You have been told that running protects your brain from dementia. The observational studies do say that. The randomised trials — which is where you find out whether something actually works — have almost all come back empty. This is the widest gap between popular claim and evidence anywhere in running, and it is worth understanding, because what running does do for your head is still substantial.
What the observational studies found
Pooling 49 of 58 studies, physically active people had about 20% lower dementia risk (RR 0.80, 95% CI 0.77 to 0.84) across 257,983 people. That is a real, large, consistently replicated association, and it is where every headline comes from.
Then someone checked whether it survives the obvious objection.
The study that changed the picture
People in the early stages of dementia stop moving — years before anyone diagnoses them. So if you measure activity once and count dementia cases a few years later, you will find that inactive people get more dementia, and you will have measured the disease causing the inactivity, not the reverse.
A 2019 analysis tested this directly across 19 cohorts, 404,840 people and 6 million person-years:
- Where dementia was diagnosed within 10 years of measuring activity: HR 1.40 (95% CI 1.24 to 1.59) — inactivity looks harmful.
- Where dementia was diagnosed 10 or more years later: HR 1.01 (95% CI 0.89 to 1.14) — nothing at all.
What makes this decisive rather than merely suggestive is the built-in control. In the very same people, using the very same activity measurements, physical inactivity stayed associated with diabetes (HR 1.42), heart disease (1.24) and stroke (1.16) — with no decay over time whatsoever.
If the dementia finding were fading because old questionnaires get less accurate, the diabetes and heart disease findings would fade too. They do not. The decay is specific to dementia, which is exactly the fingerprint of the disease causing the inactivity.
The authors’ conclusion: “When reverse causation bias was taken into account, physical inactivity was not associated with all-cause dementia or Alzheimer’s disease.”
A separate study followed 10,308 civil servants for 27 years and found the same: HR 1.00 (95% CI 0.80 to 1.24). It also mapped the prodromal decline precisely — activity in people who went on to develop dementia began dropping about nine years before diagnosis, with no difference at all between 28 and 10 years before.
The finding that should end the argument
If moving your body protected your brain, then people who move all day at work would be protected. A 2026 meta-analysis of 74 studies and 4,227,297 people looked at this by domain:
- Leisure-time activity: RR 0.76 (0.65 to 0.88) — protective
- Occupational activity: RR 1.20 (1.01 to 1.42) — higher risk
- Commuting activity: RR 1.10 (1.03 to 1.18) — higher risk
Manual labourers move far more than office workers who jog at weekends. They are not protected; the point estimate runs against them. What tracks with lower dementia risk is not physical movement — it is discretionary leisure exercise, which is also a marker of education, income, autonomy, and having the health and time to choose it.
Twin studies point the same way. Comparing 21,791 twins, the protective association was there across the whole cohort (HR 0.65) and vanished when comparing twins against their own siblings (HR 0.48, 95% CI 0.17 to 1.32 — not significant). Take away shared background and the effect goes with it.
The trials
Every adequately sized randomised trial of exercise for cognition or dementia has come back null:
- LIFE — 1,635 sedentary adults, 24 months. Cognition unchanged. Incident dementia or MCI: OR 1.08 (0.80 to 1.46).
- Generation 100 — 945 adults, five years, the longest exercise trial ever run. Cognitive score β 0.26 (−0.17 to 0.69). MCI OR 0.86 (0.66 to 1.13). Both null.
- EXERT — 296 people with mild cognitive impairment, 18 months, high adherence. Primary outcome p = 0.3. Hippocampal volume fell 0.51% in both groups.
- DAPA — 494 people with dementia. The exercise group’s cognition came out slightly worse (adjusted difference −1.4, 95% CI −2.6 to −0.2) even though fitness clearly improved.
- preDIVA — 3,526 people, six years. Dementia HR 0.92 (0.71 to 1.19).
The most-cited positive trial, FINGER, did find a benefit — 0.022 standard deviations per year — from a package that combined diet, cognitive training and vascular monitoring alongside exercise, so it cannot tell you the exercise did anything.
Worth noting in fairness: dementia prevention as a whole has almost no successful trials. Intensive blood pressure control missed its primary endpoint. Hearing aids showed a cognitive difference of 0.002. Exercise is not uniquely unsupported — the whole field is.
How big is the cognitive effect in healthy people?
An umbrella review examined 24 meta-analyses covering 109 randomised trials and 11,266 healthy participants, and watched the effect shrink as each source of bias was handled:
- Raw: d = 0.22 (0.16 to 0.28)
- Accounting for active control groups and baseline differences: d = 0.13 (0.07 to 0.20)
- Correcting for publication bias: d = 0.05 (−0.09 to 0.14) — no longer distinguishable from zero
Two details explain why. Only 2 of the 109 trials were adequately powered — detecting an effect that size needs about 650 people, and the median trial had 48. And the result moved between d = −0.05 and d = 0.47 depending purely on analytic choices.
