Published on 20 September 2026
Psilocybin for PTSD: what a 22-patient open trial actually shows, and what it does not
Journal of Psychopharmacology · 2026; 40(1): 139-148 · McGowan et al.
DOI 10.1177/02698811251362390
PMID 40883964
Scientific 61
Editorial 81
The essentials
Twenty-two adults with post-traumatic stress disorder received a single 25 mg dose of synthetic psilocybin across three research centres, with structured psychological support. The trial’s primary endpoint was safety, not efficacy: there was no control group, no randomisation and no blinding. On that primary endpoint the result is clear: 117 treatment-emergent adverse events, 70 occurring on the day of administration, of which 64 had resolved by the following day, no serious adverse event, and no participant withdrew from the trial. Two episodes of suicidal ideation were reported, both in participants with a prior history of suicidality: one began and resolved on the day of dosing and was judged related to the drug, the other occurred at week 7, resolved within a month, and was judged possibly related. The reductions on the CAPS-5 and PCL-5 are secondary endpoints measured with no comparator: they permit no conclusion about efficacy, a point the authors themselves make. The trial was sponsored and funded by Compass Pathfinder Ltd, and five of the nine authors are current or former employees of Compass Pathways Plc who hold equity in the company.
Context
Psychedelics have occupied a place in the general press for three years now that is out of proportion with the evidence available. Post-traumatic stress disorder is a natural site for this mismatch: treatments with established efficacy exist, but a share of patients respond poorly to them, and expectations are high.
A point of law first, because it frames how a trial like this should be read anywhere. In France, psilocybin and psilocin are listed as narcotics under Annex III of the Order of 22 February 1990, and hallucinogenic mushrooms, including those of the genus Psilocybe, under Annex IV of the same text. Article R. 5132-74 of the Public Health Code prohibits, absent express authorisation, the production, cultivation, manufacture, transport, import, export and possession of substances classified as narcotics. Article L. 3421-1 punishes illicit use with one year of imprisonment and a 3,750 euro fine. No marketing authorisation exists for psilocybin, in France or elsewhere in Europe; the only lawful route of exposure is participation in an authorised research protocol. Psilocybin is a controlled substance in most jurisdictions, though its exact legal status varies from one country to another, and readers should check the regulations that apply where they practise rather than assume that a clinical trial implies legal access.
This article is therefore not an announcement of a treatment. It serves another purpose: learning how to read an open-label phase 2 trial, the type of study most routinely turned into a therapeutic promise somewhere between publication and the news feed.
The study at a glance
| Question (PICO) | |
|---|---|
| Population | |
| 22 adults with DSM-5 post-traumatic stress disorder following a traumatic event that occurred in adulthood. 63.6% women, mean age 39.0 years (SD 7.91). CAPS-5 score of at least 25 at baseline, corresponding to at least moderate symptoms. 39 people assessed at screening, 22 enrolled, 22 treated, 22 followed to completion. Index trauma: physical or sexual assault 7 (31.8%), combat or war zone exposure 6 (27.3%), serious accident 1 (4.5%), other 8 (36.4%) | |
| Intervention | |
| COMP360, synthetic psilocybin, single 25 mg dose, administered with psychological support: three preparation sessions during the run-in period, one 6 to 8 hour dosing session with two therapists present and a psychiatrist available on site, three integration sessions the following day and then at weeks 1 and 2. The support model is the one designed by the sponsor for its trials. Therapists were licensed mental health professionals. Discontinuation of protocol-prohibited medications was required at least two weeks before enrolment, four weeks for fluoxetine: nine participants (40.9%) had to discontinue an antidepressant | |
| Comparator | |
| None. Single arm, no placebo, no randomisation, no blinding, either for participants or for assessing clinicians | |
| Endpoints | |
| Primary: safety and tolerability, number of participants with a treatment-emergent adverse event up to 12 weeks. Secondary: CAPS-5, PCL-5, Sheehan Disability Scale, EQ-5D-5L. Exploratory post hoc: correlations between 5D-ASC dimensions and change in CAPS-5 | |
| Design | |
| Phase 2, open-label, non-randomised, multicentre trial. Three sites: King’s College London in the United Kingdom, Mount Sinai in New York, and Sunstone Therapies in Rockville, United States. NCT05312151, conducted from June 2022 to February 2024, with 12 weeks of follow-up and clinic visits at day 2, week 1, week 2, week 4 and week 12, plus two additional visits at weeks 6 and 9, in clinic or remote. CEBM level of evidence 4. Sponsor and funder: Compass Pathfinder Ltd, London. Open-access publication under a CC BY-NC licence | |
