If you've spent an hour on the floor beside a bed, or watched the clock pass midnight while your child lies there wide awake and cheerful, you already know this isn't a discipline problem. Autistic kids often lie awake for hours after lights-out, or wake repeatedly once they're down. Some run on a body clock that seems set to the wrong time zone.
It isn't rare. Somewhere between half and two-thirds of autistic children have real trouble sleeping: long settling times and broken nights, sometimes waking for the day before dawn (Posar 2020, PMID 32520370). And a bad night doesn't stay contained. Short, fractured sleep makes the next day harder to regulate, which raises anxiety and arousal, which then wrecks the next night; researchers describe this as a self-reinforcing loop between the body clock and sleep (Carmassi 2019, PMID 31244687).
So most parents arrive with a narrow, practical question: does melatonin actually work, and is there anything worth trying before it? Here's what the trials show, the order to reach for things, and where the honest limits are.
Why the guideline says to start with the bedtime routine
The American Academy of Neurology reviewed the evidence on sleep in autistic children and adolescents and put behavioural measures first, ahead of any pill (Williams Buckley 2020, PMID 32051244). That's not a polite delaying tactic. For a lot of children the biggest gains come from dull, consistent changes: a steady wind-down, the same wake time even on weekends, a dark cool room, and screens off well before bed. It's also worth checking two quieter culprits, late-afternoon naps and the caffeine hiding in a fizzy drink or a bar of chocolate.
The guideline is useful for another reason: it says what melatonin is and isn't. It's an add-on for a child whose sleep is still a problem after those routine changes, and after a clinician has ruled out treatable medical causes. It isn't a first move, and it isn't a stand-in for the routine.
What the melatonin trials actually found
The strongest evidence is for one particular kind of melatonin: a prolonged-release formulation designed for children, not the gummies on a shop shelf. In a randomized, placebo-controlled trial of 125 autistic children and adolescents, prolonged-release melatonin taken for 13 weeks (2 mg, raised to 5 mg if needed) added about 57.5 minutes of total sleep against roughly 9 minutes on placebo, and shortened the time to fall asleep by around 40 minutes (Gringras 2017, PMID 29096777). Close to 69% of the children responded, versus about 39% on placebo, which works out to treating three or four children for one to clearly benefit.
Older, smaller work points the same way. In an open-label study of 24 autistic children, immediate-release melatonin at 1 to 3 mg shortened settling time and, for worn-out households, lowered the parents' own stress scores (Malow 2012, PMID 22160300). Children kept on prolonged-release melatonin have been followed for up to two years, with the sleep benefit holding and no sign that it changed their growth or the timing of puberty (Maras 2018, PMID 30132686; Malow 2021, PMID 31982581).
Two caveats sit on top of those numbers. The trials used pharmaceutical, prolonged-release melatonin at controlled doses, which isn't what most families pick up at the pharmacy. And even in the trial the average gain was under an hour: genuinely worth having for a depleted family, not a switch that turns a child into a sound sleeper.
Using melatonin without the avoidable problems
Melatonin is a hormone, not a vitamin, so the sensible approach is the lowest dose that works, with a clinician in the loop for a child. The American Academy of Pediatrics gives the same advice: the smallest effective amount and products checked by an independent lab, plus storage that treats it like any other medicine.
That storage point isn't fussiness. Melatonin usually arrives as sweet gummies, and US poison centres recorded a 530% rise in children's melatonin ingestions between 2012 and 2021, enough to make it the substance children were most often reported swallowing by accident (Lelak 2022, PMID 35653284). Keep it locked away.
The independent-lab part matters just as much. When researchers bought melatonin gummies off the shelf and measured them, the real melatonin content ranged from about 74% to 347% of the label, and one product contained none at all (Cohen 2023, PMID 37097362). A gummy marked 3 mg could be handing a child close to 10.
