making causal inferences about the association between exercise-induced effects on the neural pathways and depressive symptoms. Future studies need to include at least 3 time points to test whether neural changes at earlier time points (eg, immediately postintervention) prospectively mediate symptom change at later time points (eg, 3 months after intervention). Third, an important clinical implication of the opposing-pathway model is that individuals who show more favorable changes in the antidepressive pathway are likely to display greater reduction of depressive symptoms, while those showing greater changes in the prodepressive pathway are expected to display less reduction (or even increase) of depressive symptoms. Testing these hypotheses would provide further validation of the opposing-pathway model. Fourth, the activity and connectivity levels of the implicated neural networks of the antidepressive and prodepressive pathways at baseline may modulate the exercise intervention’s association with those neural systems and depressive symptoms in individuals. Future research may examine the functional patterns of the emotion and cognitive control systems during the resting state and task performance prior to the intervention and test whether these baseline neural characteristics can prospectively predict exercise effects across individuals. Fifth, the present study used a uniform exercise dosage that may not be optimized for all participants. Future research may seek individualized intervention protocols according to each individual’s prior exercise history. Another important point is that our findings were obtained from adolescents with subthreshold depression. Future replications among adolescents with MDD are necessary to establish the findings’ generalizability.

Conclusions

This secondary analysis of an RCT showed that aerobic exercise was associated with reduced subthreshold depressive symptoms in adolescents, mediated by EEG-derived neural processes. We uncovered both antidepressive and prodepressive pathways predominantly involving alpha band functional connectivity and nodal topology, which may have accounted for the bidirectional effects of exercise. These findings revealed potential mechanistic biomarkers that could inform scalable and personalized exercise-based interventions for adolescent subthreshold depression. Future intervention studies involving follow-up time points can test whether functional patterns of the key emotion and cognitive control networks at baseline and their changes immediately following the intervention prospectively predict the long-term effect of exercise in reducing depressive symptoms at the individual level. Such findings, if replicated in individuals with MDD, would inform individualized exercise interventions for depression based on each person’s preexisting neural patterns and early-phase intervention response profile.

Supplement 1.

Trial Protocol

jamanetwopen-e2621568-s001.pdf (426.3KB, pdf)
Supplement 2.

eMethods.

eResults.

eFigure 1. Change of FCs in Alpha, Beta, Delta, and Theta Frequency Bands After Exercise Intervention vs Control Intervention

eFigure 2. Change of Global Network Topology Properties in Alpha, Beta, Delta, and Theta Frequency Bands After Exercise Intervention vs Control Intervention

eFigure 3. Change of Node Network Topology Properties in Alpha Frequency Band After Exercise Intervention vs Control Intervention

eFigure 4. Change of Node Network Topology Properties in Beta, Delta, and Theta Frequency Bands After Exercise Intervention vs Control Intervention

eFigure 5. Functional Connectivity Related to Reduction of Depressive Symptoms

eFigure 6. Network Topological Properties Related to Reduction of Depressive Symptoms

eFigure 7. Relationship Between Changes in PHQ-9 Scores and Changes in Nodal Network Topological Properties That Mediate the Effect of Exercise on Depressive Symptoms

eTable 1. Definition and Functional Significance of the Global and Nodal Topology Metrics

eTable 2. Index, Full Name, and Abbreviated Name of Regions Included in the Desikan-Killiany Cortical Partition Template

jamanetwopen-e2621568-s002.pdf (1.9MB, pdf)
Supplement 3.