| Exclusions | |
| Current or past history of schizophrenia, schizoaffective disorder or another psychotic disorder, bipolar disorder, obsessive compulsive disorder, or a personality disorder, including borderline personality disorder. Complex post-traumatic stress disorder. Clinically significant suicide risk. Alcohol or substance use disorder within the past twelve months. Use of MDMA, psilocybin or another psychedelic within the past year. Primary diagnosis of major depressive episode within the past six months. Traumatic exposure within the past three months. Significant childhood physical or sexual abuse, at the clinician’s discretion |
Quality control
| Criterion | Status |
|---|---|
| Pre-registration | Solid |
| FindingNCT05312151. The registered primary endpoint is indeed the occurrence of treatment-emergent adverse events up to 12 weeks, and that is what the article reports as primary | |
| Control group | Absent |
| FindingSingle arm. No counterfactual, so no causal attribution is possible for the symptom-based endpoints | |
| Blinding | Absent |
| FindingOpen-label trial. Participants know what they received, and so do the clinicians rating the CAPS-5 | |
| Fit between design and primary objective | Solid |
| FindingTo describe the nature and time course of adverse effects following a first dose in a given population, an open-label phase 2 design is the right tool. The objection here is not to the design, it is to what is read into it | |
| Sample size | Caveat |
| Finding22 exposed participants. A rare event, by construction, cannot appear at this scale | |
| Suicidality monitoring | Solid |
| FindingC-SSRS administered at every visit, covering the period since the previous visit. Both suicidal ideation events are reported with their timing, intensity, causality assessment and the participants’ prior history. That is more than many trials in this field do | |
| Vital sign monitoring | Caveat |
| FindingThe authors themselves report limited vital sign data during the dosing session, available on the enhanced schedule for only four participants. Acute cardiovascular effects therefore remain poorly characterised here | |
| Generalisability | Caveat |
| FindingThe exclusion list removes the profiles most common in practice: complex presentations, a recent major depressive episode in the foreground, bipolar disorder, personality disorder, a substance use disorder within the past year, a history of childhood maltreatment, significant suicide risk. The safety result applies to those who were enrolled, not to those who were screened out | |
| Independence | Insufficient |
| FindingTrial funded and sponsored by Compass Pathfinder Ltd. The contract research organisation overseeing trial conduct is paid by the sponsor and works under its direction. Five of the nine authors, including the last author, are current or former employees of Compass Pathways Plc and hold shares, options or restricted stock units | |
| Transparency of reporting | Solid |
| FindingOpen access, PMCID PMC13198616, limitations stated plainly by the authors, including the absence of a comparator. The conflict of interest is structural, the reporting of the data is not | |
Results
| Endpoint | Published result |
|---|---|
| Treatment-emergent adverse events | 117 events, in 22 of 22 participants (100%). 70 (59.8%) on the day of administration, of which 64 (91.4%) resolved by the next day: 56 the same day, 8 the following day |
| PEB readingPrimary endpoint, interpretable everyone experienced something, and most of it played out and resolved on the day itself | |
| Most frequent effects | Headache 11 participants (50.0%), nausea 8 (36.4%), crying 6 (27.3%), fatigue 6 (27.3%) |
| PEB readingPrimary endpoint, interpretable crying counted as an adverse event says something about the nature of the session | |
| Severe events | 3 participants (13.6%), 8 events: euphoric mood (2), nausea, chills, muscle contractions, visual hallucination, auditory hallucination, synaesthetic hallucination. All occurred on the day of administration and were all judged related to treatment, 7 resolved the same day, 1 the next day |
| PEB readingCaveat severe does not mean serious, but one participant in seven went through an episode rated as severe in intensity | |
| Serious events and discontinuations | No serious adverse event. No participant left the trial, for any reason: all 22 participants completed the 12 weeks of follow-up |
| PEB readingPrimary endpoint, interpretable with 22 exposed, this rules out frequent acute toxicity, not rare toxicity | |
