When Does Exercise Matter Most for Dementia Risk?

Exercise in your twenties may be good for your heart, mood, bones, and ability to carry groceries without negotiating with the stairs. But a new Framingham analysis asks a narrower question: which years of adult activity best predict lower dementia risk?

The answer, with important limits, is midlife and later life. In the 2025 Framingham Heart Study analysis, people who reported higher physical activity at ages 45–64 or 65–88 had about 40% lower subsequent risk of all-cause dementia and Alzheimer disease dementia than those in the lowest activity group. Activity reported at ages 26–44 showed no statistically significant association after adjustment for health, education, smoking, and APOE ε4 status.

That does not mean exercise before 45 has no effect on the brain. It means that, in this study, current or more recent activity was the more detectable predictor of future dementia. For anyone who spent early adulthood sedentary, that is a much more useful result than a lifetime fitness report card.

What Framingham actually found

Francesca R. Marino and colleagues used the Framingham Offspring cohort to examine self-reported physical activity at three adult stages. The early-adult group included 1,526 people, measured at ages 26–44 between 1979 and 1983. The midlife group included 1,943 people, measured at ages 45–64 between 1987 and 1991. The late-life group included 885 people, measured at ages 65–88 between 1998 and 2001.

Participants completed the Physical Activity Index, a self-reported estimate based on time spent in activities of differing intensity. The researchers then tracked incident dementia through December 2023: mean follow-up was 37.2 years for the early-adult analysis, 25.9 years for midlife, and 14.5 years for late life.

After adjusting for age, sex, body mass index, education, smoking, hypertension, diabetes, hyperlipidemia, and APOE ε4, participants in the fourth and fifth activity quintiles in midlife had hazard ratios of roughly 0.55 to 0.60 versus the least-active quintile for both all-cause dementia and Alzheimer disease dementia. The late-life results were nearly identical. A hazard ratio in that range means lower observed risk over follow-up, not a guarantee that exercise itself caused the difference.

Early-adult activity did not show a statistically significant association with either dementia outcome after full adjustment. The study therefore does not support the popular idea that a busy decade of recreational sports in your twenties permanently "banks" protection regardless of what follows.

The most practical signal is leaving the bottom tier

The Framingham pattern also suggests a plateau: the advantage in the two highest activity quintiles was broadly similar. The study did not establish an exact exercise prescription, and it cannot tell us how many weekly minutes produce a particular dementia-risk reduction. Still, its shape argues more for getting out of the least-active group than for treating maximal training volume as a cognitive insurance policy.

A 2024 network meta-analysis of exercise trials in people with Alzheimer disease or dementia reached a compatible, but indirect, conclusion: moderate-intensity programs produced cognitive outcomes statistically indistinguishable from high-intensity programs. Those participants already had dementia, so this is evidence about treatment-stage exercise, not proof of the ideal prevention dose for healthy adults.

Other cohorts point in the same direction

Framingham is unusually useful because it compares activity windows across adulthood. Its overall direction fits larger observational work. A prospective UK Biobank study of 431,924 participants found that higher physical activity was independently associated with lower incident dementia. It did not divide exposure into the same early-, mid-, and late-adult windows, so it cannot settle the question of whether activity in early adulthood has an independent effect.

In Brazil, the multicenter ELSA-Brasil cohort study followed 10,121 middle-aged and older adults for a median 8.1 years and identified 396 incident cases of cognitive impairment. Higher leisure-time physical activity and estimated cardiorespiratory fitness were associated with lower cognitive-impairment risk. That endpoint is cognitive impairment, not clinical dementia, but it extends the association beyond European and U.S. cohorts.

Fitness data add another piece. A 2025 community-based longitudinal study found that higher cardiorespiratory fitness was inversely associated with dementia risk across APOE ε4 genetic-risk strata. APOE ε4 raises Alzheimer disease risk; the association between fitness and lower dementia risk was still present across those strata. That is encouraging, but it does not show that exercise erases genetic risk.

