Longevity Supplements 2026: The Evidence Tiers

The first dorsal interosseus is the small muscle that closes the space between the thumb and the index finger. It is not a muscle anyone trains, but it is easy to isolate, easy to fatigue, and easy to put inside a magnet, and so, at a medical center and a cancer research center in Seattle between March 1, 2018 and July 30, 2020, sixty-six adults between the ages of 65 and 90 were asked to contract it, over and over, until it would not contract again. The contractions were counted, and the same was done for the tibialis anterior along the front of the shin. That count — repetitions to fatigue — was a secondary endpoint. The primary endpoints were how far a person could walk in six minutes and the maximal rate at which the hand muscle could make ATP, measured by magnetic resonance spectroscopy. Mean age 71.7 years; fifty of the sixty-six were women, and all were White. Half took 1,000 mg of urolithin A daily for four months; half took placebo, and nobody knew which. The trial ran under NCT03283462.

At four months the walking test showed a mean gain of 60.8 meters on urolithin A and 42.5 meters on placebo, a difference that did not reach significance. Maximal ATP production in the hand muscle changed by 0.07 mM/s with urolithin A and 0.06 mM/s with placebo; in the shin muscle it fell slightly with urolithin A and rose slightly with placebo. Both primary endpoints, in other words, were misses. What did move was the count: at two months, repetitions to fatigue rose by a mean of 95.3 in the hand and 41.4 in the shin on urolithin A, against 11.6 and 5.7 on placebo. Plasma acylcarnitines, ceramides and C-reactive protein fell relative to placebo by four months. Adverse events did not differ. This is a genuine finding and a modest one, and the distance between what that trial measured and what the category promises is the subject of this article.

What follows ranks the supplements most often sold for longevity by one criterion: the strength of the human evidence that the compound changes an outcome a person would notice or a doctor would treat. Not the elegance of the mechanism, and not the number of pathways it touches in a cell culture. Molecules with beautiful biology and no human outcome data rank at the bottom here. Two things are true at once in this field: a handful of cheap, unglamorous compounds have randomized evidence for real functional gains, and most of the expensive ones have never been tested against an endpoint that matters.

How we ranked them

The grading rule is a hierarchy of what was measured, and in whom.

• Strongest. Randomized, placebo-controlled human trials with clinical endpoints (heart attack, fracture, cancer, death) or well-validated functional endpoints (grip strength, gait speed, chair-stand time, distance walked, a scored memory test), replicated across more than one trial or pooled in a meta-analysis.
• Middle. Randomized human trials whose endpoints are biomarkers — a lipid panel, an inflammatory marker, a methylation clock, a blood NAD concentration — with no demonstrated effect on how a person functions or how long they live.
• Weakest. Cohort and observational data, in which people who take the supplement are compared with people who chose not to; and mechanism, cell and animal work, however striking.

Two cautions belong here rather than in a footnote. First: longevity has essentially never been the endpoint of a human supplement trial. Nothing in this article tested whether a compound extends human lifespan, because no such trial exists in this evidence pool. All-cause mortality appears as a secondary or safety outcome in a few large trials, and where it does, it is flat. In VITAL, 978 deaths occurred over a median 5.3 years among 25,871 adults; the hazard ratio was 0.99 (95% CI, 0.87 to 1.12) for vitamin D and 1.02 (95% CI, 0.90 to 1.15) for marine omega-3s. A meta-analysis of 52 randomized trials and 75,454 participants put vitamin D's effect on all-cause mortality at a risk ratio of 0.98 (95% CI 0.95 to 1.02). Everything below is a proxy for aging well, not a measurement of aging.

Second: a within-group change is not a between-group difference. A supplement that improves a marker "compared with baseline" has told you almost nothing, because the placebo group usually improves too.

Do anti-aging supplements work?