Data Sharing Statement

jamanetwopen-e2621568-s003.pdf (14.8KB, pdf)

References

  • 1.Shorey S, Ng ED, Wong CHJ. Global prevalence of depression and elevated depressive symptoms among adolescents: a systematic review and meta-analysis. Br J Clin Psychol. 2022;61(2):287-305. doi: 10.1111/bjc.12333 [DOI] [PubMed] [Google Scholar]
  • 2.Pine DS, Cohen P, Gurley D, Brook J, Ma Y. The risk for early-adulthood anxiety and depressive disorders in adolescents with anxiety and depressive disorders. Arch Gen Psychiatry. 1998;55(1):56-64. doi: 10.1001/archpsyc.55.1.56 [DOI] [PubMed] [Google Scholar]
  • 3.Liu Z, Lu W, Zou W, et al. A preliminary study of brain developmental features of bipolar disorder familial risk and subthreshold symptoms. Biol Psychiatry Cogn Neurosci Neuroimaging. 2025;10(7):769-780. doi: 10.1016/j.bpsc.2024.06.005 [DOI] [PubMed] [Google Scholar]
  • 4.Lu W, Wu J, Shao R, et al. Genetic and symptomatic risks associated with longitudinal brain morphometry in bipolar disorder. Nat Ment Health. 2024;2:209-217. doi: 10.1038/s44220-023-00194-x [DOI] [Google Scholar]
  • 5.Weissman MM, Wolk S, Goldstein RB, et al. Depressed adolescents grown up. JAMA. 1999;281(18):1707-1713. doi: 10.1001/jama.281.18.1707 [DOI] [PubMed] [Google Scholar]
  • 6.Malhi GS, Bell E, Bassett D, et al. The 2020 Royal Australian and New Zealand College of Psychiatrists clinical practice guidelines for mood disorders. Aust N Z J Psychiatry. 2021;55(1):7-117. doi: 10.1177/0004867420979353 [DOI] [PubMed] [Google Scholar]
  • 7.Cuijpers P, Oud M, Karyotaki E, et al. Psychologic treatment of depression compared with pharmacotherapy and combined treatment in primary care: a network meta-analysis. Ann Fam Med. 2021;19(3):262-270. doi: 10.1370/afm.2676 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Movahed F, Heidari E, Sadeghi D, et al. Incident diabetes in adolescents using antidepressant: a systematic review and meta-analysis. Eur Child Adolesc Psychiatry. 2025;34(2):599-610. doi: 10.1007/s00787-024-02502-x [DOI] [PubMed] [Google Scholar]
  • 9.Jakobsen JC, Katakam KK, Schou A, et al. Selective serotonin reuptake inhibitors versus placebo in patients with major depressive disorder: a systematic review with meta-analysis and trial sequential analysis. BMC Psychiatry. 2017;17(1):58. doi: 10.1186/s12888-016-1173-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Wise J. NICE guidance on depression: 35 health organisations demand “full and proper” revision. BMJ. 2019;365:l2356. doi: 10.1136/bmj.l2356 [DOI] [PubMed] [Google Scholar]
  • 11.Kessler RC. The costs of depression. Psychiatr Clin North Am. 2012;35(1):1-14. doi: 10.1016/j.psc.2011.11.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Heissel A, Heinen D, Brokmeier LL, et al. Exercise as medicine for depressive symptoms? a systematic review and meta-analysis with meta-regression. Br J Sports Med. 2023;57(16):1049-1057. doi: 10.1136/bjsports-2022-106282 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Vancampfort D, Firth J, Stubbs B, et al. The efficacy, mechanisms and implementation of physical activity as an adjunctive treatment in mental disorders: a meta-review of outcomes, neurobiology and key determinants. World Psychiatry. 2025;24(2):227-239. doi: 10.1002/wps.21314 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Cheng T, Formolo DA, Zonghao M, Yau SY. The neural mechanism underlying the antidepressant effects elicited by a single bout of physical exercise. Int J Neuropsychopharmacol. 2025;28:i197. doi: 10.1093/ijnp/pyae059.342 [DOI] [Google Scholar]
  • 15.Cheng T, Douglas Affonso F, Yu J, et al. Rapid antidepressant effect of single-bout exercise is mediated by adiponectin-induced APPL1 nucleus translocation in anterior cingulate cortex. Mol Psychiatry. 2025;30(12):5760-5776. doi: 10.1038/s41380-025-03317-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Wittenborn AK, Rahmandad H, Rick J, Hosseinichimeh N. Depression as a systemic syndrome: mapping the feedback loops of major depressive disorder. Psychol Med. 2016;46(3):551-562. doi: 10.1017/S0033291715002044 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Chai Y, Sheline YI, Oathes DJ, Balderston NL, Rao H, Yu M. Functional connectomics in depression: insights into therapies. Trends Cogn Sci. 2023;27(9):814-832. doi: 10.1016/j.tics.2023.05.