Exercise for Depression: How Strong Is the Evidence Really?

Exercise helps depression, and a large 2024 synthesis in The BMJ says so clearly, but the same paper is unusually honest about how much to trust its own numbers. The fair one-line summary is that structured activity is a well-supported addition to depression care, not a proven substitute for therapy or medication. Getting from the encouraging headline to that measured conclusion is a small lesson in how to read medical evidence, and it is worth walking through.

Key points#

• A 2024 BMJ network meta-analysis pooled 218 trials and more than 14,000 people; several kinds of exercise reduced depressive symptoms.
• Dance and walking or jogging showed the biggest effects, with yoga and strength training close behind.
• The authors rated their confidence as low or very low, and only one trial met strict criteria for low risk of bias.
• No one can be blinded to whether they exercised, which keeps a ceiling on how certain any of this can be.
• Most trials added exercise to existing care, so the evidence supports it as an adjunct, especially for milder symptoms.

Start with what the trials measured#

Before the rankings, it helps to notice what was actually being asked. Michael Noetel and colleagues, writing in The BMJ in February 2024, gathered 218 randomised trials (495 treatment arms, 14,170 participants) and pooled them with network methods. A network meta-analysis is a useful trick: it lets researchers compare treatments that were never tested against each other directly, by tracing paths through the whole web of studies. So even if dance was never trialled head to head against, say, tai chi, the network can still estimate how they stack up.

In this web, the largest symptom reduction came from dance, followed by moderate reductions for walking or jogging, yoga, strength training, mixed aerobic programs, and tai chi or qigong. More vigorous activity tended to help more. When the researchers asked who stuck with the program rather than who improved most, strength training and yoga held onto participants best.

Those are not small effects. The moderate to large range here sits in roughly the same territory as an earlier umbrella review in the British Journal of Sports Medicine, which found a medium effect (a standardised reduction near 0.43) across dozens of physical-activity reviews and called it broadly comparable to psychotherapy and medication. Two separate bodies of work pointing the same way is a genuine signal, not noise.

Why the leaderboard comes with an asterisk#

Here is where a careful reader slows down. The authors themselves flagged that only one included study met the Cochrane bar for low risk of bias, and that their formal confidence, graded with CINeMA (the network version of the GRADE framework), was low for walking or jogging and very low for the other options. Low confidence is not the same as wrong. It means the true effect could sit some distance from the reported estimate, and that better trials could move the ranking around.

The hardest problem is blinding. In a drug trial you can hand the comparison group an identical dummy pill, so neither the patient nor, ideally, the person scoring symptoms knows who got the real thing. There is no dummy version of a jog. People know whether they have been running three mornings a week, and that knowledge carries expectation, attention, and the ordinary lift of having a place to be and a task to finish. Some of what these trials capture is the specific effect of movement on body and mood; some is the hopefulness that travels with any energetic, structured routine. Comparing exercise against an active control rather than a do-nothing waitlist, as this analysis did, narrows that gap, but it does not close it.

The same caution applies to the eye-catching subgroup results, such as strength training seeming to help women more, or yoga seeming to help men and older adults more. Those come from slicing the data into smaller pieces, and subgroup signals are best read as questions for future trials, not as instructions. Few of the studies followed people for a year or longer, so how durable the benefit is remains largely open.

What "adjunct" precisely means#

Calling exercise an adjunct is not a polite dodge. It is an accurate description of the experiment that was run. Most of these trials layered exercise on top of whatever care people were already receiving, or compared it against usual care, rather than pitting it against a full course of medication or therapy in people who were severely ill. That design supports one specific claim: adding structured activity tends to improve symptoms. It does not support telling a person with severe or high-risk depression to stop an effective treatment and rely on movement alone.

For milder symptoms, treating exercise as a first-line ingredient is reasonable, and some clinical guidance already does. For moderate to severe depression, the sensible framing is exercise alongside evidence-based care, with the choice of activity driven less by the leaderboard and more by what a given person will keep doing week after week. A treatment only works if it is done, which is why the tolerability data, with yoga and strength training holding people's involvement, may matter more in daily practice than which modality topped the efficacy chart.

How to read the next confident headline#

"Dance beats antidepressants" is exactly the kind of headline this study invited and does not earn. The stronger, duller truth is that a large, well-run synthesis found consistent benefit from several kinds of movement, that its authors were candid about their own low confidence, and that the impossibility of blinding keeps certainty capped. That is a reason to move more, not a reason to walk away from treatments resting on sturdier ground.

So when the next bold claim built on this same dataset crosses your feed, three questions do most of the work. What was the intervention compared against? Could anyone involved have been blinded? And how much did the authors themselves trust their numbers? Reading evidence well means holding the encouragement and the caveat in the same hand at the same time.