Does Exercise Actually Work as a Treatment for Depression?

The evidence base is large and the researchers’ confidence in it is low. A network meta-analysis of 218 randomised trials covering 14,170 participants found moderate reductions in depression for walking or jogging, yoga, strength training, mixed aerobic exercise and tai chi, compared with active controls. It concluded that exercise is an effective treatment for depression.

The same paper rates its own evidence. Only one included study met the Cochrane criteria for low risk of bias, and confidence assessed with the standard tool for network meta-analyses was low for walking or jogging and very low for all other treatments.

Both halves are true at once, and coverage of this study almost never carries the second.

What the analysis found

The study is Noetel and colleagues’ 2024 network meta-analysis in the BMJ. Its objective was to identify the optimal dose and modality of exercise for treating major depressive disorder, compared with psychotherapy, antidepressants and control conditions. Screening, data extraction, coding and risk of bias assessment were performed independently and in duplicate.

It included any randomised trial with exercise arms for participants meeting clinical cut-offs for major depression, producing 218 unique studies with 495 arms and 14,170 participants.

Compared with active controls such as usual care or a placebo tablet, moderate reductions in depression were found for walking or jogging, with a Hedges’ g of minus 0.62; yoga at minus 0.55; strength training at minus 0.49; mixed aerobic exercises at minus 0.43; and tai chi or qigong at minus 0.42.

The effects of exercise were proportional to the intensity prescribed. Strength training and yoga appeared to be the most acceptable modalities, meaning fewest participants dropped out.

What the paper says about its own evidence

This is the section the original version of this article could not write, because the confidence ratings were not retrievable.

Results appeared robust to publication bias, but only one study met the Cochrane criteria for low risk of bias. As a result, confidence in accordance with the network meta-analysis confidence tool was low for walking or jogging and very low for other treatments.

The paper adds that certainty ratings for all treatment arms were downgraded owing to high risk of bias in the underlying studies.

On publication bias specifically it is reassuring: no amount of publication bias would be sufficient to shrink effects to zero, and to reduce effects below clinical significance thresholds, studies with statistically significant results would need to be reported 58 times more frequently than studies without them.

So the concern is not that the finding is a publication artefact. It is that the trials feeding it were mostly at high risk of bias, which is a different and more ordinary problem.

There is one further number worth knowing. The active control condition itself showed a before and after change of Hedges’ g of minus 0.95 across 3,554 participants. The control groups improved more than any exercise effect measured against them. Depression fluctuates, people seek help when they feel worst, and comparisons in this field sit on top of substantial spontaneous improvement.

What it does not mean

Three things, and they matter more than the effect sizes.

It does not mean exercise replaces treatment. The paper’s own conclusion is that these forms of exercise could be considered alongside psychotherapy and antidepressants as core treatments for depression. Alongside is the operative word, and decisions about what treatment to use, add or change belong with a doctor.

It does not mean being told to exercise is helpful advice. Reduced energy and motivation are features of depression rather than failures of will, and a meta-analysis showing that an activity works when people do it says nothing about how someone in the middle of an episode is supposed to start. The trials delivered structured, prescribed interventions.

And it does not mean this applies to how you feel after a difficult week at work. The eligibility criterion was participants meeting clinical cut-offs for major depression. Extending it to ordinary low mood is an extension, not a result.

A note on the correction

The paper has a published correction, issued in the BMJ in May 2024. Anyone citing the original should be aware of it. The effect sizes reported above are from the paper as currently published and should be checked against the corrected version by anyone using them in a clinical or policy context.

If this is about you right now

Depression is a medical condition with several effective treatments, and which one suits a given person is a clinical judgement.

If you are experiencing persistent low mood, loss of interest in things you used to care about, changes in sleep or appetite, or difficulty functioning, please see a doctor. That is true regardless of what any meta-analysis concludes about walking.

If you are having thoughts of harming yourself, contact your doctor, a crisis line, or emergency services now rather than reading further. If it would help, tell someone you trust today.

If you already have a treatment plan, this research is a reason to discuss adding exercise with the person managing your care, not a reason to change anything on your own. If poor sleep is part of the picture, what a bad night costs you is worth reading alongside this.

Common questions

How strong is the evidence that exercise helps depression?

Large in volume and low in confidence, which is an unusual combination. A 2024 network meta-analysis in the BMJ included 218 unique studies with 495 arms and 14,170 participants, restricted to randomised trials in participants meeting clinical cut-offs for major depression. Compared with active controls, it found moderate reductions for walking or jogging at Hedges’ g minus 0.62, yoga at minus 0.55 and strength training at minus 0.49. But only one included study met the Cochrane criteria for low risk of bias, and confidence assessed with the standard network meta-analysis tool was low for walking or jogging and very low for all other treatments.

Should I exercise instead of taking medication or seeing a therapist?

No, and the paper says the opposite. Its conclusion is that these forms of exercise could be considered alongside psychotherapy and antidepressants as core treatments for depression. The analysis compared exercise with psychotherapy, antidepressants and control conditions in order to place exercise among treatment options, not to replace them. Decisions about starting, stopping, changing or adding a treatment are clinical decisions. If you have a treatment plan, this is a reason to raise exercise with the person managing your care rather than to alter anything yourself.

Which type of exercise is best?

Walking or jogging showed the largest effect at Hedges’ g minus 0.62, followed by yoga at minus 0.55, strength training at minus 0.49, mixed aerobic exercise at minus 0.43, and tai chi or qigong at minus 0.42. Effects were proportional to the intensity prescribed, and strength training and yoga appeared the most acceptable, meaning fewest people dropped out. Two cautions. Confidence was rated low for walking or jogging and very low for every other modality, so the ranking between them is less certain than the numbers suggest. And the trials delivered structured interventions rather than people deciding to be more active.

Why is confidence low if the sample is so large?

Because volume and quality are different things. The analysis pooled 218 trials, but only one met the Cochrane criteria for low risk of bias, and the paper states that certainty ratings for all treatment arms were downgraded owing to high risk of bias in the underlying studies. Publication bias was not the problem: the authors calculated that studies with significant results would need to be reported 58 times more often than those without to push effects below clinical significance. It is worth also knowing that the active control groups themselves improved by a Hedges’ g of minus 0.95, larger than any exercise effect measured against them, which is what makes blinding and control design so difficult in this field.

Strong evidence, easily misused

The reason to be careful with this study is precisely that it is good.

A network meta-analysis of 218 randomised trials in people with diagnosed major depression is about as solid as this kind of question gets, and its conclusion is clear. That strength is exactly what makes it available for a use it does not support: telling people who are unwell that there is a simple thing they are failing to do.

Everything in the paper points the other way. The interventions were structured and delivered. The comparison was against other treatments, not against seeking treatment. The framing is exercise alongside psychotherapy and antidepressants. And the authors’ own confidence in their findings is low for the best-supported modality and very low for the rest.

Which leaves two honest readings. If you are well, this is a reasonable argument for treating movement as maintenance rather than optional, and 10,000 steps shows how little of it the observational evidence suggests you need. If you are not well, it is a reason to have a specific conversation with a doctor, and not a reason to feel you should have sorted this out with a pair of running shoes. Diet and depression has the same shape: a real adjunctive finding, routinely reported as a replacement.