Landmark studies

The eighteen studies the rest of the site cites — the headline, the caveats and the link. Every study has limits.

37 / 37 results

Who, how long

Topic

Year

Cohort2025 · Nature

Shingles vaccine cuts dementia 20% — a natural experiment

Those born a week after the cut-off (eligible) had 20% fewer dementia diagnoses over 7 years — a quasi-randomised design, hence near-causal.

⚠ Concerns the old live vaccine (Zostavax); Shingrix data are still accumulating, but point the same way.

Cohort2024 · Sleep

Sleep regularity predicts mortality better than sleep duration

The most irregular fifth of sleepers had 48% higher mortality than the most regular — independent of sleep duration.

⚠ One week of measurement stands in for the whole sleep pattern; shift work and illness may confound.

Ongoing2024 · Nature Aging (protokoll) / AFAR

TAME: can metformin become the first official anti-ageing drug? (ongoing)

The trial targets not a disease but “ageing” as an endpoint: delaying the COMPOSITE first occurrence of cancer, heart disease, dementia and death — if it succeeds, the FDA would for the first time recognise ageing as a treatable target. Background: in cohorts, diabetics on metformin outlived non-diabetics (Bannister 2014).

⚠ Not yet fully launched; newer data weaken the hypothesis: in the ITP metformin didn't extend mouse lifespan, the Keys 2022 analysis attributes the Bannister result to methodological bias, and metformin blunts exercise adaptation (Konopka 2019).

RCT2023 · Nature Aging / Lancet Diabetes Endocrinol

CALERIE-2: two years of calorie restriction slow biological ageing

An achieved 12% calorie cut over 2 years: –10% body weight, better LDL, blood pressure and insulin sensitivity, and the DunedinPACE epigenetic clock slowed 2–3%.

⚠ Young, healthy participants; bone density dipped slightly; no direct lifespan data.

RCT2023 · NEJM

SELECT: semaglutide cuts cardiovascular events 20% in non-diabetics

Weekly 2.4 mg semaglutide: –20% major cardiovascular events, –9% body weight; the benefit emerged before the weight loss.

⚠ Manufacturer-funded; 16.6% discontinued for side effects; no detailed data on muscle loss.

Mechanistic / animal2023 · Science

Taurine deficiency as a driver of ageing

Taurine supplementation in middle-aged mice extended life 10–12% with better health markers; monkeys showed improved bone density and glucose.

⚠ No human RCT. A 2025 Science paper found taurine does NOT consistently decline with age in humans, so it isn't a good ageing biomarker.

RCT2023 · Lancet

ACHIEVE: hearing aids slow cognitive decline 48% in high-risk adults

No difference overall, but in the higher-risk ARIC subgroup hearing aids slowed cognitive decline by 48%.

⚠ The primary endpoint was negative; the subgroup result is hypothesis-generating, though pre-specified.

RCT2023 · NEJM

TRAVERSE: testosterone replacement doesn't raise cardiovascular risk

Transdermal testosterone gel didn't increase major cardiovascular events (7.0% vs 7.3%) — closing the debate opened by the 2014 FDA warning. Sexual function, mood and anaemia improved; fractures, surprisingly, increased.

⚠ Applies only to confirmed deficiency with symptoms; atrial fibrillation, pulmonary embolism and acute kidney injury were more common on treatment. It does NOT answer for healthy men with normal testosterone.

RCT2023 · NEJM

SELECT: semaglutide cuts cardiovascular events 20% in overweight non-diabetics

Weekly 2.4 mg semaglutide gave –20% heart attack/stroke/CV death, –19% all-cause mortality; –9.4% body weight. The benefit appeared before substantial weight loss — part of the effect isn't via weight.

⚠ Secondary prevention (already heart patients); 16.6% stopped for side effects (mostly gastrointestinal); 25–40% of lost weight is lean mass — sarcopenia risk without strength training and protein. Funded by Novo Nordisk.

Meta-analysis2023 · J Stud Alcohol Drugs

Stockwell: “a little alcohol protects” is a cohort artefact — 107 studies

Moderate drinkers' apparent 14% mortality advantage vanished once studies removed former and “sick quitter” drinkers from the abstainer group and measured consumption at younger ages. No protective effect at low doses; risk rises above 25 g a day.

