NMN / NAD+ precursors

NAD+ is the coenzyme every cell spends to run its mitochondria and to feed a family of enzymes — sirtuins, PARPs, CD38 — that repair DNA, regulate metabolism and signal between cells. It falls with age. NMN is one chemical step from it. Almost everyone who takes NMN takes it for that reason, and the literature behind that reason is now substantial enough to appraise on its own terms. This entry covers the ageing evidence first, because that is what the compound is for, and the cartilage question second, because that is what this site is for.

What happens to NAD+ with age

The decline is measured, not assumed, and it has been measured in people. Human pelvic skin sampled from newborns to seventy-seven-year-olds showed tissue NAD+ falling steeply with age — a correlation of about minus seven-tenths in men and minus five-tenths in women — alongside rising DNA damage. Activity of PARP, an NAD+-consuming repair enzyme, rose with age in men across the whole range, though that correlation rests on the newborn samples and does not survive their removal. In living human brain, magnetic resonance spectroscopy at seven tesla found NAD+, total NAD and the NAD+/NADH redox potential all declining with age while the reduced form rose. In plasma, mass spectrometry across ages twenty to eighty-seven found NAD+ and NADP+ down and nicotinamide, ADP-ribose and NADPH up. Different tissues, different methods, same direction. All of it is cross-sectional, so it establishes that the decline is real without establishing that it causes anything.

The salvage pathway and its drain

Mammals mostly recycle NAD+ rather than making it fresh, and the recycling line has a rate-limiting enzyme, NAMPT, and a set of consumers that spend the product. The consumer that matters most for ageing is CD38, an enzyme on the cell surface that destroys NAD+ and rises with age. Blocking it in mice with a specific inhibitor reversed tissue NAD+ decline and improved glucose tolerance, muscle function, exercise capacity and cardiac function — and the benefits vanished when NAD+ synthesis was blocked, which ties them to the molecule rather than to the drug. Where the extra CD38 comes from is partially answered: senescent cells do not express much of it themselves, but the inflammatory factors they secrete drive it up in neighbouring healthy cells.

There is a complication worth stating plainly, because it rarely appears in supplement copy. NAMPT and the NAD+ it produces also power the inflammatory secretome of senescent cells. In cell and mouse work, raising NAD+ through NAMPT amplified that secretome rather than quieting it, and the authors concluded that dietary NAD+ augmentation should be administered with precision. NAD+ is a substrate, not a drug, and substrates feed whatever is running.

The sirtuin case

Sirtuins are the reason NAD+ became a longevity story: they are deacetylases that cannot work without it, so their activity tracks NAD+ supply. What is firmly established is that they matter. What is modelled rather than demonstrated is that topping up their fuel buys you time.

The strongest mammalian result is genetic rather than nutritional. Mice overexpressing SIRT6 lived longer than their wild-type littermates in both sexes — median lifespan up 27 percent in males and 15 percent in females, maximal lifespan up 11 and 15 — and aged better with it: less frailty, preserved liver glucose output and normal blood sugar in old age, red cell counts held at young levels, and an LDL to HDL ratio that stayed low while it climbed in the controls. Two details of that experiment matter more than the headline. Overexpressing SIRT1, the sirtuin the supplement story is built around, did not extend lifespan at all. And an earlier report from the same group, in a different mouse strain, had found the benefit in males only; the sex restriction turned out to be a property of the background rather than of the gene.

The invertebrate evidence that originally made the field's case did not survive scrutiny: when a 2011 study standardised genetic background and used proper transgenic controls, the reported lifespan extension from sirtuin overexpression disappeared in both worms and flies, with the worm longevity turning out to track an unrelated second-site mutation. That paper does not rule out a sirtuin role in mammals; it removes the foundation the mammalian claims were extrapolated from.

