Published on 20 September 2026
Psilocybin and major depression: what remains at twelve months once the blind breaks down
JAMA Network Open · 2026; volume 9 (issue 5): article e2612589 · Yngwe et al.
Published title “Short-Term and Late-Term Effects of Psilocybin on Symptoms in Major Depression: A Randomized Clinical Trial”
Registration NCT04630964
DOI 10.1001/jamanetworkopen.2026.12589
PMID 42138922
Scientific 61
Editorial 82
The essentials
A single-center Swedish randomized controlled trial compared a single 25 mg dose of psilocybin with an active placebo, niacin 100 mg, in 35 adults with a recurrent major depressive episode, using identical psychotherapeutic support in both arms and a controlled follow-up of 365 days. At day 8, the difference on the clinician-rated MADRS is 7.27 points in favor of psilocybin (95% CI -12.89 to -1.65; p = 0.01), and remission at six weeks concerns 52.9% of patients versus 5.9% under active placebo. At twelve months, the difference on this same scale is no longer demonstrated (-3.68 points, 95% CI -9.30 to +1.94; p = 0.20), and no difference appears in subsequent antidepressant use (6 of 17 patients versus 7 of 18; among those who started one, a mean delay of 130.0 versus 128.0 days; p = 0.70). The main contribution of this trial is not its result but its measurement: according to the authors, it is the first psilocybin trial in major depression to formally assess the integrity of its own blind, and it finds that this blind does not hold. At twelve months, 94.1% of participants under psilocybin and 100% of niacin responders correctly guess their allocation, and at the exact time of the primary endpoint the blinded rater guesses correctly in 27 of 34 cases, or 79.4%. On a fully subjective endpoint, this configuration cannot separate the pharmacological effect from the expectation effect. The sample, 35 subjects, further precludes any precise estimate. What is solid here is the demonstration of a methodological limitation of the field, not the quantification of a benefit.
Context
The question raised by this trial reaches the consulting room before it reaches the literature. Patients bring in a news article announcing that a single dose of psilocybin can lift a depression for months, sometimes for a year. This narrative of durability was built on small trials followed by uncontrolled extensions, in which every participant knew they had received the drug and no comparator existed beyond a few weeks.
Two methodological objections have accompanied this field since its origin. The first concerns blinding: a substance that produces an intense perceptual alteration for several hours announces itself, which makes double blinding declarative rather than real. The second concerns the comparator: when the control arm is made up of volunteers who applied to receive a psychedelic and who understand within an hour that they did not receive it, its trajectory is no longer that of an ordinary placebo.
The trial analyzed here has the distinction of addressing both objections head-on, within a design that is otherwise unusually careful for the field: strictly identical psychotherapeutic co-intervention in both arms, exclusion of subjects already exposed to psychedelics, public and philanthropic funding with no conflict of interest tied to the psychedelics industry, a full protocol attached to the publication, and above all a randomized controlled follow-up maintained for twelve months, which the authors present as unprecedented in the field.
What the trial measures, and what it does not
A symptom-based assessment, a blinding measure, no biological marker
The trial does not rest on a directly tested pathophysiological hypothesis. It compares scale scores. The distinction between what is measured and what is not therefore governs the entire reading, particularly over the period where the claim of durability is decided.