For the large, hard endpoints — cardiovascular events, cancer, death — the honest answer for most of the category is no, and this has now been tested at scale. VITAL randomized 25,871 US adults, including 5,106 Black participants, to 2,000 IU/day of vitamin D3 and/or 1 g/day of marine omega-3s. Over a median 5.3 years, vitamin D did not reduce invasive cancer (HR 0.96; 95% CI, 0.88 to 1.06; P=0.47) or major cardiovascular events (HR 0.97; 95% CI, 0.85 to 1.12; P=0.69); omega-3s did not reduce major cardiovascular events (HR 0.92; 95% CI, 0.80 to 1.06; P=0.24) or cancer (HR 1.03; 95% CI, 0.93 to 1.13). DO-HEALTH randomized 2,157 adults aged 70 or older — mean age 74.9, 61.7% women — to three years of 2,000 IU/day vitamin D3, 1 g/day omega-3s, and a home strength-training program in every combination. Across six primary endpoints, including blood pressure, physical performance, cognition, fracture and infection, there were no statistically significant benefits of any intervention alone or combined. Twenty-five deaths occurred, distributed similarly across the arms.

For narrower, functional endpoints the answer changes. Creatine paired with resistance training adds lean tissue and strength in older adults. Protein and its metabolite HMB, paired with the same training, improve grip strength and gait speed in people who already have sarcopenia. A daily supermarket multivitamin produced a small improvement on memory testing in a 3,562-person substudy of a 21,442-person trial. Those are real effects on things that matter. They are also small, they require the training or the deficit to be present, and none has been shown to extend life.

The aggregate answer: a few of these compounds do something measurable, mostly to muscle and modestly to memory, while most do nothing detectable at the level of hard outcomes and a small number carry documented risk. In the largest synthesis available here — 884 randomized trials of 27 micronutrients in 883,627 participants — beta-carotene supplementation increased all-cause mortality (RR 1.10; 95% CI 1.05–1.15), cardiovascular mortality (RR 1.12; 95% CI 1.06–1.18) and stroke risk (RR 1.09; 95% CI 1.01–1.17). Vitamin C, vitamin D, vitamin E and selenium showed no effect on cardiovascular disease or type 2 diabetes risk. The same analysis found omega-3 supplementation reduced cardiovascular mortality (RR 0.93; 95% CI 0.88–0.97), myocardial infarction (RR 0.85; 95% CI 0.78–0.92) and coronary heart disease events (RR 0.86; 95% CI 0.80–0.93). A category is not one thing.

The tiers

1. Creatine monohydrate

Creatine is the least fashionable compound on this list and among the best supported. The pooled evidence in older adults comes from 22 randomized trials and 721 men and women, mean ages across studies of 57 to 70, all resistance training two to three days a week for between 7 and 52 weeks. Creatine added to that training produced a mean difference in lean tissue mass of 1.37 kg (95% CI 0.97–1.76), with standardized mean differences of 0.35 (0.16–0.53) for chest-press strength and 0.24 (0.05–0.43) for leg-press strength.

Two limitations deserve to be stated rather than smoothed over. First, every one of those trials involved resistance training; creatine is an amplifier of a training stimulus in this literature, not a substitute for one. Second, sex matters and is understudied. In older women specifically, a review of 10 RCTs and 211 participants found a significant gain in upper-body strength (7 studies, 142 participants, p = 0.04) but no effect on lower-body strength or muscle mass overall; the lower-body benefit appeared only in trials lasting at least 24 weeks. In adults under 50, pooled across 23 studies, creatine plus training raised upper-body strength by a weighted mean difference of 4.43 kg and lower-body strength by 11.35 kg, both p < 0.001 — but that literature comprised 447 male participants across 20 studies against 40 female participants across 2, and in the subgroup analysis the men gained significantly while the women did not.

2. Multivitamin–mineral (the COSMOS trials)

The COSMOS program randomized 21,442 US adults — 12,666 women aged 65 and older, 8,776 men aged 60 and older — in a two-by-two design testing a cocoa extract and a commercial multivitamin–mineral tablet. Two cognitive substudies reported positive results on different endpoints; a third was positive on episodic memory and not on global cognition.

In COSMOS-Web, 3,562 older adults took a daily multivitamin or placebo and were tested annually over three years on an internet-based neuropsychological battery. The prespecified primary outcome, immediate recall on the ModRey test at one year, favored the multivitamin (t(5889) = 2.25, P = 0.025), as did performance averaged across three years (t(5889) = 2.54, P = 0.011). The authors estimated the effect as equivalent to 3.1 years of age-related memory change. Secondary outcomes — novel object recognition, executive function — showed no significant effect. In COSMOS-Mind, 2,262 participants (mean age 73, 60% women, 89% non-Hispanic White) were assessed by telephone; on the trial’s prespecified secondary endpoint, the multivitamin improved global cognition (mean z = 0.07; 95% CI 0.02 to 0.12; P = .007), with a nominally larger estimate in participants with a history of cardiovascular disease, whose own confidence interval crossed zero, while cocoa extract did not (mean z-score = 0.03; 95% CI −0.02 to 0.08; P = .28).

