Every practical recommendation on the site, grouped by domain and ordered by how much each one actually moves mortality and healthspan. The complete checklist for a healthy midlife adult.

The healthspan long list

This is what to actually do. Each item links to the page that justifies it. Groups are ordered roughly by impact in a healthy midlife adult; items inside a group follow the same logic. Ordering within a group encodes evidence strength; the linked pages carry the formal ratings. Nobody reads a checklist front to back — skip around.

A quick orientation:

The first four groups — harm avoidance, exercise, sleep, and dietary pattern — carry the bulk of the achievable mortality reduction, and are not strictly ranked among themselves. Get these right and everything else is fine-tuning.
Mind, sensory, controlled stress, drinks add measurable but smaller increments.
Clinical screening is age-based and worth running on schedule; prescription pharmacology only earns its place when biomarkers or symptoms point at it.
Supplements are gap-fillers, not longevity drugs — five clear the bar, almost everything else is hype.
The exposome — air, radon, water, plastics — is cheap to improve and almost never audited.

1. Avoid the four biggest harms

These are the largest preventable mortality levers in a midlife adult. The other groups assume you've addressed these first.

Don't smoke; quit at any age. Combustible tobacco is the single largest preventable-mortality lever in modern medicine. Cognitive and cardiovascular trajectories improve at any quit age, though the dividend is steeply age-graded: quitting before about 40 avoids roughly 90% of the excess death risk, and quitting in your late twenties, thirties or forties returns roughly ten, nine and six years respectively. Non-combustible nicotine carries its own risks — treat it as harm reduction, not a free habit. See Smoking and nicotine.
Drink as little as possible. The "J-curve" cardio-protection has largely collapsed under genetic evidence — heart attack is the one endpoint where that evidence still fails to show harm, which is exactly the endpoint the J-curve was built on and exactly why the argument hasn't fully ended. Cancer risk rises from the lowest intakes anyone has measured, with no threshold below which it flattens. If you currently drink, less is better — Canada's 2023 guidance puts the lower-risk threshold at 2 drinks a week or fewer — about 27 g of alcohol. (The US 2025–2030 Dietary Guidelines, released January 2026, dropped numeric limits entirely; that was a policy change, not an evidence change.) See Alcohol.
Screen yourself for sleep apnea. Severe untreated obstructive sleep apnea is associated with roughly double the all-cause mortality, and is dramatically underdiagnosed in midlife. Take the STOP-Bang questionnaire, an eight-question screen you can do in a minute; if you score 3 or more — especially with snoring, high blood pressure, atrial fibrillation, or a body mass index of 30 or above — get a home sleep study. Continuous positive airway pressure, the mask that splints the airway open overnight, reliably fixes the symptoms: sleepiness, snoring, mood, accident risk. The cardiovascular payoff is a separate claim and a weaker one — the SAVE trial was null, and the lower event rates seen at four or more hours a night come from observational data. See Sleep-disordered breathing.
Don't sit through the day. Hours of essentially motionless sitting carry an independent mortality and cognitive risk that one workout doesn't undo — the enzyme that clears fat from the bloodstream loses most of its activity within hours of stillness. Break up sitting every 30–60 min, and make the break movement: in the accelerometer data, swapping long sitting bouts for short ones changed mortality not at all, so standing up and re-sitting is not the intervention. Frequency of actual movement is what the evidence supports, and the popular 20-8-2 rule — 20 minutes sitting, 8 standing, 2 moving — is a mnemonic for it rather than a tested dose. Treat incidental daily movement — the walking, fidgeting and stair-climbing you never call exercise — as its own lever. See Sitting.

2. Train: cardio, strength, mobility — and vary it

Cardiorespiratory fitness is among the best clinical predictors of all-cause mortality — a receipt of training rather than a lever in itself, and no trial has shown that raising it lowers mortality; resistance training is the best-evidenced non-drug intervention for reversing age-related muscle loss and moving bone density. Both compound.

Lift 2–3 times a week, heavy enough to matter. Compound movements (squat, hinge, press, pull), 2–4 sets × 3–6 reps for strength and 6–12 for size. Resistance training on its own tracks with about 21% lower all-cause mortality, and the dose that buys it is strikingly small — mortality bottoms out at 30–60 minutes a week of lifting and attenuates past roughly 130–150 minutes. More is for performance, not longevity. The LIFTMOR trial — twice-weekly compound lifts above 80% of a one-rep maximum plus jumping drop landings — is the bone-density evidence base: in postmenopausal women with diagnosed osteopenia or osteoporosis it produced a clinically meaningful gain at the lumbar spine (~2.9%) and a much smaller one at the hip (~0.3%), with no fragility fractures. See Resistance training, how to train it, and Bone density.
At least 3–4 hours of zone-2 aerobic per week. Conversational pace, split across 3–4 sessions of roughly an hour — walking, cycling, jogging. Builds the mitochondrial and metabolic-flexibility base that harder work sharpens. The mortality benefit keeps growing to roughly 300–600 minutes of moderate activity a week, then flattens — so treat this prescription as the floor, where most of the benefit is banked per hour spent, not the ceiling. See Zone 2.
Add one or two VO₂ max sessions weekly — work that pushes your maximum rate of oxygen uptake, the ceiling on how hard you can work aerobically. 4×4 minutes near max with 3-minute rests, or shorter sprint intervals. Across 199 cohorts, each 1-MET increment of fitness — one metabolic equivalent, roughly the difference between a brisk walk and a slow jog — is associated with about 11–17% lower all-cause mortality, though that evidence is observational and its authors graded it very low to moderate certainty. Separately, in the Cleveland Clinic cohort of 122,007 adults there was no upper ceiling: the most fit had the lowest mortality even against the merely high-fit. Read a high VO₂ max as the receipt of the training rather than the drug — genetic studies find no direct causal link between the number itself and longevity. See VO₂ max and how to train it.
Defend ~7,000 steps a day. The 10,000 figure is 1960s pedometer marketing; the dose-response curve bends around 5,000–7,000 and most of the benefit is banked by 7,000. Treat it as a working target rather than a threshold — the curve keeps rewarding steps out to about 8,000–10,000 under 60 and flattens earlier, around 6,000–8,000, from 60 onward, and for the heaviest sitters pushing toward 9,000–10,500 buys additional offset. There is no threshold below which steps stop counting. See Sitting.
Mobility and balance, 10–15 min daily. Single-leg work, hip mobility, thoracic rotation, ankle dorsiflexion, grip strength. Pooled across 108 trials, exercise cuts the rate of falls by about 23% and the number of people who fall by about 15% in older adults, The highest-certainty estimate belongs to balance-and-functional training and the largest single estimate to programmes combining balance with progressive strength work; what the evidence does not support is a fall-prevention plan built on resistance training or walking alone. Balance and power erode for decades before a fall ever happens, and they are trainable that whole time. See Mobility and balance for why it matters and how to train it.
Vary your modalities. A 2026 BMJ Medicine analysis of 111,000 adults found about 19% lower mortality at matched volume when exercise was varied across modalities — not just more of the same. See Sitting and Exercise.
Don't ignore recovery. Protein, sleep, and managing chronic stress are part of training, not separate from it. Managing chronic stress has its own page — see Stress. Chronic high-volume endurance training well above roughly 10 hours a week, sustained over decades, carries a small but real signal for atrial fibrillation (an irregular heart rhythm) and coronary calcification — it does not apply to the volumes this section prescribes.

