The Phase 2 and 3 trials behind psilocybin’s clinical dosing have public records that say nothing about autism. A preprint from King’s College London, posted to medRxiv on August 22 and not yet peer reviewed, reports for the first time, according to its authors, that the autistic brain responds differently to low-dose psilocybin; searches of PubMed and ClinicalTrials.gov found no earlier human study comparing the two groups. At 5 mg, a dose well below the 25 mg used in pivotal depression trials, two measures of communication between large brain networks moved on average in opposite directions in autistic and non-autistic adults.
The study did not test whether psilocybin helps autistic people, and its authors say so. What it probes is a design assumption: that the pharmacodynamic picture built in trials whose records do not identify autism status transfers to autistic participants. The only psilocybin treatment trial in autistic adults found on ClinicalTrials.gov uses the same 10 mg then 25 mg sequence as an earlier Imperial College trial in treatment-resistant depression.
What the study measured
The PSILAUT study enrolled 67 adults, 37 with a clinical autism diagnosis and 30 without, matched on age, sex, and IQ (ages 18 to 58, roughly half female in each group). Each participant was to attend three visits at least six days apart and receive placebo, 2 mg, or 5 mg of oral psilocybin, double-blind. The order was only partly randomized: the placebo visit could fall first, second, or third, but 2 mg always preceded 5 mg, a rule the ethics committee approved so that anyone reacting badly to the lower dose would not receive the higher one. Resting-state fMRI was acquired about 75 minutes after dosing.
Not everyone reached 5 mg. Four participants, two in each group, had elevated blood pressure after 2 mg and were withdrawn from the 5 mg visit, and one non-autistic participant left after starting an SSRI. The 5 mg analysis rests on 52 visits, 23 non-autistic and 29 autistic.
The sample was also deliberately narrow. The study excluded people with co-occurring psychiatric illness such as major mood or psychotic disorders, anyone taking serotonergic medication, people with syndromic autism of known genetic cause, and anyone with an IQ of 70 or below.
Compass Pathfinder part-funded the study and supplied its COMP360 psilocybin at no cost. Two authors are co-founders of and shareholders in N1 Bio and hold shares in Compass Pathfinder, and the corresponding author has received Compass funding for investigator-initiated studies and has consulted for N1 Bio.
What survived correction, and what did not
At baseline, on placebo, the two groups’ network connectivity did not differ after correction for multiple comparisons. That matters: the divergence under psilocybin is a difference in response, not a pre-existing difference being re-measured.
After 5 mg, the group-by-dose interaction survived false-discovery-rate correction for two between-network pairs: the default mode and frontoparietal networks (p = 0.042 corrected) and the dorsal and ventral attention networks (p = 0.015 corrected). In both, connectivity moved down on average in non-autistic participants and up in autistic participants.
The within-group pieces of that picture are weaker than the preprint’s summary suggests. Tested separately and uncorrected, the default mode-frontoparietal decrease in non-autistic participants reached significance (p = 0.030), but the increase in autistic participants did not (p = 0.066). For the attention networks it was the reverse: the autistic increase reached significance (p = 0.037) and the non-autistic decrease did not (p = 0.20). The defensible statement is that the two groups responded in opposite directions. That autistic participants showed increased integration on each measure is a looser statement than the statistics support.
Other findings were clearer. In non-autistic participants only, 5 mg reduced connectivity within the frontoparietal network (Cohen’s d = -0.71) and the limbic network (d = -0.61), both surviving correction, a pattern consistent with earlier psilocybin imaging in healthy volunteers. An aggregate index of between-network change was larger in autistic participants at both 2 mg and 5 mg. Across the whole sample, the size of the between-network increase correlated with Autism Quotient scores. Because that correlation spans two groups that differ sharply on the questionnaire, it cannot show on its own whether the relationship holds within either group.
Subjective effects did not separate the groups. On the 5D-ASC altered-states questionnaire, 5 mg raised scores in both groups with no significant group difference, at roughly half the intensity reported at high doses. The divergence the study found is in the scanner, not in how participants described the experience.
The integration premise does more work than the data
The preprint’s clinical framing rests on a link between psilocybin’s therapeutic effect and increased network integration. It concludes that because autistic participants showed integration-like changes at 5 mg, future applications of psilocybin in autistic people should consider the utility of lower doses.
