Cognitive decline isn't inevitable: the evidence-based playbook for protecting your brain

Q: Is dementia genetic, or is it modifiable?

Both, but the modifiable share is much larger than most people assume. The Lancet 2024 commission estimated that 14 modifiable risk factors - hearing loss, hypertension, education, smoking, depression, social isolation, physical inactivity, diabetes, obesity, excessive alcohol, traumatic brain injury, air pollution, visual loss, and high LDL cholesterol - collectively account for around 45% of dementia cases worldwide. APOE4 is the largest single genetic risk factor, but even APOE4 carriers show meaningful risk reduction from the same lifestyle levers. Genetics loads the gun; lifestyle pulls the trigger.

Q: Does the FINGER trial prove lifestyle can prevent dementia?

FINGER (Ngandu et al. 2015, Lancet) is the strongest single randomised trial of multi-domain lifestyle intervention for cognitive decline. 1,260 at-risk older adults aged 60-77 were randomised to either a 2-year intervention (diet + exercise + cognitive training + vascular risk monitoring) or general health advice. The intervention group showed significantly better cognition on the primary neuropsychological test battery at 24 months. It is the strongest trial evidence we have that multi-domain lifestyle change can preserve cognition in at-risk older adults - but it does not prove dementia prevention in absolute terms, because cognition is a continuous variable and the follow-up was only 2 years.

Q: Do brain-training apps actually slow cognitive decline?

Cognitive training transfers to the specific tasks you train on but generalises poorly. The FINGER trial included cognitive training as one of four components, but the other three (diet, exercise, vascular risk monitoring) carry more of the published effect size. Stand-alone brain-training apps have weak evidence for general cognitive improvement and no evidence for dementia prevention. What does show transfer: complex, novel, socially-engaging cognitive load - learning a language, an instrument, a complex skill. Repetition of trivial puzzles does not move the needle.

Q: Does omega-3 or any single supplement prevent dementia?

No single supplement has a credible randomised-trial case for dementia prevention. Omega-3, vitamin E, vitamin B-complex, ginkgo biloba, coconut oil, and curcumin have all been tested in RCTs powered for cognitive outcomes and all have failed or shown only marginal effects in subgroups. The 2023 Cochrane review of omega-3 supplementation for dementia prevention found no benefit. The honest summary: supplements are a distraction from the levers that do work - blood pressure control, hearing aids, exercise, sleep, and social engagement.

Q: Is it too late to start protecting cognition at 60?

No. FINGER, MAPT, and the Lancet 2024 commission all show meaningful risk-factor modifiability into the seventies. Hearing aid uptake, hypertension control, and physical activity show effects within months. The window narrows but never fully closes. The most powerful preventive period is midlife (40s-60s), but the next-best time is now. The 14 modifiable factors operate across the entire life course and the commission's PAF estimates assume action from midlife forward.

The unwelcome inheritance - and why "inevitable" is wrong

Most people carry a quiet, unexamined assumption: that cognitive decline is a tax you pay for living long enough. The published evidence has been steadily dismantling that assumption for two decades, and the 2024 update of the Lancet Commission on dementia prevention is the cleanest single articulation of where the science actually sits: around 45% of dementia cases worldwide are attributable to 14 modifiable risk factors [1]. That is a population-attributable-fraction estimate from a panel of dementia researchers reviewing the global cohort and trial literature - not a slogan.

Nearly half the dementia case-load that the rich-world health system currently treats as fate is, in principle, on the modifiable side of the ledger. The remaining 55% is genuinely outside the playbook - heritability, age, currently-untreatable molecular processes - but the half that is movable is more than enough to justify a serious preventive routine starting in midlife.

The FINGER trial - the strongest single RCT we have

If you are going to anchor a cognitive-prevention argument on one randomised trial, FINGER is the trial. Ngandu et al. 2015 in The Lancet reported on 1,260 Finnish adults aged 60-77 at elevated risk of cognitive decline (cardiovascular risk score above threshold, cognition at or slightly below the age-expected mean). Participants were randomly assigned to either a 2-year multi-domain lifestyle intervention or a control group receiving general health advice [2].

The intervention had four components running in parallel for two years:

Diet. Individualised counselling toward a Nordic / Mediterranean-style pattern: vegetables, fish, whole grains, lower saturated fat, limited red meat.
Exercise. Group-based and individual physical training, both aerobic and resistance components, supervised at a fitness centre, progressing in intensity across the 2 years.
Cognitive training. Group sessions on memory + executive function + computer-based training tasks at home.
Vascular and metabolic risk monitoring. Regular BP, weight, glucose and lipid checks with intervention to standard-of-care targets.

