Across the randomized trials of oral NAD⁺ precursors, the adverse events that got written down are mild and few: some stomach upset, a transient change in one blood marker, and rates that generally did not separate from placebo. The interesting part is not the list. It is how short and how small the trials producing it were, and what the ingredient manufacturer’s own animal toxicology found at doses nobody sells.
This page is the ledger for capsules and powders — nicotinamide riboside and NMN taken by mouth. If you want the format comparison instead, side effects by route covers that, and what “safe” can honestly mean covers the risk framing underneath both.
The symptoms that were actually reported
Start with the most specific count anyone has published. In a randomized, double-blind, placebo-controlled dose-escalation study of nicotinamide riboside with pterostilbene in 24 hospitalized patients with acute kidney injury, dosed twice daily for two days up to 1,000 mg NR with 200 mg pterostilbene, three of the 20 patients who received the combination reported minor gastrointestinal side effects. Every safety laboratory test — creatinine, estimated glomerular filtration rate, electrolytes, liver function and blood counts — was unchanged [1].
That is roughly the shape of the whole literature: a small number of people, mild stomach complaints, normal bloodwork. Twelve weeks of NR at 1,000 mg twice daily in 40 obese, insulin-resistant men produced no serious adverse events attributable to the supplement and normal safety blood tests [2]. An eight-week trial in overweight adults at 100, 300 and 1,000 mg daily found no significant difference in adverse events between NR and placebo, and none between the dose groups [3].
Read that last finding carefully, because it is the one most often misquoted in both directions. “No difference from placebo” does not mean nobody felt anything. It means the people on sugar pills reported symptoms at a similar rate — which is the normal result when you ask several dozen adults to keep a symptom diary for two months, and it is why a forum post describing nausea on day three is not evidence about the capsule.
The one lab value that moved
The NR-SAFE trial is the closest thing to a deliberate stress test. Twenty people with Parkinson’s disease were randomized 1:1 to nicotinamide riboside 1,500 mg twice daily (n = 10) or placebo (n = 10) for four weeks — a dose above anything previously tested in humans. All 20 completed. There were no moderate or severe adverse events and no significant difference in mild ones. But NR recipients showed a slight initial rise in serum homocysteine, with the integrity of the methyl donor pool otherwise intact [4].
Why that is worth a paragraph: NR and NMN are ultimately cleared through methylation, which consumes methyl groups your body also uses elsewhere. A transient homocysteine bump at 3,000 mg a day is not a finding that should stop anyone taking a 300 mg capsule. It is a finding that tells you where to look if long-term high-dose data ever gets collected, and it is the only signal in the oral literature that points at a mechanism rather than at a stomach.
The flushing question, answered properly
People arrive at this topic having read that vitamin B3 makes you flush. That is true of one member of the family and not the others. Nicotinic acid — plain niacin — is a vasodilator, which is why a topical dermatology formulation had to be chemically redesigned specifically to deliver nicotinic acid to skin without vasodilatation [5]. Nicotinamide riboside and NMN are not nicotinic acid, and the eight-week dose-ranging trial recorded no reports of flushing at 100, 300 or 1,000 mg daily [3].
So if a product page warns you about flushing, or promises to have solved it, check which molecule is in the bottle. The distinction is laid out in what nicotinamide actually is, and it is the sort of thing the NR product most of the published research used gets right while cheaper relabels often do not.
The animal study nobody quotes
Before NR reached the supplement shelf, its manufacturer commissioned a standard toxicology package: genotoxicity assays plus acute, 14-day and 90-day rat studies. NR was not genotoxic and there was no mortality at an oral dose of 5,000 mg/kg. The 90-day study compared NR at 300, 1,000 and 3,000 mg/kg/day against an equimolar dose of nicotinamide as a positive control, and reported that the target organs of toxicity were liver, kidney, ovaries and testes, with a lowest observed adverse effect level of 1,000 mg/kg/day and a no observed adverse effect level of 300 mg/kg/day [6].
Two honest readings of that, and you need both. The reassuring one: a rat NOAEL of 300 mg/kg/day sits far above the human exposures on sale, and that margin is exactly why regulators were satisfied. The uneasy one: the toxicology says organ effects exist somewhere above that line, and no human study has ever gone looking for them because none has run long enough or at a high enough dose. A seller citing “extensive safety testing” is usually citing this package without mentioning which organs it named.
Where the record simply runs out
The binding limitation on all of this is duration. The longest randomized exposure published is a 26-week crossover trial of NR 1,000 mg daily in patients with Werner syndrome, a rare premature-aging condition, which reported no serious adverse events [7]. A 24-week trial in long COVID gave 2,000 mg a day and logged one serious adverse event, deemed unrelated to the study drug [8]. Everything else is weeks.
That same long COVID trial carries a number worth sitting with, because it is the kind of thing a tolerability summary usually smooths over: dropout in the continuous-NR group ran 32.4% at 10 weeks and 51.4% at 20 weeks, against 14.3% at each timepoint in the group that started on placebo [8]. In a symptomatic population over months, people stopped taking it. The trial was not designed to explain why, and nobody should read a cause into it that the authors did not report — but a lopsided dropout is a tolerability observation in its own right, and it is absent from every marketing page.
For NMN specifically the picture is similar and separately documented; the NMN safety record has its own page because its trials, and its regulatory position, are not the same as NR’s.
What to do with this
Expect mild and mostly gastric if anything, expect it to settle, and expect no trial to be able to tell you what a decade looks like. Take the dose with food if your stomach objects. Tell a clinician what you are taking and at what milligram figure — not “an NAD⁺ supplement”, which is not a quantity — and raise it specifically if you are pregnant, on methylation-relevant medication, or being treated for liver or kidney disease, none of which the trials above enrolled.
And keep the comparison in view. The oral formats are the ones with an adverse-event record to read at all. The injectable versions cost several times more and have no equivalent ledger, which their own page goes through in detail.