Now the counterweight, which usually gets left out of the press release. In the same COSMOS-Mind cohort, over three years, 110 cases of mild cognitive impairment and 14 cases of dementia were adjudicated, and incidence rates did not differ by treatment assignment for either the multivitamin or the cocoa extract; the authors note directly that statistical power was low. So the finding is: a small improvement on cognitive testing, with no demonstrated effect on whether anyone actually developed cognitive impairment. Those are different claims. The cohort was 89% non-Hispanic White, which limits how far the result travels.

3. Protein and whey, with resistance training

The strongest protein evidence is not in healthy adults, it is in older adults who already meet diagnostic criteria for sarcopenia. Pooling 10 randomized trials and 1,154 such patients, whey protein against an isocaloric placebo or routine consultation increased appendicular skeletal muscle mass index (SMD 0.47; 95% CI 0.23–0.71), appendicular skeletal muscle mass (SMD 0.28; 95% CI 0.11–0.45), and gait speed (SMD 1.13; 95% CI 0.82–1.44); whey plus resistance training increased handgrip strength (SMD 0.67; 95% CI 0.29–1.04) against control. Whey also lowered interleukin-6 and raised IGF-1 and albumin, and did not change BMI, weight, or fat mass.

A second, more conservative synthesis restricted to whey given during resistance training in sarcopenic adults aged 60 and older — 7 trials, 591 participants — found a small benefit on appendicular muscle index (SMD 0.24; 95% CI 0.05–0.42; p = 0.01; I² = 0%) and a handgrip difference of +2.31 kg (95% CI 0.01–4.6; p = 0.05; I² = 81%). The authors then said the thing that matters: the mean difference did not exceed the minimally important clinical difference, and the quality of the evidence was low to very low by GRADE. Read together, the two syntheses say that protein on top of training helps a sarcopenic person measurably, and that the size of the help sits at the edge of what a person would notice.

4. HMB (beta-hydroxy beta-methylbutyrate)

HMB is a metabolite of leucine, and it is the one leucine-adjacent product with functional endpoint data in the population that needs them. In a randomized, double-blind, placebo-controlled trial of adults aged 60 and older with sarcopenia — 18 assigned to HMB, 16 to placebo, twice-daily dosing for 12 weeks, with both groups doing resistance training twice a week — the HMB group improved more than placebo in handgrip strength (4.61 kg; 95% CI 2.93–6.28; P < 0.001), gait speed (0.11 m/s; 95% CI 0.02–0.20; P = 0.014), five-time chair stand (−3.65 s; 95% CI −5.72 to −1.58; P = 0.001) and muscle quality (2.47 kg·kg⁻¹; 95% CI 1.15–3.80). Skeletal muscle mass and skeletal muscle index did not differ between groups. That last detail is the interesting one: the function improved without the mass changing.

The umbrella review of 11 meta-analyses covering 41 datasets in participants aged 23 to 79 puts the pooled effects in perspective. Muscle mass: effect size 0.21; 95% CI 0.06–0.35; p = 0.004. Muscle strength index: 0.27; 95% CI 0.19–0.35; p < 0.001. Fat-free mass: 0.22; 95% CI 0.11–0.34; p < 0.001. No significant change in fat or body mass. Of the 11 included studies, five were rated high quality, three low, three critically low. The single RCT above enrolled 34 people.

5. Omega-3 (marine n-3 fatty acids)

Omega-3 is the clearest case in this article of a compound that fails its primary endpoints and still earns a middle tier, and the reason is worth spelling out. In VITAL, 1 g/day of marine n-3 did not reduce major cardiovascular events over a median 5.3 years (HR 0.92; 95% CI, 0.80 to 1.06; P=0.24) and did not reduce invasive cancer (HR 1.03; 95% CI, 0.93 to 1.13). All-cause mortality was flat, at a hazard ratio of 1.02 (95% CI, 0.90 to 1.15). But among the prespecified secondary endpoints, total myocardial infarction fell: HR 0.72; 95% CI, 0.59 to 0.90. In DO-HEALTH, three years of 1 g/day omega-3 produced no significant benefit on blood pressure, physical performance, cognition, non-vertebral fracture or infection; the infection incidence rate ratio was 0.89 (99% CI 0.78–1.01), which did not clear the trial's prespecified threshold of P < .01.

