Longevity supplements, ranked by the evidence

There are eleven bottles on the shelf above my kettle. I know, because I counted them while writing this, which is roughly the level of scientific rigour the supplement industry is used to. Most of them arrived because a study got a headline. Two of them I can defend. The rest are, if I’m honest, a slightly expensive way of feeling like I’m doing something on days when I don’t train.

The problem with every “best longevity supplements” list I’ve read is that they’re all ranked on the same hidden axis: how interesting the mechanism sounds. Sirtuins, mitochondria, telomeres, autophagy. Beautiful stories, all of them, and the story has almost no relationship to whether anything happened to an actual human being in a controlled trial. So I ranked them on a different axis, and the leaderboard came out in an order I didn’t expect — with a €5 multivitamin sitting above several products that cost sixty times that.

The thesis, up front: rank supplements by the hardest endpoint they’ve actually moved in humans, and everything near the top turns out to be replacement therapy for a shortfall you can measure — while every compound sold specifically as a longevity supplement sits at the bottom, because there’s no deficit to correct and no hard endpoint has budged.

How I ranked them

Two questions, in this order.

First: what’s the hardest thing anyone has moved? I’m borrowing the endpoint-hardness ladder from the NMN and NAD+ post, because it generalises far beyond that one molecule. Rung one is a blood marker. Rung two is a soft functional outcome — questionnaires, walk tests, anything effort- or belief-dependent. Rung three is objective physiology: DXA scans, dynamometers, clamps, biopsies — measurements that don’t care what you believe. Rung four is aging itself: mortality, disease incidence, disability.

Exactly one thing on this list has reached rung four with a positive result, and it did so in a prescription formulation at four times a normal capsule dose in patients already on statins. Vitamin D has been measured at rung four twice, in tens of thousands of people, and came back null. For most of the shelf, nobody has run that trial at all. Moved it, measured it and found nothing, or never looked — that three-way distinction is the single most useful thing to know about this category, and no supplement ad will ever tell you which one applies.

Second: how likely are you to be in the group it works for? This is the part the lists skip, and it does most of the reshuffling. A supplement with a genuine effect in deficient people has approximately no effect in you if you’re replete — which means “does it work?” is the wrong question and “am I short of it?” is the right one. Effect size and applicability are different numbers, and multiplying them changes the order.

One caution before the list, same as always: a trial that finds nothing in an unscreened population hasn’t proved the nutrient is useless, only that topping up people who were already fine does nothing. That distinction is going to do a lot of work below.

1. Creatine monohydrate — the best-evidenced thing on the shelf

Rung three, and comfortably. According to PubMed, a 2017 meta-analysis by Chilibeck and colleagues in the Open Access Journal of Sports Medicine pooled 22 randomised trials and 721 participants aged 57 to 70, all doing resistance training for 7 to 52 weeks. Creatine on top of training produced 1.37 kg more lean tissue mass than training plus placebo (95% CI 0.97–1.76), with significant gains in chest press and leg press strength (DOI). DXA and one-rep-maxes don’t care what you believe about creatine.

And the replacement logic is clean: muscle creatine stores sit below saturation on an ordinary diet, and lower still if you don’t eat much meat or fish. You’re topping up a genuine shortfall, which is why it works at 3–5 g and why the effect doesn’t scale with heroic doses. Two honest caveats: it does nothing on its own — it’s an amplifier bolted onto resistance training — and nobody has ever run it against mortality. The full case, dosing included, is in creatine and longevity; the messier cognitive claim gets its own treatment in does creatine help the aging brain?

2. Protein powder — the one that isn’t really a supplement

Rung three, with the most precisely defined responder group on the list. According to PubMed, Morton and colleagues (British Journal of Sports Medicine, 2018) meta-analysed 49 trials and 1,863 people: protein supplementation during resistance training added 2.49 kg to one-rep-max strength and 0.30 kg to fat-free mass. Then the number that matters — beyond a total intake of about 1.62 g/kg/day, extra protein produced no further gains (DOI).

