If you have ever stood in a kitchen near midnight holding a gummy that says 10 mg, having decided that the 5 mg one stopped working, you have made the escalation decision. It is the most common thing people do with melatonin and there is no evidence behind it.
I went looking for the physiological dose — the amount your own body puts into your blood at night — expecting to find a clean number and a clean verdict. What I found was a patent, a press release from 2001, a meta-analysis from 2024 that points the other way, and one trial in exactly the right age group that measured something nobody argues about.
Where “ten times too much” comes from
The claim that retail melatonin is roughly ten times the dose you need is traceable, which is more than can be said for most supplement folklore. It comes from a comparison between a 0.3 mg dose and the 3 mg pill that was standard in American shops, publicised by MIT in 2001 alongside a trial from Richard Wurtman’s laboratory.
That trial is real and it is good. Thirty subjects, double-blind, placebo-controlled, crossover, with polysomnography on the last three nights of each treatment period and week-long washouts between them. The 0.3 mg dose restored sleep efficiency (p < 0.0001) and raised plasma melatonin to normal levels (p < 0.0008). The 3.0 mg dose also improved sleep, but it induced hypothermia and left melatonin circulating into daylight hours (Zhdanova et al., 2001).
Two things about it rarely survive into the retelling.
The first is the population. Every subject was over fifty, and half of them had actigraphically confirmed reduced sleep efficiency. If you are thirty-four and your problem is that you cannot fall asleep before one in the morning because your day ends at midnight, you were not in that study.
The second is the null. The trial also enrolled fifteen normal sleepers over fifty as controls. They had low melatonin levels too. Their sleep was unaffected by any dose — 0.1, 0.3 or 3.0 mg. In people whose sleep was not already broken, nothing happened at any dose tested.
There is a conflict of interest here and it belongs in this paragraph rather than a footnote. MIT holds patents on the use of melatonin to promote and sustain sleep, covering doses up to 1 mg, with Wurtman named as inventor. He disclosed this in his own published work, stated that his royalties are donated back to MIT, and stated that the underlying laboratory research was funded by the National Institutes of Health. Nobody concealed anything; the disclosure is a public document. But the most influential advocate of “0.3 mg is the right dose” was an inventor on a patent covering doses up to 1 mg, and you are entitled to know that before you weigh his conclusion. Funding disclosures for most of the other trials below were not stated in the sources I could reach — I have marked that rather than assuming they were clean.
The trial in your age group measured the clock, not the milligrams
Here is the study that changed how I think about this.
Burgess and colleagues built phase response curves for two doses of melatonin using the same protocol: 0.5 mg against a previously published 3.0 mg curve. Thirty-four healthy adults, sixteen men and eighteen women, aged 18 to 42. Two five-day laboratory sessions, each preceded by a week of fixed sleep times, with three days of an ultradian light–dark cycle in between and dim light melatonin onset measured at both ends. Phase shifts were calculated by subtracting each person’s shift on placebo.
The result: the largest advances came when 0.5 mg was taken in the afternoon, two to four hours before dim light melatonin onset, or nine to eleven hours before the midpoint of sleep. The optimal time is later for the lower dose. And when each dose was given at its own optimal time, the two produced advances and delays of similar size (Burgess et al., 2010).
Read that again with the bottle in your hand. Half a milligram moved the body clock as far as three milligrams did. What differed was the hour.
Nobody’s packaging says this. Every label says thirty minutes before bed, which is the instruction for a sedative, and melatonin at these doses is not really working as one.
The meta-analysis that says the opposite
If the story ended there it would be tidy, and I would distrust it.
A 2024 dose–response meta-analysis pooled 26 double-blind randomised trials published between 1987 and 2020 — 1,689 observations across insomnia patients and healthy volunteers — and found that melatonin progressively reduced sleep onset latency and increased total sleep time, with the effect peaking at 4 mg per day and the highest efficacy between 3 and 5 mg. The interval between taking it and the sleep episode was a significant predictor of onset latency (β = −0.16, p = 0.023), as was insomnia status (β = 0.50, p < 0.001).
That is not a small or obscure paper and it does not support “less is more”. Neither does a month-long crossover trial in twenty-four healthy adults over 55, which compared 0.3 mg against 5.0 mg under a forced desynchrony protocol with polysomnography: the high dose increased sleep duration, while 0.3 mg produced only a trend toward better sleep efficiency, driven by sleep during the biological day (Duffy et al., 2022).
The reconciliation is that these are measuring a different thing. The dose–response analysis pools trials of the sedative effect across mixed populations. The phase response curves measure clock shifting. Both can be right, and the practical consequence is that the supportable claim is that the high dose is unnecessary, not that the low dose works better. Those are different sentences. Only the first survives the evidence, and if you see the second stated confidently anywhere, including here, it is running ahead of the data.