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Rolle CE, Fonzo GA, Wu W, et al. Cortical connectivity moderators of antidepressant vs placebo treatment response in major depressive disorder: secondary analysis of a randomized clinical trial. JAMA Psychiatry. 2020;77(4):397-408. doi: 10.1001/jamapsychiatry.2019.3867 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Zhang Y, Wu W, Toll RT, et al. Identification of psychiatric disorder subtypes from functional connectivity patterns in resting-state electroencephalography. Nat Biomed Eng. 2021;5(4):309-323. doi: 10.1038/s41551-020-00614-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Klooster D, Voetterl H, Baeken C, Arns M. Evaluating robustness of brain stimulation biomarkers for depression: a systematic review of magnetic resonance imaging and electroencephalography studies. Biol Psychiatry. 2024;95(6):553-563. doi: 10.1016/j.biopsych.2023.09.009 [DOI] [PubMed] [Google Scholar]
  • 21.Amzica F, Lopes da Silva FH. C2Cellular substrates of brain rhythms. In: Schomer DL, Lopes da Silva FH, eds. Niedermeyer’s Electroencephalography: Basic Principles, Clinical Applications, and Related Fields. Oxford University Press; 2017:20-62. [Google Scholar]
  • 22.Biasiucci A, Franceschiello B, Murray MM. Electroencephalography. Curr Biol. 2019;29(3):R80-R85. doi: 10.1016/j.cub.2018.11.052 [DOI] [PubMed] [Google Scholar]
  • 23.Buzsáki G. Rhythms of the Brain. Oxford University Press; 2006. doi: 10.1093/acprof:oso/9780195301069.001.0001 [DOI] [Google Scholar]
  • 24.Roux F, Wibral M, Mohr HM, Singer W, Uhlhaas PJ. Gamma-band activity in human prefrontal cortex codes for the number of relevant items maintained in working memory. J Neurosci. 2012;32(36):12411-12420. doi: 10.1523/JNEUROSCI.0421-12.2012 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Nentwich M, Ai L, Madsen J, et al. Functional connectivity of EEG is subject-specific, associated with phenotype, and different from fMRI. Neuroimage. 2020;218:117001. doi: 10.1016/j.neuroimage.2020.117001 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Wirsich J, Jorge J, Iannotti GR, et al. The relationship between EEG and fMRI connectomes is reproducible across simultaneous EEG-fMRI studies from 1.5T to 7T. Neuroimage. 2021;231:117864. doi: 10.1016/j.neuroimage.2021.117864 [DOI] [PubMed] [Google Scholar]
  • 27.Hill AT, Zomorrodi R, Hadas I, et al. Resting-state electroencephalographic functional network alterations in major depressive disorder following magnetic seizure therapy. Prog Neuropsychopharmacol Biol Psychiatry. 2021;108:110082. doi: 10.1016/j.pnpbp.2020.110082 [DOI] [PubMed] [Google Scholar]
  • 28.Ulrich S, Schneider E, Deuring G, et al. Alterations in resting-state EEG functional connectivity in patients with major depressive disorder receiving electroconvulsive therapy: a systematic review. Neurosci Biobehav Rev. 2025;169:106017. doi: 10.1016/j.neubiorev.2025.106017 [DOI] [PubMed] [Google Scholar]
  • 29.Hill AT, Hadas I, Zomorrodi R, et al. Modulation of functional network properties in major depressive disorder following electroconvulsive therapy (ECT): a resting-state EEG analysis. Sci Rep. 2020;10(1):17057. doi: 10.1038/s41598-020-74103-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.World Medical Association . World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects. JAMA. 2013;310(20):2191-2194. doi: 10.1001/jama.2013.281053 [DOI] [PubMed] [Google Scholar]
  • 31.Levis B, Benedetti A, Thombs BD; Depression Screening Data (DEPRESSD) Collaboration . Accuracy of Patient Health Questionnaire-9 (PHQ-9) for screening to detect major depression: individual participant data meta-analysis. BMJ. 2019;365:l1476. doi: 10.1136/bmj.l1476 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Maddison R, Ni Mhurchu C, Jiang Y, et al. International Physical Activity Questionnaire (IPAQ) and New Zealand Physical Activity Questionnaire (NZPAQ): a doubly labelled water validation. Int J Behav Nutr Phys Act. 2007;4:62. doi: 10.1186/1479-5868-4-62 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Yan X, Pat N, Jiang Z, Wu Q, Liu Y, Halberstadt J. Name-face congruence: mechanisms and influences on social perception. Psychon Bull Rev. 2025;32(6):3088-3102. doi: 10.3758/s13423-025-02726-1 [DOI] [PubMed] [Google Scholar]