3. Sleep: duration, regularity, environment

Short, fragmented, or misaligned sleep predicts cardiovascular disease, dementia, metabolic disease, and all-cause mortality. The dose-response is U-shaped: 7–8 hours on a regular schedule is the optimum.

Anchor your wake time. Same time every day, weekends included. The nervous system adapts to consistent waking, not to consistent sleep onset. Sleep regularity outperformed duration as a mortality predictor in the 2024 Windred Sleep analysis. See Circadian rhythms.
Get morning light within an hour of waking. Light is by far the strongest circadian cue there is, and 10–30 minutes outdoors delivers it — outdoor light runs tens of thousands of lux against a few hundred indoors. See Circadian rhythms.
Aim for 7–8 hours. Against that reference, under seven hours is associated with about 14% higher mortality and nine or more with about 34% — though the long-sleep arm partly reflects underlying illness (depression, untreated apnea, frailty) rather than the sleep itself. See Sleep.
Dim the evening; cool the bedroom. Reduce screens and overhead light 1–2 hours before bed; sleep somewhere cool enough that core temperature can fall — the observed optimum in older adults is 20–25 °C, and it varies between people — with blackout curtains and no glowing electronics. The natural core-temperature drop is what initiates sleep. See Circadian rhythms.
Size the caffeine cut-off to the dose. A single cup (~100 mg) has no measurable effect on sleep even four hours before bed; a 400 mg day's worth alters sleep architecture taken within 12 hours of bed and fragments sleep within 8. A heavy total load wants a 10–12 hour buffer; one late-afternoon cup is defensible for most people. Clearance varies several-fold between people — genotype, smoking and pregnancy all shift it — so there is no single cut-off that fits everyone. See Coffee.
No alcohol within 3 hours of bed. It hastens sleep onset but fragments architecture and suppresses rapid-eye-movement sleep. See Alcohol.
For chronic insomnia, start with therapy, not pills. Cognitive behavioural therapy for insomnia outperforms sleeping pills in head-to-head trials and its effects persist after the course ends. Digital programmes (Sleepio, Somryst) have large randomised trials behind them and are the practical route when a therapist isn't available — clearly effective, though only one trial has tested them directly against in-person therapy, so treat them as a reasonable first move rather than a proven equal. The newer sleeping pills known as Z-drugs (zolpidem, zopiclone) and the older benzodiazepines — especially after age 50 — are associated with falls and fractures, with hip-fracture risk running about 90% and 52% higher respectively in pooled observational data. See Treating chronic insomnia.
Avoid over-the-counter antihistamines as sleep aids. Diphenhydramine (Benadryl, ZzzQuil) blocks acetylcholine, and it is this anticholinergic class — not the Z-drugs — that carries the observational dementia signal, alongside cognitive impairment with chronic use. See Sleep supplements.
A 20-minute power nap is fine; an obligatory 90-minute nap is a warning. In a well-rested midlife adult, a daily long nap is a warning sign for cardiovascular disease and earlier death, and more weakly for cognitive decline — most likely a marker of something upstream, such as fragmented night sleep or undiagnosed sleep apnea, rather than a cause. In people genuinely short on night sleep, the excess risk simply doesn't appear. See Daytime naps.

4. Eat the pattern, hit protein, limit the harms

Dietary pattern outperforms any individual macro or "superfood." The Mediterranean pattern, the MIND diet built from it for brain health, and DASH — Dietary Approaches to Stop Hypertension, the blood-pressure diet — all agree on the core: lots of plants and fish, minimal ultra-processed food.

