| Audience | Parents of children with autism spectrum disorder considering CBD products, pediatricians, child psychiatrists, developmental specialists, and cannabis-medicine clinicians |
| Primary Topic | CBD-rich cannabis oil and social cognition in children with autism spectrum disorder |
| Source | Read the full source |
CBD Oil and Autism: What a New Randomized Trial Shows About Social Cognition
A new post hoc subanalysis of a randomized, placebo-controlled trial found within-group improvements in facial expression recognition and theory of mind among children who received CBD-rich cannabis oil, but none of these changes differed significantly from placebo, and total theory of mind improved in both groups alike.
| Study Type | Post hoc subanalysis of a double-blind, randomized, placebo-controlled trial |
| Participants | 19 Brazilian children with autism spectrum disorder (72.7% male; mean age 7.82 years) |
| Intervention | CBD-rich cannabis oil vs. placebo for 12 weeks |
| Outcomes Measured | Facial expression recognition and theory of mind (verbal and total), assessed with neuropsychological tasks at baseline and after treatment |
| Facial Expression Recognition | Significant within-group improvement in the CBD group (p = 0.030); between-group difference after treatment not significant (p = 0.527) |
| Verbal Theory of Mind | Significant within-group improvement in the CBD group (p = 0.016); between-group difference not significant (p = 0.901) |
| Total Theory of Mind | Significant improvement within both groups; no significant between-group difference (p = 0.967) |
| Trial Registration | Brazilian Clinical Trials Registry (ReBEC), RBR-5wr2cqq |
| Journal | Autism Research |
| Published | August 26, 2026 |
| PMID / DOI | 42644517 / 10.1002/aur.70353 |
Children who received CBD-rich cannabis oil for 12 weeks showed statistically significant within-group improvement on two measures: facial expression recognition (p = 0.030) and verbal theory of mind (p = 0.016). Total theory of mind scores also improved significantly within the CBD group.
These are genuine before-after changes measured with standardized neuropsychological tasks, and they represent the kind of numbers that often appear first in coverage of a study like this.
A randomized, placebo-controlled design exists specifically to answer one question: did the treated group do better than the group that received placebo? Here it did not. Between-group comparisons after treatment showed no significant difference for facial expression recognition (p = 0.527) or verbal theory of mind (p = 0.901).
Total theory of mind told the same story from a different angle: it improved significantly within both the CBD group and the placebo group, with no significant difference between them (p = 0.967). When placebo improves as much as active treatment, within-group change cannot be attributed to the treatment itself.
This analysis included only 19 children, drawn from participants in a larger trial who completed social cognition testing at baseline and after treatment. It was a post hoc subanalysis, meaning facial expression recognition and theory of mind were not necessarily the parent trial’s primary registered endpoints.
Small post hoc analyses are prone to both false positive and false negative findings, and the authors state this directly: given the small sample size and exploratory design, the results should be interpreted cautiously and do not establish a treatment effect.
The parent study was registered with the Brazilian Clinical Trials Registry (ReBEC) under RBR-5wr2cqq and was conducted as a double-blind, randomized, placebo-controlled trial in Brazilian children with autism spectrum disorder (72.7% male, mean age 7.82 years).
Registry documentation and a double-blind, placebo-controlled parent design are meaningful methodological strengths even when a particular subanalysis lands on a null result.
Cannabinoid research in pediatric autism has expanded quickly, with growing parent and clinician interest that often outpaces the specific outcome measures being tested. Much existing evidence looks at broader behavioral or global severity scales, so a study measuring facial expression recognition and theory of mind directly addresses a narrower, more specific question than most published autism-cannabinoid research.
A null between-group result on a well-defined social cognition outcome is a useful data point precisely because it is easy for it to get lost next to studies reporting global behavioral improvement, and because parents considering CBD products for social difficulties in autism deserve to know what has, and has not, been demonstrated.
I find this study useful specifically because the authors did not spin the null between-group result. A CBD group that improves on a rating scale over 12 weeks will always sound encouraging in isolation, and it takes editorial discipline to report clearly that the placebo group improved just as much on total theory of mind, and that neither facial expression recognition nor verbal theory of mind separated the two groups statistically.