  • 34.Tadel F, Baillet S, Mosher JC, Pantazis D, Leahy RM. Brainstorm: a user-friendly application for MEG/EEG analysis. Comput Intell Neurosci. 2011;2011(1):879716. doi: 10.1155/2011/879716 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Vinck M, Oostenveld R, van Wingerden M, Battaglia F, Pennartz CMA. An improved index of phase-synchronization for electrophysiological data in the presence of volume-conduction, noise and sample-size bias. Neuroimage. 2011;55(4):1548-1565. doi: 10.1016/j.neuroimage.2011.01.055 [DOI] [PubMed] [Google Scholar]
  • 36.Wang J, Wang X, Xia M, Liao X, Evans A, He Y. GRETNA: a graph theoretical network analysis toolbox for imaging connectomics. Front Hum Neurosci. 2015;9(9):386. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Ismail LE, Karwowski W. A graph theory-based modeling of functional brain connectivity based on EEG: a systematic review in the context of neuroergonomics. IEEE Access. 2020;8:155103-155135. doi: 10.1109/ACCESS.2020.3018995 [DOI] [Google Scholar]
  • 38.Zhao D, Zhang M, Tian W, et al. Neurophysiological correlate of incubation of craving in individuals with methamphetamine use disorder. Mol Psychiatry. 2021;26(11):6198-6208. doi: 10.1038/s41380-021-01252-5 [DOI] [PubMed] [Google Scholar]
  • 39.Arnold AEGF, Protzner AB, Bray S, Levy RM, Iaria G. Neural network configuration and efficiency underlies individual differences in spatial orientation ability. J Cogn Neurosci. 2014;26(2):380-394. doi: 10.1162/jocn_a_00491 [DOI] [PubMed] [Google Scholar]
  • 40.Sun Y, Bo S, Lv J. Brain network analysis of cognitive reappraisal and expressive inhibition strategies: evidence from EEG and ERP. Acta Psycholica Sinica. 2020;52(1):12-25. doi: 10.3724/SP.J.1041.2020.00012 [DOI] [Google Scholar]
  • 41.MacKinnon DP, Fairchild AJ, Fritz MS. Mediation analysis. Annu Rev Psychol. 2007;58(1):593-614. doi: 10.1146/annurev.psych.58.110405.085542 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Sun S, Yang P, Chen H, et al. Electroconvulsive therapy-induced changes in functional brain network of major depressive disorder patients: a longitudinal resting-state electroencephalography study. Front Hum Neurosci. 2022;16:852657. doi: 10.3389/fnhum.2022.852657 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Miljevic A, Bailey NW, Murphy OW, Perera MPN, Fitzgerald PB. Alterations in EEG functional connectivity in individuals with depression: a systematic review. J Affect Disord. 2023;328:287-302. doi: 10.1016/j.jad.2023.01.126 [DOI] [PubMed] [Google Scholar]
  • 44.Kaiser RH, Andrews-Hanna JR, Wager TD, Pizzagalli DA. Large-scale network dysfunction in major depressive disorder: a meta-analysis of resting-state functional connectivity. JAMA Psychiatry. 2015;72(6):603-611. doi: 10.1001/jamapsychiatry.2015.0071 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Sporns O, Zwi JD. The small world of the cerebral cortex. Neuroinformatics. 2004;2(2):145-162. doi: 10.1385/NI:2:2:145 [DOI] [PubMed] [Google Scholar]
  • 46.Bassett DS, Bullmore E. Small-world brain networks. Neuroscientist. 2006;12(6):512-523. doi: 10.1177/1073858406293182 [DOI] [PubMed] [Google Scholar]
  • 47.Li Y, Li Y, Wei Q, et al. Mapping intrinsic functional network topological architecture in major depression disorder after electroconvulsive therapy. J Affect Disord. 2022;311:103-109. doi: 10.1016/j.jad.2022.05.067 [DOI] [PubMed] [Google Scholar]
  • 48.Yang H, Chen X, Chen ZB, et al. Disrupted intrinsic functional brain topology in patients with major depressive disorder. Mol Psychiatry. 2021;26(12):7363-7371. doi: 10.1038/s41380-021-01247-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Drevets WC, Price JL, Furey ML. Brain structural and functional abnormalities in mood disorders: implications for neurocircuitry models of depression. Brain Struct Funct. 2008;213(1-2):93-118. doi: 10.1007/s00429-008-0189-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Fornito A, Zalesky A, Bassett DS, et al. Genetic influences on cost-efficient organization of human cortical functional networks. J Neurosci. 2011;31(9):3261-3270. doi: 10.1523/JNEUROSCI.4858-10.2011 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Hamann C, Rusterholz T, Studer M, Kaess M, Tarokh L. Association between depressive symptoms and sleep neurophysiology in early adolescence. J Child Psychol Psychiatry. 2019;60(12):1334-1342. doi: 10.1111/jcpp.13088 [DOI] [PubMed] [Google Scholar]