| Element | Status in the study |
|---|---|
| Depressive symptoms, clinician-rated | Measured, at day 8, day 15, day 42 and day 365 |
| Finding Clinician-rated MADRS is the trial’s reference instrument and carries the primary endpoint. No clinician-rated measurement exists between day 42 and day 365, that is, across the entire window where the article situates its “late-term effects.” | |
| Depressive symptoms, self-rated | Measured, from day 1 to day 8, then day 15, day 42 and monthly until day 365 |
| Finding Self-reported MADRS-S fills the gap left open, but it is a secondary endpoint, not corrected for multiplicity, and the most exposed to expectation bias since it is completed by participants who almost all know what they received. | |
| Blinding integrity | Measured, in participants and in the rater |
| Finding This is the trial’s unprecedented contribution. Allocation guesses and the associated confidence level were collected, along with the intensity of the psychedelic experience. This measure quantifies a limitation that earlier trials merely mentioned. | |
| Subsequent care-seeking | Measured, over twelve months |
| Finding Time to starting an antidepressant over twelve months is a behavioral endpoint, less dependent on subjective appraisal than the scales. It is therefore particularly informative in a trial whose blind is broken. | |
| Neurobiological mechanism | Not reported in this publication |
| Finding No biomarker, imaging endpoint or mediator is presented here as an outcome. The trial therefore says nothing about the mechanism of action, in either direction. Any mechanistic reading of these results would be extrapolation. | |
| Standardized effect size | Not reported |
| Finding No Cohen’s d or equivalent accompanies the primary endpoint, which complicates comparison with other antidepressant interventions. The publication also does not report the standard deviations of the within-arm change, so a reliable reconstruction from the published data alone is not possible. | |
The study at a glance
| Population, intervention, comparator, outcome | |
|---|---|
| Population (P) | |
| Detail Swedish-speaking adults with a current recurrent major depressive episode per ICD-10, lasting 30 days or more and less than 5 years, with a MADRS of 22 or above. Recruitment through social media and online advertising: 748 candidates, 63 assessed, 35 randomized, that is 4.7% of candidates. An exclusion uncommon in this field and methodologically valuable: any prior use of psychedelics. Ongoing antidepressant treatment and ongoing psychotherapy were exclusion criteria, and 7 participants had to stop their antidepressant to enter the trial. | |
| Intervention (I) | |
| Detail Psilocybin 25 mg orally, single dose, taken with 200 mL of water. Dosing day of 7 to 10 hours, eye mask, music, supervision by psychologists, with the option to call the study physician; participants were kept for 7 hours after dosing before going home. | |
| Comparator (C) | |
| Detail Niacin 100 mg, an identical-appearing capsule, within the same day-long setup. An important point, often absent from trials in this field: the psychotherapeutic co-intervention is strictly identical in both arms, 5 sessions spread over 17 days, one preparation session on day -1, the dosing day on day 0 and three integration sessions on days 1, 8 and 15, following a standardized manual supplied by the institute that also supplied the investigational product. The comparison therefore bears on the molecule itself, not on a richer support on one side than the other. | |
| Outcome (O) | |
| Detail Primary: change in clinician-rated MADRS at day 8. Secondary: MADRS at days 15, 42 and 365; monthly self-reported MADRS-S to day 365; response and remission; time to starting an antidepressant over 12 months; GAD-7, the Sheehan Disability Scale, EQ-5D-5L, CGI. | |
| Design, sample size and analysis | |
| Detail Phase 2 randomized controlled trial, single-center, parallel-group 1:1, single dose, active placebo, declared double blind, controlled follow-up of 365 days. Enrollment from January 26, 2021 to February 19, 2024. CEBM level of evidence 2b, an individually randomized trial downgraded by its sample size and by the documented breakdown of blinding. 35 subjects randomized, 17 under psilocybin and 18 under niacin; one participant in the niacin arm provided no data after dosing, leaving 17 subjects analyzed in each arm for the primary endpoint, and 33 of the 35 randomized have self-reported data at the last assessment. Mixed model for repeated measures with baseline value as covariate, Kenward-Roger correction, univariable logistic regression and Fisher’s exact test for response and remission, Kaplan-Meier survival estimation and log-rank test for time to starting an antidepressant. |
Quality control
| Domain | Judgement |
|---|---|
| Preregistration and protocol | Complete and public |
| Finding Trial registered, full protocol attached to the publication, statistical analysis plan described in the body of the article, two appendices and thirteen supplementary result tables announced. Minor reservation: the analysis plan is not filed as a separate document. | |
| Prespecified primary endpoint | Upheld |
| Finding The announced primary endpoint is the one analyzed and reported, without substitution or redefinition during the trial. This point, unremarkable elsewhere, is not in this field. | |
| Attrition and follow-up | One lost of 35 randomized |
| Finding Only one participant lost for the primary endpoint out of 35 randomized, and 33 of the 35 still assessed by self-report at the last measurement, after twelve months of monthly follow-up. This is a remarkable trial conduct, which shields the result from attrition bias. | |