Default to a Mediterranean pattern. Vegetables, legumes, fish, nuts, olive oil, whole grains, fruit; optional moderate dairy (yogurt, cheese); modest poultry. The PREDIMED trial — the one large randomised nutrition trial with hard endpoints — cut heart attacks and strokes by roughly 30%, a result that survived being retracted and republished in 2018 over randomisation irregularities, though the olive-oil and nut industries part-funded it and the deep dive rates it Moderate rather than definitive. In UK Biobank, adults in the top fifth for adherence to plant-forward patterns like this one had life expectancy about 2–3 years longer — an association, not a demonstrated gain. See Dietary patterns.
Eat fish 2+ times a week. Fatty fish — salmon, sardines, mackerel, herring — gives you EPA and DHA, the two long-chain omega-3 fats, at a dose that makes the supplement unnecessary. See Dietary fats and Omega-3.
Hit protein targets. The number moves with activity and age: 1.2–1.6 g/kg/day if you train, 0.8–1.2 if you're sedentary in midlife, and 1.0–1.5 past 65. Per meal aim for ~0.4 g/kg, rising to 30–40 g past 65 because older muscle responds less to a given dose. The 0.83 g/kg recommended daily allowance is a floor against deficiency, not a target for healthy aging. See Protein.
Cut ultra-processed food. The highest intakes carry about 15% higher all-cause mortality versus the lowest; the per-10%-of-calories increment is contested (~3% to ~10% across the two 2025 dose-response meta-analyses). A 2025 University College London crossover trial found the minimally processed diet produced about twice the weight loss even with fat, protein, carbohydrate, salt and fibre matched — suggestive that processing harms on its own, though it ran 55 people for eight weeks and its critics point out the ultra-processed arm was also about twice as energy-dense, which would explain the result without invoking processing at all. See Ultra-processed food.
Limit added sugar — roughly 25 g/day for women and 36 g for men, and keep any single meal under ~10 g. The daily figures are the American Heart Association's; the per-meal cap is from the 2025–2030 US Dietary Guidelines, and it matters separately because it's the single large hit of sugar that overwhelms the liver. (The World Health Organization sets its ceiling differently, at 10% of calories.) Added sugar dose-responsively accelerates validated biological-age clocks, and sugar-sweetened drinks are the worst single delivery vehicle. Swapping to "diet" is a smaller win than it looks — non-nutritive sweeteners beat sugar and roughly match water in trials, while the long-term cohort signals still argue for water; and erythritol and xylitol, the two sugar alcohols that dominate keto products, carry a cardiovascular signal strong enough that anyone with established heart disease should avoid using them routinely. See Sweeteners and Sugar substitutes.
Limit red meat; minimise processed meat. Cap unprocessed red meat near ~3 servings/week (the World Cancer Research Fund upper bound, roughly 350–500 g cooked); treat processed meat (bacon, sausage, ham, deli, hot dogs) as occasional, not weekly — there's no established safe threshold and the dementia, cardiovascular and colorectal-cancer signals all align. Each 100 g/day of unprocessed red meat is associated with about 11% higher cardiovascular risk and about 10% higher type 2 diabetes risk — though that unprocessed signal is population-dependent, showing up in US cohorts and often absent in European and Asian ones, and genetic studies have so far found no causal link. The processed-meat case is the robust one: each 50 g/day is associated with 16–18% higher colorectal cancer risk, and as little as a quarter-serving a day tracks with about 13% higher rates of new dementia diagnoses. Replace with fish, poultry, legumes, nuts, or eggs — not refined carbs. If you grill, marinate and don't char; eat it alongside high-fibre foods. See Red and processed meat and Foods to limit.
Get the fat type right. Eliminate industrial trans fats (check for "partially hydrogenated oil"; mostly already gone in regulated markets). Anchor cooking on extra-virgin olive oil. Replace saturated fat with polyunsaturated or monounsaturated fat, not refined carbohydrates — the substitution effect dominates (swapping saturated fat for polyunsaturated fat is associated with roughly 30% fewer cardiovascular events; swapping it for refined carbohydrate does nothing). Don't fear linoleic acid in a home kitchen; the "seed oils are toxic" framing fails against a 2025 umbrella review pooling 150 cohorts. The full lipid picture, including the omega-6:omega-3 ratio, is under Dietary fats.
Eggs are fine for most adults. Whole eggs at up to one a day / 5–7 per week are compatible with cardiovascular health for healthy adults. The geographic paradox (US/Western cohorts mildly elevated, Asian cohorts neutral-to-protective) reflects the companion foods, not the egg. Eggs are also the richest food source of choline, a nutrient the liver and brain both need — in a form the body absorbs about four times better than the synthetic version in supplements. People with diabetes should stay nearer 3–4 eggs a week. See Dietary fats.
Eat 25–30 g of fibre a day. Getting from the typical 16 g to 25 matters far more than getting from 28 to 30. The strongest single-nutrient signal for preventing cardiovascular disease and colorectal cancer — though it is an observational signal, and no trial has shown that adding fibre prevents disease; one supplement trial actually increased the return of precancerous colon polyps. So the target is a reason to eat legumes and whole grains rather than to buy a supplement. Fibre also blunts the spike in TMAO, a compound gut bacteria make from red meat that helps drive artery plaque. See Fiber.
Front-load eating. Larger breakfast and lunch, lighter dinner, last meal by 18:00–19:00 where the schedule allows and at minimum 2–3 hours before bed. Early time-restricted eating improves fasting insulin and body composition versus matched-calorie late eating. See Fasting.
Include fermented dairy. Habitual yogurt, kefir, and traditionally aged cheese track with lower all-cause, cardiovascular, and cancer mortality across pooled cohorts — the most consistent signal in this literature, though modest at roughly 6–7%, and for yogurt specifically the cardiovascular association missed statistical significance. The food matrix matters more than live-culture counts. See Fermented foods.
Eat carbs last; choose sourdough; add vinegar. Sequencing vegetables and protein before starches reliably flattens the post-meal glucose rise without cutting carbohydrates — though whether flattening it buys anything in people without diabetes is unproven, and the founding trial was eleven participants who had type 2 diabetes. Individual glucose responses to the same food vary enormously between people. See Glycemic index.
Skip the keto/carnivore experiment for general longevity. Short-term metabolic improvements in some populations, but no long-term outcome data and an unfavourable 2025 mouse-lifespan signal for low-carb high-protein — in DNA-repair-deficient mice, a sensitised model rather than a normal one. The pattern with the data is Mediterranean. See Nutrition.
Forget "detoxes." No clinical evidence; the liver and kidneys handle this. See Nutrition.
For a concrete weekly menu that hits the protein, fibre, fish, legume, fermented-food, and meal-sequencing targets simultaneously, see a sample longevity week.

5. Look after the mind, mood, and meaning

Loneliness, chronic stress, and lack of purpose track with immune dysregulation and higher mortality, and the social-connection signal is one of the largest in epidemiology — though how much of it, and of the purpose and epigenetic-clock findings, is causal rather than a marker of underlying health is unsettled. The biology of aging has a name for this dimension, which traditional accounts left out: psychosocial adaptation.