  • 52.Disner SG, Beevers CG, Haigh EA, Beck AT. Neural mechanisms of the cognitive model of depression. Nat Rev Neurosci. 2011;12(8):467-477. doi: 10.1038/nrn3027 [DOI] [PubMed] [Google Scholar]
  • 53.Zhang B, Rolls ET, Wang X, Xie C, Cheng W, Feng J. Roles of the medial and lateral orbitofrontal cortex in major depression and its treatment. Mol Psychiatry. 2024;29(4):914-928. doi: 10.1038/s41380-023-02380-w [DOI] [PubMed] [Google Scholar]
  • 54.LeMoult J, Gotlib IH. Depression: a cognitive perspective. Clin Psychol Rev. 2019;69:51-66. doi: 10.1016/j.cpr.2018.06.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Menon V. 20 Years of the default mode network: a review and synthesis. Neuron. 2023;111(16):2469-2487. doi: 10.1016/j.neuron.2023.04.023 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Raichle ME. The brain’s default mode network. Annu Rev Neurosci. 2015;38(1):433-447. doi: 10.1146/annurev-neuro-071013-014030 [DOI] [PubMed] [Google Scholar]
  • 57.Utevsky AV, Smith DV, Huettel SA. Precuneus is a functional core of the default-mode network. J Neurosci. 2014;34(3):932-940. doi: 10.1523/JNEUROSCI.4227-13.2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 58.Sheline YI, Barch DM, Price JL, et al. The default mode network and self-referential processes in depression. Proc Natl Acad Sci U S A. 2009;106(6):1942-1947. doi: 10.1073/pnas.0812686106 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Zhao JL, Jiang WT, Wang X, Cai ZD, Liu ZH, Liu GR. Exercise, brain plasticity, and depression. CNS Neurosci Ther. 2020;26(9):885-895. doi: 10.1111/cns.13385 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.Riemann D, Krone LB, Wulff K, Nissen C. Sleep, insomnia, and depression. Neuropsychopharmacology. 2020;45(1):74-89. doi: 10.1038/s41386-019-0411-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Wang S, Tian Y, Deng X, Jülich ST, Lei X. Impaired emotional processing in insomnia: REM sleep beta power and frontal cortex activation. BMC Med. 2025;23(1):608. doi: 10.1186/s12916-025-04437-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.Gee B, Orchard F, Clarke E, Joy A, Clarke T, Reynolds S. The effect of non-pharmacological sleep interventions on depression symptoms: a meta-analysis of randomised controlled trials. Sleep Med Rev. 2019;43:118-128. doi: 10.1016/j.smrv.2018.09.004 [DOI] [PubMed] [Google Scholar]
  • 63.Morawetz C, Riedel MC, Salo T, et al. Multiple large-scale neural networks underlying emotion regulation. Neurosci Biobehav Rev. 2020;116:382-395. doi: 10.1016/j.neubiorev.2020.07.001 [DOI] [PubMed] [Google Scholar]
  • 64.Jiang J, Ferguson MA, Grafman J, Cohen AL, Fox MD. A lesion-derived brain network for emotion regulation. Biol Psychiatry. 2023;94(8):640-649. doi: 10.1016/j.biopsych.2023.02.007 [DOI] [PubMed] [Google Scholar]
  • 65.Ho TC, Walker JC, Teresi GI, et al. Default mode and salience network alterations in suicidal and non-suicidal self-injurious thoughts and behaviors in adolescents with depression. Transl Psychiatry. 2021;11(1):38. doi: 10.1038/s41398-020-01103-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 66.van den Boogert F, Klein K, Spaan P, et al. Sensory processing difficulties in psychiatric disorders: a meta-analysis. J Psychiatr Res. 2022;151:173-180. doi: 10.1016/j.jpsychires.2022.04.020 [DOI] [PubMed] [Google Scholar]
  • 67.Liu M, Ma J, Fu CY, et al. Dysfunction of emotion regulation in mild cognitive impairment individuals combined with depressive disorder: a neural mechanism study. Front Aging Neurosci. 2022;14:884741. doi: 10.3389/fnagi.2022.884741 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 68.Canbeyli R. Sensory stimulation via the visual, auditory, olfactory and gustatory systems can modulate mood and depression. Eur J Neurosci. 2022;55(1):244-263. doi: 10.1111/ejn.15507 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplement 1.

