RightRep EN · PT Open the app
BLOG

Does Exercise Prevent Dementia? The 45% Audit: The Disease Where the Lancet's Own Standing Commission Puts Inactivity on the List of 14 Modifiable Factors Behind 45% of Cases, the Brazilian-Led Meta-Analysis of 4,227,297 People Separating Chosen Movement (Leisure-Time RR 0.76) From Compelled Movement (Occupational RR 1.20 — the Wrong Direction), the UK Biobank Steps File at HR 0.49 for 9,826 Steps and 0.38 for a 112-Step Cadence, FINGER's Multidomain RCT Separating at 0.022 Points a Year, Erickson's Trial Where a Year of Aerobics Regrew the Aging Hippocampus by 2%, the Vancouver Barbell Trial That Lifted Executive Function 12.6% While Whole-Brain Volume Fell, and the Honesty Twins — LIFE's 1,635-Person Two-Year Null and EXERT's MCI Trial Where Neither Arm Declined

By RightRep AI Assistant · · 18 min read

There are questions this library answers with a trial, and questions it answers with a commission. For most conditions, the audit means pulling randomized trials; for dementia — the diagnosis patients say they fear more than any other, the one families spend decades watching the driveway for — the Lancet convenes a standing commission of leading specialists, refreshed every few years, whose 2024 headline number is the most quoted statistic in brain aging: 45% of dementias are potentially preventable by addressing 14 modifiable risk factors, and physical inactivity is on the list [1]. This library has already audited what training does for a healthy memory, the Parkinson's motor file next door, and the blood-pressure chapter of that same risk list; this is the dementia audit: what the file actually says exercise does about the disease — in risk ratios across millions of lives, regenerated hippocampal volume, Stroop points, steps per day, and the honest nulls.

One fence before any of it, and it is load-bearing. Dementia is a physician's diagnosis, and the medical layer stays first. Memory slips that start interfering with daily function — missed medications, lost routes, unpaid bills, personality change — are a reason to see a doctor this month, not a reason to buy walking shoes. Sudden confusion is an emergency, full stop: that is delirium or stroke territory, and the response is emergency services, not a training question. A diagnosis of mild cognitive impairment or dementia puts a neurologist or geriatrician in charge of evaluation and treatment; exercise in this file is the adjunct layer with its own measured effects — real ones, quantified below — stacked on top of medical care, never in place of it. And one more word of honesty about the title: "prevent" is the file's word, not a promise. Everything below is risk ratios — lower odds across populations — and the difference between those two things is exactly what this audit is for.

The short answer: the prevention file is strong, the treatment file is honest, and the difference is the whole post

Strip the file to its spine and three findings remain. First, the prevention file is the strongest in this entire library: across 74 studies and 4,227,297 people, high physical activity runs at RR 0.80 for all-cause dementia, 0.79 for Alzheimer's, 0.71 for vascular dementia — and when the analysis splits by where the movement happens, leisure-time activity runs at 0.76 while occupational activity points the wrong way at 1.20 [2]. Second, the brain responds like a muscle in the trials that measured it directly: a year of aerobic training regrew the aging hippocampus by 2% and reversed one to two years of age-related shrinkage [5], twelve months of once-weekly lifting improved executive-function test performance by 12.6% while the control group drifted down [6], and the field's flagship prevention RCT separated — by a small, honest margin [4]. Third, the boundary: in the biggest, longest, best-designed trials, exercise did not beat its own control arms on cognitive scores — 1,635 older adults over 24 months, null [7]; 296 people with diagnosed MCI over 18 months, no difference between aerobic and stretching [8]; the Cochrane in established dementia pooling 17 trials and calling the cognition evidence inconclusive at P=0.08 [9]. Prevention is one sport and management is another, and the honest version of this post coaches them separately.

What dementia actually is — and why the risk list reads like a training plan

The mechanics in one paragraph: dementia is not one disease but a clinical syndrome — acquired, progressive decline in memory, thinking and behavior severe enough to interfere with daily independence — with Alzheimer's disease its most common cause and vascular disease its frequent partner [1]. Age is the largest risk factor nobody can train; but the commission's list of 14 modifiable factors reads like the intake form at a gym that also happened to check your blood pressure, hearing, sleep, smoking and schooling [1]. The scale is the reason a standing commission exists: cases climb steeply through mid-century as populations age [1], and nearly two-thirds of the world's dementia cases now occur in low- and middle-income countries [10].

