Noetel · BMJ (Clinical research ed.) 2024 · systematic review and network meta-analysis · n=14170 participants (218 studies)

Effect of exercise for depression: systematic review and network meta-analysis of randomised controlled trials.

Cited 592 times in the scientific literature.

Level 1 - systematic review of randomized trials

Systematic review and network meta-analysis of randomized controlled trials

PubMed 38355154 · doi:10.1136/bmj-2023-075847 · record verified 2026-08-27

What was done

The authors conducted a systematic review and Bayesian multilevel network meta-analysis of randomized controlled trials from Cochrane Library, Medline, Embase, SPORTDiscus, and PsycINFO. Trials evaluated exercise modalities against active controls, psychotherapy, or antidepressants in participants meeting clinical thresholds for major depression. Quality of evidence was graded using the CINeMA tool.

What was found

Across 218 unique studies (495 arms, 14,170 participants), exercise modalities showed moderate reductions in depression compared with active controls: - Walking or jogging: Hedges' g -0.62 (95% credible interval -0.80 to -0.45; n=1210, k=51) - Yoga: g -0.55 (95% CrI -0.73 to -0.36; n=1047, k=33) - Strength training: g -0.49 (95% CrI -0.69 to -0.29; n=643, k=22) - Mixed aerobic exercises: g -0.43 (95% CrI -0.61 to -0.24; n=1286, k=51) - Tai chi or qigong: g -0.42 (95% CrI -0.65 to -0.21; n=343, k=12) Effects were proportional to prescribed exercise intensity. Yoga and strength training were the most acceptable modalities.

Why it matters

This network meta-analysis offers comparative effect estimates across specific exercise types, suggesting that walking/jogging, yoga, and strength training can serve as core clinical treatments for major depression alongside standard care.

Limits

Only one study met Cochrane criteria for low risk of bias. Overall confidence via CINeMA was low for walking or jogging and very low for all other modalities. Participants and staff could not be blinded in most exercise protocols, leaving potential for expectancy effects.

Cited by