Seven minutes
Whatever dose you land on, the size of the effect deserves saying plainly.
A meta-analysis of nineteen studies in 1,683 adults and children with primary sleep disorders found melatonin reduced sleep latency by a weighted mean of 7.06 minutes (95% CI 4.37 to 9.75, p < 0.001) and increased total sleep time by 8.25 minutes (95% CI 1.74 to 14.75, p = 0.013). Sleep quality improved with a standardised mean difference of 0.22. The authors’ own framing was that the effects are modest and the absolute benefit smaller than other pharmacological options, though the side-effect profile is milder (Ferracioli-Oda et al., 2013). Trial duration and dose had no significant effect on sleep quality — a null the coverage tends to drop.
Seven minutes is a real effect. It is not a different life. Anyone selling you a transformed working day on the back of it is selling something.
And there is a cleaner null worth sitting with. A randomised crossover study in sixty adults with self-reported poor sleep compared 3.0 mg, 0.5 mg and placebo using wearable-measured sleep duration. Averaged sleep duration did not differ significantly from baseline for either dose — p = .70 for 3.0 mg, p = .90 for 0.5 mg — and the heterogeneity indices were low, meaning responses did not vary much between individuals either. That study’s primary outcome was the usability of its own trial platform, not efficacy, so it was not powered to detect a small effect. But it is the kind of null that should make you hold the whole category loosely. When a trial has the right population and a clean comparison, a null is an answer rather than a puzzle to be explained away.
The part where the bottle stops being a known quantity
Two laboratory analyses have measured what is actually in melatonin products.
The first tested 31 supplements from sixteen brands, bought in Guelph, Ontario, using ultra-high performance liquid chromatography. Content ranged from 83% below to 478% above the label. Lot-to-lot variation within a single product reached 465%. More than 71% of products fell outside ten per cent of their label claim, and serotonin was detected in 26% of them. The greatest variability was in chewable formulations (Erland and Saxena, 2017). That is a Canadian sample, and it should not be presented as a description of the American market.
The second is American and narrower: 25 unique melatonin gummy brands obtained in September 2022, measured at 74% to 347% of labelled melatonin, with only 12% falling within ten per cent of the label. Actual content per serving ran from 1.3 mg to 13 mg. The study tested one sample per brand and the authors flagged that limitation themselves (Cohen et al., 2023). Their disclosures: research support from Consumers Union and PEW Charitable Trusts, royalties from UpToDate, and a civil suit brought against one author by a supplement company in which the jury found in the author’s favour; a co-author reported grants from the FDA, NIH and USDA. The Council for Responsible Nutrition, the dietary supplement industry’s trade association, published a response arguing the findings overstated the risk and that overages reflect shelf-life and potency requirements. That is a trade body defending its members’ products, and you should weigh it as such — but you should also know it exists.
There is a physiological reason this matters more than it would for, say, magnesium. Four healthy men aged 21 to 32, given the same 500 µg oral dose, reached peak plasma concentrations ranging from 480 to 9,200 nanograms per litre — roughly a twenty-fold spread. Oral bioavailability across the four ranged from 10% to 56%, and the authors traced it to first-pass hepatic extraction (Di et al., 1997). Four men is a very small study and I would not build a protocol on it alone. But a systematic review of 22 pharmacokinetic studies puts oral bioavailability at roughly 15%, ranging from 9% to 33% across studies, with a half-life around 45 minutes and peak concentration about 50 minutes after an immediate-release dose — and notes that melatonin pharmacokinetics are altered by age, caffeine, smoking, oral contraceptives and feeding status.
Stack those two facts. The label is unreliable, and the conversion from milligrams swallowed to nanograms circulating varies several-fold between people. The precision of the dose argument is largely imaginary.
Safety: what the last year added, and what it can show
In November 2025, a retrospective analysis was presented at the American Heart Association Scientific Sessions. Using the TriNetX Global Research Network, investigators identified adults with an insomnia diagnosis, defined exposure as at least one melatonin prescription with 365 or more exposure-days, excluded anyone with prior heart failure or other prescription hypnotics, and propensity-matched 65,414 users against 65,414 controls across demographics, fifteen comorbidities, cardiometabolic drugs, laboratory values, vitals and healthcare utilisation. Over five years: incident heart failure hazard ratio 1.89 (4.6% versus 2.7%), heart failure hospitalisation 3.44, all-cause mortality 2.09 (Nnadi et al., abstract 4371606).
Now the limitations, in the same breath, because a risk figure without its design is a rumour.
This is a conference abstract. The American Heart Association states plainly in its own release that abstracts presented at its meetings are not peer-reviewed and that findings are preliminary until published in full. It is observational and cannot establish causation; the authors say so.