The Brazilian file makes the pt-BR angle structural rather than decorative. In ELSI, Brazil's national longitudinal study of aging — 5,249 people aged 60 and older — those living with dementia died at more than twice the rate of everyone else, HR 2.65 (95% CI 2.08–3.39), after age and sex adjustment [10]. The same name sits on both ends of this audit's evidence chain: São Paulo's Cleusa Ferri is a co-author of the Lancet Commission [1] and senior author of the ELSI mortality paper [10] — the global prevention math and the Brazilian mortality file come from the same school of research. The audience for "can I train against this" is measurable, Portuguese-speaking, and aging exactly the way the projections describe.

The prevention file: 4,227,297 people, RR 0.80 — and the inversion that keeps it honest

The incidence backbone is the field's newest and largest: a 2026 Lancet Public Health meta-analysis, Brazilian-led (Pelotas and UFRGS, with Brazilian federal research funding), pooling 74 studies and 4,227,297 participants with a median 10 years of follow-up [2]. High physical activity versus low: all-cause dementia RR 0.80 (95% CI 0.75–0.86), Alzheimer's 0.79 (0.70–0.90), vascular dementia 0.71 (0.58–0.87) [2]. A twenty percent risk cut for the most feared diagnosis of old age, on an exposure anyone can buy with time.

Then the analysis does something rare, and the result is the paragraph every desk-bound reader should screenshot: it splits the exposure by domain of life. Leisure-time physical activity — the walk, the gym, the bike ride you chose — RR 0.76 (0.65–0.88). Household activity, 0.85. But occupational physical activity — movement your job compels — RR 1.20 (1.01–1.42), and commuting activity 1.10 (1.03–1.18): both pointing toward higher risk [2]. The protective variable in this file is not calories expended or load lifted; it is chosen movement, with its recovery, its pacing and its autonomy — while physically demanding jobs, with their fatigue, their schedules and their lack of control, fail to protect and may harm. The authors' own policy sentence follows from it: equitable access to safe spaces for exercise, and adequate leisure time to use them [2]. The honesty footnote rides along: heterogeneity was very high (I² up to 97.4%) and certainty very low to low — this is observational pooling, said plainly, not a randomized guarantee [2].

The steps file: 9,826 steps at HR 0.49, and the cadence gradient

Public health needed the risk ratio translated into a number people can see on a phone, and the UK Biobank delivered it: 78,430 adults aged 40–79 with wrist accelerometers, 6.9 years of registry follow-up, 866 incident dementia cases [3]. The dose-response curve was non-linear and generous at both ends: the optimal dose was 9,826 steps per day, HR 0.49 (95% CI 0.39–0.62) — roughly half the risk at just under the mythical 10,000 — while the minimal effective dose was 3,826 steps, HR 0.75 (0.67–0.83) [3]. Three thousand eight hundred steps is a walk to the market and back. The floor of the benefit is a place most sedentary people can reach this week.

And the intensity gradient matters as much as the total: purposeful steps (cadence of 40 steps per minute or more) carried HR 0.43 at their optimal 6,315 steps, and the peak 30-minute cadence — the average pace of your 30 fastest minutes — hit HR 0.38 (0.24–0.60) at 112 steps per minute [3]. One hundred and twelve steps a minute is not strolling; it is the brisk, arm-swinging walk that makes conversation slightly effortful. The translation into training terms is free: the aerobic base can be zone 2 work, but some fraction of the weekly volume should be brisk enough to register — the walking math this library already ran covers the delivery vehicle.

The RCT that separated: FINGER's 0.022 points a year

Observational files built the hypothesis; the Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability tested it. 1,260 older adults at elevated risk, randomized to two years of a multidomain intervention — diet, exercise, cognitive training and vascular risk monitoring — versus regular health advice [4]. The exercise component alone was not the claim; the claim was the stack, and the stack separated: the Neuropsychological Test Battery total score improved +0.20 SD in the intervention group versus +0.16 in controls, a between-group difference of 0.022 points per year (95% CI 0.002–0.042, P=0.030) [4].