The pooled picture is more favorable than either single trial. Across 884 randomized micronutrient trials in 883,627 participants and 4,895,544 person-years, n-3 supplementation was associated with lower cardiovascular mortality (0.93; 95% CI 0.88–0.97), myocardial infarction (0.85; 95% CI 0.78–0.92) and coronary heart disease events (0.86; 95% CI 0.80–0.93) — clinical endpoints, pooled across a very large evidence base.

One caveat. Omega-3 is the compound here with randomized data on a biological-aging biomarker, and the effect is smaller than the marketing around methylation clocks would suggest: in a post hoc analysis of 777 DO-HEALTH participants, omega-3 alone slowed PhenoAge, GrimAge2 and DunedinPACE over three years, with vitamin D and exercise adding to the PhenoAge effect, and standardized effects ranging from 0.16 to 0.32 standardized units, or 2.9 to 3.8 months. That is three to four months of clock age accumulated over three years, on a measure that has not been shown to respond to an intervention and then predict what happens to the person whose clock it is.

6. Vitamin D

Vitamin D has been tested more thoroughly than any other supplement in this article, and the result is a narrow one. All-cause mortality: a meta-analysis of 52 randomized trials and 75,454 participants found a risk ratio of 0.98 (95% CI 0.95 to 1.02), with no heterogeneity. Cardiovascular mortality: 0.98; 95% CI 0.88 to 1.08. Cancer mortality is where something appears — risk ratio 0.84; 95% CI 0.74 to 0.95, a 16% relative reduction. VITAL, the largest single trial, found no reduction in invasive cancer incidence (HR 0.96; 95% CI, 0.88 to 1.06; P=0.47) but a hazard ratio for death from cancer of 0.83; 95% CI, 0.67 to 1.02 across 341 cancer deaths, which points in the same direction without reaching significance on its own. A later systematic review and individual-patient-data meta-analysis did not confirm it: across 14 randomized trials and 104,727 participants, cancer mortality fell by 6% and the reduction was not statistically significant (RR 0.94; 95% CI 0.86 to 1.02), with a 12% reduction appearing only in the 10 trials that used a daily rather than a bolus regimen (RR 0.88; 95% CI 0.78 to 0.98). The cancer-mortality result is inconsistent across meta-analyses.

DO-HEALTH tested 2,000 IU/day in adults 70 and older for three years and found no benefit on blood pressure, physical performance, cognition, non-vertebral fracture or infection. That trial enrolled generally healthy older adults with good mobility and cognition, and it still found nothing on its primary endpoints. The reasonable reading: the compound has not been shown to change how an older adult moves, thinks or falls.

7. Cocoa flavanols

COSMOS randomized 21,442 older adults to a cocoa extract supplying 500 mg/day of flavanols, including 80 mg of (−)-epicatechin, or placebo, over a median 3.6 years. The primary composite of total cardiovascular events was missed: 410 events on cocoa versus 456 on placebo, HR 0.90; 95% CI 0.78, 1.02; P = 0.11. Among secondary endpoints, cardiovascular death was lower, HR 0.73; 95% CI 0.54, 0.98 — a 27% relative reduction. All-cause mortality was HR 0.89; 95% CI 0.77, 1.03. A per-protocol analysis censoring at nonadherence gave HR 0.85; 95% CI 0.72, 0.99. There were no safety concerns.

A 27% relative reduction in cardiovascular death is a large claim resting on a secondary endpoint in a trial whose primary endpoint was missed, and per-protocol analyses that censor at nonadherence are not randomized comparisons. That is the honest frame. Separately, in COSMOS-Mind, the same cocoa extract had no effect on cognition (mean z-score = 0.03; 95% CI −0.02 to 0.08; P = .28). Cocoa is not a nootropic.