The specific problem is the exposure definition. Melatonin is sold over the counter in the United States. The database spans countries where it requires a prescription, such as the United Kingdom, and countries where it does not. An unknown share of the “non-melatonin” control group almost certainly took melatonin bought off a shelf, with no record of it anywhere. Independent commentary has also questioned how well the underlying database represents any defined population. And confounding by indication is obvious: more severe insomnia, depression and anxiety plausibly drive both long-term use and cardiac risk.
Pointing the other way, a 2025 systematic review reported positive effects of melatonin in heart failure patients, and earlier reviews have discussed it as a candidate protective therapy — though that literature rests substantially on experimental and animal work, and none of it should be read as a human clinical finding.
Two things that are not in dispute. At 3.0 mg, melatonin induced hypothermia and left plasma elevated into the next day; at 0.1 and 0.3 mg it did not (Zhdanova et al., 2001). And paediatric ingestions reported to American poison control centres rose 530% between 2012 and 2021, totalling 260,435 cases, with melatonin accounting for 4.9% of all paediatric ingestions in 2021 against 0.6% in 2012; five children required mechanical ventilation and two died. A later surveillance report attributed roughly 11,000 emergency department visits among children under five between 2019 and 2022, frequently involving flavoured gummies, with about three-quarters of documented cases involving a bottle being accessed. Child-resistant packaging is not required for these products. If there is a small child in your house, this is the paragraph that matters most, and it has nothing to do with your sleep.
What we do not have: any long-term randomised trial of nightly melatonin in healthy adults. The longest human evidence is the observational abstract above. How long is safe, nightly, at any dose, in someone your age — data not specified, and that phrasing is deliberate.
How to check a dose claim yourself
This is the transferable part, and it generalises well beyond melatonin.
When someone tells you a supplement dose, ask which effect the dose refers to. Melatonin has two: it shifts the body clock, and it mildly promotes sleep onset. They have different optimal doses, different optimal timing, and different evidence bases. Almost every popular dosing guide answers one question and then gives advice that only makes sense for the other.
Then ask who was in the trial. “Adults over fifty with insomnia” and “healthy thirty-year-olds with a late schedule” are different species for this purpose, and the gap between them is where most confident advice goes wrong.
Then check whether the recommended dose came from a regulator or a press release. The European Food Safety Authority concluded that 1 mg taken close to bedtime substantiates a claim of reduced sleep onset latency, and that 0.5 to 5 mg substantiates a jet lag claim. France, the most permissive European Union state on this, caps food supplements at 2 mg per daily dose. Those are checkable documents with reasoning attached.
Which produces an arithmetic that needs none of the contested premises: a 10 mg American gummy contains five times the maximum daily amount France permits in a supplement, and ten times what European regulators concluded is needed to claim it helps you fall asleep faster.
What to change, in order
First, move the hour, not the milligrams. If your problem is a body clock that runs late, the phase response data says to take it in the afternoon or early evening — roughly nine to eleven hours before the midpoint of your sleep — not thirty minutes before bed. For a midnight-to-eight sleeper, that is late afternoon. This costs nothing and you can do it tonight. The honest caveat: that trial measured circadian phase, not how anyone felt the next day. Nobody has run this in knowledge workers and measured next-day performance. No direct evidence exists for that specific question.
Second, stop escalating. No evidence retrieved here supports going above the 3 to 5 mg region for the sedative effect, and the dose–response analysis found the curve peaks around 4 mg rather than continuing upward. Above that you are buying side effects. If a dose is not working, taking more of it is the one move the data actively fails to support.
Third, if you are using it to shift a clock, low dose is reasonable. Not because low doses are proven superior — they are not — but because 0.5 mg produced the same size of shift as 3.0 mg at its own optimal time, without leaving melatonin in your blood the next morning.
Fourth, buy on measurement, not brand. Given that a quarter of one tested sample was outside its label by more than a factor of two, a third-party tested product in a plain tablet you can split beats a flavoured gummy whose contents were the least predictable of any format tested.
The dose debate has been running for twenty-five years and it is still not settled. The timing question was answered in 2010, in thirty-four adults aged 18 to 42, and almost nobody acts on it.
About this article
Written by Drew Anton. Drew covers compound analysis, dose arithmetic, and what the label does and does not tell you. Not a physician or research scientist — reads the primary literature closely and refuses to round up.
Medical review: None. NeuriFuel does not currently have a licensed clinician on the editorial team, and this article has not been medically reviewed. We state this rather than implying an authority we do not have. See our About page for our full methodology.