In mice, the precursors themselves do work. Twelve months of oral NMN in normally ageing mice suppressed weight gain, raised energy expenditure, improved insulin sensitivity, tear production and bone mineral density and preserved photoreceptor function, with no observed toxicity. Glucose tolerance did not move, and the effect on movement went the other way at the higher of the two doses, which cut rearing activity. That is the result the human trials were built to follow up. Genetic proof that a sirtuin matters and a supplement that raises its fuel are two different claims, and the gap between them is where this entry sits.

What oral NMN does to blood NAD+

This is the part that replicates. Raising blood NAD+ with an oral precursor is reliable, dose-dependent and reproducible across independent laboratories. The dose-response is not open-ended, though. The one trial to run three doses side by side — eighty healthy middle-aged adults on 300, 600 or 900 mg a day for sixty days — found 600 mg raised blood NAD+ more than 300 did, and 900 mg no more than 600, which is where its authors put the ceiling.

The one pharmaceutical-grade formulation tested here, at a gram twice daily for twenty-eight days in overweight and obese adults aged forty-five and over, substantially raised circulating NAD and its metabolites; that regimen was chosen because earlier pharmacokinetic work had put it above a 200 percent rise in blood NAD. At 250 mg a day for twelve weeks, whole-blood NAD+ rose significantly against placebo in healthy volunteers. The sister precursor nicotinamide riboside shows the same curve more precisely, and shows where it runs out: against their own baselines, 100, 300 and 1,000 mg a day raised whole-blood NAD+ by 22, 51 and 142 percent respectively within two weeks. By week eight the top two doses were still there and the lowest had fallen back to ten percent, having tracked placebo the whole way.

The most informative trial is the newest. A four-arm randomised study in sixty-five healthy adults compared nicotinamide riboside at 1,000 mg, NMN at 1,000 mg and plain nicotinamide at 500 mg over fourteen days. NR and NMN both roughly doubled whole-blood NAD+ and did so comparably — differences against placebo of 49.4 micromolar for NR and 43.1 for NMN — while nicotinamide did nothing. The surprise was the route: ex vivo work with human gut microbiota traced the effect to microbial conversion of NR and NMN into nicotinic acid, and in whole blood it was nicotinic acid, not NR, NMN or nicotinamide, that raised NAD+. If that holds, the two expensive precursors may be working through a cheap gut metabolite.

The human trials at their real endpoints

Blood NAD+ is target engagement, not benefit. Here is what happens when trials measure something a person would notice.

Aerobic capacity. Forty-eight amateur runners trained five to six times weekly for six weeks on 300, 600 or 1,200 mg of NMN a day or placebo. VO2max did not differ between any arm. What rose dose-dependently were ventilatory thresholds: the percentage of VO2max reached at the first ventilatory threshold improved 2.1 percent on placebo, 6.5 at 600 mg and 10.3 at 1,200 mg. Twelve people per arm, and the supplement and placebo were both supplied by an NMN manufacturer.

Muscle function and walking. The longest functional trial gave 250 mg a day for twenty-four weeks to fourteen men over sixty-five with diabetes and reduced grip strength or gait speed. It was null on both co-primary endpoints: grip changed 1.25 kg on NMN against minus 0.44 on placebo, walking speed 0.033 metres per second against 0.014, with confidence intervals straddling zero throughout. A sixty-person trial at the same dose was null on its primary endpoint — a stepping test — but found a shorter four-metre walking time at twelve weeks, partly because the placebo group's time got worse. That trial was funded entirely by a food company, three of whose employees are among the four authors. Pooled across the trials, none of this survives: gait speed, grip strength, muscle mass and the chair stand test are all null.

Sleep and fatigue. A hundred and eight older Japanese adults took 250 mg in the morning or the afternoon for twelve weeks. Afternoon dosing gave the largest improvements in the five-times sit-to-stand test and in self-reported drowsiness, with effect sizes of 0.72 and 0.64 — but the sit-to-stand test improved in all four arms, placebo included. The sixty-person trial above found better global sleep-quality and daytime-dysfunction scores at twelve weeks.

