Effectiveness of different exercise interventions on depressive symptoms among college students: a network meta-analysis

Depression is highly prevalent globally and particularly common among college students (≈25%), impairing quality of life, academic performance, and social functioning. While antidepressants are effective, they carry risks (dependence, withdrawal), and exercise is a promising low-cost, adherent non-pharmacologic option. Prior studies show exercise reduces depressive symptoms, with effects varying by type, intensity, and frequency. Some evidence suggests specialized training and traditional rhythmic practices (e.g., Qigong) can be beneficial. However, most reviews focus on single interventions and lack comparative evidence across multiple exercise modalities, limiting clinical optimization. This study uses network meta-analysis to compare multiple exercise interventions simultaneously, rank their efficacy for depressive symptoms in college students, and provide guidance for targeted prescriptions and future mechanistic research.

Protocol and reporting: Registered in PROSPERO (CRD420251027876); reported per PRISMA 2020. Databases and search: PubMed, Web of Science, Cochrane Library, EMBASE, SCOPUS, ScienceDirect; extended to Chinese databases (CNKI, VIP, Wanfang). Searches from inception to July 1, 2024; English and Chinese languages; MeSH and free-text terms for exercise, college students, and depression; manual reference screening. Screening and selection: Two-stage independent screening with third-party arbitration for disagreements; inclusion criteria: (1) full-time college students aged 18–30 without severe illness; (2) aerobic or resistance exercise interventions vs no-exercise controls; (3) RCTs with described randomization, blinding, allocation concealment, and attrition reporting; (4) primary outcomes: standard depression scales (SDS, CES-D), secondary: HAMD, BDI; (5) analyzable data (group sizes, means, SDs, CIs, effect sizes). Exclusions: non-eligible populations, non-exercise interventions, severe illness, non-RCTs or unclear randomization/blinding, non-standard outcomes, incomplete analyzable data, duplicate/fraudulent data, poor compliance/high dropout. Data extraction: Two independent extractors recorded author, year, country, intervention content and parameters (type, intensity, frequency, session duration, total duration), outcomes, sample sizes, means/SDs, effect sizes with SE and 95% CI; contacted authors for missing critical data or excluded if unavailable; non-critical missing coded as NA. Quality assessment: Cochrane risk-of-bias tool via Review Manager 5.4; domains included random sequence generation, allocation concealment, blinding (participants/personnel, outcome assessors), incomplete outcome data, selective reporting, other bias; categorized studies by number of low-risk ratings. Effect size and statistics: Standardized mean difference (SMD) calculated using pooled SD; heterogeneity assessed with Q-test and I² (fixed-effects if I²<50% and P>0.1; random-effects if I²≥50% or P≤0.1; subgroup/sensitivity analysis if I²>75%); pooled effects via random-effects weighted average. Network meta-analysis: Conducted in Stata 14.0 using node-splitting for inconsistency; closed-loop networks formed across six intervention categories: Aerobic Exercise Group (AEG), Dance and Rhythmic Movement Group (DRMG), Moderate Intensity Exercise Group (MIG), Strength and Resistance Training Group (SRTG), High Intensity Training Group (HITG), Special Training Unit (STU). SUCRA used to rank probabilities of intervention efficacy. Publication bias: Egger tests and funnel plots assessed symmetry.

- Studies included: 31 RCTs; total participants: 1,169 college students. Six intervention categories: AEG, DRMG, MIG, SRTG, HITG, STU. - Network consistency: Node-splitting global/local inconsistency tests showed P>0.05 across key comparisons (e.g., AEG vs DRMG P=0.67; AEG vs SRTG P=0.47; AEG vs HITG P=0.65; AEG vs STU P=0.54; DRMG vs MIG P=0.51; DRMG vs SRTG P=0.89), indicating good agreement between direct and indirect evidence. - SUCRA rankings (probability of being most effective): STU 65.1%, DRMG 64.8%, AEG 61.3%, SRTG 60.9%, HITG 26.2%, MIG 21.7. - Comparative effects: STU and DRMG had the highest probabilities of efficacy for reducing depressive symptoms; HITG and MIG were least effective. Pairwise network estimates often had wide 95% CIs overlapping zero (e.g., DRMG vs STU SMD=0.23, 95% CI −3.67 to 4.13; DRMG vs HITG SMD=−1.51, 95% CI −3.90 to 0.88). - Heterogeneity: Pairwise meta-analyses showed substantial heterogeneity (>50%) across most comparisons, attributed to differences in participant characteristics, exercise protocols, and outcome scales; sensitivity analyses confirmed robustness of main findings. - Publication bias: Funnel plots were approximately symmetric, suggesting low risk of publication bias/small-study effects.

The network meta-analysis addresses the comparative efficacy of multiple exercise modalities for depressive symptoms in college students. Findings indicate that interventions combining physical activity with psychological components (STU) and rhythmic/traditional movement-based practices (DRMG) are most effective, followed by aerobic (AEG) and strength/resistance training (SRTG). Mechanistically, STU may act via enhanced neuroplasticity (BDNF/TrkB–PI3K/Akt–MAPK/ERK, CREB), monoaminergic modulation (5-HT, DA, NE), HPA axis regulation (GR function), and anti-inflammatory effects (NF-κB downregulation, IL-10 upregulation). DRMG may leverage myokine signaling (IL-6–BDNF, irisin), improved metabolic and mitochondrial function (AMPK, PGC-1α), and antioxidant defenses (SOD, GPx, CAT). AEG likely involves β-endorphin/μ-opioid pathways, dopaminergic reward circuitry, neurogenesis/angiogenesis (Wnt/β-catenin, VEGF, IGF-1), and epigenetic modulation. SRTG may improve self-efficacy/body image and engage monoamine and neurogenesis pathways. In contrast, HITG may produce excessive oxidative stress and HPA axis hyperactivation, while MIG may be subthreshold to induce optimal neurobiological adaptations. Overall, the results support exercise as an adjunctive treatment for depression, with modality and intensity carefully tailored to maximize benefits and minimize adverse stress responses. The findings are relevant for university mental health services, informing targeted, acceptable, and scalable exercise prescriptions.

This network meta-analysis demonstrates that exercise interventions effectively alleviate depressive symptoms among college students, with special training units (combining physical activity and psychological components) and dance/rhythmic movement programs showing the highest efficacy. The work provides comparative evidence to guide tailored, evidence-based exercise prescriptions in university settings. Practical applications include integrating STU-like programs and DRMG courses into campus mental health services, personalized to students’ symptom severity, preferences, and fitness. Future research should: (1) examine effects of STU and DRMG across varying depression severities; (2) incorporate long-term follow-up (≥12 months) to evaluate durability; (3) include objective neurobiological measures (e.g., BDNF, inflammatory markers, EEG/fMRI); (4) test multimodal combinations (exercise plus CBT); (5) develop customized plans for subgroups (e.g., comorbid anxiety or trauma).

- High heterogeneity across included RCTs in participant demographics (gender, age, baseline severity), exercise protocols (type, intensity, frequency, duration), and assessment tools, affecting generalizability. - Predominantly short-term interventions with limited or no long-term follow-up, leaving durability of effects unclear. - Reliance on subjective scales without integration of objective neuroimaging or biomarker data, constraining mechanistic inference. - Insufficient exploration of specific mechanisms and comparative advantages across exercise modalities; more refined, head-to-head mechanistic trials are needed.