Why exercise could help the brain

There are plausible human mechanisms. A 2024 meta-analysis of randomized trials found that at least four weeks of aerobic exercise training significantly increased hippocampal volume in healthy older adults. The hippocampus is heavily affected in Alzheimer disease, so preserving or increasing its volume is biologically relevant. It remains a surrogate outcome: a larger hippocampus does not automatically translate into fewer dementia diagnoses.

Similarly, a randomized trial in healthy older adults found aerobic exercise-related white matter plasticity. White matter supports communication among brain regions and is sensitive to vascular health. A separate 2025 prospective cohort linked leisure-time activity across life with slower epigenetic-aging acceleration and found the strongest cognitive benefits in mid-to-old-life windows. These findings make the Framingham timing pattern plausible, but they do not demonstrate one settled biological route from running shoes to dementia prevention.

The EXERT randomized trial adds direct biomarker evidence in adults with mild cognitive impairment. It tested aerobic exercise and assessed cognition along with Alzheimer-related cerebrospinal-fluid and plasma biomarkers. Because participants already had mild cognitive impairment, the trial informs disease-stage biology rather than primary prevention in cognitively healthy people.

Why the 40% figure is not a promise

The Framingham estimates are associations. People who exercise more often differ from less-active people in ways that standard statistical adjustment cannot fully capture: diet quality, social activity, income, healthcare use, sleep, and underlying health can all travel together.

A 2024 Cooper Center cohort study illustrates the problem. Midlife cardiorespiratory fitness and dietary patterns acted as co-exposures: people with higher fitness and healthier diets had better dementia outcomes than fitness alone would predict. Framingham adjusted for several major cardiovascular and genetic factors, but not dietary quality or social engagement. Its hazard ratios may therefore overstate the part attributable specifically to physical activity.

Reverse causation also matters most in the late-life analysis. Dementia pathology can reduce activity years before a diagnosis, so lower activity may sometimes be an early consequence of disease rather than its cause. The late-life cohort also included only people who survived and remained available for assessment at ages 65–88, which can distort comparisons.

Early-adult activity has the opposite timing problem. Framingham measured it once, then followed participants for more than 37 years on average. Someone highly active at 30 could become sedentary at 50; someone inactive at 30 could begin exercising at 45. Those unmeasured changes can wash out a real early-life association. The null result is informative, but it is not proof that early exercise is neurologically irrelevant.

Randomized trials have not closed the case

Randomized evidence has not yet shown that starting an exercise program in cognitively normal midlife or later life prevents dementia diagnoses. That would require a very large trial with long follow-up.

The most relevant experimental warning comes from a one-year randomized trial in cognitively normal older adults. Tarumi and colleagues tested aerobic exercise training and reported inconsistent evidence for neurocognitive benefit. This does not negate the observational studies or the imaging findings. It does mean that brain-structure and biomarker changes have not reliably produced measurable cognitive gains in the prevention-eligible population over that trial's duration.

What to do with this evidence

If you are in midlife or later life, the Framingham result supports a simple priority: become and remain physically active now. The strongest observed associations came from moderate-to-high activity categories, with no clear reason in these data to chase extreme volume for dementia prevention.

Midlife and later-life activity consistently track with lower dementia risk, but the studies do not prove that exercise alone produces the full observed reduction.

Aerobic activity has the best direct support for brain-structure outcomes, and improving cardiorespiratory fitness is a sensible target. Pair it with the unglamorous confounders that also matter—diet quality, blood-pressure control, sleep, social connection, and medical care—because real lives do not separate those variables as neatly as regression models do.

The useful conclusion is neither "your twenties determined your fate" nor "exercise guarantees a dementia-free old age." Framingham suggests that activity in midlife and later life remains meaningfully associated with lower risk, even after earlier decades have passed. That makes consistency more relevant than nostalgia.