Exercise for Depression: What the Best Evidence Actually Shows

Exercise for depression has real evidence behind it, and that evidence is weaker than the headlines suggested. A 2024 network meta-analysis in The BMJ pooling 218 randomised trials and 14,170 participants found moderate-to-large reductions in depressive symptoms for walking or jogging, yoga, strength training and dance — with effects broadly comparable to cognitive behaviour therapy. The same authors rated their confidence in those findings as low for walking or jogging and very low for everything else, because almost none of the underlying trials were methodologically clean. Exercise is best understood as a genuine adjunct to depression treatment, not a substitute for it.

Before anything else: depression is a medical condition that warrants professional care. Nothing on this page is medical advice, and nobody should start, stop or change a prescribed treatment on the strength of an article. If you are in crisis or having thoughts of harming yourself, contact emergency services, your doctor, or a crisis line now — in the US you can call or text 988, and Find a Helpline lists services in most countries.

The headline study: 218 trials in The BMJ

The centrepiece of the modern evidence is Noetel and colleagues, published in The BMJ in February 2024: a Bayesian network meta-analysis of 218 randomised trials, 495 treatment arms and 14,170 participants, all meeting clinical cut-offs for major depression. A network design ranks treatments never tested head to head by connecting them through shared comparators.

Effects are Hedges’ g against active controls such as usual care or a placebo tablet, with 95% credible intervals. Negative numbers mean fewer symptoms; roughly 0.2 is small, 0.5 moderate and 0.8 large.

Which modalities performed best

• Dance — g −0.96 (95% credible interval −1.36 to −0.56), from just 5 trials and 107 participants.
• Walking or jogging — g −0.63 (−0.80 to −0.46), from 51 trials and 1,210 participants.
• Yoga — g −0.55 (−0.73 to −0.36), from 33 trials and 1,047 participants.
• Strength training — g −0.49 (−0.69 to −0.29), from 22 trials and 643 participants.
• Mixed aerobic exercise — g −0.43 (−0.61 to −0.25), from 51 trials and 1,286 participants.
• Tai chi or qigong — g −0.42 (−0.65 to −0.21), from 12 trials and 343 participants.

Dance topped the ranking, and the authors immediately walked it back: five small trials, 107 people, mostly young women, and biased designs mean the finding “prohibits us from recommending dance more strongly,” in their words.

Dropout data pointed a different way. Strength training (odds ratio 0.55, 0.31 to 0.99) and yoga (0.57, 0.35 to 0.94) were the only arms people were significantly less likely to quit than control.

How it compared with SSRIs and psychotherapy

Within the same network, cognitive behaviour therapy alone produced g −0.55 (−0.75 to −0.37) across 20 trials and 712 participants; SSRIs alone produced g −0.26 (−0.50 to −0.01) across 16 trials and 432 participants. On paper, several exercise modalities sat between those two.

The authors are careful here, and so should we be. Their review was not built to find every trial of psychotherapy or antidepressants, so those estimates “should not usurp” reviews designed for that question — which report roughly g −0.67 for psychotherapy and g −0.30 for pharmacotherapy.

The combination results are arguably more useful. Exercise added to SSRIs gave g −0.55 (−0.86 to −0.23) across 11 trials, and aerobic exercise added to psychotherapy gave g −0.54 (−0.76 to −0.32) across 15 trials. That is the adjunct model — the evidence most relevant to someone already in treatment.

The intensity signal

A clear dose-response curve appeared for prescribed intensity. Light activity such as walking or hatha yoga still produced clinically meaningful effects (g −0.58, −0.82 to −0.33); vigorous exercise such as running or interval training produced stronger ones (g −0.74, −1.10 to −0.38).

This was not explained by total weekly energy expenditure — the curve for METs per week was unclear, with wide credible intervals. Longer programmes did not do better either: 10-week interventions (g −0.53) edged out 30-week ones (g −0.37), with substantial uncertainty. Effects held steady up to six months after treatment ended (g −0.63, −0.87 to −0.40).

What the Cochrane review adds — and subtracts

Cochrane reviews are the conservative benchmark in evidence-based medicine, and the January 2026 update (Clegg and colleagues, 73 randomised trials, at least 4,985 participants) is the best reality check on the BMJ numbers.

Across 57 trials and 2,189 participants comparing exercise with no treatment or a control, the pooled standardised mean difference was −0.67 (95% CI −0.82 to −0.52) — close to the BMJ estimate. Cochrane graded it low-certainty evidence: exercise “may” reduce depressive symptoms.

Then the important part. Restricted to the seven trials (447 participants) with adequate allocation concealment, intention-to-treat analysis and blinded outcome assessment, the effect shrank to −0.46 (−0.88 to −0.04) — an interval that only just clears zero. In the 2013 version, the equivalent subset gave −0.18 (−0.47 to 0.11), not statistically significant at all.

Cochrane also found little to no difference between exercise and psychological therapy (SMD 0.03, −0.16 to 0.23; 10 trials, 414 participants; moderate certainty) or between exercise and pharmacological treatment (SMD −0.11, −0.33 to 0.10; 5 trials, 330 participants; low certainty). Long-term follow-up evidence was very uncertain (SMD −0.53, −1.11 to 0.06; 9 trials, 405 participants).

The reviewers’ own summary is worth quoting for its restraint: adding 35 new trials “has had very little effect on the estimate of the benefit of exercise on symptoms of depression.” Thirteen more years of research barely moved the number, and did not move the certainty rating.

The wider evidence base

An umbrella review in the British Journal of Sports Medicine (2023) pooled 97 systematic reviews covering 1,039 trials and 128,119 participants, finding a median effect size of −0.43 (IQR −0.66 to −0.27) for depression. It echoed the BMJ patterns — higher intensity did more, longer interventions did less — while flagging that 77 of its 97 reviews scored “critically low” on the AMSTAR-2 quality tool.

