Where Pediatric Cannabinoid Evidence Is Strong, Where It Is Thin, and Where It Does Not Exist At All

In a placebo-controlled randomized trial in tuberous sclerosis complex, cannabidiol at 25 mg per kilogram per day reduced seizures by 48.6 percent compared with 26.5 percent on placebo, a 30.1 percentage point difference. Nearly one in five children taking cannabidiol had elevated liver transaminases and none on placebo did.

In the largest randomized trial of cannabinoids in autism, 150 participants aged 5 to 21, the registered primary behavioral outcome did not differ between groups. The investigators described the efficacy evidence as mixed and insufficient.

A parent researching cannabis for a child with seizures, autism, chronic pain, or attention problems will find the same confident tone applied to all four, and only one of them has randomized evidence behind it. Sorting those apart is the difference between an informed specialist conversation and an expensive detour.

The safety information is also frequently understated. Cannabidiol in the pediatric epilepsy trials was not a gentle intervention: it produced diarrhea, somnolence, liver enzyme elevations, and trial discontinuations at rates that any parent deserves to see before starting.

The evidence that changed pediatric practice came from randomized, placebo-controlled trials of a pharmaceutical cannabidiol oral solution in specific, severe, drug-resistant epilepsies. In Dravet syndrome, Orrin Devinsky and colleagues randomized 120 children and young adults to cannabidiol at 20 mg per kilogram per day or placebo, added to existing antiepileptic treatment. Median convulsive seizure frequency fell from 12.4 to 5.9 per month on cannabidiol, compared with 14.9 to 14.1 on placebo, an adjusted median difference of 22.8 percentage points.

The trial in tuberous sclerosis complex, led by Elizabeth Thiele and published in JAMA Neurology, randomized 224 patients aged 1 to 65 with a median age of 11.4 years to cannabidiol at 25 or 50 mg per kilogram per day or placebo for 16 weeks. Seizure reduction was 48.6 percent at the lower dose, 47.5 percent at the higher dose, and 26.5 percent on placebo. The reduction relative to placebo was 30.1 percentage points for the 25 mg per kilogram group.

These results are real and they are also specific. The conditions are rare and severe, the product is a standardized pharmaceutical preparation rather than a cannabis extract, dosing is by body weight, and treatment is added to existing antiepileptic medication under neurology supervision. Every one of those conditions is part of what was tested.

The epilepsy trial evidence is covered in more detail in our companion piece on recognizing seizures, linked below, so this page focuses on where the evidence stops.

Autism is where parents most often encounter enthusiastic claims, and it is where the best-designed trial produced the most equivocal result. Adi Aran and colleagues at Shaare Zedek Medical Center in Jerusalem randomized 150 participants aged 5 to 21 with autism spectrum disorder to a whole-plant cannabis extract with cannabidiol and THC in a 20 to 1 ratio, a purified cannabinoid preparation at the same ratio, or placebo, for 12 weeks, followed by washout and crossover.

The registered primary outcome, change in total score on the Home Situation Questionnaire adapted for autism, did not differ between groups. On the co-primary Clinical Global Impression measure, disruptive behavior was rated much or very much improved in 49 percent on whole-plant extract against 21 percent on placebo. Median Social Responsiveness Scale scores improved by 14.9 points on whole-plant extract against 3.6 on placebo. Somnolence was reported by 28 percent and decreased appetite by 25 percent on whole-plant extract, against 8 percent and 15 percent on placebo.

The authors’ own conclusion is the fairest summary available: the preparations were well tolerated, evidence for efficacy was mixed and insufficient, and further testing is recommended. A parent should read a split between a null primary outcome and positive secondary outcomes as a reason to keep studying the question, not as a demonstration that it works.

The same research group examined sleep outcomes in the same 150 participants using the Children’s Sleep Habits Questionnaire and found that cannabinoid treatment was not superior to placebo on any measured aspect of sleep, including bedtime resistance, sleep-onset delay, and sleep duration. Sleep improvement is one of the most commonly claimed benefits in this population and it did not separate from placebo.

For pediatric attention deficit hyperactivity disorder, no adequately powered randomized trial of cannabis or cannabinoids exists. Claims that cannabis improves concentration or reduces hyperactivity in children rest on case reports, parent surveys, and extrapolation from adult data that is itself thin.

For pediatric chronic pain, the same gap applies. Cannabinoids have been studied in adults with chronic pain extensively and with mixed results, and that literature does not transfer to children, whose pain conditions, developmental stage, and risk profile differ substantially.

For symptom control in pediatric cancer, including nausea and appetite, the evidence base is small and largely observational. This is an area where a pediatric oncology or palliative care team may reasonably consider a cannabinoid in a specific case, and that decision belongs inside that team rather than being made by a family independently.

The absence of evidence in these areas is not proof of absence of benefit. It does mean that any product used for these purposes in a child is being used without the kind of testing the epilepsy indications received, and a parent should know that is the situation rather than assume the research exists somewhere.

In the tuberous sclerosis trial, diarrhea occurred in 31 percent of the 25 mg per kilogram group and 56 percent of the 50 mg per kilogram group, against 25 percent on placebo. Somnolence occurred in 13 and 26 percent respectively, against 9 percent. Eight patients in the lower-dose group and ten in the higher-dose group discontinued because of adverse events, compared with two on placebo. Twenty-eight patients taking cannabidiol, or 18.9 percent, had elevated liver transaminases, and none on placebo did.