⚠ A meta-analysis of heterogeneous studies; Mendelian randomisation (Millwood 2019, n=512,715) reached the same conclusion by an independent method.

Meta-analysis2022 · Br J Sports Med

Strength training and mortality: meta-analysis of 16 cohorts

30–60 min/week of strength training: –10–17% all-cause mortality, cardiovascular disease and cancer; –40% combined with aerobic activity. J-curve: no extra benefit beyond 130 minutes.

⚠ Self-reported activity; most cohorts are US and Japanese.

RCT2022 · AJCN / Alzheimer's & Dementia

COSMOS: cocoa flavanols –27% cardiovascular death; multivitamin improves memory

500 mg cocoa flavanols: primary endpoint (all CV events) –10%, not significant; CV death –27% (significant). A daily multivitamin (Centrum Silver) improved global cognition by the equivalent of ~2 years of ageing versus placebo (COSMOS-Mind, -Web).

⚠ The primary endpoint was missed — mortality is a secondary analysis. The multivitamin effect didn't appear in an earlier large RCT (PHS II). Co-funded by Mars Inc. and Pfizer.

Meta-analysis2022 · Lancet Public Health

Paluch: 15 cohorts — the step-count benefit peaks at 6,000–8,000 after 60

The top quarter of steppers had 40–53% lower mortality than the bottom. Benefit plateaus at 6,000–8,000 steps over 60 and 8,000–10,000 below; the steepest gain lies between 2,000 and 6,000. Step pace added nothing beyond step count.

⚠ Observational data: sick people walk less (reverse causation), though excluding the first 2 years didn't change the result. A single 1–7-day measurement represents the whole follow-up.

RCT2021 · Science

NMN improves muscle insulin sensitivity in prediabetic women

250 mg/day NMN improved insulin-stimulated glucose uptake in muscle by 25% — the first placebo-controlled human NMN data.

⚠ Very small sample; body weight, lipids and blood pressure unchanged; the effect size is clinically modest.

RCT2021 · NEJM

SSaSS: potassium-enriched salt substitute cuts mortality 12% in 21,000 people

Replacing 25% of table salt with potassium chloride gave –14% stroke, –13% major cardiovascular events, –12% all-cause mortality; hyperkalaemia didn't rise.

⚠ A high-salt (11 g/day), rural Chinese population; kidney patients were excluded — potassium can be dangerous for them.

RCT2020 · NEJM

SAMSON: 90% of statin side effects appear on placebo too

Each patient took 4 months each of atorvastatin, placebo and nothing in random order, blinded. The symptom score on placebo was almost identical to that on statin (nocebo ratio 90%); 90% of symptoms came from taking a pill, not the drug. Six months later half the participants restarted statins.

⚠ Small sample; rare, genuine statin myopathy (with CK elevation) exists — for the individual a blinded trial decides, not the statistics.

RCT2019 · NEJM

VITAL: vitamin D and omega-3 — no blanket protection, but subgroups benefit

2,000 IU D3 and 1 g fish oil did not reduce overall cancer or cardiovascular events; fish oil gave –19% events in low fish eaters, D3 signalled –17% cancer mortality and –22% autoimmune disease.

⚠ Most participants weren't vitamin D deficient — the trial doesn't answer for deficient people.

RCT2019 · NEJM

REDUCE-IT: high-dose EPA cuts cardiovascular events by 25%

4 g/day icosapent ethyl: –25% major cardiovascular events and –20% cardiovascular death.

⚠ The placebo was mineral oil, which worsened LDL — part of the effect may stem from that. STRENGTH (mixed EPA/DHA) was neutral.

Meta-analysis2019 · Lancet

Fibre and health: 185 cohorts and 58 RCTs

Highest fibre intake: –15–30% all-cause mortality, cardiovascular death, type 2 diabetes and colorectal cancer; RCTs showed lower weight, blood pressure and cholesterol.

⚠ The dose-response keeps rising beyond 25–29 g/day; most Europeans eat 15–20 g.