Cardiometabolic outcomes. Thirty overweight or obese adults on 1,000 mg twice daily for twenty-eight days lost 1.9 kg more than placebo, dropped diastolic pressure by 7.0 mmHg more and LDL cholesterol by 18.73 mg/dL more. Insulin sensitivity, liver and abdominal fat, muscle strength, aerobic capacity and stair-climbing power did not differ. Nine of those thirty were the placebo group, and the trial was funded by the company that owns the formulation.

Arteries. Thirty-six healthy middle-aged adults on 250 mg a day for twelve weeks saw pulse wave velocity fall by 25.1 cm/s — a between-group p-value of 0.097, which is not a result.

Insulin sensitivity. The trial that put NMN in Science gave 250 mg a day for ten weeks to twenty-five prediabetic postmenopausal women. Insulin-stimulated glucose disposal rose 25 percent from its own baseline on NMN and did not move on placebo, and the between-group evidence is a three-way interaction at p=0.022. Everything else held still: body composition, liver and abdominal fat, blood pressure, glucose, insulin, lipids, and both hepatic and adipose insulin sensitivity. Muscle NAD+ itself did not rise, though NMN metabolites in muscle did, which the authors read as faster turnover rather than a bigger pool.

The trial's own baseline table carries the reason to hold it loosely. Intrahepatic triglyceride content was 6.3 percent in the thirteen women given NMN and 14.8 percent in the twelve given placebo, a group difference at p=0.003 — a randomisation that came apart on exactly the variable the mechanism runs through, in a trial of twenty-five people. The result has not been replicated.

What the pooled analyses find

Four pooled analyses now exist, and they agree with each other more than the individual trials do.

On glucose and lipids, 342 participants gave nothing: no significant effect on fasting glucose, fasting insulin, HbA1c, insulin resistance or any lipid fraction. On muscle the answer is the same, and it is worth stating carefully, because muscle is where the healthspan case is usually made. Pooling the trials of NMN and nicotinamide riboside in adults whose mean ages ran from sixty-one to eighty-three found no change in skeletal muscle index, handgrip strength on either side, gait speed or the five-time chair stand test; the narrative arm added no change in knee extension strength, the physical performance battery or thigh muscle mass. Its conclusion is that current evidence does not support either precursor for preserving muscle mass or function past sixty.

Blood pressure is the one place something moves reliably, and not by much. Ten trials in 349 adults gave a diastolic reduction of 2.15 mmHg, with a systolic reduction of 3.94 mmHg confined to people aged sixty and over. A separate safety and metabolism synthesis of fifteen trials at 250 to 2,000 mg a day for up to twenty-four weeks found the same diastolic effect at 2.43 mmHg, and no effect on body weight, BMI, fasting glucose, HbA1c, lipids or systolic pressure — alongside no increase in adverse events, serious adverse events, withdrawals or liver enzymes.

Read together they describe a compound that hits its biochemical target reliably and its clinical targets rarely. The one durable exception is small, sits in diastolic pressure, and turns up in two independent pools.

Nicotinamide riboside, the sister precursor

NR has the larger and older human literature, so it is the best guide to what an NAD+ precursor can and cannot do, and it is worth reading alongside NMN rather than instead of it.

The nulls are informative because several are independently funded. A gram of NR a day for six weeks in healthy overweight and obese adults, in a crossover, raised fat-free mass slightly and muscle acetylcarnitine substantially, and left insulin sensitivity, mitochondrial function, liver and muscle lipid, cardiac energy status, blood pressure and inflammatory markers all unchanged — small mechanistic movement, nothing a person would notice. A gram a day before and after an experimental muscle injury in thirty-two adults aged fifty-five to eighty, combined with pterostilbene, did not affect muscle stem cell recruitment, fibre area or any other regeneration measure. Ten weeks at a gram a day in twenty older adults with mild cognitive impairment raised blood NAD+ 2.6-fold and left cognition unchanged, with an exploratory and uncorrected reduction in epigenetic age.