| Suicidal ideation | 2 events. The first, of moderate intensity, began and resolved on the day of administration in a participant who went on to be a responder in remission at week 4: it was judged related to the drug. The second, of mild intensity, occurred at week 7 in a non-responder and resolved within a month: it was judged possibly related. Both participants had a history of suicidality on the C-SSRS |
| PEB readingCaveat the signal is there, and in one case it is dated to the day of administration itself. Reminder: participants with clinically significant suicide risk had been excluded | |
| CAPS-5, week 4 | Baseline score 47.5 (9.78). Change −29.9 (14.06), 95% CI −36.1 to −23.7; Cohen’s d 2.13. Response 81.8%, remission 63.6% |
| PEB readingNot interpretable secondary endpoint measured with no comparator. None of this change can be attributed to the drug | |
| CAPS-5, week 12 | Change −29.5 (15.43), 95% CI −36.4 to −22.7; Cohen’s d 1.92. Response 77.3%, remission 54.5%. Response defined as a decrease of at least 15 points, remission as a score of 20 or less |
| PEB readingNot interpretable apparent durability changes nothing: without a control arm, there is no way to know what the same support package would have produced without the drug | |
| PCL-5 | Baseline score 52.1 (13.68). Day 2: −33.5 (14.32), 95% CI −39.8 to −27.1; d 2.34. Week 12: −34.3 (18.13), 95% CI −42.4 to −26.3; d 1.89 |
| PEB readingNot interpretable self-report questionnaire completed by participants who all knew exactly what they had received, from two days after the session onward | |
| 5D-ASC, correlations with CAPS-5 | Spearman correlations with change in CAPS-5. Oceanic boundlessness: rs = −0.442 at week 4 and rs = −0.394 at week 12, a more intense experience going with a greater reduction. The three other moderate-strength correlations run in the opposite direction: anxious ego dissolution rs = 0.396 at week 4, reduction of vigilance rs = 0.425 at week 4, visionary restructuralisation rs = 0.327 at week 12, a more intense experience going here with a more modest reduction. The remaining coefficients are below 0.3 |
| PEB readingNot interpretable exploratory post hoc analysis, on 22 observations, with no p-value reported. At this sample size, a single atypical participant shifts the coefficient. And the pattern is not uniform: keeping only the dimension that runs in the expected direction, while setting aside the three that run the other way, would be a selective reading | |
Why the efficacy figures do not read as evidence
A −29.9 change on the CAPS-5 and a Cohen’s d of 2.13 look spectacular on paper. The instinct is to compare this figure with those from controlled trials. That instinct is wrong, and it is worth explaining exactly why, because the same reasoning will apply elsewhere.
A within-group before-after change in a single arm bundles together at least six things that cannot be separated from one another.
- The natural course of the disorder over twelve weeks, which is not zero.
- Regression to the mean, which is mechanical once people are selected on a high score at a single point in time.
- Participant expectancy: these were people who applied to receive a psychedelic, who knew they had received it, and who felt it very clearly.
- Rater expectancy: the clinician scoring the CAPS-5 knew who had been treated and when.
- The support package itself, which is far from incidental: three preparation sessions, six to eight hours with two therapists, a psychiatrist on site, three integration sessions. No comparison here tells us what this same package would produce without the drug.
- The pharmacological effect of psilocybin, which is precisely what is being sought.
The published confidence intervals around these changes do not help. They describe the precision with which a change observed in this group was measured. They say nothing about its cause. A narrow interval around a non-interpretable quantity remains a non-interpretable quantity. The authors themselves state that they conducted no inferential statistical analysis: no test, no p-value, only descriptive summaries.
One figure from the same article makes this concrete: on the PCL-5, the drop is 33.5 points by day 2, roughly two thirds of the baseline score, forty-eight hours after a six to eight hour session. Such a kinetic is compatible with a rapid pharmacological effect. It is equally compatible with the state of a patient the day after an intense experience lived through in a supportive setting, asked to fill in a self-report questionnaire. Telling the two apart requires a control arm. There is none.
The authors, to their credit, do not cross that line. Their conclusion is that 25 mg psilocybin administered with psychological support “may be safe, well tolerated and associated with symptomatic improvement” in adults with post-traumatic stress disorder. Associated, not responsible. Their first stated limitation is the absence of a comparator arm, and they write that larger controlled studies are needed to understand safety and efficacy. PEB does not harden that conclusion, and rejects the efficacy reading that will be made of it elsewhere.