For families asking me about CBD for social difficulties in autism, this is exactly the kind of study I want to be able to point to honestly: a well-intentioned, methodologically serious, small trial that does not yet give us the answer. I would not use this paper to justify starting or stopping CBD oil for social cognition concerns in a child with autism, and I would not let anyone tell a parent otherwise based on this data alone.
How to Read a Small Post Hoc Subgroup Analysis
A study can be well-designed at the trial level and still produce a subanalysis that cannot support a confident clinical claim. The distinction usually comes down to what was actually tested, how many participants were involved, and which comparison the headline number is drawn from.
This paper offers a clean example of all three issues at once, which makes it a useful teaching case even though its own conclusion is appropriately cautious.
Four distinctions that protect interpretation
Primary endpoint versus post hoc analysis
Facial expression recognition and theory of mind were assessed in a subset of the parent trial’s participants after the fact, not necessarily as the trial’s original primary outcome, which changes how much weight the result should carry.
Within-group change versus between-group difference
A significant p-value comparing a group’s own before-and-after scores does not establish efficacy; only the comparison against the placebo group’s change can do that, and here it did not separate the groups.
Small samples produce unstable statistics
With 19 children total, both true effects and chance findings can look similar on paper, and a single additional participant or missing data point can shift a p-value across the 0.05 threshold.
Statistical significance versus clinical meaning
Even where a within-group p-value was significant, the study was not powered or designed to tell us whether the size of that change would matter in a child’s daily functioning.
From a Promising Within-Group Signal to an Honest Null Result
Eight perspectives on why a small post hoc subanalysis of CBD-rich cannabis oil in autism produced encouraging within-group numbers but no evidence of benefit over placebo.
Improvement Doesn't Automatically Mean the Treatment Caused It
It’s understandable to focus on the fact that children in the CBD group got measurably better at recognizing facial expressions and reading verbal theory-of-mind tasks over 12 weeks. Those are real, statistically significant changes within that group, and it is easy to read them as proof that the oil worked.
But the placebo group in this trial also improved, enough that the two groups could not be told apart statistically on any of the three outcomes measured. That means the study cannot currently tell a family whether a child’s improvement, if it happens, comes from the CBD itself or from something else, such as time, repeated testing, or the general experience of being in a supportive trial.
A Result to Discuss Honestly, Not Dismiss or Oversell
When a family raises this study, the most useful response is neither dismissal nor endorsement. The trial was double-blind, randomized, and placebo-controlled at the parent-study level, and the investigators reported the negative between-group findings as clearly as the positive within-group ones, which reflects careful, transparent reporting.
At the same time, nothing here supports adjusting a treatment plan around CBD oil for facial expression recognition or theory of mind specifically. The honest clinical position is that this is a well-conducted pilot-scale signal that has not yet cleared the bar of showing benefit over placebo, and that more participants and a prespecified design would be needed before that could change.
Nineteen Children Is Not Enough to Answer This Question
A sample of 19 children split across two arms is small enough that a study can easily miss a real effect or produce a within-group p-value that would not survive replication in a larger cohort. Post hoc subanalyses compound this risk further, since they were not the trial’s original, prespecified focus.
The fact that between-group comparisons for all three outcomes came back non-significant, including a total theory of mind result where both groups improved almost identically, is the more statistically reliable signal here. Skepticism in this case does not mean assuming the authors did anything wrong; it means recognizing that a small, exploratory subanalysis simply cannot carry the weight some readers may want to put on it.
The Study Itself Draws the Right Boundary
One of this paper’s strengths is that its own authors do not overstate their findings. They explicitly describe the design as exploratory and post hoc, and they state directly that the results should be interpreted cautiously and do not establish a treatment effect of CBD on social cognition in children with autism.
That built-in caution is a model for how findings like this should be discussed publicly. A critical read does not need to find a hidden flaw in the paper; it needs to make sure the finding is not repackaged elsewhere as stronger evidence than the authors themselves claim it to be.
A Narrower Question Than Most Cannabinoid-Autism Studies Ask
Much of the existing research on cannabinoids in autism has focused on broader outcomes such as global behavioral severity, irritability, or overall functioning, rather than a specific cognitive domain like facial expression recognition or theory of mind. This subanalysis narrows the lens considerably, testing a more precise and arguably harder question.