8. Berberine

Berberine has the largest effect on a laboratory number of anything here, and the least evidence that the number matters. Pooling 46 randomized trials in patients with type 2 diabetes, berberine alone or added to standard therapy reduced HbA1c by a mean difference of −0.73; 95% CI −0.97 to −0.51, fasting plasma glucose by −0.86; 95% CI −1.10 to −0.62, and two-hour postprandial glucose by −1.26; 95% CI −1.64 to −0.89. Insulin resistance measures improved: fasting insulin −2.05; 95% CI −2.62 to −1.48, HOMA-IR −0.71; 95% CI −1.03 to −0.39, BMI −1.07; 95% CI −1.76 to −0.37. Triglycerides fell −0.5; 95% CI −0.61 to −0.39 and HDL rose 0.17; 95% CI 0.09 to 0.25.

They are also entirely surrogate. No trial in this evidence pool tested whether berberine prevents a heart attack, a stroke, kidney failure or a death, and the trials that produced these numbers enrolled people who already had type 2 diabetes, not healthy adults taking it for aging. Berberine is a plausible metabolic agent with pharmacologic-sized effects and no outcome data.

9. Magnesium

Magnesium is sold hard for sleep. The direct randomized evidence in older adults with insomnia amounts to three trials and 151 people across three countries. Pooled, sleep onset latency after the intervention was 17.36 minutes (95% CI −27.27 to −7.44; p = 0.0006) shorter with oral magnesium than with placebo. Total sleep time improved by 16.06 minutes, which was not statistically significant. Every trial was at moderate-to-high risk of bias, and the outcomes were supported by low to very low quality evidence — the authors' own conclusion is that the literature is substandard for making a recommendation, tempered by the observation that magnesium is cheap and widely available, the reviewers’ own suggested ceiling being under 1 g given up to three times daily.

Seventeen minutes of sleep onset latency is a real number and a small one, drawn from 151 people in studies the reviewers themselves rated poorly. If magnesium helps you fall asleep, that is compatible with the evidence. It is not an intervention on aging.

10. Collagen peptides

The collagen literature is unusual here because the endpoints are instrumented rather than self-reported. In a randomized, double-blind, placebo-controlled trial, 64 participants took 1,000 mg/day of a low-molecular-weight collagen peptide — tripeptide content above 15%, including 3% Gly-Pro-Hyp — or placebo, for 12 weeks, with assessment at baseline, 6 weeks and 12 weeks. Skin hydration values were significantly higher than placebo at both 6 and 12 weeks. At 12 weeks the collagen group improved against placebo on the visual assessment score and on three parameters of skin wrinkling. On elasticity, one of three parameters improved from baseline within the group and two of three were significantly higher than placebo. No adverse symptoms related to the test material were reported.

What this shows: oral collagen peptides change measurable skin biophysics over three months in a small trial. What it does not show: any effect on musculoskeletal function, joint outcomes, or anything beyond the surface of the skin. The endpoints are cosmetic in the literal sense.

11. Urolithin A

Urolithin A has the most interesting mechanism in this article and, at present, the widest gap between mechanism and outcome. In the 4-month trial in adults aged 65 to 90 described at the top of this piece, both primary endpoints — six-minute walk distance and maximal ATP production in hand muscle — came back non-significant, while muscle endurance and plasma metabolomic and inflammatory biomarkers moved.

A 2026 systematic review settled the question of how much is actually known. Five randomized trials, 236 participants total. Quantitative pooling was possible for exactly one outcome, the six-minute walk test, from two trials: pooled mean difference +17.03 m (95% CI −5.33 to 39.40; p = 0.135; I² = 0%), with a sensitivity analysis giving +18.80 m (95% CI −3.24 to 40.85; p = 0.095). GRADE certainty: low. Every other outcome — strength, endurance, aerobic capacity, mitochondrial-related biomarkers — was too heterogeneous to pool and was reported as exploratory signal rather than reproducible effect. The reviewers' conclusion is that larger, longer, standardized trials in better-defined populations are required before firm clinical recommendations can be made. That is where urolithin A sits: a compelling target, a trial that missed both of its primary endpoints, and one poolable outcome that did not reach significance.

12. NMN

NMN raises blood NAD. That much is settled. In a randomized, multicenter, double-blind, placebo-controlled trial of 80 healthy middle-aged adults given placebo or 300, 600, or 900 mg daily for 60 days, blood NAD rose significantly in every NMN group at day 30 and day 60 (all p ≤ 0.001), highest at 600 and 900 mg, with no safety signals and all 80 participants completing. Six-minute walk distance increased more than placebo in all three dose groups at both timepoints (all p < 0.01). HOMA-IR did not differ from placebo. A blood "biological age" derived from a web calculator rose in the placebo group and stayed flat in the NMN groups, which is a difference in a calculator output, not in a person.

