Which Type of Exercise Actually Helps Anxiety

"Exercise helps anxiety" is true but not very actionable on its own — which type, and how much? A 2026 network meta-analysis set out to answer exactly that. It has a clearer answer than most reviews in this space, and also more caveats than the headline version carries.

The evidence base

Lei and colleagues, publishing in BMJ Open Sport & Exercise Medicine, ran a frequentist network meta-analysis of 30 randomized controlled trials in 1,421 participants, searching five databases to March 2025.¹ They deliberately split diagnosed anxiety disorders from elevated anxiety symptoms that don't meet a diagnostic threshold — a distinction most previous reviews blurred, and one that turns out to change the answer. The effectiveness network for diagnosed anxiety disorders covered 760 participants; the network for anxiety symptoms covered 578; the combined-conditions network covered 1,372.

A network meta-analysis (NMA) differs from a standard trial by pooling every direct and indirect comparison across all interventions in the literature — resistance vs. control, mind-body vs. control, aerobic vs. resistance, and so on — into one model, then ranking every option on the same scale, here against a waitlist reference. Each ranking carries a probability score (SUCRA) and a GRADE-style certainty rating (CINeMA), telling you not just what came out on top but how much to trust that ranking. Two honest limits before the numbers: on the Cochrane risk-of-bias tool, only 4 of the 30 trials were rated low risk, 19 had some concerns and 7 were high risk, and nearly half didn't use intention-to-treat analysis. And several arms in the tables below rest on zero direct head-to-head comparisons — their estimates come entirely from indirect evidence through the network.

What came out on top

For diagnosed anxiety disorders, resistance training produced a moderate-certainty effect size of SMD −0.79 (95% CI −1.18 to −0.40) — essentially on par with third-wave cognitive behavioural therapy (SMD −0.75, −1.15 to −0.36) in this analysis, with mind-body exercise close behind (−0.71, −1.23 to −0.19). In the wider comparison across both diagnosed and subclinical anxiety, resistance training (−0.80, −1.24 to −0.36) and mind-body exercise (−0.78, −1.12 to −0.44) were the two strongest performers, both at moderate certainty. Network meta-regression found that diagnostic status did not alter the treatment effects, with low heterogeneity (I²=2%).

One finding the headline version of this study tends to lose: resistance training's first-place ranking did not survive either sensitivity analysis. Excluding trials with unclear diagnostic tools, mind-body exercise rose to the top (SMD −1.04, 95% CI −1.51 to −0.56) with resistance training second (−0.95, −1.52 to −0.39). Excluding high-risk-of-bias trials, resistance training's effect size and ranking declined further (−0.84, −1.53 to −0.14) while mind-body exercise held first place (−0.89, −1.41 to −0.37). So the fair reading isn't "lifting beats yoga": it's that resistance training and mind-body practice are the two moderate-certainty leaders, and which of them ranks first depends on which trials you're willing to include.

The full ranking: where aerobic exercise (and everything else) placed

Most public health advice defaults to "cardio" as the anxiety-fighting exercise. This particular ranking doesn't back that up — though, as the intervals show, several of the lower-ranked arms can't be distinguished from the waitlist reference at all. Here's the complete order for diagnosed anxiety disorders (n=760), with confidence intervals:

A few things stand out. Aerobic exercise — the default "go for a run" advice — placed fifth, but it is one of only four arms whose interval clearly excludes zero, and at moderate certainty: on that basis it has stronger support than standard CBT's fourth-place point estimate, which doesn't. Stretching's small estimate and the inactive-placebo arm's positive one are both statistically indistinguishable from the waitlist reference, so neither supports a conclusion in either direction — which is why we removed the claim that the placebo result showed "waiting and hoping" makes anxiety worse. This network also didn't isolate high-intensity interval training as its own category; HIIT protocols in the underlying trials were most likely folded into the aerobic or multicomponent groupings, so no separate HIIT ranking exists here.

