Exercise, asked three ways
90 min
Two hosts talk the lesson through. The voices are synthetic; the script was written from this lesson and checked against it, and asserts nothing the lesson does not.
- State what the whole literature, the best-controlled trials and the head-to-head comparisons each show about exercise for depression, with their certainty
- Trace a press release's claim back through its paper and say what changed at each step
- Say what the evidence shows for exercise and anxiety, and what NICE does and does not recommend
If there's one piece of mental-health advice everyone has heard, it's that exercise helps. This lesson asks how much, compared with what, and how sure anyone can be, and finds that one question gets three different answers depending on which trials you look at.
This course is education, not care. If you're thinking about suicide or self-harm, or don't feel able to keep yourself safe, contact emergency services (911 in the US and Canada, 999 in the UK, 112 across the EU, 000 in Australia) or a crisis line: call or text 988 in the US and Canada, call Samaritans on 116 123 in the UK and Ireland, or Lifeline on 13 11 14 in Australia. Elsewhere, findahelpline.com lists free, confidential lines by country.
What NICE says
Start with the guideline, because it's more specific than the advice. For less severe depression, NG222's table of first-line options includes group exercise, which it defines like this: it "Uses a physical activity programme specifically designed for people with depression", "Usually consists of more than 1 session per week for 10 weeks", and "Includes moderate intensity aerobic exercise."1
That's a treatment: a designed programme, in a group, for weeks. NG222 also says something more general: "Advise people that doing any form of physical activity on a regular basis (for example, walking, jogging, swimming, dance, gardening) could help enhance their sense of wellbeing."1 The page calls that a consensus recommendation, meaning the committee's agreed view rather than a finding graded from trials, and its verb is "could".
For generalised anxiety disorder, lesson 4 showed that NICE's CG113 names no exercise treatment at all. Its one line about exercise is for panic disorder, from 2004: "The benefits of exercise as part of good general health should be discussed with all people with panic disorder as appropriate."2 That's advice to talk about exercise as part of general health, not a treatment for panic.
One question, three answers
The current Cochrane review of exercise for depression, by Clegg and colleagues, was published in 2026, with searches to November 2023. It included 73 trials. This course read its abstract, which answers the question three ways.3
Against no treatment or a control, all the trials. "For the 57 trials (2189 participants) comparing exercise with no treatment or a control intervention, the pooled SMD for depressive symptoms at the end of treatment was -0.67 (95% confidence interval (CI) -0.82 to -0.52; low-certainty evidence), showing that exercise may result in a reduction in depressive symptoms."3 SMD is the standardised mean difference that lesson 3 used, and the minus sign means depression went down. "Low-certainty" is the review's grade on the evidence, not on exercise, the kind of grade lesson 3 explained.
Against a control, only the best-run trials. "When we included only the seven trials (447 participants) with adequate allocation concealment, intention-to-treat analysis and blinded outcome assessment, the pooled SMD was smaller (SMD -0.46, 95% CI -0.88 to -0.04)."3
Before you read on. Those seven trials hid which group people would be in until they were enrolled, analysed everyone they randomised, and used assessors who didn't know who had exercised. Why might their answer be smaller than the whole literature's?
Show the answer
This course's reading of why: each of those three things closes a route by which a trial's result can be pushed up without the treatment doing anything more. People get steered into a group, drop-outs are quietly left out, an assessor who knows expects improvement. When those routes are closed, what's left is probably closer to the effect itself.7
Notice the interval too: from -0.04, barely above nothing, to -0.88, large. Seven trials can't pin the size down.
The earlier version of the same review, in 2013, ran the same check on six trials and found that "the pooled SMD for this outcome was not statistically significant (-0.18, 95% CI -0.47 to 0.11)."3 Since then, the review says, "The addition of 35 RCTs (at least 2526 participants) to this update has had very little effect on the estimate of the benefit of exercise on symptoms of depression."3 So the whole-literature figure has hardly moved in thirteen years, and the best-trials figure has gone from not significant to smaller but clear of zero.