On the primary outcome - a comprehensive neuropsychological test battery at 24 months - the intervention group's composite Z-score was 25% better than the control group's (p = 0.03). Executive function showed an 83% greater improvement and processing speed a 150% greater improvement than control. These are continuous-cognition results, not a binary "did they get dementia" endpoint - but they are the cleanest randomised demonstration that multi-domain lifestyle intervention can preserve cognition in at-risk older adults.

The MAPT trial (Andrieu et al. 2017, Lancet Neurology) ran a parallel multi-domain intervention in 1,680 community-dwelling French adults aged 70+ [3]. The primary endpoint at 3 years was negative overall, but pre-specified subgroup analyses showed benefit in participants at higher baseline cognitive or vascular risk - a finding that mirrors FINGER (the at-risk respond; the low-risk have less to gain). The combined FINGER + MAPT evidence is the substrate for the international FINGERS network now running parallel trials in 40+ countries.

The Lancet 2024 commission - 14 risk factors, ~45% attributable

The 2020 Livingston commission identified 12 modifiable risk factors. The 2024 update added two - untreated visual loss and high LDL cholesterol - and revised the population-attributable-fraction estimates upward [1]. The current 14, ranked roughly by attributable share:

Hearing loss (~7% PAF) - the largest single modifiable factor. Untreated mid-life hearing loss is associated with substantially elevated dementia risk; hearing aid use in observational data is associated with attenuated decline.
High LDL cholesterol (~7% PAF, new in 2024) - the cardiovascular-brain link, formalised.
Low education (~5% PAF) - fewer years of formal schooling, less cognitive reserve built.
Social isolation (~5% PAF) - chronic loneliness associated with faster cognitive decline.
Depression (~3% PAF) - both a prodrome marker and a modifiable driver.
Hypertension (~2% PAF) - sustained mid-life hypertension is one of the cleanest dementia risk factors.
Physical inactivity, obesity, diabetes, smoking, excessive alcohol, traumatic brain injury, air pollution, untreated visual loss - each contributing 1-3% PAF.

The PAFs don't sum cleanly to 45% (overlapping confounders), but the commission's modelled total attributable fraction lands in that band. The single largest cluster is cardiovascular - hypertension, diabetes, obesity, high LDL, physical inactivity, smoking - which collectively dominate the modifiable picture. If you had to pick one lever, it would be cardiovascular health.

The cardiovascular-brain link - your brain is a vascular organ

The brain is 2% of body mass and consumes ~20% of resting cardiac output. Every cognitive moment is a delivery problem: oxygen and glucose moving through ~600 km of cerebral microvasculature. Hypertension narrows and stiffens those vessels. Diabetes glycates them. High LDL drives endothelial dysfunction. The downstream consequence is silent ischaemic injury accumulating across the decades - small-vessel disease visible on MRI as white matter hyperintensities, lacunar infarcts, microbleeds - that adds together into clinically-meaningful cognitive impairment.

The cohort evidence here is strong. SPRINT-MIND randomised 9,361 hypertensive adults to either intensive (target SBP <120 mmHg) or standard (<140 mmHg) BP control; the intensive arm showed a significant reduction in mild cognitive impairment (HR 0.81) [4]. The Yaffe et al. work in the CARDIA cohort showed that mid-life cardiovascular risk factors - particularly hypertension and diabetes from the late 30s and 40s onward - predict cognitive decline 25 years later [5]. The window for cardiovascular protection of the brain is mid-life. Waiting until cognitive symptoms appear is treating the consequence.

Exercise plugs directly into the same machinery - and into one more. Erickson et al. 2011 in PNAS randomised 120 older adults to either aerobic exercise or stretching for one year; the aerobic group's hippocampal volume increased by ~2% on MRI while the stretching group's decreased by the expected age-related ~1.4%. The change tracked with serum BDNF (brain-derived neurotrophic factor) and correlated with improved spatial memory [6]. Sleiman et al. 2016 in eLife identified the molecular link: the ketone body β-hydroxybutyrate, produced during exercise and fasting, directly upregulates BDNF transcription via class I histone deacetylase inhibition [7]. Cardio doesn't just defend the cerebral vasculature, it inputs into the molecular program that maintains hippocampal plasticity.

Sleep - the glymphatic clearance window

The brain has no traditional lymphatic system. Until about a decade ago, this looked like an evolutionary oversight: every other organ uses lymphatics to flush metabolic waste, and the metabolically-hungriest organ has none. The Iliff et al. 2012 + Xie et al. 2013 work resolved the puzzle by characterising the glymphatic system - a perivascular CSF-flow network, driven by aquaporin-4 channels on astrocytic endfeet, that flushes interstitial waste from the brain. Critically, glymphatic flow is roughly 60% higher during sleep than wake, and the waste species it clears include amyloid-β [8]. Bad sleep is, in a mechanistic sense, a failure to take out the trash.

