Exercise & Depression: The Effect Sizes, Honestly

Ask the literature whether exercise treats depression and the answer comes back with numbers that look too good — effect sizes larger than many drug trials report. Then you ask what the control group was doing, and they shrink. This page walks the meta-analyses, the sertraline comparison, and why the honest effect is moderate, not spectacular. Depression is clinical territory: this page informs the conversation with a clinician, never replaces one.

Why the Headline Numbers Look Too Good

Depression trials usually report a standardized mean difference, or SMD: how far the exercise group moved relative to the spread in the data, in units that let different scales be pooled. By convention, 0.2 is small, 0.5 moderate, 0.8 large. The big meta-analyses here report pooled SMDs of roughly 0.9 to 1.1 against non-active controls — genuinely large. Look at what they were measured against. An SMD is a distance between two groups, one of them a control condition the researchers chose. A trial against a wait-list is not asking "does exercise treat depression?" It asks whether a structured program with trainers and a schedule beats waiting while knowing you got nothing.

There is also a structural problem no design can fix: you cannot blind someone to exercise. Participants know they are moving, expectancy is a real ingredient of mood change, and most outcomes are self-rated scales — so expectancy lands on the instrument. The effects aren't fake; the honest size emerges only when you sort trials by how fair their comparison was.

The Meta-Analyses, Ranked by Strictness

Four analyses carry most of the weight, and the oldest has just been replaced. The Cochrane review is the reference standard here: last fully updated in 2013 by Cooney and colleagues, replaced in January 2026 by Clegg and colleagues. Its pooled estimate for exercise against no treatment or a control condition — 57 trials, 2,189 participants — is an SMD of −0.67 (95% CI −0.82 to −0.52), rated low-certainty evidence. The strictest filter, the seven trials with adequate allocation concealment, blinded outcome assessment and intention-to-treat analysis, gives a smaller −0.46 (95% CI −0.88 to −0.04): still favouring exercise, but ending at the edge of significance. The 2013 version of that filter returned −0.18 (95% CI −0.47 to 0.11), crossing zero. Same literature, stricter lens, half the effect. Exercise is also statistically indistinguishable from psychological therapy (SMD 0.03, 95% CI −0.16 to 0.23) and from medication (SMD −0.11) in the few head-to-head trials.

Schuch and colleagues (2016) pooled 25 trials totaling 1,487 adults against non-active controls, reporting it as a positive SMD of 0.98 (95% CI 0.68 to 1.28) in exercise's favour — and 1.11 (0.79 to 1.43) after trim-and-fill adjustment for publication bias, with a fail-safe number of 1,057 null studies. Heissel and colleagues (2023), the largest synthesis restricted to diagnosed or above-threshold depression, pooled 41 trials: SMD −0.95 (95% CI −1.18 to −0.71) overall, −1.03 (−1.28 to −0.77) for supervised programs, −0.67 (−0.99 to −0.34) in low-risk-of-bias trials, and a number needed to treat of 2 overall, 2.8 in that low-risk subset. The same review's clinical units: a mean difference of −4.7 Hamilton points (95% CI −6.3 to −3.2), past the three-point change NICE calls clinically meaningful. Finally Noetel and colleagues (2024, BMJ) networked 218 trials with 14,170 participants against active controls — usual care, placebo pills: walking or jogging −0.62 (95% credible interval −0.80 to −0.45), yoga −0.55, strength training −0.49, with effects proportional to prescribed intensity. Their candor is the story: exactly one trial met Cochrane low-risk-of-bias criteria, and formal confidence rated low for walking or jogging, very low for everything else.