Ten trials set exercise against psychological therapy, and five set it against medication. Before you read the results: clearly better, clearly worse, or about the same?
Show the answer
About the same, in both, though neither result is sure. The review's words are "probably little to no difference" against therapy, at moderate certainty, and "may be little to no difference" against medication, at low certainty. The quotations follow.
Head to head. "Ten trials (414 participants) compared exercise with psychological therapy, finding there is probably little to no difference in their effect on depressive symptoms at the end of treatment (SMD 0.03, 95% CI -0.16 to 0.23; moderate-certainty evidence)." And against medication: "Five trials (330 participants) compared exercise with pharmacological treatment, finding there may be little to no difference in their effect on depressive symptoms at the end of treatment (SMD -0.11, 95% CI -0.33 to 0.10; low-certainty evidence)."3 In these two rows the number is a gap between two treatments, so zero means exercise and the other treatment did about as well as each other, not that nothing worked.
The four answers from the 2026 Cochrane abstract, drawn to one scale.3 Dots are the estimates and bars their 95% intervals. The top two rows are exercise against a control; the bottom two are exercise against another treatment, where zero means no difference between them. The chart is this course's drawing of the review's figures.7
So: moderate against a control across everything, on low-certainty evidence, smaller and less certain in the best trials, and about the same as therapy or medication where they were compared directly. The review's conclusion puts it in one line: "Exercise may be moderately more effective than a control intervention for reducing symptoms of depression. Exercise appears to be no more or less effective than psychological or pharmacological treatments, though this conclusion is based on a few small trials."3 Two more things from the same abstract. The nine trials with long-term follow-up "provided very uncertain evidence about the effect of exercise on depressive symptoms", and "There did not appear to be a difference between exercise and other interventions in terms of treatment acceptability, as measured by participants completing the study (moderate to low certainty evidence)." Its conclusion adds: "Long-term follow-up was rare."3
Why every trial is at risk
One sentence in the review applies to all 73 trials: "we judged all studies to be at high risk of performance bias." Performance bias is what happens when the people in a trial, and the people running it, know who got the treatment, and that knowledge changes how they behave or what they report. And the reason every trial has it: "Blinding of those receiving and those delivering the interventions is inherently difficult".3 It's hard to give someone a placebo run. People know whether they're exercising, and so do the people running the sessions, and that knowledge can shape both how people feel and how they report it. That's also why the review adds that "Many trials used participant self-report rating scales, which have the potential to bias findings."3
Hard isn't the same as impossible. One of the reviews below ends: "To mitigate expectancy effects, future studies could aim to blind participants and staff."4 It's a proposal for future trials. Cochrane's judgement on all 73 trials says the trials so far haven't managed it.
The review also reports harms, which matters for a treatment people start on their own: "Adverse events were not common in any comparison, but included musculoskeletal injuries and depression affecting those undertaking exercise, and diarrhoea, sexual dysfunction and fatigue reported by those receiving sertraline."3 Sertraline is an antidepressant.
The same question, in other reviews
Other reviews read the same literature more warmly, and their case deserves its own words. Heissel and colleagues (this course read the abstract) compared exercise with non-exercising controls and found "large effects (standardised mean difference (SMD)=-0.946, 95% CI -1.18 to -0.71) favouring exercise interventions", and "moderate effects when analyses were restricted to low risk of bias studies (SMD=-0.666, 95% CI -0.99 to -0.34, k=12, NNT=2.8 (95% CI 1.94 to 5.22))". Here k is the number of studies, and NNT, the number needed to treat, is how many people would need the treatment for one more of them to improve. Their conclusion is firm: "Exercise is efficacious in treating depression and depressive symptoms and should be offered as an evidence-based treatment option focusing on supervised and group exercise with moderate intensity and aerobic exercise regimes." That's close to what NG222's group exercise row describes. They add: "The small sample sizes of many trials and high heterogeneity in methods should be considered when interpreting the results."4
Noetel and colleagues (abstract read here) ran a network meta-analysis of 218 studies, the kind of analysis lesson 2 described, which links trials through the comparison groups they share. "Compared with active controls (eg, usual care, placebo tablet), moderate reductions in depression were found for walking or jogging", and for yoga and strength training. Their conclusion: "Exercise is an effective treatment for depression, with walking or jogging, yoga, and strength training more effective than other exercises, particularly when intense." And: "These forms of exercise could be considered alongside psychotherapy and antidepressants as core treatments for depression." In the same abstract: "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 CINeMA was low for walking or jogging and very low for other treatments."4 CINeMA is a system for grading confidence in a network meta-analysis.