Trial Protocol

jamanetwopen-e2621568-s001.pdf (426.3KB, pdf)
Supplement 2.

eMethods.

eResults.

eFigure 1. Change of FCs in Alpha, Beta, Delta, and Theta Frequency Bands After Exercise Intervention vs Control Intervention

eFigure 2. Change of Global Network Topology Properties in Alpha, Beta, Delta, and Theta Frequency Bands After Exercise Intervention vs Control Intervention

eFigure 3. Change of Node Network Topology Properties in Alpha Frequency Band After Exercise Intervention vs Control Intervention

eFigure 4. Change of Node Network Topology Properties in Beta, Delta, and Theta Frequency Bands After Exercise Intervention vs Control Intervention

eFigure 5. Functional Connectivity Related to Reduction of Depressive Symptoms

eFigure 6. Network Topological Properties Related to Reduction of Depressive Symptoms

eFigure 7. Relationship Between Changes in PHQ-9 Scores and Changes in Nodal Network Topological Properties That Mediate the Effect of Exercise on Depressive Symptoms

eTable 1. Definition and Functional Significance of the Global and Nodal Topology Metrics

eTable 2. Index, Full Name, and Abbreviated Name of Regions Included in the Desikan-Killiany Cortical Partition Template

jamanetwopen-e2621568-s002.pdf (1.9MB, pdf)
Supplement 3.

Data Sharing Statement

jamanetwopen-e2621568-s003.pdf (14.8KB, pdf)

Articles from JAMA Network Open are provided here courtesy of American Medical Association

RESOURCES