The integration evidence it relies on comes from a different setting. The central paper, Daws and colleagues in Nature Medicine in 2022, measured network modularity in patients with depression one day after a 25 mg dose in one trial and three weeks after the second 25 mg dose in the other, and found that reduced modularity tracked antidepressant response. That is a subacute, post-treatment measure in depressed patients at a therapeutic dose. PSILAUT measured acute connectivity about 75 minutes after 2 or 5 mg in adults without co-occurring psychiatric illness. An earlier Imperial study of psilocybin for treatment-resistant depression noted that post-treatment brain changes differ from psilocybin’s acute effects.
The preprint’s result complicates the idea that one direction of connectivity change reads the same way in every brain. It does not establish that acute low-dose integration in autistic adults predicts therapeutic response, and the authors state their study did not evaluate clinical efficacy.
What the preprint cannot establish
The study was not designed to assess safety, and the UK medicines regulator confirmed it was not a clinical trial of an investigational medicine. The four blood-pressure withdrawals after 2 mg are an observation in 67 people, not a safety signal or its absence. There were no pharmacokinetic measurements, so the authors cannot rule out that autistic participants metabolized psilocybin differently or peaked at a different time than the 75-minute scan. Autistic participants moved more in the scanner, although the difference was not significant at any single dose and head motion was included as a covariate. The analysis was described as exploratory, and the registry lists EEG and spectroscopy outcomes from the same study that this preprint does not report. Those outcomes, from the same visits, are the nearest test of whether the divergence appears in a second modality.
Thirty-seven psychiatrically well autistic adults with IQs above 70 are not the autistic population that clinical trials would enroll, and the result has not been peer reviewed or replicated. The parts of the analysis on the weakest footing, the within-group results and the whole-sample Autism Quotient correlation, are the ones review would need to settle.
Where autism sits in psilocybin trials
The human evidence on psychedelics in autistic adults is thin. Searches of PubMed and ClinicalTrials.gov surfaced one earlier interventional psychedelic study, a 12-person placebo-controlled pilot of MDMA-assisted therapy for social anxiety published in 2018. Definium’s open-label Phase 2a of DT402, an R-enantiomer of MDMA, is recruiting autistic adults. MDMA acts mainly by releasing monoamines, not by direct serotonin 2A agonism, so neither study bears on psilocybin dosing.
The only psilocybin treatment trial in autistic adults found on ClinicalTrials.gov carries the depression-trial dosing template into that population. The Centre for Addiction and Mental Health’s Phase 1 study (NCT06731621) is an open-label trial in autistic adults with treatment-resistant depression, with an estimated enrollment of 20: a 10 mg psilocybin session with therapy for safety, then a 25 mg treatment session. Its primary outcomes are feasibility, subjective effects, treatment-related adverse events, and the feasibility of repeated MRI, with brain-network change as a secondary outcome. It is no longer recruiting, and primary completion is estimated for December 2026. The PSILAUT result does not show that CAMH’s doses are wrong. It does show that CAMH’s 10 mg safety session is twice the highest dose PSILAUT tested, so the preprint says nothing directly about how autistic adults respond at either of CAMH’s doses.
The two studies also cover different populations. CAMH requires treatment-resistant depression, the kind of co-occurring condition PSILAUT excluded. Findings from psychiatrically well autistic adults may not transfer to depressed autistic adults whose medications have been tapered. If CAMH’s readout reports its MRI and subjective-effect data by dose, it would be a first look at how autistic adults with depression respond at 10 mg and 25 mg.
A search of ClinicalTrials.gov on September 29 for Phase 2 and Phase 3 studies listing psilocybin as an intervention returned 168 records, including Compass’s COMP005 and COMP006 depression trials, its Redefine PTSD trial, and Usona’s uAspire. None mentions autism in its listed eligibility criteria, as an inclusion, an exclusion, or a screening requirement. Registry entries summarize protocols and do not always list every criterion, so this establishes what the public records say, not what every protocol screens for. It is consistent with the preprint’s own observation that mental-health trials do not typically screen for autism.
On the public record, then, autistic participants are neither systematically excluded nor identified, and the registry entries behind psilocybin’s dosing carry no information on which participants were autistic. Whether psilocybin sponsors begin recording autism status in depression and PTSD trials, as a stratification variable or even a demographic field, will decide whether the question PSILAUT raises, how autistic participants respond at therapeutic doses, can be answered from programs already running or needs studies of its own.