This course's summary of the three: they agree on the direction, that exercise helps against a control. They differ on how big, how sure, and what to call it, and none of them contradicts the Cochrane head-to-head finding.7
A claim, traced
Now follow one claim from a paper to a headline. Every step of it is free to read.
Step one, the paper. Singh and colleagues, in 2023, collected 97 reviews covering 1,039 trials. This course read the abstract and the Discussion. The abstract reports that "Most reviews (n=77) had a critically low A MeaSurement Tool to Assess systematic Reviews score", and that "Physical activity had medium effects on depression (median effect size=-0.43, IQR=-0.66 to -0.27), anxiety (median effect size=-0.42, IQR=-0.66 to -0.26) and psychological distress (effect size=-0.60, 95% CI -0.78 to -0.42), compared with usual care across all populations." Its conclusion is strong: "Physical activity should be a mainstay approach in the management of depression, anxiety and psychological distress."5
Three terms there. A critically low score means most of the reviews it collected were judged poorly conducted by a standard quality tool. The median effect is the middle one when all the reviews' effects are lined up in order. The IQR, the interquartile range, is the span of the middle half of them, so half the reviews found a depression effect between -0.27 and -0.66.
Step two, the discussion. In the body of the paper: "The effect size reductions in symptoms of depression (−0.43) and anxiety (−0.42) are comparable to or slightly greater than the effects observed for psychotherapy and pharmacotherapy (SMD range=−0.22 to −0.37)." And the next sentence: "Future research to understand the relative effectiveness of PA compared with (and in combination with) other treatments is needed to confirm these findings."5 PA is physical activity; the journal's reference numbers between the two sentences are left out.
Step three, the press release. The university's release: "physical activity is 1.5 times more effective than counselling or the leading medications."5
Where does 1.5 come from? The parts of the paper this course read do not print it. But the middle of the range -0.22 to -0.37 is about -0.295, and 0.43 divided by 0.295 is about 1.46. So the ratio looks like arithmetic on the paper's own figures. That arithmetic is this course's.7
Put the three steps side by side. What changed between the discussion and the press release, and why can't the paper's comparison support "more effective"?
Show the answer
A comparison across different studies, hedged as "comparable to or slightly greater than" and followed by a call for head-to-head research, became a single ratio and a verdict, "more effective". The paper's own caveat, that the comparison needed confirming, didn't reach the release.
It can't support "more effective" because of lesson 2's question: what did the comparison group get? Singh's -0.43 was measured against usual care. The -0.22 to -0.37 came from other reviews, of other trials, against their own controls. Dividing one by the other compares two different baselines as much as two treatments.
Singh's comparison was not itself a direct test. The trials that did set exercise against therapy and medication directly are in the Cochrane review, and they found "little to no difference". Sleep lesson 7 showed claims hardening on the way into a headline; here the same hardening happens to a comparison across studies. This course reports the wording at each step and reaches no judgement about anyone's intentions.7
And anxiety
The evidence on exercise for anxiety is much thinner than for depression, and the numbers show how much.6
Depression has a 73-trial Cochrane review. Before you read on, guess: how many people with an anxiety disorder or raised anxiety are in the biggest review of exercise for anxiety this course found?