The Deflation Culprits, One by One

• 🎭 Blinding is impossible — participants always know they are exercising, and expectancy flows straight into the self-rated scales.
• ⏳ The wait-list artifact — controls assigned to nothing know it, and the comparison partly measures hope versus its absence.
• 🧑‍🤝‍🧑 Volunteers and adherers — trial samples are people who raised a hand for an exercise study; those who finish are the ones it suited.
• 📏 Small trials, hot heterogeneity — Heissel 2023 reported I² near 82%, many trials are small, and funnel-plot asymmetry flagged likely publication bias.
• 📆 End-of-treatment snapshots — week-12 scores dominate; the current Cochrane long-term pool (nine trials) finds −0.53, an interval crossing zero, very low certainty.

Exercise versus Medication: The SMILE Trial

The most useful single trial is SMILE — Blumenthal and colleagues (2007), 202 adults with major depressive disorder randomized for 16 weeks to supervised group exercise, home-based exercise, sertraline (50–200 mg), or placebo pill. At four months, remission rates were 45% supervised exercise, 40% home exercise, 47% sertraline, 31% placebo. Exercise and medication were statistically indistinguishable; the active treatments together only trended above placebo (p=.057). That placebo arm is the honesty dividend: nearly a third remitted on expectations, contact, and time — the machinery exercise trials absorb.

The one-year follow-up of the same cohort (Hoffman and colleagues, 2011) complicates the simple version. Remission among available participants had risen to 66%, and neither original assignment nor follow-up antidepressant use predicted being well. What did predict it — in within-trial observational analysis, association rather than randomized proof — was regular exercise during the follow-up year, concentrated between zero and roughly 180 minutes a week.

What Guidelines Actually Do With This

No major guideline says "exercise instead of treatment." NICE's 2022 depression guideline lists a supervised group exercise programme among its first-line options for less severe depression, and includes it for more severe depression too — with the explicit caveat that treatments offering more therapist contact should be considered first. The 2024 network analysis reaches the same place: these forms of exercise "could be considered alongside psychotherapy and antidepressants." Two findings matter for acting on it. Modality: walking and jogging carry the most evidence, strength training and yoga close behind — the Walking pillar covers the lowest-barrier entry, and strength training after 40 covers the iron. Intensity and supervision: effects scaled with prescribed intensity, and supervised or group programs outperformed unsupervised ones — one reason structured plans like the cardio conditioning and resistance training protocols beat vague intentions. For the mechanisms, see BDNF and the hippocampus trials; for the pairing, meditation and mindfulness, plus the quarterly life-stress load audit.

Safety: Where This Page Stops

Depression is a clinical condition, and this page exists to inform the conversation with a clinician — never to replace care. Certain signals change the project entirely: thoughts of suicide or self-harm, inability to function at work or in caregiving, symptoms worsening week over week, psychosis, or a spiral of alcohol or substance use deserve professional evaluation now, not after a training block. If you are in the United States and in crisis, call or text 988 (the Suicide & Crisis Lifeline); elsewhere, contact your local crisis line or emergency services, and go to an emergency department if danger feels immediate.

Two more boundaries. Never stop or adjust antidepressant medication in order to start exercising — medication changes belong with the prescriber, and unsupervised discontinuation is where relapse risk concentrates. Treat inability to exercise as data, not failure: severe depression drains the motivation and energy a program requires. Treatment first, movement as capacity returns. Watch, too, for exercise turning compulsive or entangled with guilt — worth naming to a clinician.

Questions, Answered Briefly

• 💊 "Is exercise as good as antidepressants?" In the few head-to-head trials, comparable — with wide uncertainty, and no reason to drop a medication that works.
• 🤔 "Why did my doctor call the evidence weak?" Effect size and certainty are different axes: moderate effects carried by trials with low certainty ratings.
• 🚶 "What kind, and how hard?" The best-attested menu is unglamorous: brisk walking or jogging, strength training, or yoga at moderate-to-vigorous intensity, supervised or in a group where possible — those were the formats with the larger effects.
• Severity? Most trials recruited volunteers well enough to exercise, with severe, disabling depression largely excluded; that needs clinical care first — see the sibling limits page.

Related Topics