The positives are narrower and real. Twenty-one days of NR in aged men lowered circulating inflammatory cytokines and moved the NAD+ metabolome in biopsied skeletal muscle — though what rose there was a marker of NAD+ turnover rather than muscle NAD+ itself, and mitochondrial function was unchanged. It is still the clearest human evidence that an oral precursor reaches a target tissue. Six weeks of NR in forty COPD patients cut sputum interleukin-8 by 52.6 percent against placebo, an effect still present twelve weeks after treatment stopped. And in ninety people with peripheral artery disease, six months of NR improved six-minute walk distance by 17.6 metres over placebo — a clinically meaningful margin, in a trial whose result the authors ask to see confirmed.

How NMN reaches a cell, and whether it reaches a joint

Two things are unsettled here. The first is transport. A 2019 paper reported that the gene Slc12a8 encodes a dedicated NMN transporter carrying the molecule into cells intact, chiefly in the small intestine; a formal rebuttal the following year argued the data do not establish that, and that the older model — NMN dephosphorylated to nicotinamide riboside before entry — remains standing. The original authors replied, the exchange closed unresolved, and the paper carries a published correction. The head-to-head trial above adds a third possibility, that the gut microbiome converts both precursors to nicotinic acid before anything is absorbed at all.

The second is tissue. Human muscle has been biopsied after NR and the NAD+ metabolome moved there, by way of a downstream marker rather than a rise in muscle NAD+. Nobody has sampled human cartilage or synovium after any NAD+ precursor. Cartilage is avascular and fed by diffusion through synovial fluid, so distribution to it cannot be assumed from a blood level.

The chain from NAD+ to cartilage

Every link in the joint story has been tested, and every one of them is animal or human tissue rather than a human trial.

NAD+ is low in osteoarthritic cartilage. Measured in human tissue: a 2025 multi-model study found NAD+ depleted in OA cartilage, with the NAD+-consuming enzyme PARP14 implicated as one culprit. Cross-sectional.

Ageing drains the machinery. Cartilage SIRT3 — the mitochondrial sirtuin that keeps antioxidant defences working — falls substantially with age in rats and mice, and its damage signature turns up in human OA cartilage. CD38 rises with age in mouse tissues and is upregulated in human OA cartilage. SIRT6 sits in the same position: it is NAD+-dependent, its activity falls as NAD+ falls, and in chondrocytes it governs mitochondrial number and membrane integrity, limits reactive oxygen species and supports mitochondrial DNA transcription. The review that assembles that case is also the clearest about where it stops — SIRT6 targeting has never been verified in a person, no clinical trial has tested a SIRT6 activator for efficacy or safety, and SIRT6 is not a validated biomarker of anything.

Chondrocytes need it. Mice lacking SIRT1 in cartilage develop accelerated osteoarthritis under load and ageing; whole-body SIRT3 deletion accelerates knee OA. Mouse genetics prove the pathway matters, not that topping it up helps.

Restoring NAD+ protects rodent joints. NAD+ precursors including NMN rescued cartilage in rodent OA models. Oral nicotinamide riboside reduced cartilage damage in rats with chemically induced knee OA, in groups of five animals; hydrolysed collagen, given alongside at the same dose, cleared the same bar, and the two were never tested against each other. NMN paired with a CD38 blocker preserved aged mouse cartilage, and NMN rescued stressed human disc cells and protected rat discs.

Two findings pull the other way and belong here. Extracellular NAMPT, the enzyme that makes NMN inside the body, is actively catabolic when it acts on cartilage from outside the cell, driving prostaglandin release and matrix degradation. And CD38 — elevated in aged and osteoarthritic joints — is the main enzyme degrading NMN itself. Notably, the largest joint effects in mice came from blocking CD38 rather than from adding precursor.

The resveratrol pairing

A popular protocol pairs NMN with resveratrol on the logic that resveratrol activates SIRT1 and NMN supplies the NAD+ it burns. Two parts of that are solid: SIRT1 does require NAD+, and NMN does raise it. The third part is contested at the bench. The original activation signal appeared with a fluorophore-tagged laboratory substrate; independent biochemistry found no activation against native substrates, while a 2013 study identified a single SIRT1 residue required for activation by every activator class tested and showed the metabolic effects disappeared in cells carrying an activation-defective enzyme. Both sides are appraised in the resveratrol entry.

