Exercise for depression: what 218 mostly weak trials show about walking, yoga and strength training

Being told to go for a run is among the least welcome things a depressed person can hear, and for a long time it was also among the least well supported. Exercise was something doctors mentioned alongside real treatment, the antidepressants and talking therapies that had trials behind them.

That has changed. There are now more than two hundred randomized trials of exercise in people with diagnosed depression, and in 2024 a team led by Michael Noetel at the University of Queensland pulled them together in The BMJ. The headline result was strong: several kinds of exercise reduced symptoms by about as much as psychotherapy did. The small print is just as important, because the trials themselves are mostly weak, and reviewers who admit only the best ones reach a much cooler conclusion.

What exercise treatment for depression means in the trials

The BMJ review counted any randomized trial that had exercise arms for participants meeting clinical cut-offs for major depression. That excludes studies of people who are merely stressed or low. It also means the “exercise” in question is a program: a supervised walking or jogging group, a yoga class, a schedule of weight training, usually running for a matter of weeks.

In all, the authors found 218 unique studies with a total of 495 arms and 14,170 participants. An arm is one group within a trial. Most trials compared exercise with what the reviewers call active controls, such as usual care or a placebo tablet, and some compared it with antidepressants or therapy directly.

Because few trials compared one form of exercise with another, the team used a network meta-analysis, a statistical method that links treatments through their shared comparison groups to estimate how each would fare against the others.

What the trials of exercise for depression show

Five types of exercise produced what the authors classed as moderate reductions in depression compared with active controls.

Type of exercise	Trial arms	Participants	Size of effect
Walking or jogging	51	1,210	Moderate
Yoga	33	1,047	Moderate
Strength training	22	643	Moderate
Mixed aerobic exercise	51	1,286	Moderate, slightly smaller
Tai chi or qigong	12	343	Moderate, slightly smaller
Dance	5	107	Large, but few trials

For comparison, cognitive behavior therapy scored in the same range as walking, jogging and yoga in this analysis, and antidepressants of the SSRI type, the most widely prescribed class, scored lower. The authors warn against reading much into that last comparison. Their search was built to find exercise trials, and they write that these estimates should not usurp these directly focused systematic reviews of drugs and therapy.

Three patterns ran across the types. Harder exercise did more: the abstract states that the effects of exercise were proportional to the intensity prescribed, with running and interval training ahead of gentle walking, though light activity still showed a worthwhile effect. The weekly amount did not clearly matter. And the benefit held up, with effects about the same when measured up to six months after a program ended.

Some results were unexpected. Programs that gave participants a choice over what they did and how often tended to work less well than programs that prescribed it. Strength training and yoga appeared to be the most acceptable modalities, meaning fewer people dropped out of them than out of the comparison groups.

What does “moderate” look like in people? A well-known trial gives a sense. James Blumenthal and colleagues at Duke University assigned 202 adults with major depression to supervised group exercise, exercise at home, the antidepressant sertraline or a placebo pill. After four months, 45% of the supervised exercisers, 40% of the home exercisers and 47% of those on the drug were in remission, against 31% on placebo.

How does exercise compare with antidepressants and therapy?

That Duke trial is one of very few to put exercise beside a drug. Its authors concluded that the efficacy of exercise in patients seems generally comparable with patients receiving antidepressant medication.

The Cochrane Collaboration, which produces systematic reviews to a fixed and strict method, looked at all such head-to-head trials in 2013. Seven trials compared exercise with psychological therapy and four compared it with drug treatment, and in neither case was there a clear difference. The reviewers attached a warning: compared with those therapies exercise appears to be no more effective, though this conclusion is based on a few small trials.

The BMJ analysis adds that exercise combined with an SSRI, and aerobic exercise combined with psychotherapy, each helped by a similar margin. None of this amounts to a test of exercise as a replacement for medication, and no trial has shown that stopping treatment in favor of exercise is safe. Do not stop or change a prescribed medicine without talking to your doctor.

Why might exercise ease depression?

Nobody is sure. The BMJ authors looked for trials that had tested a mechanism and found only a handful, with inconsistent answers. “Our review did not uncover clear causal mechanisms,” they write.

They suspect there is no single one. Their hypothesis is that a combination of social interaction, mindfulness or experiential acceptance, increased self-efficacy, immersion in green spaces, neurobiological mechanisms, and acute positive affect is at work, with different kinds of exercise drawing on different parts of the list. A yoga class leans on mindful attention, a walking group on company and being outdoors, a weights program on the visible evidence of getting stronger. Self-efficacy is a person’s belief that they can do what they set out to do, and positive affect is the psychologists’ term for good mood.

How strong is the evidence that exercise treats depression?

This is where the reviews part company, and the disagreement is about which trials to believe.

An exercise trial cannot be blinded in the way a drug trial can. Participants know whether they have been sent to the gym, they often rate their own symptoms, and those who expected to feel better may report that they do. The Duke trial showed how much expectation can do: almost a third of patients given a placebo pill went into remission, and its authors remarked that a considerable portion of the therapeutic response is determined by patient expectations and by attention.

Noetel’s team graded every trial against the Cochrane criteria. They report that only one study met the Cochrane criteria for low risk of bias. As a result, their confidence in the results was low for walking or jogging and very low for the other treatments. They also found that studies with funding reported stronger effects than those without.

Stricter reviewers have taken the next step and set the weaker trials aside. The 2013 Cochrane review found a moderate effect across 35 trials, the same size the BMJ analysis would report for walking or jogging a decade later. When the reviewers included only the six trials with the soundest designs, the effect shrank to less than a third of that and could no longer be distinguished from chance. Their summary was that analysis of the methodologically stronger trials only shows a smaller effect in favor of exercise.

A Danish group led by Jesper Krogh went further in 2017. Across 35 trials they too found a clear benefit. When they restricted the analysis to the four trials that seemed less affected by bias, the effect vanished. Their conclusion was that “trials with less risk of bias suggested no antidepressant effects of exercise”.

The most recent large review, by Andreas Heissel and colleagues in 2023, lands in between. In their analysis the benefit was still of middling size when analyses were restricted to low risk of bias studies, of which they counted twelve.

So the size of the benefit depends on how many trials a reviewer is willing to trust. With all trials counted, exercise looks about as good as therapy. With only the most rigorous, the benefit lies somewhere between that and nothing.

What guidelines say about exercise for depression

Guideline writers have mostly settled on including exercise without ranking it first. The BMJ review notes that clinical practice guidelines in the US, UK, and Australia recommend physical activity as part of treatment for depression, with differences in detail. British guidance recommends group exercise programs, the American Psychiatric Association recommends any dose of aerobic exercise or resistance training, and guidance for Australia and New Zealand suggests strength and vigorous aerobic exercise two or three times a week.

Noetel and colleagues think that is too timid. They argue that treatment guidelines may be overly conservative in offering exercise mainly when drugs or therapy have failed or been declined. Krogh’s group, reading largely the same trials, would not agree.

Which treatment suits any one person depends on how severe their depression is and on the rest of their circumstances, which is assessed by their doctor.

What the research on exercise and depression has not settled

The trials say little about who benefits. People who volunteer for an exercise study are those who feel able to attempt one, and depression at its worst removes exactly that capacity. Whether a program that works for volunteers can be delivered to patients who did not ask for it is largely untested.

The evidence also thins out with time. Most trials ran for weeks, and the BMJ analysis found some sign that longer programs did less well than shorter ones, with too much uncertainty to be sure. Harms were rarely recorded at all.

Exercise has more trials behind it as a depression treatment than most people realize, and fewer good ones than its strongest advocates suggest.