Show the answer
675, in 15 studies. The other review below has 262 adults in six trials. The course's search found no Cochrane review of exercise for anxiety in adults.
A 2018 review by Aylett and colleagues (abstract read here) found "Fifteen studies were identified with a total of 675 patients. Nine trials had participants with diagnosed anxiety disorders and six trials had participants with raised anxiety on a validated rating scale." It compared exercise with a waiting list or with low-intensity exercise, and reported that "Aerobic exercise was effective in the treatment of raised anxiety compared to waiting list control groups (effect size - 0.41, 95% CI = - 0.70 to - 0.12)", and that "High intensity exercise programmes showed greater effects than low intensity programmes." Its conclusion: "Exercise programmes are a viable treatment option for the treatment of anxiety." And its limitation: "Conclusions were limited by the small number of studies and wide variation in the delivery of exercise interventions."6
A 2017 meta-analysis by Stubbs and colleagues (abstract read here) describes the field before it with one word: "The literature regarding exercise for people with established anxiety disorders is equivocal." Equivocal means the results pointed both ways. It pooled "six randomized control trials (RCTs) including 262 adults", comparing exercise with usual treatment or control conditions, and found a moderate effect, a standardised mean difference of -0.582. Its conclusion: "Our data suggest that exercise is effective in improving anxiety symptoms in people with a current diagnosis of anxiety and/ or stress-related disorders."6 One number in that abstract is wrong. Its printed confidence interval does not contain its own estimate, so one of the two is a typing error. The printed p-value of 0.02 fits an estimate of that size, so the interval is the likely misprint, and this course does not reproduce it. That reading is this course's.7
Singh's anxiety median, -0.42, is not a figure for anxiety disorders. Its abstract says the populations "included healthy adults, people with mental health disorders and people with various chronic diseases".5
This course's comparison: for people with an anxiety disorder or raised anxiety, the evidence is a few hundred people per review, in small trials, mostly against waiting lists, and lesson 2 showed that a waiting list may flatter a treatment. This course's search found no Cochrane review of exercise for anxiety in adults. For depression, there's a 73-trial Cochrane review. The direction for anxiety is favourable; the evidence is a fraction of the size.7
Four things people get wrong
"Exercise is 1.5 times more effective than therapy or medication." It's a memorable number from a university press release, which is why it spread. The paper said "comparable to or slightly greater than", across different studies, and asked for head-to-head trials. The head-to-head trials found little to no difference.
"Exercise is proven to treat depression." Against a control it helps, on low-certainty evidence, and less in the best-run trials. Heissel and Noetel do use firm words, "efficacious" and "an effective treatment", but "proven" goes beyond the Cochrane review's certainty grade, and those two abstracts carry the qualifiers quoted above.
"Exercise doesn't really help; it's all placebo." The opposite overreach. Even the seven best-run trials found a benefit, and the reviews agree on the direction. What's hard is blinding exercise, so how much of the effect is expectation hasn't been separated out.
"Any exercise counts as the treatment NICE lists." NG222's group exercise is a programme designed for depression, more than once a week for about 10 weeks, at moderate aerobic intensity. Its advice that any regular activity "could help" with wellbeing is a separate consensus recommendation.
Practice
Take 25 minutes.
Both are free. Open the university's press release and the Singh paper at PubMed Central. You've already seen the "1.5 times" sentence traced, so pick a different one.
Copy one other sentence from the release that says what the study found, word for word.
Find the passage in the paper it comes from: the abstract, the results or the discussion. If you can't find one, write that down; it's a finding too.
Write down every word that changed, and what each change does to the claim. Check the comparison group as well: what does the paper say the exercisers were compared with, and does the release say?
One line: would the paper's own sentence support the release's as written? Name the words that decide it.
Take 15 minutes.
Find the last claim you met about exercise and mood: an article, an app, a gym poster, a friend. Write down what it implies the exercisers were compared with, and whether it says. Then say which of this lesson's three answers the claim is closest to: against a control, in the best trials, or head to head.