Stay socially connected. Pooling 148 cohorts, people with strong relationships had roughly 50% greater odds of surviving follow-up — comparable to or larger than quitting smoking, and one of the largest non-drug mortality signals in epidemiology. Depth of ties, not headcount — though this is observational, and an association that large in cohort data is not the same as a lever you can pull to that size. See Purpose.
Defend purpose, especially between ages ~63 and 70. Keep a reason to get up — work, caregiving, volunteering, a craft — through the retirement transition, which is where the data cluster. Wisconsin 28-year data identify this as the critical psychosocial window, and purpose held through it tracked with about 15% lower dementia odds at 80. But the effect is modest, the design observational, and the causal direction genuinely unsettled — the honest reading leans against the flattering interpretation. A 2024 analysis of the Health and Retirement Study found that declining health drives declining purpose rather than the reverse, though its authors note that null may be under-powered for slow-developing conditions; separately, the epigenetic-age signal attenuated to non-significance once health behaviours were adjusted for. No trial has shown that raising purpose extends life. Worth doing; not yet a proven lever. See Purpose.
Slow-paced breathing, 10–20 min a day. About 6 breaths per minute is the highest-leverage short autonomic intervention; several weeks of practice raises baseline heart-rate variability and lowers resting sympathetic tone, though long-term durability is less certain than the acute effect. See Stress.
Mindfulness, if you'll do it consistently. Mindfulness-Based Stress Reduction and its cognitive-therapy variant reliably and safely reduce anxiety, depression, and pain; against gold-standard active controls the effect is small-to-moderate — about a third of a standard deviation for anxiety at 8 weeks, fading over months — and no better than other active treatments. The much larger figures sometimes quoted (nearly a full standard deviation) are uncontrolled before-and-after estimates in diagnosed patients, not the real effect. See Stress.
Learn something genuinely new — a language, an instrument, a skill — rather than doing more crosswords, and keep socialising alongside it. In one three-month randomised study, learning new skills improved memory where familiar leisure did not — a mechanism demonstration, not a hard-endpoint result. Social connection is the strongest late-life reserve marker in the life-course meta-analysis, ahead of cognitive activity, so the two aren't substitutes. What not to believe: the much-cited ACTIVE trial's "25% lower dementia" held only for speed-of-processing participants who also did booster sessions, was never significant in a whole-arm comparison, and carries the field's only harm signal for a cognitive intervention — among participants with 12 years of education or less, memory and speed training tracked higher long-term mortality; the crossword result came from people who already had mild cognitive impairment, with no do-nothing control and a null primary brain-imaging outcome; and the claim that bilingualism delays dementia by four to five years traces to a proponents' commentary that the best prospective meta-analysis and a 2026 cohort both fail to replicate. See Cognitive engagement.
Treat depression, regardless of when. Depression is both a risk factor for and a prodrome of dementia, and one of the Lancet Commission's 14 modifiable factors — carrying the largest risk increase of the fourteen for anyone who has it, ranking below hearing loss overall only because far fewer people are affected. See Dementia prevention.
Ask about the shingles vaccine at the age it's offered. A natural experiment in Wales, where a birthdate cut-off made the vaccine abruptly available to people born on or after 2 September 1933 but not to those born a week earlier, found dementia diagnoses over the following seven years fell by 3.5 percentage points in absolute terms — about a fifth in relative terms. The birthdate design mimics randomisation and sidesteps the healthy-vaccinee bias of ordinary observational work, and an independent Australian analysis found the same direction. It is quasi-experimental rather than a randomised trial, and it rests on the older live-attenuated vaccine — data for the now-standard recombinant one are still accruing — but it is the cleanest natural experiment in dementia prevention. See Dementia prevention.

6. The sensory and dental levers nobody mentions

Three underrated domains — hearing, vision and the mouth — sit at very different evidence tiers. Hearing loss is, jointly with high LDL cholesterol, the largest modifiable midlife dementia risk factor; vision is one of the smaller ones; and oral health earns its place on diabetes and pneumonia grounds rather than on the cardiovascular association, which the genetic evidence does not support as causal. All three are cheap and unevenly addressed.

Get audiometry between 40 and 50; use hearing aids if prescribed. Hearing loss ties with high LDL ("bad") cholesterol as the largest modifiable midlife dementia risk factor in the Lancet Commission 2024 update — and it leads on how common it is, not on how dangerous it is per person. Be careful with the treatment claim: the only randomised trial, ACHIEVE, was null on its primary endpoint, and the encouraging 48% slowing came from a higher-risk subgroup. The ASPREE 2026 target-trial emulation — an observational design, not a trial — found hearing-aid users had a 7-year dementia risk of 5.0% versus 7.5% without: about a third lower, or roughly 25 fewer cases per 1,000 people over seven years. Anyone telling you the question is settled is overselling it. If you struggle in noisy rooms despite a "normal" audiogram, ask for extended high-frequency (EHF) audiometry, which tests above the standard 8 kHz ceiling; one analysis suggests it explains about four times as much of the age-related difficulty understanding speech in noisy rooms as the standard test does — though that comparison comes from a single 263-adult non-peer-reviewed preprint, so treat the numbers as provisional. See Hearing.
Get a comprehensive dilated eye exam at midlife. Treat any correctable loss promptly — cataract surgery, glasses, low-vision rehabilitation, macular-degeneration management. Visual impairment was added to the Lancet 2024 list, but it is a much smaller dementia lever than hearing — explaining only about 2% of dementia cases across the population — and causality remains unresolved. Follow the age-tiered screening cadence; retinal "aging clock" analysis of fundus photos is a promising research tool, not yet a clinically validated test. See Vision.
Feed the macula — the light-sensitive centre of the retina. Leafy greens daily for lutein and zeaxanthin — the dietary pattern, not the pill. Higher macular pigment density tracks better visual performance (contrast sensitivity, dark adaptation, glare recovery), but the decisive cognitive test was null: an AREDS2 sub-trial randomised more than 3,000 older adults to lutein and zeaxanthin and found no effect on cognition. See Vision.
Brush twice daily, floss or use interdental brushes once daily, and see a hygienist every 6 months — every 3–4 months if you have gum disease, diabetes, or cardiovascular disease. Oral hygiene is strongly associated with lower cardiovascular mortality, but genetic studies and a 2025 American Heart Association review find the evidence does not support a causal heart link; the firmest interventional gains are narrower — better blood-sugar control in diabetes and preventing aspiration pneumonia in frail elders. Advanced gum disease is still a chronic low-grade inflammatory state affecting the whole body, and the main gum-disease bacterium has been found in the brains of Alzheimer's patients. See Oral health.
Don't routinely use chlorhexidine mouthwash. It kills the tongue bacteria that produce most of the body's nitric oxide after midlife. The blood-pressure consequence is not established — a 2026 meta-analysis found no clear effect, with systolic pressure 1.6 mmHg higher on mouthwash, a difference well inside the range expected from chance — so the reason to skip it is the absence of any benefit for a healthy mouth plus the depletion of those nitrate-reducing bacteria, not a proven pressure rise. Reserve for short-course or post-surgical use. See Oral health.

7. Controlled stress: sauna yes, sun yes, cold maybe

The principle that mild, controlled stressors trigger adaptive responses which leave you stronger — hormesis — is the framework that ties together sauna, sun, cold, fasting, and exercise itself. The evidence is sharply different across them.