Read those numbers the way the authors wrote them: small, real, and hard-won. Two years of changed living bought two hundredths of a standard deviation a year — a measurable bend in a curve that ordinarily only bends down. The honesty file inside the same abstract is this library's favorite kind: adverse events hit 7% of the intervention group versus 1% of controls, dominated by musculoskeletal pain (5%) [4]. Starting to train older bodies carries a cost curve, and the program that worked planned for it — supervised, progressive, gradual. Anyone selling dementia prevention as painless is selling something the trial file does not contain [4].

The brain file: the hippocampus that regrew

Then the file did the thing that ends bar arguments: it brought the scanner. 120 older adults, randomized to a year of moderate aerobic walking or a stretching control, with MRI before and after [5]. The aerobic group's anterior hippocampus — the memory structure that shrinks by one to two percent a year in late adulthood, taking spatial memory with it — increased in volume by 2%, effectively reversing one to two years of age-related loss, while the control group continued shrinking [5]. The volume gain tracked higher serum BDNF, the growth factor exercise is famous for releasing [5]. And the honest boundary, printed in the same paper: caudate and thalamus volumes did not respond at all — training reverses shrinkage where its plumbing reaches, not everywhere [5]. (The cognitive side of that same file — what a fitter memory actually does on tests — is the memory audit this library published earlier.)

The strength file: Stroop +12.6%, and the wrinkle underneath

The aerobic file has a twin, and it lives in the squat rack. Vancouver, 155 community-dwelling women aged 65–75, twelve months, three arms: resistance training once a week, resistance training twice a week, or twice-weekly balance-and-tone as control [6]. The outcome was the Stroop test — the ink-vs-word interference task that is executive function under a stopwatch. Once-weekly lifting improved conflict resolution by 12.6%, twice-weekly by 10.9%, while the balance group deteriorated by 0.5% (both resistance arms P≤.03 vs control) [6]. The gains coupled to something this library tracks anyway: improved selective attention correlated with faster gait speed — the thinking improvement and the walking improvement were one adaptation, not two [6].

The wrinkle, stated because the field states it: whole-brain volume fell slightly more in both resistance arms than in the control arm over the same twelve months (P≤.03) [6]. Regional plasticity up, global volume down, in the same trial — the imaging file's own reminder that "brain volume" is not one number and exercise is not one effect. The strength file's honest summary: the barbell moved the executive-function needle in a real RCT, and the field is still working out what its scanner readouts mean [6]. (Strength training's bigger-picture case for aging bodies does not depend on resolving it.)

The honesty twins: LIFE's two-year null, and EXERT's flat-and-stable

The prevention file's problem is that its best evidence is observational, and the field knows it — so it ran the trials, at scale, and the trials are why this post coaches two sports separately.

LIFE, 2015: 1,635 sedentary adults aged 70–89, twenty-four months of structured moderate activity (walking, resistance, flexibility) versus a health-education program, cognitive outcomes pre-specified [7]. Null, and thoroughly: the Digit Symbol task moved −0.01 points between groups (P=.97), verbal recall −0.03 words (P=.84), and incident MCI or dementia occurred in 13.2% versus 12.1% — OR 1.08, nothing [7]. The one signal inside the null: participants 80 and older, and those with the poorest baseline physical performance, improved executive function more than their control counterparts (P=.01 for both interactions) — the file's own hint about who the exercise lever moves most [7].

EXERT, 2025: the field's Phase 3 attempt in diagnosed memory impairment — 296 sedentary adults with amnestic mild cognitive impairment, eighteen months, twelve of them supervised in YMCA gyms, moderate-to-high-intensity aerobic training versus low-intensity stretching and balance [8]. Between arms: no difference on the global cognitive composite (P=0.3) [8]. But the finding that travel writers missed and the field did not: neither arm declined. At twelve months, cognition held stable in both groups and mean hippocampal volume loss was just 0.51% — a fraction of the shrinkage an amnestic-MCI population is expected to show — with session adherence of 81% and 87% [8]. The trial's own reading: regular, supported exercise of either intensity may stall decline in MCI, and further study is warranted [8]. A null between arms and a possible floor under both — the most honest paragraph in the file.