The most rigorous NMN result is narrower and better. In a 10-week randomized, placebo-controlled, double-blind trial in postmenopausal women with prediabetes who were overweight or obese, insulin-stimulated glucose disposal measured by hyperinsulinemic-euglycemic clamp — the reference method — increased with NMN and did not change with placebo, alongside increased skeletal muscle AKT and mTOR phosphorylation. That is a real physiological effect in a well-defined population, obtained with an unimpeachable technique. It is also insulin sensitivity in prediabetic postmenopausal women over ten weeks. Extrapolating from it to healthy people taking NMN for longevity is not supported.

13. Glucosamine

Glucosamine is the cleanest illustration in this article of why observational data ranks last. In UK Biobank, 495,077 adults with a mean age of 56.6 were followed for a median 8.9 years, during which 19,882 deaths occurred; 19.1% reported regular glucosamine use at baseline. Adjusted hazard ratios for glucosamine users were 0.85; 95% CI 0.82 to 0.89 for all-cause mortality, 0.82; 95% CI 0.74 to 0.90 for cardiovascular mortality, 0.73; 95% CI 0.66 to 0.81 for respiratory mortality and 0.74; 95% CI 0.62 to 0.90 for digestive-disease mortality. Those are large, consistent and widely cited.

In NHANES, 38,021 US adults were followed through 2015 with 4,905 deaths. Glucosamine and chondroitin use appeared inversely associated with mortality in minimally adjusted models, and the association vanished in multivariable models: HR 1.02; 95% CI 0.86–1.21 for glucosamine and HR 1.04; 95% CI 0.87–1.25 for chondroitin. Cardiovascular-specific mortality showed a suggestive but non-significant inverse association. A third analysis, also drawn from NHANES but covering 16,686 participants surveyed between 1999 and 2010 and followed for a median 107 months, went the other way: after adjustment for age, sex, race, education, smoking status and physical activity, regular glucosamine/chondroitin use remained associated with lower all-cause mortality (HR 0.73; 95% CI 0.57 to 0.93) and lower cardiovascular mortality (HR 0.42; 95% CI 0.23 to 0.75). Same exposure, different cohorts, conflicting conclusions once confounders were handled. People who take glucosamine regularly differ from people who do not in ways that are hard to measure and easy to mistake for a drug effect. There is no randomized mortality trial of glucosamine in this evidence pool.

What we would skip, and why

Spermidine. The SmartAge trial is the reason to skip it, and it is a good trial. One hundred adults aged 60 to 90 with subjective cognitive decline — mean age 69, 49% women — were randomized to 12 months of a wheat-germ-derived supplement providing 0.9 mg spermidine daily or to microcrystalline cellulose placebo; 89% completed. The primary outcome, mnemonic discrimination on the Mnemonic Similarity Task, showed a between-group difference of −0.03; 95% CI −0.11 to 0.05; P = .47. Secondary outcomes showed no significant changes. Exploratory analyses hinted at effects on inflammation and verbal memory, which is what exploratory analyses do. Adverse events were balanced. This is a well-conducted twelve-month null.

Fisetin. The senolytic story is genuinely exciting in cell and animal work, and the human evidence in this pool is a study protocol. The trial described is triple-blind, randomized and placebo-controlled, enrolling adults aged 50 and older to 100 mg of oral fisetin or placebo once daily for seven weeks, with a primary outcome of the between-group change in plasma soluble urokinase plasminogen activator receptor and a secondary outcome of adverse events. Frailty, physical function and cognition are exploratory. The protocol's own framing is that randomized human evidence for fisetin is limited. A compound whose human evidence base is a seven-week biomarker trial that has not yet reported is a purchase made on hope.

Taurine. In 2023, a paper in Science reported that circulating taurine declines with age in mice, monkeys and humans, and that taurine supplementation increased healthspan and lifespan in mice and healthspan in monkeys, with mechanistic effects on senescence, mitochondrial function, DNA damage and inflammation. The paper's own conclusion was that clinical trials in humans seem warranted. In 2025, that hypothesis was tested against human data: in 137 physically inactive and physically active men aged 20 to 93, circulating taurine showed no association with age, muscle mass, strength, physical performance or mitochondrial function. That is one cross-sectional study in men only and it does not close the question, but it is the direct human test of the animal claim and it came back negative.