Critical appraisal
| Domain | Judgement |
|---|---|
| Level of evidence | Low |
| FindingCEBM 4. Prospective open-label case series of 22 exposed cases. This is the floor above which interpretable clinical research begins, not a result on which to base practice | |
| What the trial establishes | Solid within its limits |
| FindingIn 22 selected adults, a single 25 mg dose in a research setting produced no serious event and no dropout, and most of the acute manifestations resolved within one to two days. That is the contribution, and it is real: the post-traumatic stress disorder population had until now been poorly represented in psilocybin trials | |
| What the trial cannot establish | Out of scope |
| FindingEfficacy, comparison with a reference treatment, the respective contribution of the drug and of the support package, safety beyond twelve weeks, safety for rare events, safety in the excluded patients | |
| Trauma-specific risk | Caveat |
| FindingA psychedelic experience can bring traumatic material back to the foreground. The companion qualitative study conducted on the same cohort describes precisely this kind of engagement with the material, both direct and indirect, through affective, bodily and self-transcendent pathways. This may be exactly the intended mechanism. Without a frame, it is also the scenario for an unsupported, uncontained reactivation. Crying in 27.3% of participants and suicidal ideation occurring on the day of administration are not details of tolerability | |
| The support package is part of the treatment | Caveat |
| FindingStrict selection, preparation, two therapists present for six to eight hours, a psychiatrist on site, integration sessions, C-SSRS at every visit. The safety result belongs to this entire package. It does not transfer to a standalone dose, still less to a dose taken outside of care | |
| Independence | Insufficient |
| FindingSponsor, funding, oversight of trial conduct, and the majority of the authors, five of nine, belong to the same industry player. This does not invalidate the data, but it requires taking them for what they are: a step in a product’s development, reported by those developing it | |
| Novelty claim | Caveat |
| FindingThe authors write that the effects of psilocybin in this disorder were previously unknown. We looked for an earlier counter-example predating their publication of 29 August 2025 and found none: the older work we located consists of protocols or narrative reviews. Since then, an open-label pilot trial in 12 veterans has been published, with two doses and a different design. The literature remains very thin | |
A word on what this trial does not say but will be made to say. No serious event among 22 exposed participants bounds the risk from above very loosely. The rule of three, a back-of-the-envelope calculation applied to trials where no event occurred, gives an upper bound of roughly 14% for the true frequency. This figure is not in the article, it is an arithmetic deduction from its result. In other words, an adverse event occurring in one patient in twenty would be perfectly compatible with what was observed. Writing “no serious event” is accurate. Concluding “it is safe” is not.
Level of evidence
PEB assessment: low confidence in the short-term safety of a single 25 mg dose administered in a tightly controlled research setting, in selected adults without major comorbidity. Zero confidence, in the strict sense, in efficacy, which was not evaluated here and cannot be evaluated with this design. The scientific score of 61 reflects this gap: the method is adequate for the question it asks, but that question is narrow, the sample is small and independence is weak. The editorial score of 81 reflects something else. This study is a good teaching case, and that is why we are publishing it.
The colleague test
What an experienced colleague would say if handed this study in thirty seconds, between two consultations.
“Twenty-two patients, one dose, no control, primary endpoint is tolerability: on that front, nothing serious, and most of what happened on the day of the session had settled by the next day. What’s left is two episodes of suicidal ideation, one on the day itself and one at week 7. The CAPS-5 drops, I don’t even look at them, there’s nothing to compare them with. What interests me is that we now have something to say to the patient who read an article in the news. And the answer is that the drug is a controlled substance, there’s no authorisation, and there is not yet the slightest proof that it works.”
Translated for practice: when the primary endpoint is safety, the efficacy figures in the article are groundwork for the next trial, not results.
What you can do with this on Monday morning. Nothing to change in prescribing. A good deal to say in consultation, because the question does come up.