That’s the whole ranking philosophy in one meta-analysis. Below the threshold it works; above it, nothing. Whey is a convenience, not a compound — it’s food that fits in a shaker. If you’re already hitting your protein intake from meals, a scoop buys you precisely zero, and I’d file it under groceries rather than supplements. If you’re not, it’s the cheapest fix on this page.

3. Omega-3 — one real hard-endpoint win, in a narrow group

The only rung-four win on the list — and then read the fine print, because it’s all fine print. The VITAL trial (NEJM, 2019) randomised ~26,000 people to 1 g/day of fish oil and missed its primary cardiovascular and cancer endpoints. REDUCE-IT (NEJM, 2019) then used 4 g/day of pharmaceutical-grade EPA in statin-treated patients with high triglycerides and cut major cardiovascular events by 25%, including cardiovascular death. That is disease incidence in humans: a genuine hard endpoint, genuinely moved. Note what it isn’t, though — a lifespan result, or a finding in healthy people, or a fish-oil capsule.

Then the awkward bit, which most rankings leave out. According to PubMed, the STRENGTH trial (Nicholls et al., JAMA, 2020) ran essentially the same experiment with a different omega-3 formulation in 13,078 high-risk patients — and was halted early for futility, hazard ratio 0.99, with more gastrointestinal side effects than the corn-oil comparator (DOI). Two big trials, same nutrient class, opposite results. Add DO-HEALTH, which missed its primary endpoints too, and the honest summary is: a specific prescription-dose formulation, in a specific high-risk group, produced a real effect — and generalising that to a supermarket fish-oil capsule in a healthy person is a leap the data doesn’t support. So: the highest rung anything on this page has reached, attached to the narrowest responder group on it. That’s exactly why the criterion has two halves — on endpoint hardness alone omega-3 tops the list, and on your odds of being the person it worked for it drops to third. Details in omega-3 and longevity.

4. Vitamin D — the cleanest proof that “am I short?” is the whole question

Vitamin D is the entry that made me write the post this way, because you can run the thesis as a controlled experiment.

Screen for deficiency and you find effects: Martineau’s 2017 individual-participant analysis in the BMJ (25 trials, 11,321 people) found supplementation reduced acute respiratory infections, with the strongest protection — odds ratio 0.70 — in those most deficient at baseline.

Don’t screen, and it vanishes. According to PubMed, the D-Health trial (Neale et al., Lancet Diabetes & Endocrinology, 2022) gave 60,000 IU monthly to 21,315 unscreened older Australians for five years and found no mortality benefit: hazard ratio 1.04 (95% CI 0.93–1.18). Placebo-group blood levels averaged 77 nmol/L — these people were already fine. The authors went further, noting their point estimates were consistent with a slightly increased cancer mortality risk and that this dosing “might not be appropriate in people who are vitamin D-replete” (DOI).

Same nutrient. Opposite answers. The deciding variable is a blood test, not the size of the trial. If you live somewhere with a Munich winter, test before you decide — the practical version is in vitamin D and longevity.

5. The boring multivitamin — the surprising placement

Here’s the one I didn’t see coming. The most-mocked product in the pharmacy has better randomised cognitive evidence than every premium longevity supplement combined.

According to PubMed, COSMOS-Mind (Baker et al., Alzheimer’s & Dementia, 2022) randomised 2,262 older adults to a commercial multivitamin-mineral or placebo for three years and found a significant benefit on global cognition (mean z = 0.07, p = 0.007), with effects on memory and executive function too — while the cocoa-extract arm in the same trial did precisely nothing (DOI). A 2024 meta-analysis of three COSMOS cognitive substudies (Vyas et al., American Journal of Clinical Nutrition), covering 5,203 non-overlapping participants, confirmed benefits on global cognition and episodic memory and sized the effect as equivalent to reducing cognitive aging by about two years (DOI).