A dose-response meta-analysis in JAMA Psychiatry (2022) pooled 15 prospective cohorts totalling 191,130 participants. Against adults reporting no activity, those at half the recommended volume had 18% lower risk of developing depression (95% CI 13% to 23%) and those at the full recommended volume 25% lower risk (18% to 32%). But that study is about prevention, not treatment, and it is observational: cohort data cannot rule out reverse causation, since early depression itself reduces activity.

What the clinical guidelines say

NICE guideline NG222 (published 29 June 2022) lists a structured group exercise programme among the first-line options for a new episode of less severe depression, alongside guided self-help, CBT and behavioural activation, with the choice matched to the person’s needs and preferences.

For more severe depression it still appears on the menu, but with an explicit caution: “the potential advantages of providing other treatment choices with more therapist contact should be carefully considered first.” NICE also lists adding a group exercise intervention when antidepressant medication alone has produced a limited response.

The WHO’s position is similar: psychological treatments are the first treatments for depression, combined with antidepressants in moderate and severe cases. It estimates 332 million people worldwide live with depression, and that in high-income countries only about one third receive treatment — much of why a low-cost adjunct is of interest at all.

What this evidence does not establish

This is the section most coverage skipped, and the one the authors wrote most carefully.

Almost no trial was methodologically clean

Of the 218 studies in the BMJ network, exactly one met Cochrane criteria for low risk of bias; for most, risk of bias was unclear or high. Applying the CINeMA framework, the authors rated confidence low for walking or jogging and very low for every other modality — including the yoga, strength training and dance results that travelled furthest in the press.

You cannot blind someone to exercise

Participants know whether they were assigned to a running programme or a waiting list, and so do the staff delivering it. Most depression outcomes are self-reported scales. That combination means effect sizes “could include expectancy effects, among other biases” — people who believe they are getting a promising treatment report feeling better.

The Cochrane update judged every included study to be at high risk of performance bias for exactly this reason. It is a structural limit on what exercise research can currently prove, and why the quality-restricted estimates matter more than the headline pooled ones.

Publication bias is present

The BMJ authors found statistically significant funnel plot asymmetry (multilevel Egger’s test, F = 23.93, P < 0.001). They also showed the effect survives it: no plausible amount of publication bias reduced the pooled effect to zero, and significant studies would need publishing 58 times more often than non-significant ones to push it below clinical relevance. Both facts belong in an honest summary.

Who was actually studied

Modalities were not evenly distributed across people. Tai chi trials skewed older (mean age 59) while dance trials skewed young and female (mean age 31, 88% women), so differences between modalities partly reflect who was recruited. The age and sex interactions reported are study-level, not individual-level, so they cannot tell you what will work better for you.

Two more caveats. The BMJ paper carries a published correction (BMJ 2024;385:q1024) about how effect sizes were standardised, which the authors state did not materially change the findings. And funded trials reported larger effects than unfunded ones — a moderator the authors pre-registered and reported rather than buried.

Where that leaves you

Exercise is not an alternative to depression treatment, and this article deliberately gives no programme, dose or schedule. That is a clinical decision, depending on your diagnosis, severity, physical health and what you already take.

What the evidence supports is raising it as a conversation. Ask your GP, psychiatrist or therapist whether adding physical activity alongside your current treatment makes sense, and what form would be realistic for you. Worth raising too: adverse events were uncommon but not zero — Cochrane logged musculoskeletal injuries and, in some participants, worsening depression.

And the point that bears repeating: do not stop or reduce prescribed medication or therapy because a study looked encouraging. Antidepressant discontinuation in particular needs medical supervision.

The honest bottom line

Across a very large body of randomised evidence, exercise reduces depressive symptoms by a moderate amount, roughly on par with psychotherapy and no worse than antidepressants head to head. Guidelines already recognise it, and the effect does not vanish under scrutiny.

But the people who assembled that evidence rate it low certainty, the trials are small and unblindable, and the estimate shrinks toward the margin of significance in the best studies. “Promising adjunct with a shaky evidence base” is the accurate description. “Exercise instead of treatment” is not, and the researchers behind these numbers do not say it either.

Sources

• Noetel M, Sanders T, Gallardo-Gómez D, et al. Effect of exercise for depression: systematic review and network meta-analysis. BMJ. 2024;384:e075847. Funding: none received; no competing interests declared. Full text (PMC10870815)
• Correction: Effect of exercise for depression. BMJ. 2024;385:q1024. Correction notice (PMC11131084)
• Clegg AJ, Hill JE, Mullin DS, et al. Exercise for depression. Cochrane Database of Systematic Reviews. 2026;1:CD004366. No grant funding; some authors part-funded by NIHR ARC North West Coast. DOI 10.1002/14651858.CD004366.pub7
• Cooney GM, Dwan K, Greig CA, et al. Exercise for depression. Cochrane Database of Systematic Reviews. 2013;(9):CD004366 — the previous version. Full text (PMC9721454)
• Singh B, Olds T, Curtis R, et al. Effectiveness of physical activity interventions for improving depression, anxiety and distress: an umbrella review. British Journal of Sports Medicine. 2023;57(18):1203-1209. No competing interests declared. Full text (PMC10579187)
• Pearce M, Garcia L, Abbas A, et al. Association between physical activity and risk of depression. JAMA Psychiatry. 2022;79(6):550-559. Full text (PMC9008579)
• National Institute for Health and Care Excellence. Depression in adults: treatment and management. NICE guideline NG222, 29 June 2022. Guideline NG222
• World Health Organization. Depressive disorder fact sheet. WHO fact sheet