Beta-carotene, and by extension "antioxidant" formulas. In the 884-trial micronutrient synthesis, beta-carotene supplementation increased all-cause mortality, cardiovascular mortality and stroke risk. This is the clearest evidence in the pool that a supplement can do harm at scale, and a correction to the assumption that a nutrient found in healthy diets is safe when isolated in a capsule. Food evidence is not supplement evidence.

What's actually in the bottle

Everything above assumes the capsule contains what the label says, in the amount the label says, without anything else. That assumption is the reason pharmacopeial standards exist. The United States Pharmacopeia and its Herbal Medicines Compendium define quality for botanical articles through interlocking tests: chromatographic identification procedures — high-performance liquid chromatography and high-performance thin-layer chromatography — specifically to detect species substitution and other confounders; quantitative assays to determine the content of the constituents considered bioactive or used as analytical markers; and purity specifications that limit contaminants including toxic elements, pesticides and fungal toxins. Naming, reference materials, packaging and storage conditions are part of the monograph too, because a correctly manufactured product can still degrade.

Note what that architecture implies. A chromatographic identity test exists to detect species substitution. Contaminant limits exist to cap toxic elements, pesticides and fungal toxins. A content assay exists because a species name on a label says nothing about how much of the active constituent is in the capsule. A product carrying a compound name on the front of the bottle has not thereby demonstrated identity, content or purity.

The consequence that shows up in hospitals is liver injury. Drug-induced liver injury is diagnosed by exclusion: there is no biomarker for it, diagnosis rests on the temporal relationship between elevated liver tests and a recently started drug or herbal and dietary supplement, on normal hepatobiliary imaging, and on ruling out competing causes. The management is to stop the implicated agent and observe closely; there is no specific therapy. Among patients who present with jaundice, roughly 10% face liver-related mortality or the need for liver transplantation. Supplements are named alongside prescription drugs in that framework, not as a lesser category. If you start something new and your liver enzymes rise, the supplement is a suspect.

Best longevity supplements for women over 50

The right way to answer this is to ask what the trials enrolled, and to refuse to go further.

What that adds up to: for a woman over 50 the best-evidenced interventions in this article are creatine paired with resistance training, which the female-specific review ran for 24 weeks or longer; whey if she meets the sarcopenia criteria the trials used; and, from age 65, a daily multivitamin if the goal is the small cognitive effect COSMOS measured. Note the qualification on creatine — in the pooled analysis of adults under 50, male participants outnumbered female participants roughly eleven to one, males gained significantly and females did not, and the review's own conclusion was that greater benefit is likely in males. The evidence in older women is more encouraging than that, but it is 211 people, and the honest statement is that creatine has been studied far less in women than in men and the female literature is thinner and shorter.

We will not extrapolate the NMN result beyond postmenopausal women with prediabetes who are overweight or obese, because that is the only population in which it was demonstrated. Nothing here was tested for menopausal symptoms, bone density in perimenopause, or hormone-related outcomes.

Best longevity supplements for men over 40

Men in their forties are the demographic least represented in this evidence base. The large outcome trials enrolled people much older: VITAL required men to be 50 or older, DO-HEALTH 70 or older, COSMOS 60 or older. The creatine literature in adults under 50 is overwhelmingly male — 447 male participants across 20 studies, against 40 female participants across 2 — and it is the one place whose enrolled age band includes men in their forties. There, creatine plus resistance training raised upper-body strength by a weighted mean difference of 4.43 kg and lower-body strength by 11.35 kg, with men gaining significantly.

The middle-aged trials of the newer compounds are small. The entire randomized urolithin A literature is five trials and 236 participants across all ages. The NMN dose-ranging trial enrolled 80 healthy middle-aged adults for 60 days and measured blood NAD, walking distance and questionnaire scores. Neither is a basis for a decade-long habit.

So: creatine with a training program is the defensible choice, and it is defensible specifically because the enrolled age band extends to 50. Everything else routes to the same answer as for everyone else — omega-3 has pooled clinical-endpoint data but null primaries in the two large prevention trials, and nothing here was tested for testosterone, prostate outcomes, or male fertility. If a product is marketed to men over 40 on the strength of a mechanism, the relevant question is which trial enrolled men over 40 and what it measured.