Sauna 2–3× a week as a floor, 4–7 where the numbers peak. Four to seven sessions a week is associated with roughly 50% lower cardiovascular mortality and 40% lower all-cause mortality versus once weekly; two to three a week already tracks with meaningfully lower all-cause death. Sessions over 19 minutes predicted lower cardiac death but not lower all-cause mortality, so read the duration figure as cardiac-specific. The 80–100 °C figure describes the traditional Finnish sauna those cohorts used — temperature was never tested as a dose. This is the largest mortality signal of any lifestyle intervention on this site, but observational and essentially one cohort: the one independent replication found about 53% lower dementia over its first 20 years and 19% across the full follow-up, against the original cohort's 66%; that same cohort followed to 27.8 years shows the mortality advantage attenuate to about 14%; and a meta-analysis of 20 randomised passive-heating trials was null on every vascular and metabolic endpoint except systolic pressure in whole-body heating. Hydrate; cool down before driving. Avoid with severe aortic stenosis, unstable angina, recent heart attack, cardiac arrhythmia, impaired sweating, acute febrile illness, or alcohol; consult a clinician if pregnant. See Sauna.
Brief, sub-burning sun on most days. Strict avoidance tracks with roughly double the all-cause mortality of the highest-exposure group in the Melanoma in Southern Sweden cohort of about 29,500 Swedish women — observational, with self-reported exposure, and judged by the leading 2025 systematic review too variable to justify changing sun-protection guidance. Calibrate the dose to the UV index rather than the clock: roughly 15–20 minutes at UV index 3–5 for an intermediate skin tone with arms and legs uncovered, 10–15 at 6–7, under 10 above that. That's enough for vitamin D and for the ultraviolet-driven nitric-oxide release that nudges blood pressure down — real, but on the order of 1 mmHg across a seasonal swing, not a treatment. Past that window, broad-spectrum sunscreen and UV-blocking sunglasses. Never burn; never use tanning beds. Hard contraindications: very pale skin (Fitzpatrick types I–II), melanoma history, a dense crop of atypical moles, immunosuppression, photosensitising drugs — manage vitamin D orally instead. See Sun exposure.
Cold plunging is optional. Reliable short-term physiology — a surge of adrenaline-family hormones, a rebound into the body's rest-and-digest state, a real mood lift — modest adaptations, and no longevity outcome data. 11–15 °C for 10–15 min for recovery between hard endurance sessions — not after lifting, see below; 1–3 min cold shower or 2–5 min at 10–15 °C for mood and alertness. Avoid within ~4 hours of resistance training if muscle growth is the goal — post-training cold blunts the adaptation. Hard contraindications: cardiac arrhythmia (including long QT syndrome or prior cardiac arrest), severe Raynaud's, cold urticaria, pregnancy, recent heart attack; beta-blockers, blood-pressure drugs and diuretics all interact with the cold response, so raise it with a clinician. Never breath-hold while submerged. See Cold exposure.

8. Drinks

What you drink daily is a small lever pulled thousands of times. Coffee and tea are net-positive to neutral; the temperature you drink them at matters more than which one you choose.

2–4 cups of coffee a day. Across large cohorts, drinkers in that range have roughly 15–30% lower all-cause mortality than non-drinkers — consistent, but observational, and the genetic evidence doesn't confirm it. Associations run the same direction for diabetes, Parkinson's, liver disease and several cancers; caffeinated and decaf both show signal. Paper filtering removes the oils that raise LDL cholesterol slightly — worth choosing if your cholesterol is already elevated, otherwise either method qualifies. Watch evening timing; skip the added sugar. European and US regulators converge on up to 400 mg of caffeine a day as safe for healthy adults, 200 mg in pregnancy. See Coffee.
2–3 cups of tea if you enjoy it — where most cohort signals converge, though the dementia signal peaks lower, at 1–2 cups. Modest associations with lower mortality and cardiovascular mortality, observational and not confirmed by genetic analyses, plus about 14% lower dementia risk at 1–2 cups/day in the 2026 JAMA Nurses' / Health Professionals analysis — a figure the authors reported without a confidence interval — the range that shows how precise an estimate is — so read it as approximate. Avoid concentrated green-tea extract supplements (rare liver toxicity). See Tea.
Let hot drinks cool below about 60 °C before drinking. The cheapest item on this list, and it applies to coffee and tea alike — it is the temperature, not the drink. In the Golestan cohort of about 50,000 adults, with tea temperature measured objectively rather than self-reported, drinking at 60 °C or hotter carried 41% higher risk of cancer of the oesophagus — the tube from throat to stomach — rising to roughly 90% higher at high daily volumes; the International Agency for Research on Cancer classifies any beverage above 65 °C as probably carcinogenic for the same reason — thermal injury to the oesophageal lining. Typical Western drinking temperatures land around 56–60 °C, at the boundary rather than safely below it, which is why 60 °C is the number worth acting on. The cancer is uncommon in Western countries, so 41% more of a small number is still a small number — but this costs nothing, which is what makes it worth doing. The habit to change is drinking a fresh pour straight away. See Tea.
Sip water through the day rather than downing large amounts at once — but let thirst set the volume. Roughly 2.0–3.0 L/day of total fluids including the water in food for most adults; pale-yellow urine and rare thirst are the real biomarkers, and the international consensus on exercise-associated hyponatraemia is that drinking ahead of thirst is the wrong instruction, because over-drinking dilutes blood sodium and sports drinks don't prevent it. Sipping steadily probably beats chugging (~75% versus ~55% capture in one small post-exercise trial — plausible physiology, thin evidence). Blood sodium chronically above 142 mmol/L made people up to 50% more likely to score as biologically older than their actual age, in a cohort of 15,752 adults. The only clinical outcomes that randomised trials have actually shown drinking more water improves are kidney-stone recurrence and weight loss. See Water.
Filter for persistent industrial chemicals and microplastics. Reverse osmosis with remineralisation is the protocol of choice for most modern municipal supplies. Aim for roughly 10 mg/L magnesium and 20–30 mg/L calcium in the remineralised output — in a cohort of 26,733 postmenopausal Swedish women, higher-magnesium municipal water tracked with about 13% lower ischaemic stroke risk, with no effect on heart attacks, which is the reason not to run reverse osmosis without a remineralisation stage. See Water.
Skip "structured", alkaline and marine-plasma waters. Alkaline water shows no measured differences against plain mineral water; "structured" water is physicochemical pseudoscience; marine-plasma products drew a Health Canada contamination warning. Hydrogen water is the one functional-water exception with plausible biology and small trials — and no hard outcome data yet. See Water.
Skip "bulletproof" / butter coffee. Calorie-dense, no benefit over plain coffee, meaningful saturated-fat load. See Coffee.

9. Numbers and labs

The annual untargeted physical doesn't move all-cause mortality, but a short, evidence-anchored midlife panel does change clinical decisions and lets you track interventions over time.