The authors’ conclusion: “caution in claims and recommendations linking regular physical exercise to cognitive benefits in the healthy human population until more reliable causal evidence accumulates.”
A Cochrane review supplies the cleanest single refutation. Across 12 trials, eight successfully improved cardiorespiratory fitness — and “this improvement did not coincide with improvements in any cognitive domains assessed.” The intervention worked on the proposed mechanism and produced nothing downstream.
About the famous hippocampus study
You have seen the claim that exercise grows your hippocampus and reverses one to two years of brain ageing. It comes from a 2011 trial of 120 analysed participants. Two things about it are rarely mentioned:
- The memory comparison was null. The paper states it plainly: the exercise group “did not improve performance above that achieved by the stretching control group” (P < 0.40). Both groups improved, with no statistically significant difference between them. The stretching group’s raw gain was larger, but it also started from a lower baseline, and the interaction test is the null one quoted above — so this is not evidence that stretching beat exercise.
- The BDNF comparison was also null (P < 0.23).
So the study that founded the “exercise grows your brain” story found a volume change with no accompanying memory benefit.
The replication picture is worse. A 2021 meta-analysis of 23 trials and 1,204 participants — by the original author’s own group — found a pooled effect of g = 0.13 that decomposed unhelpfully: the control groups declined significantly (−0.72%) while the exercise groups’ own change was not statistically significant (p = 0.300). Neither hemisphere reached significance alone. One bias test was significant and another was not, and the trim-and-fill correction actually nudged the pooled estimate up (g = 0.14, 95% CI 0.04 to 0.25) rather than down. Removing the 2011 study eliminated statistical significance rather than the effect itself.
On BDNF, the protein everyone invokes: exercise does raise it in your blood. Whether blood levels reflect brain levels in humans has never been established — the paper always cited for it contains no human data at all (rats, pigs and mice), and about 99% of circulating BDNF sits in platelets, which do not cross into the brain. A post-run rise is entirely compatible with platelets releasing their contents and nothing happening in your head.
What running genuinely does do
This is not an article arguing you should stop. Three things hold up, and two of them hold up well.
Mood. In a network meta-analysis of 218 studies and 14,170 participants, walking or jogging came out at g = −0.62 (95% CrI −0.80 to −0.45) for depression — comparable to cognitive behavioural therapy (−0.55) and larger than SSRIs alone (−0.26) in the same comparison.
Be honest about the caveat: only one of those 218 studies met the criteria for low risk of bias, and you cannot blind someone to whether they are exercising. When Cochrane restricted to well-conducted trials, the effect fell from −0.67 to −0.46 (95% CI −0.88 to −0.04) — still real, but smaller and only just clearing zero. The best pragmatic trial, which successfully got 361 primary-care patients moving more, found no improvement in depression at all.
And there is a genuinely useful prevention finding: across 33,908 adults followed 11 years, about 12% of future depression cases might have been prevented by one hour of activity a week. Most of that protection appeared at low levels, regardless of intensity, with no further gain from doing more. It protected against depression but not anxiety.
Sleep. Exercise improves sleep quality by 2.87 points on the Pittsburgh index (95% CI −3.95 to −1.79), on a scale where above 5 counts as poor sleep. That is a meaningful change, though it is self-reported.
Living longer. This is the best-evidenced benefit of all, and it is covered in our guide on running and longevity.
What this means in India
The Indian picture is genuinely different, and it matters.
Dementia affects 7.43% of Indians over 60 — about 8.8 million people, projected to reach 16.9 million by 2036. Prevalence rises from 2.94% at 60–64 to 25.41% above 85.
The dominant risk factor is not physical. It is education: prevalence is 10.29% among those with no schooling and 1.54% among those with middle school or more. When researchers adjusted for age and education, the apparent sex difference and most state differences disappeared.
An Indian analysis calculated population attributable fractions using relative risks measured in India rather than imported from the West:
- No education: 22.0%
- Vision impairment: 14.1%
- Physical inactivity: 12.5%
- Social isolation: 8.2% · Hypertension: 5.0% · Being underweight: 4.6%
Three things to take from that. Physical inactivity matters more in India than the global Lancet Commission estimate of 2% — but here “inactive” means hardly ever or never doing any moderate or vigorous activity. That is about getting a completely sedentary person to move at all, which anyone reading a running club’s website has already done. The marginal brain benefit of your next 10 km is essentially unmeasured.
Second, being underweight is a risk factor here, not overweight — the opposite of the Western story.
Third, Western risk weights demonstrably do not transfer: applying high-income-country figures to the same Indian sample gave 59.3% against the India-derived 70.3%.
And in India’s own national survey of 59,813 adults, vigorous physical activity showed no association with cognitive impairment at all (APR 1.07, 95% CI 0.94 to 1.22). There is no Indian randomised trial and no Indian prospective cohort on exercise and cognitive decline.