How to know it's working

The trials above did not measure how anyone felt. They measured specific things, mostly with a stopwatch, a dynamometer or a scored test, and those are the things a reader can track. Measure at baseline before starting anything, then at the interval the relevant trial used, and expect effects in the range the trials found rather than the range the label implies.

Three things are not on that list, deliberately. Methylation clock age: the only post hoc randomized supplement data here moved it by the equivalent of 2.9 to 3.8 months over three years, and no trial has shown that moving a clock changes what happens to the person whose clock it is. Blood NAD: it rises reliably on NMN and has not been linked to any clinical outcome in humans. A web-calculator "biological age": the NMN trial reported one, and it is a formula applied to a blood panel, not a measurement of aging.

If nothing on the trackable list has moved after the interval the trials used — 12 weeks for HMB and collagen, 4 months for urolithin A, 24 weeks or longer for creatine in older women, 3 years for the multivitamin — you have no individual evidence that it is working, which is a reasonable point at which to stop paying for it. The strongest interventions in this article share one feature: they were tested alongside resistance training, and the training was doing the heavy part.

Questions people actually ask

Top 10 longevity supplements This page does not rank ten. It sorts the commonly sold compounds into three tiers by the strength of the human evidence: creatine, whey protein, HMB and a daily multivitamin–mineral in the top tier; omega-3, cocoa flavanols, vitamin D, berberine, magnesium and collagen peptides in the middle; urolithin A, NMN, glucosamine, spermidine, fisetin and taurine at the bottom.

What is the best supplement for anti-aging? Nothing in this evidence base was tested against aging or lifespan, so there is no compound the evidence names as best for that. The best-evidenced compounds here are creatine and whey protein taken alongside resistance training, which add lean tissue and strength, and a daily multivitamin–mineral, which produced a small improvement on memory testing in COSMOS.

Do any anti aging supplements work? For the hard endpoints — cardiovascular events, cancer, death — the large randomized trials were null. For narrower functional endpoints a few do something measurable: creatine and whey with resistance training, HMB in adults who already have sarcopenia, and a multivitamin on a scored memory test.

Does creatine shorten lifespan? No trial in this evidence set tested lifespan, in either direction, so nothing on this page can answer that. The creatine literature here measured lean tissue mass and strength over 7 to 52 weeks.

Can you use creatine long term? The pooled trials in older adults ran between 7 and 52 weeks, and the female-specific review's longer subgroup was 24 weeks or more. No trial in this evidence set tested multi-year use, so there is no basis here for a statement about what happens beyond roughly a year.

What is fisetin supplement? Fisetin is a naturally occurring flavonoid with anti-inflammatory, antioxidant and senolytic effects in cell and animal models. The only human item in this evidence set is a study protocol: 100 mg of oral fisetin or placebo once daily for seven weeks in adults aged 50 and older, with plasma soluble urokinase plasminogen activator receptor as the primary outcome. It has no results.

Does fisetin work? Unknown from this evidence base. The pool's only human fisetin item is a study protocol with no results, and the protocol itself states that randomized human evidence for fisetin is limited.

Is fisetin safe to take? The protocol lists the occurrence and severity of adverse events as its secondary outcome, which means the human safety data it was designed to produce do not exist yet. There is no reported human safety result for fisetin in this evidence set.

How much omega 3 is beneficial? The two large randomized trials here, VITAL and DO-HEALTH, both used 1 g/day of marine n-3, and both missed their primary endpoints. This evidence set contains no dose-comparison trial for omega-3, so it cannot say which amount, if any, is the beneficial one.

What is the best supplement to take for aging skin? The only skin evidence on this page is oral collagen peptide: 64 adults took 1,000 mg/day or placebo for 12 weeks, with higher skin hydration than placebo at 6 and 12 weeks and improvement on visual assessment and three wrinkle parameters at 12 weeks. This page does not cover topical products or any other skin evidence.

Urolithin a side effects In the four-month trial in 66 adults aged 65 to 90, adverse events did not differ significantly between urolithin A and placebo, and the authors reported the supplement as safe and well tolerated in that population. That is one small trial over four months; the entire randomized literature is five trials and 236 participants.

Related on Magellan: Creatine for healthy aging · Urolithin A and muscle · NMN vs. NR.

Educational, not medical advice.