The Dravet trial reported the same pattern: diarrhea, vomiting, fatigue, fever, somnolence, and abnormal liver function tests occurred more often on cannabidiol than placebo, and there were more withdrawals from the cannabidiol group.

Those numbers are from a monitored trial with a standardized product, scheduled laboratory testing, and specialist oversight. A retail product used at home has none of those safeguards. Cannabidiol also interacts with antiepileptic medications, which is one reason liver monitoring is part of the protocol rather than an optional extra.

Separately, unintentional exposure is a documented pediatric hazard that has grown quickly. A retrospective analysis of National Poison Data System records published in Pediatrics found 7,043 reported exposures to edible cannabis products in children under 6 between 2017 and 2021, rising from 207 cases in 2017 to 3,054 in 2021. Of cases followed to a known outcome, 70 percent involved central nervous system depression, and 22.7 percent of all reported cases were admitted to hospital. Nearly all exposures, 97.7 percent, occurred in a residential setting.

If the child has drug-resistant epilepsy, particularly Dravet syndrome, Lennox-Gastaut syndrome, or tuberous sclerosis complex, the conversation belongs with a pediatric epileptologist, and the product in question is a prescribed pharmaceutical cannabidiol formulation with weight-based dosing and scheduled liver monitoring. That is a specialist referral, and it is a reasonable one to ask for.

If the child has autism, attention problems, chronic pain, or another condition, the honest framing is that a family would be trying something that has not been shown to work in a properly controlled trial, using a product that was not the one tested, without the monitoring the trials included. That may still be a conversation worth having with the child’s specialist, and it should happen with that specialist rather than around them.

Bring the specific product to any such conversation. Content varies between preparations, cannabidiol interacts with several medications a child with a neurological condition is likely to be taking, and the trial products were standardized in ways retail products generally are not.

And regardless of any of the above, store cannabis products where a young child cannot reach them. The poison center data describe thousands of preventable hospitalizations in children under six, almost all of them occurring at home.

The epilepsy evidence is strong and unusual for this field: multiple independent randomized, placebo-controlled trials, hundreds of participants, prespecified primary endpoints, and results that supported regulatory approval. Very little else in cannabinoid medicine reaches that standard.

The autism evidence is moderate in quality and weak in result. A 150-participant randomized crossover trial is a serious study, and its registered primary outcome was null. That combination is more informative than a positive uncontrolled study, and it points away from confident claims.

Positive secondary outcomes in a trial whose primary outcome was null should be treated as hypothesis-generating. The autism trial reported improvement on clinician global impression and on a social responsiveness measure while the registered primary measure did not separate, which is a pattern that frequently fails to replicate.

The epilepsy trials also tested a specific pharmaceutical preparation at defined weight-based doses in narrowly defined syndromes. Generalizing from a 20 mg per kilogram per day standardized oral solution in Dravet syndrome to an over-the-counter product in a child with a different diagnosis is a much larger leap than it appears.

These trials do not show that cannabis helps children with attention deficit hyperactivity disorder, chronic pain, or most other pediatric conditions, because those trials have not been conducted at adequate size and rigor. Absence of evidence is not evidence of absence, and it is also not a basis for treatment.

They also do not show that cannabidiol is a low-risk intervention in children. Roughly one in five participants in the tuberous sclerosis trial developed elevated liver transaminases, and discontinuation for adverse events was several times higher than on placebo.

Pediatric cannabinoid medicine is a useful case study in how a specific, well-run trial program can establish a genuine therapy while simultaneously creating a halo around a whole product category. The trials tested one molecule, in one formulation, in three syndromes. The marketing that followed tested nothing.

The distance between those two facts is where most pediatric cannabis decisions go wrong, and closing it is largely a matter of asking which product, which dose, and which trial.

When a parent comes to me about a child, my first question is which condition and my second is which product. Those two answers determine almost everything, and they are usually the two things missing from whatever the family read before the appointment.

For a child with Dravet syndrome or Lennox-Gastaut or tuberous sclerosis, I am pointing toward a pediatric epileptologist and a prescription formulation with liver monitoring, and I say so without hedging. That pathway exists, it is well studied, and a family should not have to improvise it.

For autism, and this is the conversation I have most often, I try to be honest about the Aran trial rather than quote the part parents want to hear. The primary outcome was negative. Some secondary outcomes were not. That is a reason to want a better trial, and it is not a reason to start a child on a product at home.

The other thing I bring up whether or not anyone asks: lock the products away. The poison center numbers describe thousands of small children hospitalized after finding an edible at home, and that harm is entirely preventable.

Pharmaceutical cannabidiol has randomized, placebo-controlled evidence in Dravet syndrome, Lennox-Gastaut syndrome, and tuberous sclerosis complex, as an add-on therapy under specialist supervision with liver monitoring. Outside those syndromes the evidence thins quickly: the largest autism trial had a null primary outcome, sleep outcomes did not separate from placebo, and for pediatric attention problems, chronic pain, and most other indications no adequate trial exists. The pathway for any child is specialist evaluation, never a product purchased on a parent’s own research.

The word doing the most work in the strong evidence is pharmaceutical. What was tested was a standardized cannabidiol oral solution, dosed by body weight, added to existing antiepileptic medication, with scheduled laboratory monitoring, in three named syndromes. Strip away any of those conditions and you are no longer inside the evidence, whatever the product label says.