Run the annual midlife panel. Apolipoprotein B (apoB), fasting insulin and HbA1c (your average blood sugar over about three months), high-sensitivity C-reactive protein, ferritin, uric acid, a full blood count and a metabolic panel. Once-in-a-lifetime: lipoprotein(a). Symptom-driven: thyroid-stimulating hormone, two morning total-testosterone draws, estradiol and follicle-stimulating hormone. See Midlife labs.
Skip the low-yield tests. Untargeted lipoprotein particle profiles when apoB is available, IgG food panels, salivary "adrenal stress" panels, routine insulin-like growth factor 1 in healthy adults, provoked heavy-metal challenges. These add cost and overdiagnosis, not signal. See Midlife labs.
Treat high blood pressure aggressively in midlife. Aim for a home-average systolic under 130 if tolerated. SPRINT — 9,361 high-risk adults over 50, excluding those with diabetes or prior stroke — treated to a below-120 clinic target and cut major cardiovascular events by 25% and all-cause mortality by 27% — a clinic-style reading, which is why the practical home-average target lands nearer 130; its companion SPRINT-MIND cut mild cognitive impairment by 19%. Two things the headline usually loses: the survival advantage began to fade 2–4 years after the trial ended once pressures drifted back, so this is lifelong management rather than a finite win; and under 130/80 is the target, not the threshold to start a drug — in the 130–139/80–89 range, medication is gated on a 10-year cardiovascular risk of 7.5% or more on the PREVENT calculator. Home monitoring beats one-off clinic readings; holding a wall-sit lowers systolic by ~8 mmHg, comparable to a first-line drug; bedtime dosing is not superior to morning (TIME and BedMed both negative). See Blood pressure.
Lower apoB early, not just eventually — artery damage is dose multiplied by time. Atherosclerosis biology is a dose × time integral; genetic studies show up to 3× greater benefit per unit of cholesterol lowered when intervention starts earlier. ApoB is at least as good a predictor as LDL cholesterol and better where the two disagree — about 25% of adults are discordant — though the case for acting on apoB rather than LDL rests on observational cohorts, not outcome trials. Statins, ezetimibe, bempedoic acid and PCSK9 inhibitors (injectable antibody drugs) reach the same final pathway by four different molecular routes, and each delivers benefit in proportion to how far it lowers apoB. In SAMSON, a crossover study in 60 adults with documented severe statin intolerance, about 90% of the symptom burden was also present on placebo. A coronary calcium scan is the highest-value single test if a statin decision is genuinely borderline. See Lipid management.
Test lipoprotein(a) once. It is more than 90% genetically determined, stable from about age five, and untouched by diet, exercise or statins, so a single lifetime measurement is all you need. A high result changes how hard you pull the other levers — nothing yet shows that lowering lipoprotein(a) itself reduces events, which is what the HORIZON trial of pelacarsen is powered to answer, and it had not reported as of mid-2026. See Lipid management.
Run the high-value cancer screens. Biennial mammography from 40 to 74; colorectal screening from 45 to 75; primary human papillomavirus (HPV) testing every 5 years from 30 to 65; low-dose computed tomography (CT) annually at 50–80 with 20+ pack-years for people who still smoke or quit within the last 15 years — outside those criteria it isn't justified; one-time hepatitis C antibody at 18–79; one-time abdominal aortic aneurysm ultrasound for men 65–75 who ever smoked. Mean life-extension across the six common screens is measured in days not years — the value is in the tail. See Cancer screening cadence.
Don't chase "subclinical" findings. Subclinical hypothyroidism in older adults rarely warrants levothyroxine (the TRUST trial); about 58% of cases revert to normal on their own over a median three years, and thyroid-stimulating hormone runs seasonally high, so a single elevated reading is usually noise. An estimated 72–94% of US papillary thyroid cancer diagnoses between 1991 and 2019 were overdiagnoses driven by neck ultrasound. See Thyroid management.
Get adequate iodine — but don't megadose. The 150 µg/day recommended intake, from iodised salt and seafood, is the dose with the cohort data behind it. The American Thyroid Association advises against chronic supplementation above 500 µg/day, and kelp or Lugol's solution can push past the ~1,100 µg/day upper limit and cause the disease they're marketed against. See Thyroid management.
Optionally, a third-generation epigenetic clock. DunedinPACE estimates the rate of biological aging and is reliable enough on repeat testing to serve as a personal longitudinal endpoint. But it is a stand-in measurement, not an outcome anyone actually experiences — that any individual reading should change a decision is not established — and every serial sample must go through the same provider, because scores aren't comparable across platforms and switching vendors destroys the only thing the test was for. See Midlife labs.
Track your resting heart rate — and don't try to lower the number directly. It is the cheapest measurement on this site and one of the most replicated: across more than a million people, mortality rises roughly 9–17% for every 10 beats per minute, and the association survives adjustment for directly measured fitness. Two caveats decide how to use it. A single reading against the textbook 60–100 range is close to meaningless — healthy individual averages span 40 to 109 beats per minute, so the only useful comparison is your own trend over weeks. And it is a marker, not a lever: lowering it with a drug helped in heart failure, did nothing in 19,102 people with coronary disease, and has never been tested in healthy adults. A sustained rise is a prompt to look for the cause — new medication, poor sleep, more alcohol, a thyroid problem, deconditioning, a developing illness. Aerobic training is what lowers it durably, over months rather than weeks. See Resting heart rate.

10. Prescription pharmacology, when indicated

Most levers on this site are behavioural, but a small set of prescription interventions delivers effect sizes lifestyle can't match — in the populations where they actually matter. A recurring pattern across the group, named Wilder's Law of Initial Value by the testosterone registry that reported it, is that benefit tends to be proportional to how far someone has already drifted from a healthy baseline. Read that as a heuristic for reading the evidence, not a finding in its own right: its three usual exhibits sit at three different evidence tiers.