One more trap is baked into how sleep supplements get formulated. Many melatonin products, including several in our own catalogue, combine it with 5-HTP, a serotonin precursor. Layering serotonin-raising ingredients, especially on top of an SSRI or SNRI that a child or teen may already take, can drive serotonin too high, and serotonin syndrome is a real danger from that kind of mix (Mikkelsen 2023, PMID 37309284). If you use melatonin, use a plain one, and tell the prescriber it's on board.
The gentler options, and how thin their evidence is
Plenty of sleep and calm products lean on magnesium, glycine or GABA. The honest summary: for sleep in autistic children the evidence is thin to absent for all three, and where trials do exist they were run in adults, on outcomes that aren't quite "does this child sleep better."
Magnesium is the most defensible of the three, mainly as a wind-down mineral. In adults who were short on magnesium, 300 mg a day cut perceived stress by about 42% (Pouteau 2018, PMID 30562392). That was a stress trial rather than a sleep trial, and it wasn't done in children, so read it as support for settling before bed, not as evidence of better sleep. Glycine has a single thread: 3 g before bed left adults less foggy and quicker on next-day tests after their sleep was deliberately cut short (Bannai 2012, PMID 22529837), again an adult study, and about the next day rather than the night itself. Oral GABA is the weakest of the set; whether much of a swallowed dose even reaches the brain is still debated, and the human calming data are small and soft. Treat it as a mild evening ritual, with no pretence that it has much evidence behind it.
How this shapes our protocol
Our Rest set is built to track this evidence and to be honest about its edges. For adults it uses one evening lever, not a stack: either Magnesium Night (300 mg of magnesium with tart cherry, California poppy and lavender), the dose that matched the adult stress trial above, or Glycine Powder at 3 g before bed, the amount used in the glycine study. One or the other, because doubling up on evening sedatives is how you earn a groggy morning. GABA Lozenges sit alongside as an optional evening calm, with no claim that oral GABA is well supported.
For children the set is deliberately smaller: Glycine Powder, practitioner-directed, with no hormone in it. Melatonin is the awkward gap. It has the best pediatric evidence of anything on this page, and yet it's the one thing our Rest set leaves out. Our child-suitable melatonin lozenge is waiting on confirmation of the exact strength printed on its label, and every other melatonin product we carry is combined with 5-HTP, which rules it out for a child. Until that label is confirmed, the responsible move is to leave melatonin off the shelf rather than guess at a dose.
The limits deserve stating plainly. Almost all of the supplement evidence here comes from adult trials of single ingredients; the magnesium and glycine doses in our products match those trials, but the trials weren't about children's sleep, and no combination like this has been tested as a whole. The intervention with the best evidence for autistic children, prolonged-release melatonin used after behavioural measures, is a clinician's decision and, for now, one we can't fully supply. If your child's sleep is a real problem, start with the routine and talk to your clinician about melatonin; treat these wind-down products as the mild help they are. You can see exactly what's in the set on the Rest protocol page.
Sources
- Posar A, Visconti P. 2020. Pediatr Ann. PMID 32520370
- Carmassi C, et al. 2019. Front Psychiatry. PMID 31244687
- Williams Buckley A, et al. 2020. Neurology. PMID 32051244
- Gringras P, et al. 2017. J Am Acad Child Adolesc Psychiatry. PMID 29096777
- Malow B, et al. 2012. J Autism Dev Disord. PMID 22160300
- Maras A, et al. 2018. J Child Adolesc Psychopharmacol. PMID 30132686
- Malow BA, et al. 2021. J Am Acad Child Adolesc Psychiatry. PMID 31982581
- Lelak K, et al. 2022. MMWR Morb Mortal Wkly Rep. PMID 35653284
- Cohen PA, et al. 2023. JAMA. PMID 37097362
- Mikkelsen N, et al. 2023. Basic Clin Pharmacol Toxicol. PMID 37309284
- Pouteau E, et al. 2018. PLoS One. PMID 30562392
- Bannai M, et al. 2012. Front Neurol. PMID 22529837