Where the picture shifts

For subclinical anxiety symptoms specifically — not a diagnosed disorder — mind-body exercise ranked highest, though only at low certainty, and resistance training showed no significant effect in that group. The type of exercise that helps most may depend on how severe the anxiety actually is — a distinction we unpack separately in why anxiety the feeling and anxiety the disorder aren't the same thing.

In that subclinical-symptoms network (n=578), the ordering was: mind-body exercise (SMD −0.84, 95% CI −1.48 to −0.20, low certainty), psychoeducation (−0.65, −2.10 to 0.81, very low), resistance training (−0.64, −2.34 to 1.05, very low), aerobic exercise (−0.62, −1.82 to 0.58, very low) and multicomponent exercise (−0.40, −1.20 to 0.41, very low).¹ Only mind-body exercise reached significance; every other arm's interval crossed zero, and this network carried high heterogeneity (τ²=0.36). Resistance training's point estimate there isn't trivial — but with an interval running from −2.34 to 1.05, the analysis cannot treat it as a reliable effect once anxiety drops below a diagnosable threshold. That's a genuinely different conclusion than the diagnosed-disorder ranking above, and it's the kind of nuance that gets lost when "exercise helps anxiety" gets compressed into a headline. For where non-exercise options sit in the same hierarchy, see what actually works for anxiety, in order.

The exercise type that ranks highest for a diagnosed anxiety disorder is not necessarily the one that ranks highest for everyday anxious symptoms — severity changes the answer, and so does which trials you include.

Why type might matter: the mechanisms

None of this happens by magic, and the trial authors point to a few candidate biological pathways — none proven as the definitive mechanism, but plausible enough to explain why different exercise types might work through different routes.

For resistance training, the leading hypotheses involve insulin-like growth factor 1 (IGF-1), which recent meta-analyses indicate rises significantly after resistance training in women but not in men, and which is thought to have anxiolytic effects partly by regulating FKBP5, a gene involved in how stress-hormone receptors function.¹ Resistance training is also linked to hippocampal neurogenesis — new neuron growth in a region tied to stress regulation and fear memory — plus the simpler psychological route of visible strength gains improving self-efficacy and body image. The review's authors lean on the sex-specific IGF-1 finding deliberately, because 88.6% of participants in the resistance-training comparisons were women. A separate, well-established line of research backs part of this picture: a 2015 meta-analysis of 29 studies and 1,111 participants found exercise reliably raises brain-derived neurotrophic factor (BDNF) — a moderate effect after a single session (Hedges' g=0.46), a larger one when that session follows a regular training programme (g=0.59), and a small effect on resting BDNF after regular exercise (g=0.27).² Its one significant moderator was sex: studies with more women showed less BDNF change, which sits awkwardly alongside the IGF-1 story above rather than neatly confirming it.

Mind-body exercise appears to work through a different route: parasympathetic activation and increased GABA levels in the thalamus, alongside a possible dampening of HPA-axis (hypothalamic-pituitary-adrenal) reactivity — though the authors note the HPA-axis evidence specifically remains inconsistent and needs more work.¹ The cortisol part of that story is more tangled than it's usually told, as we cover in the cortisol story and in what magnesium can and can't do for the stress axis. There's also overlap with interoceptive exposure — a well-established idea from anxiety treatment more broadly, not a claim from this analysis — where controlled, low-threat exposure to physical sensations resembling anxiety symptoms helps reduce fear of those sensations over time. Yoga and tai chi pair slow, deliberate breathing with sustained postures, which is one reason they may work even though the review classifies mind-body exercise as low-intensity activity combined with mindfulness and meditation.