That specificity is valuable methodologically, but it also means this single, small subanalysis should not be read as either confirming or contradicting broader findings from differently designed studies measuring different outcomes. It occupies its own narrow evidentiary lane.
What This Means for Families Weighing CBD Oil Today
For a family already considering CBD-rich cannabis oil for a child with autism, this study neither strengthens nor eliminates that option specifically for concerns about facial expression recognition or theory of mind. It simply does not provide evidence that the oil produces a measurable social cognition benefit beyond what placebo produced in this trial.
Any decision about CBD oil should still involve a clinician who can discuss formulation, dosing uncertainty, drug interactions, and realistic expectations, and it should be framed around outcomes with supporting evidence rather than this particular measure, where the current evidence is null.
What a Stronger Follow-Up Trial Would Need
A future study designed to answer this question directly would need a larger sample specifically powered for facial expression recognition and theory of mind as prespecified primary outcomes, rather than testing them as a post hoc subanalysis of a trial built around other endpoints.
Longer follow-up, standardized dosing, and replication in additional cohorts beyond the original Brazilian sample would also help determine whether the encouraging within-group signal in the CBD arm reflects a real, reproducible effect or simply the kind of change that also occurred in the placebo group here.
Transparent Null Results Deserve to Be Reported, Not Buried
As interest in cannabinoid products for pediatric neurodevelopmental conditions grows, the research and clinical communities benefit when studies like this one are published and discussed even though their headline finding is a null between-group result. Suppressing or downplaying negative comparisons would distort the overall evidence base that families and regulators rely on.
Educational materials and product marketing aimed at parents of children with autism should reflect this level of nuance rather than citing within-group improvement numbers in isolation. Accurate public communication about a study like this means naming both the encouraging signal and the clear limitation together.
Join the Conversation
Have a question about how this applies to your situation? Ask Dr. Caplan
Want to discuss this topic with other patients and caregivers? Join the forum discussion
Frequently Asked Questions
What did this study test?
It tested whether 12 weeks of CBD-rich cannabis oil, compared with placebo, changed facial expression recognition and theory of mind in 19 Brazilian children with autism spectrum disorder, using standardized neuropsychological tasks.
How many children were included?
Nineteen children were included, 72.7% male, with a mean age of 7.82 years, drawn from participants in a larger double-blind, randomized, placebo-controlled parent trial.
Did CBD improve facial expression recognition?
Children in the CBD group showed a significant within-group improvement (p = 0.030), but the improvement was not significantly different from the placebo group after treatment (p = 0.527).
Did CBD improve theory of mind?
Verbal theory of mind improved significantly within the CBD group (p = 0.016), but this was not significantly different from placebo (p = 0.901). Total theory of mind improved significantly in both groups, with no significant difference between them (p = 0.967).
Does this study prove CBD oil helps children with autism read facial expressions or understand others' thoughts?
No. The authors state directly that given the small sample size and exploratory post hoc design, the findings should be interpreted cautiously and do not establish a treatment effect of CBD on social cognition in children with ASD.
What is a post hoc subanalysis, and why does it matter here?
A post hoc subanalysis examines an outcome that was not necessarily the trial's original, prespecified primary focus, using a subset of participants who completed the relevant testing. It carries a higher risk of false positive and false negative results than a prespecified primary analysis.
Why did the placebo group also improve?
The study did not identify a specific cause. Improvement in a placebo group can reflect factors such as repeated exposure to the testing tasks, ongoing development over the 12-week period, or nonspecific effects of trial participation.
How long did the trial last?
Participants received CBD-rich cannabis oil or placebo for 12 weeks, with facial expression recognition and theory of mind assessed at baseline and after treatment.
Was this trial registered?
Yes. The parent trial was registered with the Brazilian Clinical Trials Registry (ReBEC) under RBR-5wr2cqq.
Should parents change their child's care based on this study?
Not based on this analysis alone. It is a small, exploratory subanalysis with a null between-group result, and any decisions about CBD oil for a child with autism should be made with a clinician, using outcomes with stronger supporting evidence.