What the pairing now has is a randomised human test, though not with NMN and not with a joint endpoint. The peripheral artery disease trial above ran a third arm: nicotinamide riboside plus 125 mg of trans-resveratrol daily. At six months NR alone beat placebo by 17.6 metres on the six-minute walk; NR plus resveratrol beat it by 3.7 metres and was not significant. At three months both arms beat placebo, at 22.4 and 20.6 metres. The combination arm also reported nausea or vomiting in 36 percent against 14 percent on NR alone, and only 52 percent of its participants took at least three-quarters of their pills against 75 and 76 percent in the other arms — so the failure to add benefit may be pharmacology or may be adherence. It is the only randomised evidence there is on the combination in people.

The pairing has been tested directly once, in mice. A single oral dose of NMN with resveratrol left NAD+ 1.59-fold higher in heart and 1.72-fold higher in skeletal muscle than NMN alone, measured across the six hours after dosing in three animals per group at each timepoint. Six tissues were sampled and none of them was a joint. The authors name that sample size as a limit on their own statistics. What it establishes is that resveratrol changes where an NMN dose ends up — which is the question a pharmacokinetic study can answer, and it is a different question from whether the pairing does anything for a knee.

Dosing, safety, and regulatory status

Human trials have used 250 to 2,000 mg a day. Blood NAD+ rises reliably from about 250 to 300 mg, and doubling a dose does raise exposure: 1,000 mg twice daily produced roughly twice the exposure of once daily. But the curve is not uniformly steep. The one trial to compare three doses head to head found 600 mg a day raised blood NAD+ more than 300 did and 900 mg no more than 600, and put its own ceiling at 600 mg.

At 1,000 mg a day the direct safety evidence is short-term but no longer thin. The pooled analysis of fifteen randomised trials at 250 to 2,000 mg a day for up to twenty-four weeks found no increase in adverse events, serious adverse events, withdrawals, or ALT and AST. The highest dose in that range, 2,000 mg a day for twenty-eight days, produced adverse events at the same rate as placebo. What remains unmapped is the long term: no trial has run past twenty-four weeks.

Two theoretical concerns are worth naming because neither has been tested at supplement doses in people. The first is the senescence work above — that NAD+ feeds the inflammatory secretome as readily as it feeds repair. The second is newer and more concrete. Atherosclerosis-prone mice on a high-cholesterol diet given high-dose nicotinamide riboside developed larger aortic plaques and higher TNF-alpha, IL-6 and LDL cholesterol than unsupplemented controls. Liver and plasma NAD+ did not rise; the terminal metabolite 4PY did, SIRT1 fell and CD38 rose, which the authors read as NAD+ metabolism being diverted away from the sirtuins and towards CD38 and PARP1. It is a mouse study at dietary doses and it reports plaque area rather than events, but the mechanism it describes is the same CD38 competition that the joint literature keeps running into, and its authors advise caution in people with atherosclerosis.

The US regulatory position is a genuine oddity: since late 2022 the FDA's stated view has been that NMN is excluded from the dietary-supplement definition because it was first investigated as a drug. It remains widely sold.

What would move this

The healthspan literature has arrived at a stable position: oral NMN raises NAD+ reliably, is well tolerated at the doses people take, and produces small, inconsistent functional effects that are largest where the sponsor is the manufacturer and the endpoint is secondary. That is a better-evidenced supplement than most things in this section, and it is why people take it.

The cartilage question is a different one, and it is still open in a way the ageing question is not. The mechanism there is unusually well mapped — human-tissue NAD+ depletion, a named drain, genetic proof the pathway matters, rodent rescue from several independent groups — and what it lacks is a single human joint measurement. Two results would move this tier: a randomised NMN or NR trial reporting knee pain, function or MRI cartilage outcomes, or human evidence that an oral precursor reaches joint tissue at all. The second may be the easier study, and it would tell you whether the first is worth running.