How much, and how often

Knowing "resistance training helps" isn't very actionable without a dose. Across the pooled trials, the average session ran 43.2 minutes, three times weekly, over an 8-week period; the resistance-training programmes averaged 6.4 weeks, and study durations ranged from 2 weeks to over 12, with the anxiety-symptom trials spanning 4–16 weeks.¹ Intensity was predominantly moderate — 3.5 to 6.5 METs — and for resistance training specifically the review prioritised the original reports' figures of roughly 60–70% of one-repetition maximum. Around 80% of participants started out untrained, so these effects reflect beginners starting from close to zero, not athletes refining an existing programme. The sample was 72.3% female with a mean age of 36.3, worth flagging as a limit on generalizing to older adults or men specifically. Only 12 of the 30 trials reported any follow-up data at all, so how durable these effects are once training stops is still an open question.

What this doesn't mean

This is comparative trial evidence on exercise types, not a claim that exercise treats or replaces treatment for an anxiety disorder — a moderate-certainty SMD is a real, meaningful signal, not a guarantee for any one person, and only 4 of these 30 trials were low risk of bias. Anyone with a diagnosed anxiety disorder should treat this as one input alongside clinical care, not a substitute for it. Supplements sit further down the same list again — see magnesium and anxiety and what the chamomile and passion flower trials actually found for how much thinner that evidence is.

How this lines up with what helps depression

Proco's companion piece on exercise and depression covers a separate, much larger network meta-analysis — Noetel and colleagues, published in The BMJ in 2024, pooling 218 studies, 495 arms and 14,170 participants.³ Compared with active controls, dance came out on top (Hedges' g −0.96, 95% credible interval −1.36 to −0.56), ahead of walking or jogging (−0.63, −0.80 to −0.46), yoga (−0.55, −0.73 to −0.36), strength training (−0.49, −0.69 to −0.29) and tai chi or qigong (−0.42, −0.65 to −0.21). Two caveats belong with the dance figure, because it's the one that travels furthest online: it rests on 5 studies and 107 participants, and only one of the 218 included studies met Cochrane's criteria for low risk of bias, so CINeMA confidence was low for walking or jogging and very low for everything else, dance included. The authors' own conclusion names walking or jogging, yoga and strength training — not dance — as the modalities more effective than others. Even with those caveats, it's a genuinely different leaderboard: resistance training tops the anxiety-disorder ranking above, but for depression it lands fourth, and dance isn't a category in the anxiety network at all. Mind-body practice is the one throughline, ranking strongly in both. Our full write-up of that analysis goes through the dose-response findings.

The practical takeaway

If exercise is going into a mental health toolkit and there's a choice about what kind, the trial evidence favours resistance training and structured mind-body practice over generic "just get moving" advice — the type of exercise chosen isn't incidental to the effect, even if the exact ordering isn't settled. As a starting dose, something in the neighbourhood of three 40-to-45-minute sessions a week at moderate intensity, sustained for at least 8 weeks, is roughly what generated the effect sizes above — a realistic bar for someone starting from scratch, not an elite training block. And if you're weighing exercise against a bottle more generally, our honest comparison of the two reaches much the same conclusion for cognition.

This article is for informational purposes only and is not medical advice. It is not intended to diagnose, treat, cure, or prevent any disease, including anxiety disorders. Speak with a healthcare provider about your specific situation, especially if you have a diagnosed anxiety condition.

References

1. Lei EF, Wan K-W, Dai Z, Tam BT (2026). Effectiveness and acceptability of exercise treatments for adults with anxiety disorders: a systematic review and network meta-analysis. BMJ Open Sport & Exercise Medicine, 12(1):e002781. pmc.ncbi.nlm.nih.gov/articles/PMC12820870
2. Szuhany KL, Bugatti M, Otto MW (2015). A meta-analytic review of the effects of exercise on brain-derived neurotrophic factor. Journal of Psychiatric Research, 60:56-64. pubmed.ncbi.nlm.nih.gov/25455510
3. Noetel M, Sanders T, Gallardo-Gómez D, et al. (2024). Effect of exercise for depression: systematic review and network meta-analysis of randomised controlled trials. The BMJ, 384:e075847. pubmed.ncbi.nlm.nih.gov/38355154

Want to know what's actually in what you take? Join the list — no spam, just it straight.