The cohort evidence catches up with the mechanism. Sabia et al. 2021 in Nature Communications followed nearly 8,000 Whitehall II cohort participants for 25 years and found that those reporting persistent short sleep (≤6 hours) at age 50, 60, and 70 had a 30% higher risk of dementia compared to those sleeping 7 hours [9]. The signal is consistent across cohorts, but the relationship is U-shaped: very long sleep is also associated with elevated risk, partly because long sleep at older ages is often a marker of poor sleep quality, undiagnosed apnoea, or depression.

For a deeper read on the sleep biology, see our sleep stages article - particularly the glymphatic + deep-sleep section. The practical summary for cognitive protection: 7-8 hours, consistent timing, treat apnoea aggressively if present.

Cognitive reserve - the buffer hypothesis

One of the most replicated findings in dementia epidemiology is the cognitive reserve effect: at any given level of underlying neuropathology, people with more years of education, more cognitively-complex occupations, and more lifelong learning engagement show fewer clinical symptoms. Stern's 2012 review in Lancet Neurology is the canonical framing [10]: reserve gives the brain richer compensatory networks, so a given amount of pathology produces less functional impairment.

The harder question is what builds reserve in adulthood. Education in early life is the strongest single proxy, but lifelong learning continues to add reserve in midlife and beyond - with the strongest evidence sitting on complex, novel, effortful activity rather than passive consumption. A second language. A musical instrument. A skill with a real evaluation function. Social-cognitive complexity counts: regular complex conversation with people whose minds you don't already predict.

What does not show transfer is repetitive easy puzzles. The Owen et al. 2010 Nature study put 11,430 participants through 6 weeks of online brain training and found the trained tasks improved but no transfer occurred to general cognitive ability [11]. Sudoku does not earn its reputation.

APOE4 - the gene most people are scared of

The apolipoprotein-E ε4 allele is the largest single common genetic risk factor for late-onset Alzheimer's disease. One copy roughly triples lifetime risk; two copies multiply it by ~10x. The frequency in European-ancestry populations is ~14%, so most carriers don't know they are. Direct-to-consumer genetics has made APOE4 status increasingly visible and increasingly anxiety-producing.

The evidence is consistent and reassuring on one point: the same modifiable risk factors that work in the general population work in APOE4 carriers. Lourida et al. 2019 in JAMA followed 196,383 UK Biobank participants for 8 years and stratified by both APOE4 status and a lifestyle score. The relative dementia-risk reduction from a healthy lifestyle was similar in carriers and non-carriers - roughly 32% in both groups [12]. APOE4 raises the baseline; lifestyle still moves the needle by the same relative amount, which means the absolute risk reduction is actually larger for carriers. The worst response to a positive APOE4 result is fatalism; the best response is the same playbook with extra urgency on cardiovascular and metabolic control.

What the evidence does not support

The brain-health market has run far ahead of the trial data. Three claims to delete from the playbook.

"Omega-3 / fish oil prevents cognitive decline." Large RCTs of DHA-EPA at 800-2000 mg/day have found no meaningful benefit on cognitive outcomes in healthy older adults, and the 2023 Cochrane review reached the same conclusion. The cohort evidence linking fish consumption to lower dementia risk is real, but the supplementation trials that should have confirmed it failed. Eating fish is fine; the capsule has not earned its reputation.

"Coconut oil / MCT oil reverses Alzheimer's." A persistent claim built on a single anecdotal case report and small poorly-controlled studies. Multiple subsequent trials failed to replicate the effect. The mechanism (ketone bodies as alternative brain fuel) is real, but the dose required for sustained ketosis and the trial evidence at that dose simply don't exist. Coconut oil is a high-saturated-fat food, not a neuroprotective intervention.

"Ginkgo biloba protects memory." The GEM (Ginkgo Evaluation of Memory) trial randomised 3,069 adults aged 75+ to either 120 mg ginkgo twice daily or placebo for a median 6.1 years. The trial showed no reduction in dementia incidence and no reduction in cognitive decline. Ginkgo is one of the few supplements with a large, well-conducted, long-duration RCT pointing at the right endpoint, and the answer was negative.

The practical weekly playbook

Pulling the published evidence into something you can actually do, ordered by approximate effect size from the trial + cohort data:

1. Treat your cardiovascular risk like brain protection. Know your blood pressure, fasting glucose, and LDL. Target sustained BP <130/80 from mid-life onward (SPRINT-MIND supports tighter targets for cognitively-vulnerable adults). If you're on statins for cardiovascular indication, the cognitive evidence is now neutral-to-positive, not negative. The single biggest lever in the entire playbook is the cardiovascular cluster.