Where the field genuinely disagrees
We should not pretend this is settled. The strongest counter-argument comes from a 2022 meta-analysis of 58 studies which found the association survived at long follow-up — RR 0.79 (0.71 to 0.87) beyond 20 years — with no evidence that follow-up length moderated it. Its authors concluded they found no sign that reverse causation explained their results.
But note what those same authors concede against their own position: “higher cognitive ability or other unmeasured confounding factors, and not leisure-time PA, may drive the association,” because few studies adjusted for baseline cognitive ability — which is the single best predictor of later cognition.
So there are two separate arguments, and they should not be conflated. Reverse causation is the better explanation for the short-follow-up signal. Confounding by education and cognitive reserve is the better candidate for the long-follow-up one. Both point away from running being the active ingredient.
The honest bottom line
Run because it will very likely help you live longer, because it is one of the better-evidenced things you can do for low mood, because it reliably improves your sleep, and because you enjoy it.
Do not run because you have been told it will save you from dementia. That claim outran its evidence, and every trial that tested it directly came back empty.
If dementia is what worries you — for yourself or a parent — the Indian data points somewhere specific and it is not the pavement: education and mental engagement, getting eyes tested and cataracts fixed, staying socially connected, treating blood pressure, and not being underweight. Those are the big levers here. Getting a genuinely sedentary relative moving at all is worth a great deal. Your next long run, for your brain specifically, is worth much less than you have been led to believe.
References
- Kivimäki M, et al. Physical inactivity, cardiometabolic disease, and risk of dementia: an individual-participant meta-analysis. BMJ, 2019;365:l1495. doi:10.1136/bmj.l1495
- Sabia S, et al. Physical activity, cognitive decline, and risk of dementia: 28 year follow-up of Whitehall II cohort study. BMJ, 2017;357:j2709. doi:10.1136/bmj.j2709
- Feter N, et al. Domain-specific physical activity levels and risk of dementia: a systematic review and meta-analysis. Lancet Public Health, 2026. doi:10.1016/S2468-2667(26)00142-8
- Ciria LF, et al. An umbrella review of randomized control trials on the effects of physical exercise on cognition. Nat Hum Behav, 2023;7(6):928–41. doi:10.1038/s41562-023-01554-4
- Young J, et al. Aerobic exercise to improve cognitive function in older people without known cognitive impairment. Cochrane Database Syst Rev, 2015;4:CD005381. doi:10.1002/14651858.CD005381.pub4
- Sink KM, et al. Effect of a 24-month physical activity intervention vs health education on cognitive outcomes (LIFE). JAMA, 2015;314(8):781–90. doi:10.1001/jama.2015.9617
- Zotcheva E, et al. Effects of 5 years aerobic exercise on cognition in older adults: the Generation 100 study. Sports Med, 2022;52(7):1689–99. doi:10.1007/s40279-021-01608-5
- Baker LD, et al. Effects of aerobic exercise on cognition in older adults with mild cognitive impairment (EXERT). Alzheimers Dement, 2025;21(4):e14586. doi:10.1002/alz.14586
- Lamb SE, et al. Dementia And Physical Activity (DAPA) trial. BMJ, 2018;361:k1675. doi:10.1136/bmj.k1675
- Erickson KI, et al. Exercise training increases size of hippocampus and improves memory. PNAS, 2011;108(7):3017–22. doi:10.1073/pnas.1015950108
- Wilckens KA, et al. Exercise interventions preserve hippocampal volume: a meta-analysis. Hippocampus, 2021;31(3):335–47. doi:10.1002/hipo.23292
- Noetel M, et al. Effect of exercise for depression: systematic review and network meta-analysis. BMJ, 2024;384:e075847. doi:10.1136/bmj-2023-075847
- Clegg AJ, et al. Exercise for depression. Cochrane Database Syst Rev, 2026;1:CD004366.
- Harvey SB, et al. Exercise and the prevention of depression: results of the HUNT cohort study. Am J Psychiatry, 2018;175(1):28–36. doi:10.1176/appi.ajp.2017.16111223
- Banno M, et al. Exercise can improve sleep quality: a systematic review and meta-analysis. PeerJ, 2018;6:e5172. doi:10.7717/peerj.5172
- Iso-Markku P, et al. Physical activity as a protective factor for dementia and Alzheimer’s disease. Br J Sports Med, 2022;56(12):701–9. doi:10.1136/bjsports-2021-104981
- Baumeister SE, et al. Physical activity and risk of Alzheimer disease: a Mendelian randomization study. Neurology, 2020;95(13):e1897–905. doi:10.1212/WNL.0000000000010013
- Livingston G, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. Lancet, 2024;404(10452):572–628.
- Lee J, et al. Prevalence of dementia in India: national and state estimates from a nationwide study. Alzheimers Dement, 2023;19(7):2898–912. doi:10.1002/alz.12928
- Angrisani M, et al. Modifiable risk factors for dementia in India. BMJ Public Health, 2024;2(2):e001362. doi:10.1136/bmjph-2024-001362