Connections
Back. Lesson 2's question, what the comparison group got, splits this lesson's one question into three. Sleep lesson 7 gave you the shapes a claim takes on the way into a headline, and lesson 4 showed that NICE's GAD guideline names no exercise treatment.
Forward. Lesson 6 is mindfulness, behavioural activation and breathing, where the enthusiasm is similar and the comparisons matter just as much. Lesson 7 takes three popular ideas, resilience, a ratio and a mindset, where a claim is compressed even further on its way to the public.
Go deeper
- The Cochrane review, 2026. This course read the abstract. Its "Synthesis of results" paragraph is the whole lesson in a page.
- Noetel and colleagues, 2024, free at PubMed Central. Abstract only here. A favourable reading that compares types of exercise, with its qualifier in the same abstract.
Nothing in this lesson is a reason to start, stop or swap a treatment. NG222 lists group exercise among the first-line options for less severe depression. For more severe depression it comes last in the table, and its row adds: "In more severe depression, the potential advantages of providing other treatment choices with more therapist contact should be carefully considered first."1
If you take an antidepressant and are thinking of stopping it, NICE's advice is to "talk with the person who prescribed their medication" first, because "it is usually necessary to reduce the dose in stages over time". It adds that "most people stop antidepressants successfully", and a later recommendation, 1.4.14, warns that "withdrawal can sometimes be more difficult" for some people, with symptoms lasting weeks or occasionally months.1
Sources
- National Institute for Health and Care Excellence, NG222, 2022: Table 1's group exercise row, Table 2's group exercise row, and recommendations 1.4.12, 1.4.14 and 1.4.40. Read. Supports: the definition of group exercise, the note for more severe depression, the advice on stopping an antidepressant, and the consensus recommendation on physical activity.
- National Institute for Health and Care Excellence, CG113. Read. Supports: no exercise treatment for GAD (lesson 4), and recommendation 1.3.11 on exercise in panic disorder.
- A. J. Clegg and colleagues, "Exercise for depression", Cochrane Database of Systematic Reviews 2026, CD004366, doi 10.1002/14651858.CD004366.pub7; and its 2013 version, G. M. Cooney and colleagues, doi 10.1002/14651858.CD004366.pub6. Read: the abstracts. Supports: the three answers and the chart's figures, the 2013 comparison, long-term follow-up and acceptability, the conclusion, performance bias, self-report, and adverse events.
- A. Heissel and colleagues, British Journal of Sports Medicine 57(16), 2023, doi 10.1136/bjsports-2022-106282; and M. Noetel and colleagues, BMJ 384, 2024, e075847, doi 10.1136/bmj-2023-075847. Read: the abstracts. Supports: their figures, conclusions and qualifiers, Noetel's controls, and its sentence on blinding.
- B. Singh and colleagues, "Effectiveness of physical activity interventions for improving depression, anxiety and distress: an overview of systematic reviews", British Journal of Sports Medicine 57(18), 2023, doi 10.1136/bjsports-2022-106195. Read: the abstract and the Discussion. And the University of South Australia press release, "Exercise more effective than medicines to manage mental health", 23 February 2023, on EurekAlert. Read in full. Supports: the three steps of the trace, and the populations behind the anxiety median.
- E. Aylett and colleagues, BMC Health Services Research 18(1), 2018; and B. Stubbs and colleagues, Psychiatry Research 249, 2017. Read: the abstracts. Supports: the anxiety figures, comparators, conclusions and qualifiers, and the misprinted interval.
- This course's own constructions, labelled where they appear. The reading of why the best-run trials give a smaller answer, the chart, the one-line summary of what the reviews agree on, the arithmetic behind "1.5", the reading of the press release against the paper, which number in Stubbs's abstract is the misprint, and the comparison of the anxiety and depression evidence are this course's.
Check your understanding
This lesson has a 6-question quiz. Pass it and the questions come back on a schedule in Review, so what you learned stays learned. Your progress is saved in your browser; no account needed.