2. Get 150-300 minutes of zone 2 cardio per week, plus 2 sessions of resistance training. Erickson's hippocampal-volume work used 40 minutes 3x/week. Cardio drives BDNF, defends the cerebral microvasculature, and improves sleep quality. Resistance training adds insulin sensitivity. Our zone 2 piece covers the cardio dose in detail.

3. Defend 7-8 hours of sleep, with consistent timing. Treat snoring + observed apnoeas seriously - undiagnosed obstructive sleep apnoea is the single sleep variable most strongly linked to cognitive decline. The glymphatic clearance argument means short and fragmented sleep is mechanistically - not just associatively - bad for brain protein clearance.

4. Get your hearing checked at 50 and at 60. Use the aid if you need it. Untreated hearing loss is the single largest modifiable risk factor in the Lancet commission. The social-isolation pathway and the cognitive-load pathway are both plausible. Hearing aid stigma has cost more cognitive years than almost any other modifiable behaviour in the rich-world data.

5. Build cognitive reserve through complex, effortful, novel learning. A second language. A musical instrument. A skill with a real evaluation function. The complexity matters more than the time spent. Two hours per week of effortful new learning is more protective than ten hours of crossword puzzles.

6. Stay socially engaged. Regular complex conversation. Volunteering. Group activities that require remembering people and contexts. Social isolation is the third-largest modifiable factor in the commission's estimate.

7. Skip the brain-health supplement aisle. Omega-3, vitamin E, vitamin B-complex, ginkgo, and coconut oil have all been tested at the dementia endpoint and have all failed. The money + cognitive bandwidth is better spent on the first six items.

The takeaway

The Lancet 2024 commission's estimate that around 45% of dementia is attributable to 14 modifiable risk factors is the cleanest single summary of where the prevention evidence sits. FINGER is the strongest single demonstration that multi-domain lifestyle intervention preserves cognition in at-risk older adults. The cardiovascular cluster is the largest lever and starts paying off when you act in midlife. Sleep is the glymphatic clearance window. Cognitive reserve is built by complex, effortful, novel learning - not puzzle apps. APOE4 raises the baseline but does not blunt the relative benefit of the playbook. The supplement-based shortcuts have, in the RCTs designed to test them, mostly failed.

The unfashionable truth is that cognitive protection looks almost identical to cardiovascular protection plus sleep, hearing, social engagement, and lifelong learning. There is no separate brain-only intervention with strong trial evidence. If you want your cardiovascular markers, sleep, recovery, and cognitive-test trends tracked together and rolled into a coherent bio-age picture, have a look at Thier.

Frequently asked questions

Is dementia genetic, or is it modifiable?

Both, but the modifiable share is much larger than most people assume. The Lancet 2024 commission estimated that 14 modifiable risk factors - hearing loss, hypertension, education, smoking, depression, social isolation, physical inactivity, diabetes, obesity, excessive alcohol, traumatic brain injury, air pollution, visual loss, and high LDL cholesterol - collectively account for around 45% of dementia cases worldwide. APOE4 is the largest single genetic risk factor, but even APOE4 carriers show meaningful risk reduction from the same lifestyle levers. Genetics loads the gun; lifestyle pulls the trigger.

Does the FINGER trial prove lifestyle can prevent dementia?

FINGER (Ngandu et al. 2015, Lancet) is the strongest single randomised trial of multi-domain lifestyle intervention for cognitive decline. 1,260 at-risk older adults aged 60-77 were randomised to either a 2-year intervention (diet + exercise + cognitive training + vascular risk monitoring) or general health advice. The intervention group showed significantly better cognition on the primary neuropsychological test battery at 24 months. It is the strongest trial evidence we have that multi-domain lifestyle change can preserve cognition in at-risk older adults - but it does not prove dementia prevention in absolute terms, because cognition is a continuous variable and the follow-up was only 2 years.

Do brain-training apps actually slow cognitive decline?

Cognitive training transfers to the specific tasks you train on but generalises poorly. The FINGER trial included cognitive training as one of four components, but the other three (diet, exercise, vascular risk monitoring) carry more of the published effect size. Stand-alone brain-training apps have weak evidence for general cognitive improvement and no evidence for dementia prevention. What does show transfer: complex, novel, socially-engaging cognitive load - learning a language, an instrument, a complex skill. Repetition of trivial puzzles does not move the needle.

Does omega-3 or any single supplement prevent dementia?

No single supplement has a credible randomised-trial case for dementia prevention. Omega-3, vitamin E, vitamin B-complex, ginkgo biloba, coconut oil, and curcumin have all been tested in RCTs powered for cognitive outcomes and all have failed or shown only marginal effects in subgroups. The 2023 Cochrane review of omega-3 supplementation for dementia prevention found no benefit. The honest summary: supplements are a distraction from the levers that do work - blood pressure control, hearing aids, exercise, sleep, and social engagement.

Is it too late to start protecting cognition at 60?