Now, restraint. That’s a small effect, on cognitive tests rather than dementia diagnoses, in a cohort that was 89% non-Hispanic White, and COSMOS found no mortality benefit. It is not a longevity drug. But it is a rung-three-adjacent, adequately powered, repeatedly replicated randomised finding — and it fits the thesis exactly, because a multivitamin is nothing but replacement therapy for whatever micronutrient gaps your diet has. Fifth place, for about €5 a month, ahead of things costing sixty times more. Fill the gaps you can’t see; don’t expect it to do more than that.

6. Magnesium — real, modest, and only if your intake is short

Rung three on one specific outcome. A 2016 meta-analysis by Zhang and colleagues in Hypertension pooled 34 randomised, double-blind trials (~2,000 people, median 368 mg/day) and found systolic blood pressure fell by about 2 mmHg and diastolic by about 1.8 mmHg. Modest — but it’s a properly controlled causal finding on a real risk factor, which is more than most of this list can say. The mortality signal, from Fang’s 2016 BMC Medicine dose-response analysis of 40 cohorts and over a million people, is observational and should be read as such.

It ranks sixth rather than higher because the effect size is genuinely small and, again, contingent: you get it by correcting a shortfall, and a lot of people don’t have one. Sleep is where magnesium gets oversold — the trial evidence is one small 46-person study — and I’ve pulled that apart properly in magnesium vs melatonin. The general case is in magnesium and longevity.

Everything above this line is replacement therapy for a measurable shortfall. Everything below it is enhancement — adding a compound to a body that isn’t short of it. Notice what happens to the evidence.

7. Collagen — decent trials, for things that aren’t aging

Collagen is the honest edge case, and I’d rather flag that than pretend the line is tidy. It has real randomised evidence: Proksch’s 2013 trial found objectively measured improvements in skin elasticity, and König’s 2018 year-long trial in 131 postmenopausal women found gains in spine and femoral-neck bone density by DXA. That last one is a legitimate rung-three result.

So why below the line? Because there’s no collagen deficiency to correct — it’s amino acids you’re already eating, in a form with a nice story attached — and because the endpoints, real as they are, are skin, joints and one bone-density finding, not aging. It’s a decent skin product wearing a longevity costume. The full accounting is in does collagen do anything for aging?

8. Resveratrol — the mechanism that ate a decade

Steelmanning it first, because it deserves that. According to PubMed, Delpino and Figueiredo’s 2021 systematic review in Critical Reviews in Food Science and Nutrition pooled 30 studies and found significant reductions in insulin resistance (SMD −0.34) and HbA1c (SMD −0.64), with fasting glucose improving in people who already had diabetes (DOI). That’s not nothing.

It’s also not what resveratrol was sold as. Note the heterogeneity in that analysis — I² of 70 to 90%, which means the studies disagree with each other a great deal — and note that these are diabetes parameters, not aging outcomes. The sirtuin-activation story that launched the category has already had its ending written: GSK bought Sirtris for around $720 million in 2008 and shelved the lead compound by 2010. Eighteen years on, the human aging evidence is still zero.

9. NMN and NAD+ boosters — does exactly what it says, and what it says doesn’t matter

Rung one reliably; rung three exactly once. NAD+ precursors do raise blood NAD+ — that part is real and well replicated — and then the effects thin out at soft functional outcomes and mostly evaporate at objective physiology. Mostly, not entirely, and the exception deserves its due: Yoshino and colleagues (Science, 2021) gave 25 prediabetic postmenopausal women 250 mg of NMN for ten weeks and found genuinely improved insulin-stimulated glucose disposal by clamp, with matching muscle signalling. Hard measurement, independent academic group, positive result. It’s also 25 people, one narrow population, ten weeks, and a measure of insulin signalling rather than of aging. Best card in the deck; small card.