Sources: Fifteen, comprising five randomised controlled trials with polysomnographic, circadian or wearable measurement, three meta-analyses or systematic reviews, two laboratory product analyses, one human pharmacokinetic study, two public health surveillance reports, one set of regulatory opinions, and one non-peer-reviewed conference abstract, which is labelled as such wherever it appears. Funding and conflict-of-interest disclosures were retrievable for one study and are stated inline; for the remainder they were not available in the sources consulted, and that absence is marked rather than assumed favourable. Values that could not be established are marked “data not specified” in the text, and conflicts between sources are presented rather than resolved silently.
Corrections: Found an error? Write to hello@neurifuel.com with a source and we will fix it and log the correction.
Last updated: 31 July 2026
References
- Zhdanova IV, Wurtman RJ, Regan MM, Taylor JA, Shi JP, Leclair OU (2001). Melatonin Treatment for Age-Related Insomnia. Journal of Clinical Endocrinology & Metabolism 86(10):4727–4730. DOI: 10.1210/jcem.86.10.7901. Author disclosure: MIT holds patents on melatonin use to promote and sustain sleep, with Wurtman named as inventor; royalties donated to MIT; laboratory research funded by the NIH.
- Zhdanova IV, Wurtman RJ, Morabito C, Piotrovska VR, Lynch HJ (1996). Effects of low oral doses of melatonin, given 2–4 hours before habitual bedtime, on sleep in normal young humans. Sleep 19(5):423–431. DOI: 10.1093/sleep/19.5.423. PMID: 8843534.
- Burgess HJ, Revell VL, Molina TA, Eastman CI (2010). Human Phase Response Curves to Three Days of Daily Melatonin: 0.5 mg Versus 3.0 mg. Journal of Clinical Endocrinology & Metabolism 95(7):3325–3331. DOI: 10.1210/jc.2009-2590. PMID: 20410229.
- Optimizing the Time and Dose of Melatonin as a Sleep-Promoting Drug: A Systematic Review of Randomized Controlled Trials and Dose–Response Meta-Analysis (2024). Journal of Pineal Research. DOI: 10.1111/jpi.12985. PMID: 38888087.
- Ferracioli-Oda E, Qawasmi A, Bloch MH (2013). Meta-Analysis: Melatonin for the Treatment of Primary Sleep Disorders. PLOS ONE 8(5):e63773. DOI: 10.1371/journal.pone.0063773. PMID: 23691095.
- Duffy JF et al. (2022). High dose melatonin increases sleep duration during nighttime and daytime sleep episodes in older adults. Journal of Pineal Research. DOI: 10.1111/jpi.12801.
- A Series of Personalized Melatonin Supplement Interventions for Poor Sleep: Feasibility Randomized Crossover Trial. PMC12468169.
- Erland LAE, Saxena PK (2017). Melatonin Natural Health Products and Supplements: Presence of Serotonin and Significant Variability of Melatonin Content. Journal of Clinical Sleep Medicine 13(2):275–281. DOI: 10.5664/jcsm.6462. PMID: 27855744.
- Cohen PA, Avula B, Wang Y-H, Katragunta K, Khan I (2023). Quantity of Melatonin and CBD in Melatonin Gummies Sold in the US. JAMA 329(16):1401–1402. DOI: 10.1001/jama.2023.2296. PMID: 37097362. Disclosures stated inline in the text.
- Di WL, Kadva A, Johnston A, Silman R (1997). Variable Bioavailability of Oral Melatonin. New England Journal of Medicine 336:1028–1029. DOI: 10.1056/NEJM199704033361418.
- Harpsøe NG et al. (2015). Clinical pharmacokinetics of melatonin: a systematic review. PMID: 26008214.
- Nnadi E, Masara M, Offor R, Unal S, Rebah R, Atere M, Nigussie B, Graham-Hill S (2025). Abstract 4371606: Effect of Long-term Melatonin Supplementation on Incidence of Heart Failure in Patients with Insomnia. Circulation 152(suppl_3). DOI: 10.1161/circ.152.suppl_3.4371606. Conference abstract; not peer-reviewed.
- Lelak K, Vohra V, Neuman MI, Toce MS, Sethuraman U (2022). Pediatric Melatonin Ingestions — United States, 2012–2021. MMWR 71(22):725–729. An erratum to this report has been published.
- Freeman DI, Lind JN, Weidle NJ, Geller AI, Stone ND, Lovegrove MC (2024). Emergency Department Visits for Unsupervised Pediatric Melatonin Ingestion — United States, 2019–2022. MMWR 73(9):215–217. DOI: 10.15585/mmwr.mm7309a5.
- EFSA Panel on Dietetic Products, Nutrition and Allergies (2011). Scientific Opinion on melatonin and reduction of sleep onset latency. EFSA Journal 9(6):2241. And (2010), melatonin and alleviation of subjective feelings of jet lag. EFSA Journal 8(2):1467. DOI: 10.2903/j.efsa.2010.1467.