RCT2019 · NEJM

D2d: vitamin D doesn't prevent diabetes — except in the deficient

In the whole group diabetes incidence fell 12%, not significant. In those with low baseline levels (< 12 ng/ml), though, –62%; the 2023 individual-data meta-analysis (3 RCTs, n=4,190) found –15% progression to diabetes in prediabetes when serum levels rose above 50 ng/ml.

⚠ 80% of participants started with sufficient vitamin D — the “give everyone” model failed again, while “test, then supplement” was vindicated.

RCT2019 · Cell Metabolism

Hall: an ultra-processed diet causes +500 kcal of overeating a day — inpatient RCT

Two weeks of ultra-processed and two of unprocessed food, matched for calories, sugar, fat, fibre and macronutrients, eaten ad libitum. During the UPF period participants ate 508 kcal more per day, ate faster, and gained 0.9 kg; on unprocessed food they lost 0.9 kg.

⚠ Small sample, short duration, artificial setting; the mechanism (energy density, eating rate, protein dilution, flavour enhancers) can't be separated. It remains the field's only controlled feeding RCT.

Meta-analysis2019 · Cochrane Database Syst Rev

Sherrington (Cochrane): exercise programmes mean 23% fewer falls in older people

Exercise programmes gave –23% fall rate and –15% people falling; balance and functional exercises –24%, combined with strength training –34%, tai chi –19%. 3+ hours a week and balance challenge are the keys to effectiveness; walking alone doesn't reduce falls.

⚠ Evidence for fractures and hospital admission is weaker (fewer studies); applies to community-dwelling, independently mobile older people, not those in care homes.

Cohort2019 · JAMA Netw Open

Alimujiang: lack of life purpose means 2.4× mortality after 50

The fifth with the lowest purpose scores had 2.4× the mortality of the highest, adjusted for age, sex, education, health status and depression; the link is strongest for cardiovascular and digestive deaths. The effect is dose-dependent and appears within 4 years.

⚠ Observational: hidden illness may lower the sense of purpose (reverse causation), though adjustment and the 4-year lag temper this. “Raising purpose” as an intervention isn't yet proven in an RCT.

RCT2018 · NEJM

PREDIMED: Mediterranean diet cuts cardiovascular events by 30%

A Mediterranean diet supplemented with extra-virgin olive oil or nuts reduced heart attack, stroke and cardiovascular death by 30% versus a low-fat control.

⚠ The 2013 version was retracted over randomisation flaws and re-analysed (2018) — the result held. Mediterranean population.

Cohort2018 · JAMA Network Open

No upper limit to fitness: 122,007 exercise tests

The fittest 2.3% had 80% lower mortality than the least fit; low fitness was a bigger risk than smoking, diabetes or heart disease.

⚠ Observational: patients referred for exercise testing aren't a random sample; reverse causation (sick people are unfit) may explain part.

Cohort2018 · AJCN

Spermidine intake and mortality in the Bruneck cohort

The highest-intake third had mortality as low as if they were 5.7 years younger.

⚠ Food-frequency questionnaire; spermidine-rich foods (whole grains, legumes, cheese) also mark a healthy diet. The SmartAge RCT (2022) found no memory benefit from supplements.

Cohort2018 · Genetics

Ruby: lifespan heritability is only ~7% — a 400-million-person family tree

Spouses' lifespans correlated almost as strongly as siblings' — the earlier 20–30% heritability estimates came from assortative mating. Filtering that out, true genetic heritability is under 7%; lifespan is overwhelmingly environment and lifestyle.

⚠ A historical cohort (infections, war, child mortality) — in today's era of chronic disease genetics may matter more; past 100 it demonstrably does.

Cohort2018 · Circulation

Li: five habits add 12–14 years of life at 50

Not smoking, BMI 18.5–25, 30+ minutes of daily moderate activity, moderate alcohol, a good diet score: people keeping all five had 14 (women) and 12 (men) more years of life expectancy at 50 than those keeping none; –74% all-cause, –82% CV, –65% cancer mortality.