Above that, nothing. No trial of any NAD+ precursor, in any human, has ever measured mortality, disease incidence or disability. Rung four is untouched, ten years and dozens of trials in.

It ranks ninth rather than tenth only because it has been studied properly and appears safe at the doses tested. The full evidence review — including which trials were funded by whom, and why the effect size tracks how soft the endpoint is — is in do NMN and NAD+ boosters do anything?

10. Taurine — the newest one, and a cautionary tale in real time

Last place, and this is the entry I’d hand to anyone who wants to understand how longevity supplements are actually built.

In 2023, Singh and colleagues published a genuinely impressive paper in Science: circulating taurine declined with age in mice, monkeys and humans, and reversing that decline extended health span in mice and monkeys and lifespan in mice (DOI). Supplement brands launched within weeks. Fair enough — it’s a strong paper, and the authors were explicit that human clinical trials were needed to test whether any of it applied to us.

Then, in June 2025, a team at the US National Institute on Aging (Fernandez et al., Science) went looking for the premise. Measuring taurine longitudinally — the same people repeatedly over time — across three human cohorts plus monkeys and mice, they found circulating taurine increased or stayed flat with age, and concluded that a decline “is not a universal feature of aging” (DOI). (Both according to PubMed.)

So the deficiency the supplement corrects may not exist. Note the sequence: mouse result, product launch, human premise checked afterwards and found wanting. That’s not a scandal, it’s just how the pipeline runs — and it’s why “there’s a Science paper on it” is a much weaker recommendation than it sounds.

The line down the middle

Stand back and the leaderboard tells one story, and it isn’t the one on the labels.

Everything with defensible human evidence is replacement therapy. Creatine tops up stores that sit below saturation. Protein fills a gap up to ~1.6 g/kg/day and then stops mattering. Vitamin D works in the deficient and does nothing measurable in the replete — the same trial programme demonstrated both. A multivitamin fills micronutrient gaps. Magnesium helps if your intake is short. In every case the active ingredient is the shortfall, and the size of your benefit is the size of your gap.

Everything sold specifically as a longevity supplement is enhancement, and none of it has moved an aging endpoint. Resveratrol, NMN, taurine: no deficit to correct, beautiful mechanism, mouse data, and rung four completely empty — not measured and found wanting, simply never measured. Between them they have one small positive clamp study, in 25 women, over ten weeks. Not fraud, then — real compounds doing real biochemistry — but the thing they do has never been connected to the thing you want.

Which produces the ranking’s most useful implication: your gaps, not the shelf, determine what’s worth taking. Two people can be handed the same honest list and correctly buy different things from it. That’s not a weakness of the evidence. It’s the finding.

How to actually use this

Three questions, in order, for anything you’re considering:

1. Is there a shortfall I can measure? A blood level (25-hydroxyvitamin D, ferritin, B12), a dietary intake you can count (protein, magnesium), or a training-adjusted requirement (creatine). No measurable gap, no expected benefit.
2. Has anyone moved a hard endpoint in humans? Not a blood marker, not a questionnaire, not a six-minute walk test. A DXA scan, a clamp, a disease event. If the wins are all on soft outcomes, you’re mostly buying expectation.
3. Would food fix it more cheaply? Usually yes — and it comes with the rest of a decent dietary pattern attached.

A defensible shelf for most people, on that basis: creatine 3–5 g daily if you lift; protein to close the gap to ~1.6 g/kg if your meals don’t; vitamin D through the winter if a test says you’re low; a cheap multivitamin if your diet is patchy. That’s four items, well under €20 a month, and it’s the whole list I’d defend.

And keep the proportions straight. Even the best supplement on this page is an amplifier for training you’re already doing — which is why the supplement drawer came ninth out of nine when I ranked longevity habits by healthspan per hour. The bottles are the seasoning. Lifting, walking and sleeping are the meal.

I’ve since binned four of my eleven. It’s quite freeing, actually.