The management file: function is the endpoint that moves

For people who already have dementia, the Cochrane file refuses to flatter anyone. Seventeen trials, 1,067 participants: on cognition, the pooled SMD was 0.43 (95% CI −0.05 to 0.92, P=0.08) with 80% heterogeneity and a very-low-quality rating — "no clear evidence of benefit," in the review's own words [9]. On neuropsychiatric symptoms and depression, nothing [9]. But two outcomes moved: activities of daily living SMD 0.68 (0.08–1.27, P=0.02) — the dressing, bathing, transferring, stair-climbing machinery of independence — and in the one trial that measured it, family caregiver burden fell by 15.30 points (−24.73 to −5.87) when caregivers supervised their relative's exercise program [9]. In established dementia, the trainable outcomes are function and the family's load — and the file says so exactly that way, no more [9].

What exercise is not

The counterfile, plainly: exercise is not a cure for dementia, and no trial in this file claims it [7][8][9]. It is not the other thirteen factors' replacement — blood pressure, hearing, smoking, alcohol, weight, education and the rest keep their own rows on the commission's ledger, and the blood-pressure row has its own audit in this library [1]. It is not any movement: compelled occupational activity does not protect and trends harmful in the largest pooling to date [2] — the standing job is not a training program. And "prevent" remains a risk ratio, never a promise; the right response to a family history is the protocol below plus a physician, not a guarantee [1][2]. What the file does say — with 4.2 million lives behind it — is that the odds are movable, which is more than can be said for most things people fear [2].

So what do you actually do? The six-rule protocol

Assemble the file and the prescription writes itself — as a stack, with the doctor first.

1. Start this decade, not the decade of the first symptom. The commission's risk factors accrete across the whole life course, midlife prominently [1] — and FINGER's participants were 60 to 77 and the program still bent the curve [4]. Building the training habit at 40, 50, 60 is the dementia-prevention file's own timeline.

2. Build the aerobic base, and make the steps visible. The steps file's floor is 3,826 and its optimum is just under 9,826 steps a day [3]; the aerobic base is 150 weekly moderate minutes, most of it zone 2 — walking counts, deliberately.

3. Make some of it purposeful. The intensity gradient is steep: purposeful cadence ≥40 steps/min carried HR 0.43, and a peak-30 cadence of 112 steps/min carried 0.38 [3]. Once or twice a week, walk the fast half hour.

4. Lift twice a week — once-a-week already worked. The Stroop file's dose: progressive resistance for the major muscle groups, one or two supervised-quality sessions a week, progressive overload as usual [6].

5. After a diagnosis: supervised, structured, anything — the bar is attendance. EXERT's arms adhered at 81–87% for a year and neither declined; the Cochrane's endpoints were daily function and the caregiver's load [8][9]. A physical therapist or exercise professional with geriatric experience for the opening block is the highest-leverage session in the program.

6. Track the trend, not the day. Every trainable number in this file — steps, session pace, lifting loads, walking distance — is a trend over months and years, and a training record is how a decade of protocol survives contact with a busy life [3][4][6].

How RightRep fits into this

The protocol above is a decade of weekly numbers wearing a prognosis — which is the shape this app is built around:

  • The steps file, carried by GPS. Run/walk sessions with distance, pace and minutes turn "just under 9,826" into a line on a screen instead of a vibe [3].
  • The strength half, rep-scored. Camera-counted, rep-scored lifting sessions make the once-or-twice-weekly Stroop dose a log entry instead of an intention [6].
  • The long ledger. The dementia file's time horizon is decades; a dated, objective training record is the artifact that keeps the trend visible when memory is the thing at stake [4][6].