- State the legal framework plainly and without judgement. Psilocybin and psilocin are listed as narcotics under Annex III of the Order of 22 February 1990 in France, and hallucinogenic mushrooms, including those of the genus Psilocybe, under Annex IV. Possession is prohibited absent express authorisation (Article R. 5132-74 of the Public Health Code), and illicit use is punishable by one year of imprisonment and a 3,750 euro fine (Article L. 3421-1). No marketing authorisation exists. A patient who asks about this is usually first trying to find out whether it is allowed: the answer is no, outside an authorised research protocol. Readers outside France should check their own jurisdiction’s classification, since it varies from one country to another.
- Separate the two questions that press coverage tends to merge. The safety question: in a research setting, in 22 carefully screened people, a single dose caused no serious event. The efficacy question: this study did not ask it, and no data today allow it to be answered. Patients generally hear the two as one.
- Explain why the support package is not a backdrop. The 22 participants were selected, prepared over three sessions, supported for six to eight hours by two therapists with a psychiatrist available, seen three more times for integration, and assessed with the C-SSRS at every visit. The tolerability result belongs to this whole package. A dose taken alone, at home, or during a commercial retreat abroad, reproduces none of these elements.
- Name the risk specific to trauma. A psychedelic experience can bring traumatic material back to the foreground. In this trial, more than one participant in four cried to the point that it was logged as an adverse event, and one episode of suicidal ideation occurred on the day of administration itself. People with a complex presentation, significant suicide risk, a psychotic or bipolar history, a personality disorder, a substance use disorder within the past year, or a history of childhood maltreatment had been excluded: for them, nothing is known, which is not reassuring, it is simply unknown.
- Document any reported use without moralising. Note the substance, the presumed dose, the context, current medications. Worth knowing: the protocol required stopping prohibited psychotropics at least two weeks beforehand, four weeks for fluoxetine. A patient on a serotonin reuptake inhibitor who used psilocybin would not match any of the conditions studied here.
- Redirect toward treatments with established efficacy for post-traumatic stress disorder, foremost among them trauma-focused psychotherapies, in line with current guidelines. Those guidelines are external to the study analysed here and do not rest on it.
- Take away the reading grid. Three questions suffice for any open-label trial: what was the primary endpoint, was there a comparator, who scored the outcomes while knowing what.
Frequently asked questions
Does this trial show that psilocybin treats post-traumatic stress disorder?
No, and it was not designed to. Its primary endpoint is safety. The reductions on the CAPS-5 and PCL-5 are secondary endpoints observed in a single arm, with no placebo and no blinding. It is not possible to know how much of the change is due to the drug, the support package, expectancy, or the natural course of the disorder. The authors themselves write that controlled studies are needed.
Isn’t a Cohen’s d of 2.13 huge?
Numerically, yes. But an effect size calculated on a before-after change within a single group and an effect size calculated between two randomised arms do not measure the same thing and cannot be compared. The first quantifies the magnitude of an observed change, the second the magnitude of a difference attributable to treatment. Putting the two on the same axis is the most common error made with this type of article.
Is psilocybin legal?
That depends on the jurisdiction: psilocybin is a controlled substance in most countries, and its precise legal status varies from one to another. In France, a case that has been independently verified, it is classified as a narcotic under Annex III of the Order of 22 February 1990. Possession is prohibited absent express authorisation, and illicit use is a criminal offence punishable by one year of imprisonment and a 3,750 euro fine. No marketing authorisation exists there. Only participation in an authorised research protocol allows lawful exposure. Readers should verify the rules that apply in their own country rather than assume a clinical trial implies legal access.
Can we call this safe, since there was no serious adverse event?
No. Twenty-two exposed people is too few to detect a rare event, and zero cases observed among 22 remains compatible, by the rule of three, with a true frequency of up to roughly 14%. On top of that, three participants had events rated severe, two suicidal ideation events occurred, and the authors themselves judge the vital sign data collected during the session to be insufficient.
Is there a risk specific to trauma?
Yes, and it is specific to this indication. The psychedelic experience can bring traumatic material back to the foreground, as documented by the qualitative study conducted on the same cohort. In a trial, this unfolds with two therapists present and integration sessions planned. Outside that frame, the same resurfacing can occur with no one there to contain it. The highest-risk profiles had, in fact, been excluded from the trial.
What would it take to draw any conclusion about efficacy?
A randomised trial against a comparator, with blinded raters, an adequate sample size, an identical psychological support package in both arms, and follow-up beyond three months. Maintaining blinding remains a genuinely difficult problem with a drug whose subjective effects are this pronounced, which makes the choice of comparator central to the design of such trials.
