October brought a new randomized trial of psilocybin in obsessive-compulsive disorder and the registry results of an older one. On October 5, Johns Hopkins researchers posted a randomized trial of 37 people as a preprint. On October 16, Yale posted to ClinicalTrials.gov the results of its 28-person randomized trial, which it had first reported in a January preprint.
Together with an Arizona trial published in May, at least 82 people have now been randomized to psilocybin or a comparison in three OCD trials with public results. That is a meaningful increase for a disorder where, two years ago, the published trial evidence consisted of one nine-person study from 2006.
What the new evidence does not resolve is the problem that has followed psychedelic trials in every indication. Patients usually know when they have taken a psychedelic. In OCD, where the main outcome is a clinician’s rating of what patients report, that matters. The trials that separate most clearly from their comparisons cannot rule out expectation, and the one with the longest blinded active-placebo comparison did not separate.
Yale: the result was already out
Yale’s randomized result is not new this month. In January the investigators posted a preprint of a trial run from 2018 to 2023, in which 28 adults were randomized to a single dose of psilocybin at 0.25 milligrams per kilogram or to niacin, a vitamin that causes flushing, used as an active placebo. Forty-eight hours after dosing, scores on the acute version of the Yale-Brown Obsessive Compulsive Scale had fallen 9.83 points more with psilocybin than with niacin (95 percent confidence interval 5.04 to 14.62). One week after dosing, 9 of 13 psilocybin participants met the preprint’s response threshold of a 35 percent fall in symptoms, against none of 14 on niacin. The preprint has not been peer reviewed, and no peer-reviewed efficacy paper from the trial has appeared.
The preprint names two primary outcomes: the acute scale through 48 hours and the standard weekly scale through 12 weeks. Only the first week was controlled. Participants learned their assignment after the 48-hour assessment, the independent raters stayed blinded through the week-one assessment, and the niacin group was then offered open-label psilocybin; all 14 took it. The 12-week result, a mean of 15.2 on the standard scale among the 12 originally assigned to psilocybin who completed follow-up, has no comparison group. Seven of the nine one-week responders still met the response threshold at 12 weeks.
The authors write that functional unblinding “was evident in our study,” but the preprint reports no data on what participants guessed and no measure of expectancy. Six of the 28 participants had taken psilocybin more than a year before enrolling, four of them in the psilocybin arm; use in the previous 12 months was an exclusion. That does not show expectation drove the result; it does mean some participants had a reference point for what the drug feels like. When the niacin group later took psilocybin knowing what it was, 5 of 14 met the response threshold at one week, against 9 of 13 in the blinded phase. The authors read the gap as a sign that factors beyond the drug and the support model shape outcomes. It is a comparison across phases, not a randomized one.
The trial’s one serious adverse event occurred in the psilocybin arm. According to the preprint, a participant with chronic passive suicidal ideation at baseline developed suicidal ideation about 24 hours after dosing, which became active suicidal ideation with intent before the 48-hour assessment. The investigators rated the event severe and possibly related to study participation, withdrew the participant and extended the hospital stay. The suicidality resolved, and the participant was discharged 72 hours after dosing with close outpatient monitoring. The protocol kept every participant on an inpatient research unit for two days before dosing and one day after. One event among 14 participants cannot establish how often this happens, and a single case does not show that the drug caused it.
The ClinicalTrials.gov results add a few things the preprint does not contain. The registry lists four primary outcomes, not two: the acute and standard scales, a self-rated visual analog scale for OCD symptoms and the Columbia suicide rating scale. The preprint reports results for neither of the last two. On the self-rated scale, psilocybin participants rated their obsessions, compulsions and anxiety lower than niacin participants at 48 hours, with no between-group test reported. On the suicide scale, the registry records one new case of suicidal ideation with psilocybin and none with niacin, and no suicide attempts or deaths in either arm. The registry classifies the trial as Phase 1; the preprint calls it phase 2.
The two documents also report different statistics for the same 48-hour comparison. The registry gives p = 0.012 and labels the analysis both a linear mixed model and a two-sided t-test. The preprint, using a mixed model on the same scores, reports p below 0.001. The public record does not explain the difference. Both describe a statistically significant separation.
Johns Hopkins: randomized, but against a waiting list
The Johns Hopkins preprint, also not peer reviewed, randomized 37 adults to start psilocybin at once or after an eight-week wait. Both groups received a 20 milligram dose and, two weeks later, 30 milligrams if the first was well tolerated, with preparation and support from two facilitators. Thirty participants completed the study and were analyzed, 16 who started at once and 14 on the waiting list.
One week after the second dose, the immediate group’s mean on the standard OCD scale was 13.4, down from 25.9. The waiting-list group’s was 25.7 at the matching time, from a baseline of 25.4. The difference was large (Cohen’s d of 1.92 in the post-hoc comparison), and the authors report that it held up when missing data were imputed under unfavorable assumptions. They report no serious adverse events.
The design rules out the comparison that matters most. Everyone knew they would receive psilocybin, and only the timing was randomized. The authors write that the design “may therefore have amplified expectancy-related effects” and that the waiting group’s lack of improvement “should not be interpreted as evidence of a limited placebo response.” The trial was funded by philanthropic donors. The ClinicalTrials.gov entry lists an actual enrollment of 35, two fewer than the preprint randomized.