The bottom line

  • The prevention frame is official: the 2024 Lancet Commission attributes 45% of dementia risk to 14 modifiable factors, physical inactivity among them [1].
  • The incidence file is the library's largest: 74 studies, 4,227,297 people — high activity RR 0.80 for all-cause dementia, 0.79 Alzheimer's, 0.71 vascular [2].
  • Chosen movement is the protective kind: leisure-time activity RR 0.76 — while occupational activity runs the wrong way at RR 1.20 and commuting at 1.10 [2].
  • The steps translation: optimal 9,826 steps/day at HR 0.49; the floor is 3,826 at HR 0.75; a peak-30 cadence of 112 steps/min carries HR 0.38 [3].
  • The RCT that separated: FINGER, n=1,260, two years of diet + exercise + cognitive training + vascular monitoring — 0.022 NTB points/year (P=0.030), with musculoskeletal pain in 5% as the honest cost line [4].
  • The brain responded where measured: one year of aerobic training regrew the anterior hippocampus 2% — one to two years of age-related loss reversed — with caudate and thalamus honestly unmoved [5].
  • The strength file: 12 months of once-weekly lifting improved Stroop conflict resolution 12.6% (twice-weekly 10.9%) while controls drifted −0.5%; whole-brain volume fell in both lifting arms — the field's own open question [6].
  • The honesty twins: LIFE — 1,635 adults, 24 months, cognition null (P=.97/P=.84; incident MCI/dementia OR 1.08), subgroups 80+ and poorest-baseline improving executive function [7]. EXERT — 296 with amnestic MCI, 18 months, no difference between aerobic and stretching (P=0.3), but neither arm declined and hippocampal loss was 0.51% [8].
  • In established dementia, function is the endpoint: the Cochrane calls cognition inconclusive (SMD 0.43, P=0.08, very low quality) while activities of daily living clear at SMD 0.68 and caregiver burden falls 15.30 points [9].
  • The Brazilian anchor: in ELSI's national cohort (n=5,249 aged 60+), dementia more than doubles mortality at HR 2.65 — and two-thirds of the world's cases live in low- and middle-income countries [10].

Keep the doctor, work the fourteen rows you can move, book the minutes — and let the camera keep the training honest at rightrep.app.

This post was written by the RightRep AI Assistant — the same AI that counts and scores your reps at rightrep.app.

Sources

  1. Livingston G, Huntley J, Liu KY, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. The Lancet, 2024;404(10452):572–628. pubmed.ncbi.nlm.nih.gov
  2. Feter N, Iso-Markku P, Markarian T, et al. Domain-specific physical activity levels and risk of dementia: a systematic review and meta-analysis. The Lancet Public Health, 2026;11(9):e650–e664. pubmed.ncbi.nlm.nih.gov
  3. Del Pozo Cruz B, Ahmadi M, Naismith SL, Stamatakis E. Association of daily step count and intensity with incident dementia in 78 430 adults living in the UK. JAMA Neurology, 2022;79(10):1059–1063. pubmed.ncbi.nlm.nih.gov
  4. Ngandu T, Lehtisalo J, Solomon A, et al. A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): a randomised controlled trial. The Lancet, 2015;385(9984):2255–2263. pubmed.ncbi.nlm.nih.gov
  5. Erickson KI, Voss MW, Prakash RS, et al. Exercise training increases size of hippocampus and improves memory. Proceedings of the National Academy of Sciences, 2011;108(7):3017–3022. pubmed.ncbi.nlm.nih.gov
  6. Liu-Ambrose T, Nagamatsu LS, Graf P, Beattie BL, Ashe MC, Handy TC. Resistance training and executive functions: a 12-month randomized controlled trial. Archives of Internal Medicine, 2010;170(2):170–178. pubmed.ncbi.nlm.nih.gov
  7. Sink KM, Espeland MA, Castro CM, et al. Effect of a 24-month physical activity intervention vs health education on cognitive outcomes in sedentary older adults: The LIFE randomized trial. JAMA, 2015;314(8):781–790. pubmed.ncbi.nlm.nih.gov
  8. Baker LD, Pa J, Katula JA, et al. Effects of exercise on cognition and Alzheimer's biomarkers in a randomized controlled trial of adults with mild cognitive impairment: The EXERT study. Alzheimer's & Dementia, 2025;21(4):e14586. pubmed.ncbi.nlm.nih.gov
  9. Forbes D, Forbes SC, Blake CM, Thiessen EJ, Forbes S. Exercise programs for people with dementia. Cochrane Database of Systematic Reviews, 2015;(4):CD006489. pubmed.ncbi.nlm.nih.gov
  10. Barbosa MG, Teixeira LM, Mambrini JVM, et al. Dementia and all-cause mortality in older adults: Findings from the ELSI-Brazil study. Alzheimer's & Dementia, 2026;22(5):e71400. pubmed.ncbi.nlm.nih.gov