GLP-1 receptor agonists — the Ozempic and Wegovy class — in overweight or obesity with established cardiovascular disease. SELECT cut major cardiovascular events by 20% in non-diabetic adults with overweight and established cardiovascular disease — which shows the effect exists there, not that it is largest there. Semaglutide has also moved validated epigenetic aging clocks in a randomised trial, but read the design first: a post-hoc, exploratory analysis of 45 participants on drug and 39 on placebo, in adults selected for accelerated biological aging, with the clocks scored by the vendor that sells them, in an analysis that has not been peer-reviewed — and one clock in the trial didn't move at all. It places this drug class alongside, not above, the rest of the geroprotector field. The catch: 25–40% of weight lost is fat-free mass, and hip bone density fell about 2.6% over 52 weeks in a trial of adults already at elevated fracture risk. Without 1.6–2.2 g/kg of protein a day — above the general midlife target, because the drug's calorie deficit attacks muscle — and 2–4 resistance sessions a week, you arrive at sarcopenic obesity. Also worth knowing before starting: nausea in about 44% versus 16% on placebo, gallbladder disease up about 37% and more than doubled in weight-loss trials specifically, and a signal for sudden painless loss of vision in one eye — any new visual loss means same-day ophthalmology, and the drug should not be restarted. See GLP-1 receptor agonists.
Testosterone replacement in symptomatic, biochemically confirmed deficiency. TRAVERSE (2023, n=5,204) settled the cardiovascular fear — heart attack, stroke and cardiovascular death occurred at essentially equal rates on drug and placebo, and the US Food and Drug Administration removed its strongest ("boxed") safety warning in 2025. But in the same action the FDA added a class-wide blood-pressure warning, TRAVERSE itself showed more atrial fibrillation, acute kidney injury and pulmonary embolism, and a pre-specified substudy found no reduction in progression to type 2 diabetes. About 25% of men start treatment without a baseline workup, the most common avoidable mistake. Optimise sleep, sleep apnea, alcohol, exercise and endocrine-disrupting-chemical exposure first. Check haematocrit — the share of your blood made up of red cells — every 6–12 months; above 54% is the bright line: cut the dose, lengthen the interval, switch to gel, or give blood. Testosterone also halts sperm production, so preserving fertility needs a different protocol. See Testosterone therapy.
Menopausal hormone therapy, within the 10-year window. The indications are hot flashes, the vaginal dryness and urinary symptoms of menopause, where local vaginal estrogen is the highest-leverage intervention in the field, and early postmenopausal bone loss. Estrogen initiated within 10 years of menopause preserves endothelial function; observational cohorts, carrying a healthy-user bias no adjustment removes, associate early initiation with 30–50% lower coronary heart disease, and the early-initiation subgroup of the trial meta-analysis shows about 30% lower all-cause mortality (relative risk 0.70; Benkhadra 2015, against a null overall). KEEPS and ELITE confirm the timing-dependent vascular response but used surrogate endpoints and reduced no hard events, and the 2024 twenty-year review of the Women's Health Initiative is explicit that hormone therapy should not be used to prevent chronic disease at any age. Initiated more than 10 years out, the cardiovascular signal flips. Transdermal beats oral for clot safety; micronized progesterone beats the synthetic progestins for breast safety. Fracture odds rebound toward never-user levels in the 1–10 years after stopping, so the cessation plan matters. See Menopausal hormone therapy.
Treat type 2 diabetes hard. Well-controlled diabetes is a different thing from uncontrolled diabetes for cognitive outcomes. Metformin remains the established first-line drug; GLP-1 drugs and SGLT2 inhibitors — a class that makes the kidneys excrete surplus glucose — have proven heart and kidney benefits in people who already have diabetes or kidney disease. See Dementia prevention and Geroprotectors.
Target the middle of the testosterone reference range, not the top. Genetic data on lifelong higher testosterone show about 17% higher coronary risk in men — in absolute terms a lifetime risk of roughly 7.3% rising to about 8.5% — mediated largely by blood pressure, and null in women. That is compatible with TRAVERSE, because late-life restoration is a different thing from cumulative lifelong exposure, but it is a strong argument against "optimizing" a healthy adult toward 25-year-old levels. See Testosterone therapy.
Avoid compounded GLP-1s and compounded female testosterone. Both bypass regulatory oversight on dose accuracy, sterility, and salt formulation. Global consensus discourages compounded female testosterone specifically. See GLP-1 receptor agonists and Testosterone therapy.
Don't microdose GLP-1 "for longevity" if you're healthy. Mechanistically plausible; zero trial evidence; compounded supply chain. Speculative. See GLP-1 receptor agonists.

11. Supplements: a short useful list, a longer pointless list

No supplement on the market has been shown to extend life or prevent disease in healthy middle-aged adults at anything like the magnitude of basic lifestyle interventions. Five clear a usable bar; almost everything else is gap-specific or hype; a few actively cause harm. Test before you supplement anything fat-soluble, mineral, or chronically dosed — with one caveat about the most popular test: both the US Preventive Services Task Force and the Endocrine Society advise against routine vitamin D screening in healthy adults, so testing earns its keep when you want to dose deliberately rather than guess, not as a population screen.

The five that earn their place

Vitamin D3, if your blood level is low. Target 75–125 nmol/L (30–50 ng/mL). 1000–2000 international units (IU) a day usually does it; up to 4000 IU/day for severe deficiency. Take with a fat-containing meal. Don't routinely megadose — above 4000 IU/day without monitoring, raised blood calcium is rare but real. See Vitamin D.
EPA and DHA omega-3, about 1 g/day, if you don't eat fish. Two-plus servings of fatty fish per week makes the supplement unnecessary. DO-HEALTH produced a randomised signal that 1 g/day slows DNA-methylation aging clocks — a post-hoc readout in 777 of its 2,157 participants, in a trial that missed all six of its pre-specified primary endpoints, and vitamin D and exercise moved the clocks too. The trials show a dose-dependent excess of atrial fibrillation — about 25% overall, but roughly 50% above 1 g/day against a smaller though real 12% at or below it, which is why 1 g is the default and going to 2 g should have a triglyceride or cardiovascular reason behind it. Refrigerate after opening; look for International Fish Oil Standards or Labdoor marks. Algae oil raises blood omega-3 as well as fish oil does, and works for vegans. See Omega-3.
Magnesium, 200–400 mg/day — noting that the top of that range sits above both regulatory ceilings for supplemental magnesium (250 mg/day in the EU, 350 mg/day in the US), so it isn't "more is better" territory. About half of adults don't meet the recommended intake from food; supplemental magnesium lowers blood pressure mainly where there's a deficit to correct — about 7.7 mmHg systolic in people treated for hypertension, somewhat less in the magnesium-deficient, and no significant effect at all in people already normotensive. Glycinate or threonate in the evening if you want the sleep effect: it is real but small and low-certainty, on the order of one extra responder per seven people treated; citrate any time. Be cautious in chronic kidney disease; separate doses from quinolones, tetracyclines, and bisphosphonates. See Magnesium.
Vitamin B12 in specific groups. Vegetarians and vegans; anyone on metformin or chronic acid-suppressing drugs; adults past about 60, where age-related absorption decline becomes common. 25–250 µg/day for prevention; 1000 µg/day for confirmed deficiency. Don't trust serum B12 alone — confirm with a methylmalonic acid test, which catches a true functional shortfall that serum B12 can miss. See B vitamins.
Creatine monohydrate, 3–5 g/day, paired with resistance training. Strong for muscle and strength when you actually lift — creatine alone, without the training, does little; moderate for cognition under metabolic stress and in older adults. It has one of the deepest safety records in nutrition: a 2025 review across 685 trials in 12,839 participants, at doses to 30 g/day for up to 14 years, found no excess of serious adverse events. Stick with monohydrate — hydrochloride, nitrate and ethyl ester cost more and don't outperform. ~10 g/day for brain effects. See Creatine.