Arizona: the active-placebo trial did not separate
Francisco Moreno’s group at the University of Arizona published its trial in the Journal of Psychopharmacology in May. Eighteen people completed baseline assessment. Fifteen finished a randomized, double-blind first phase of four weekly sessions, five each on 300 or 100 micrograms per kilogram of psilocybin or on 1 milligram of lorazepam, a sedative used as active placebo. Two more left after a single randomized session, so at least 17 were randomized. The trial’s primary outcomes were safety and tolerability, and the authors write that it was not powered to detect differences between groups.
On scores taken a week after each session, the three conditions did not differ significantly (p = 0.092), with no significant condition-by-time interaction. Scores taken the day after each session did differ by condition overall, though only the high-dose group fell significantly within itself. The two psilocybin groups combined showed a significant decline over four weeks (p = 0.044). The paper reports no formal assessment of blinding.
The abstract’s headline figures, 73.3 percent response and 40 percent remission at eight weeks, come after a second phase in which everyone received high doses with no comparison group. One participant included in them left after four low-dose sessions and never received a high dose. The trial reported no serious adverse events.
Imperial: the study that measured the blind
The fourth controlled study is not randomized but is the most informative about masking. Imperial College London researchers gave 19 adults 1 milligram and then 10 milligrams of psilocybin, four weeks apart, in fixed order; 18 completed. Participants were told they would receive “up to 10mg” each time and that one dose would be lower, without being told the order. Only the outcome assessor was blinded. Most participants stayed on their selective serotonin reuptake inhibitors, unlike in the American trials.
One week after dosing, the 10 milligram dose produced a larger fall in OCD scores than 1 milligram (Cohen’s d of 0.82, p = 0.002). At two weeks the difference was a trend (p = 0.06), and by four weeks it was gone.
The study asked participants the day after each session to guess their dose. After 10 milligrams, 12 of the 13 who answered, or 92 percent, chose the correct category, and their average confidence was also 92 percent. After 1 milligram, 5 of 17 guessed correctly. With a fixed order and a known lower dose, the second session was partly deducible. Either way, participant blinding largely failed. The study was published in Comprehensive Psychiatry in 2025.
How the comparisons line up
Set side by side, the studies make the pattern plain.
Yale used an active placebo that produces a body sensation but not a psychedelic one, kept participants blind for 48 hours and reported no data on what they guessed. Its separation was large.
Johns Hopkins had no blind to preserve. Its separation was also large.
Arizona used a sedative as active placebo, kept the blind for four weeks and did not assess it. Its randomized groups did not separate on the planned weekly comparison.
Imperial compared two doses of the same drug, measured the blind and found participants could mostly tell. Its separation was moderate and lasted a week.
The lineage begins with Moreno’s 2006 study of nine patients, which the authors called modified double-blind: doses rose in fixed order, with a very low dose inserted at a random point. Symptom scores fell within 24 hours of active doses, with a significant effect of time but no significant effect of dose.
The treatment-resistance criteria also vary. Yale’s registered minimum was one failed standard treatment, while the preprint describes eligibility as two or more failed adequate trials, and every participant in its baseline table had failed at least two medication trials. The public record reviewed here does not explain the difference. Johns Hopkins and Arizona required one prior treatment attempt. These are people with moderate to severe OCD who have tried something, for whom first-line care is a serotonin reuptake inhibitor or clomipramine alongside exposure and response prevention, and the drug options after that are limited.
The case for the drug
The strongest reading in psilocybin’s favor is not trivial. Four research groups on two continents have now seen OCD scores fall quickly after psilocybin, within a day to a week. Yale’s 48-hour difference is large, and it held within a blinded window against an active placebo. Johns Hopkins adds a randomized trial in which the waiting group did not improve at all, and the authors note that OCD has a relatively low placebo response. Analyses from Yale and Johns Hopkins both found that improvement tracked the intensity of the experience, which fits a drug effect. The Yale authors argue that the size of the benefit exceeded typical placebo responses in OCD. On this reading, expectation alone would have to be unusually powerful to produce changes this large.
The case for caution
Each of those observations is also compatible with a large contribution from expectation. A low placebo response in conventional OCD trials comes from pill-versus-pill designs where patients cannot tell what they took; it says little about trials where they can. The one trial with an active placebo held for weeks did not separate its groups. The one study that measured participant blinding found it largely failed. The biggest headline numbers come from open-label phases. And two of the three randomized trials are reported only in preprints and a registry entry, not in peer-reviewed efficacy papers. In total, the randomized comparisons rest on fewer than 75 analyzed participants.
What would settle it
Yale’s own randomized trial of repeated doses lists 28 participants and a February 2026 primary completion date, with no results posted. Like Johns Hopkins, it compares psilocybin with a waiting list. What would answer more than another trial of the same kind is a comparator that participants cannot easily distinguish from psilocybin, a blinded comparison measured in weeks rather than days, and published data on what participants guessed. The problem has been examined in our coverage of trial design; in OCD it has not yet been solved.
The evidence has moved from one small study to several randomized ones in under two years. It has not yet moved past the question of whether patients know what they took.