⚠ Health professionals, self-report, observational; the “moderate alcohol” category is no longer considered protective given the J-curve debate. The magnitude, though, was replicated in other cohorts (UK Biobank).

Cohort2018 · Aging

Levine: PhenoAge — 9 routine lab values predict death better than age

Albumin, creatinine, glucose, CRP, lymphocyte percentage, MCV, RDW, ALP, white cell count plus age: every +1 year of the resulting “phenotypic age” (above calendar age) means 9% higher mortality, across every disease group. The DNAm PhenoAge epigenetic clock built on it predicts mortality better than the Horvath clock.

⚠ A population-level predictor; at the individual level day-to-day lab variation shifts it ±2–3 years. It's unproven that lowering PhenoAge (e.g. by pushing down CRP) yields longer life.

RCT2015 · NEJM

SPRINT: intensive blood-pressure control cuts mortality by 27%

A <120 mmHg systolic target: –25% major cardiovascular events and –27% all-cause mortality versus <140; SPRINT-MIND found –19% mild cognitive impairment.

⚠ More hypotension, syncope and acute kidney injury in the intensive arm; diabetics were excluded.

Cohort2015 · JAMA Internal Medicine

Sauna bathing and mortality in the Finnish KIHD cohort

4–7 saunas a week: –50% cardiovascular death and –40% all-cause mortality versus once a week; sessions over 19 minutes were better.

⚠ Men only; frequent sauna-goers may be wealthier and healthier; no RCT.

RCT2015 · Lancet

FINGER: a multidomain lifestyle programme slows cognitive decline

Diet plus exercise plus cognitive training plus vascular risk management together: the overall cognitive score improved 25% more, executive function 83% more and processing speed 150% more than in controls receiving advice only.

⚠ The effect is statistically significant but modest in absolute terms; the trial was too short for dementia incidence. The global WW-FINGERS network (30+ countries) is replicating it.

RCT2013 · JAMA

AREDS2: lutein and zeaxanthin slow macular degeneration

Replacing beta-carotene with lutein (10 mg) + zeaxanthin (2 mg) in the AREDS formula cut progression to late AMD by 18%; beta-carotene doubled lung cancer in former smokers. Omega-3 (1 g) had no effect. The 10-year follow-up confirmed lutein's advantage.

⚠ Studied in existing AMD — no evidence for prevention in people with healthy eyes. The formula also contains vitamins C, E, zinc and copper; the effect belongs to the whole.

RCT2011 · PNAS

Erickson: a year of walking grows the hippocampus 2% after 65

Three 40-minute moderate walks a week grew hippocampal volume 2% (reversing 1–2 years of age-related shrinkage) while the stretching control lost 1.4%; spatial memory improved in proportion to BDNF levels.

⚠ Small sample; memory improved in both groups (practice effect); MRI volume is a surrogate endpoint, not dementia.

Meta-analysis2011 · JAMA

Studenski: gait speed as the “sixth vital sign” of survival

Every +0.1 m/s of gait speed meant 12% lower mortality. Above 0.8 m/s life expectancy reaches the average, above 1.2 it exceeds it; gait speed predicted survival as well as age, sex and chronic diseases combined — measurable in a 4-metre corridor.

⚠ Slow walking may signal illness rather than cause it — but interventional trials (strength training) improve both speed and survival.

Meta-analysis2010 · PLoS Medicine

Social relationships and mortality: meta-analysis of 148 studies

People with strong social relationships had 50% higher odds of survival — an effect comparable to quitting smoking and larger than obesity.

⚠ Observational; sick people may withdraw socially (reverse causation), though the authors controlled for it.

Mechanistic / animal2009 · Nature

Rapamycin extends lifespan even in old mice (ITP)

Rapamycin in the feed: +14% (female) and +9% (male) median lifespan — the first drug to work in mammals, started late, across three labs simultaneously.

⚠ Mice. In humans, immunosuppressive doses have known side effects; low, intermittent dosing looks safe (PEARL 2024), with no lifespan data.

Last reviewed: 6 September 2026

This site is educational and does not replace medical advice. Start any supplement or drug only after consulting a doctor or pharmacist.