Situational add-ons (case-by-case, not universal)

Multivitamin, no iron. A small improvement on a telephone-administered cognitive test in COSMOS-Mind — a test score, not a dementia diagnosis, on donated supplements — against a US Preventive Services Task Force "insufficient evidence" rating. A gap-filler for a poor diet, not a longevity intervention. See Dementia prevention and Supplements.
Vitamin K2 (menaquinone-7), 90–180 µg/day paired with D3. It activates the proteins that route absorbed calcium into bone and away from the artery wall. Two trials are positive on surrogate endpoints — arterial stiffness, coronary-calcium progression — against an equal weight of nulls, including a larger higher-dose trial and the kidney-disease trials where the biomarker moves but the outcome doesn't; and every positive trial used product from the dominant commercial manufacturer, so the independent evidence is thinner than the trial count suggests. Cheap and safe, but unproven on hard outcomes. Directly opposes warfarin, though not the newer direct oral anticoagulants. See Vitamin K.
Calcium from food, 1,000–1,200 mg/day — supplement only the gap. Dietary calcium is cardioprotective; large isolated pills are the opposite, tracking with a probable rise in heart attacks, and the fracture benefit is confined to the deficient and institutionalised elderly. If you do supplement, never more than ~500 mg elemental at once, always with a meal, citrate or hydroxyapatite. See Calcium.
Folic acid 400 µg/day for women capable of pregnancy. See B vitamins.
Coenzyme Q10, 100–200 mg/day — adjunct in heart failure; mixed evidence for statin muscle pain. No mortality benefit in the general population — and about 45% higher all-cause mortality among users with obesity in cohort data. See CoQ10.
Iron only with documented deficiency (ferritin below 30 µg/L in premenopausal women; men over 50 sometimes need less, not more). The body has no route to excrete it, so high-normal stores are a risk signal rather than a safety margin. See Iron and aging.

Things to skip or be careful with

Most "longevity molecules" — nicotinamide mononucleotide (NMN), resveratrol, urolithin A, spermidine, lithium orotate — currently rest on animal data, surrogate biomarkers, or short-term industry-funded trials. No hard outcomes in healthy adults. See Geroprotectors.
If you use melatonin, dose it to the purpose, and keep it short-term rather than an open-ended nightly habit. Use 0.3–1 mg to shift the clock. If the goal is simply falling asleep — on arrival after a long flight, say — up to 5 mg of the immediate-release form is reasonable, and a dose-response pooling of 26 randomised trials puts the sleep-onset optimum near 4 mg, given about three hours before bed.^[1] Doses of 10 mg are not justified by either target, and higher doses are more likely to leave you groggy the next day, and the 2017 American Academy of Sleep Medicine guideline actually recommends against melatonin for ordinary insomnia; the evidence is strongest in older adults whose own melatonin output has fallen. Dose accuracy is a real problem: over-the-counter melatonin in the US ranges from 83% below to 478% above the labelled amount^[2], and a US Food and Drug Administration survey of 110 children's products found content anywhere from 0% to 667% of the declared amount.^[3] Prolonged-release prescription melatonin (Circadin 2 mg in Europe) is far more reliably dosed — but do not reach for it for jet lag specifically, where the Cochrane review found slow-release 2 mg relatively ineffective.^[4] And long-term prescription users carry a heart-failure and mortality signal — observational, from a conference abstract that still hasn't been peer-reviewed, and describing people whose insomnia was severe enough to be prescribed a drug for a year or more rather than someone taking an occasional low-dose tablet. See Sleep supplements.
Over-the-counter antihistamine sleep aids, benzodiazepines, and Z-drugs as a regular habit — and from 65, the formal guidance says avoid the antihistamines outright. Falls and fractures for all three; the dementia signal specifically attaches to the anticholinergic antihistamines. Below 65 there is no age threshold in the evidence — the reason to skip diphenhydramine earlier is that its sedative effect is gone within about four days of nightly use. See Supplements to avoid and Sleep supplements.
Megadose single nutrients. Beta-carotene in current or former smokers (cancer increase, in the CARET and ATBC trials); high-dose vitamin E (mortality increase in pooled meta-analyses); chronic high-dose zinc (copper depletion); vitamin B6 above the 12 mg/day European ceiling (nerve damage in the hands and feet — check your B-complex label, many carry 50–100 mg); calcium supplements above ~1,000 mg/day in non-deficient adults. See Supplements to avoid.
Cheap multivitamins that use the least bioavailable forms. Look for USP Verified, NSF Certified for Sport, or Informed Sport on the bottle. See Supplements.

12. Clean up your environment

Estimates of the inherited share of lifespan cluster around 15–25%, leaving roughly three-quarters to four-fifths non-genetic — though the range is contested in both directions and this site doesn't pick a number (see Foundations). Whatever the exact split, lifestyle is one part of the non-genetic share; the exposome — air, water, household chemicals, psychosocial stress — is the other. Specific chemicals (cadmium, lead, phthalates, bisphenols, and the perfluoroalkyl "forever chemicals") track measurably with faster epigenetic-aging clocks at population scale, and substitution interventions drop urinary biomarkers of the non-persistent ones within weeks.

Test your home for radon. The second leading cause of lung cancer after smoking, with about 16% higher risk per 100 becquerels per cubic metre and no threshold below which it stops mattering. A test kit costs about what a meal out does, and a bad result has a cheap engineering fix — the best evidence-to-cost trade in this section. It combines multiplicatively with smoking, so smokers benefit most; but for a never-smoker it is still plausibly the single highest-yield environmental hazard there is. See Environmental toxins.
Filter indoor air if you live near major roads. Each 5 µg/m³ of long-term fine-particulate exposure tracks with about 8% higher dementia incidence — moderate certainty, with socioeconomic confounding hard to rule out. A bedroom filter is the intervention with the cleanest randomised data. See Environmental toxins.
Get plastic out of food contact. Glass, stainless steel, ceramics. Don't microwave plastic; don't drink hot liquids from plastic-lined cups. See Environmental toxins.
Skip synthetic fragrance. Phthalates and other endocrine disruptors track measurably with urinary biomarkers, and substitution interventions drop those biomarkers within weeks. See Environmental toxins.
Wear seatbelts and helmets; avoid repeated head impacts. Contact sports and motorcycling without a helmet carry a real dementia signal mediated by traumatic brain injury. See Dementia prevention.

That covers most of the practical landscape. Each pillar page on this site goes deeper on its own group; each topic article holds the trial-level evidence behind the headline numbers.