Annotated bibliography
McGowan NM, Rucker JJ, Yehuda R, Agrawal M, Modlin NL, Simmons H, Tofil-Kaluza A, Das S, Goodwin GM (2026). Investigating the safety and tolerability of single-dose psilocybin for post-traumatic stress disorder: a nonrandomized open-label clinical trial. Journal of Psychopharmacology, 40(1), 139-148. DOI 10.1177/02698811251362390 · PMID 40883964 · PMCID PMC13198616. Source study analysed here, first published online on 29 August 2025, open access under a CC BY-NC licence. Trial funded and sponsored by Compass Pathfinder Ltd, London, with the contract research organisation paid by the sponsor and working under its direction. Five of the nine authors are current or former employees of Compass Pathways Plc and hold shares, options or restricted stock units. The remaining authors declare ties to several companies in the psychedelic sector and institutional research funding.
Modlin NL, Williamson V, Goodwin GM, Malievskaia E, Atli M, Elek Z, Gaillard A, Koelpin D, Cleare A, Agrawal M, Yehuda R, Kirlic N, Rucker JJ (2025). Investigational psilocybin treatment for post-traumatic stress disorder: a qualitative study of participant experience, trauma engagement, and differences from standard treatment. eClinicalMedicine, 90, 103692. DOI 10.1016/j.eclinm.2025.103692. Qualitative study nested in the same NCT05312151 trial, 21 adults interviewed before treatment, the day after, and at week 12. It describes engagement with traumatic material that is both direct and indirect, through affective, bodily and self-transcendent pathways, where reference treatments rely on direct confrontation with the memory. Qualitative work: it illuminates lived experience, it measures no effect, and it does not compensate for the absence of a control arm in the main trial.
Armstrong SB, Levin AW, Sepeda ND, Shaub H, Hunter T, Douglas A, Lancelotta R, Davis AK (2026). Safety, feasibility, and preliminary clinical outcomes of psilocybin-assisted therapy for veterans with severe, treatment-resistant PTSD: an open-label pilot clinical trial. Communications Medicine, 6, 411. DOI 10.1038/s43856-026-01767-4. Published on 30 July 2026, therefore after the trial analysed here. Twelve veterans completed the protocol out of thirteen enrolled, two doses of 15 mg then 25 mg two to three weeks apart, roughly 16 hours of psychotherapy. No serious adverse event, no reported increase in suicidal ideation. Cited to situate the state of the literature: two open-label trials with no comparator, totalling fewer than forty patients, do not constitute an evidence base.
ClinicalTrials.gov, NCT05312151. The safety and tolerability of COMP360 in participants with post-traumatic stress disorder. clinicaltrials.gov. Registry consulted to verify the registered primary endpoint, the single-arm design, the three sites, the final sample of 22 participants and the exclusion criteria. The published primary endpoint matches the registered primary endpoint.
Order of 22 February 1990 listing substances classified as narcotics, consolidated version published by the French National Agency for the Safety of Medicines and Health Products (ANSM). ansm.sante.fr. Psilocin and psilocybin appear in Annex III, hallucinogenic mushrooms, notably of the genera Stropharia, Conocybe and Psilocybe, in Annex IV. Since February 2022, classification decisions have been the responsibility of the agency’s director general, whose rulings amend this order, which remains in force.
French Public Health Code, Articles R. 5132-74 and L. 3421-1. legifrance.gouv.fr. Article R. 5132-74, in its version in force since 1 March 2022 under Decree No. 2022-194 of 17 February 2022, prohibits, absent express authorisation, the production (including cultivation), manufacture, transport, import, export and possession of substances classified as narcotics, with authorisations governed by Articles R. 5132-75 to R. 5132-77. Article L. 3421-1 punishes illicit use with one year of imprisonment and a 3,750 euro fine.
Editorial collections
Tags
Verified on August 11, 2026 against the full text of the publication and its supplementary material where available. This analysis underwent an independent double reading. The English version was checked for conformity on September 20, 2026, against the figures of the French version and against the source. How we verify what we publish
This analysis is intended for healthcare professionals. It does not constitute a prescribing recommendation and does not replace individual clinical judgment.
Analysis from Psychiatry Evidence Base, evidence-based psychiatry, explained with rigor.