| Allocation concealment | Adequate |
| Finding Computer-generated sequence, identical capsules prepared and labeled by the product supplier, allocation key held by the statistician alone until the end of the trial. Reservation: participants 31 to 36 were to be assigned so as to reproduce the allocation pattern of the first thirty; since enrollment stopped at 35, five participants fall under this deterministic allocation, that is 14% of the sample. | |
| Blinding maintained | Measured, and broken |
| Finding The fact of having measured it is an act of rigor. The result of this measurement is the trial’s main limiting factor, detailed below. | |
| Evaluator independence | Not independent |
| Finding The MADRS raters were the study physicians, also responsible for recording adverse events. On a fully subjective endpoint, this dual role compounds the blinding problem. | |
| Baseline comparability | Marked imbalances |
| Finding Five prognostic variables tilt against the active placebo arm: baseline MADRS 21.1 versus 24.6, lifetime SSRI exposure 9 of 17 versus 17 of 18, history of suicide attempt 1 versus 4, psychiatric hospitalization 1 versus 4, chronic pain 0 versus 3. One variable tilts the other way, lifetime number of depressive episodes, 8.2 versus 6.7. On 35 subjects, this kind of imbalance is expected, and the authors mention it themselves as a limitation, but it is not neutral for binary endpoints with a fixed threshold. | |
| Multiplicity control | Absent, acknowledged |
| Finding The authors explicitly state that no correction for multiple comparisons was applied to the secondary and exploratory endpoints, which cover seven instruments, some twenty measurement time points and thirteen supplementary tables. | |
| Funding and conflicts of interest | Above the field’s standard |
| Finding Public and philanthropic funding, the Swedish Research Council and the Norrsken Mind foundation, with a declared nil role for the funders. No conflict of interest tied to the psychedelics industry is declared. Use of a generative artificial intelligence tool for language editing is declared, with the analyses and scientific content attributed to the authors. Classification reservation: the institute that supplied the investigational product, the treatment manual and the online training appears in the acknowledgments rather than in the conflicts of interest, which is not the expected place for an intervention supplier. Personal fees from a pharmaceutical company, outside the submitted work, are also declared for the last author. | |
Results
The primary result is positive in the short term, and this must be said without trimming it. At day 8, the clinician-rated MADRS falls 7.27 points more under psilocybin than under niacin (95% CI -12.89 to -1.65; p = 0.01). The gap widens at day 15 then holds at six weeks. What the confidence interval allows, however, ranges from a very large effect to a clinically modest one, the trial not being sized to decide between the two.
| Outcome | Reported result |
|---|---|
| Clinician-rated MADRS, day 8 (primary) | -7.27 points (95% CI -12.89 to -1.65); p = 0.01 |
| Finding Prespecified endpoint, met, unmodified. No standardized effect size accompanies this figure. The observed effect is markedly below the 11-point effect on which power was calculated. | |
| Clinician-rated MADRS, days 15 and 42 | -11.03 (95% CI -16.65 to -5.42) then -8.33 (95% CI -13.94 to -2.71) |
| Finding The maximum separation between the two arms occurs at two weeks. These endpoints are secondary and not corrected for multiplicity. | |
| Clinician-rated MADRS, day 365 | -3.68 points (95% CI -9.30 to +1.94); p = 0.20 |
| Finding On the trial’s reference instrument, the difference is no longer demonstrated at twelve months. The exact wording matters: this is an absence of demonstration, not a demonstration of absence. The interval remains compatible with a genuine residual benefit as well as with a slight disadvantage, and 35 subjects cannot discriminate between the two. | |
| Self-rated MADRS-S, day 102 | -6.60 points (95% CI -13.01 to -0.19); p = 0.04 |
| Finding This is the result on which the claim of an effect duration beyond three months rests. A secondary, self-reported endpoint, not corrected for multiplicity, with one bound of the interval 0.19 point from the threshold of non-significance, in participants nearly all of whom know their allocation. No clinician-rated measure supports it over this window, since none exists between day 42 and day 365. | |
| Remission at day 42 | 52.9% (9 of 17) versus 5.9% (1 of 17); OR 9.00 (95% CI 1.45 to 175.40); NNT 2.1 |
| Finding This is the trial’s most spectacular and most fragile figure. The width of the confidence interval, on the order of one to a hundred, says by itself what a proportion calculated on 17 subjects per arm is worth. Above all, the test used, Fisher’s exact test and univariable logistic regression, includes no adjustment for baseline value, whereas the baseline MADRS in the psilocybin arm was 3.5 points lower: with the remission threshold set at a score below 10, this meant losing on average more than 11.1 points on one side against more than 14.6 on the other. The NNT is our own calculation from the published proportions. | |
| Remission at day 365 | 52.9% (9 of 17) versus 41.2% (7 of 17); OR 1.29 (95% CI 0.39 to 4.36); not significant |
| Finding At twelve months, the two arms converge, mainly through a rise in the active placebo arm. This convergence is consistent with the natural course of a supported depressive episode, and with subsequent care-seeking in both groups. | |
| Antidepressant use over 12 months | 6 of 17 versus 7 of 18; mean time to start 130.0 (44.6) versus 128.0 (96.8) days; p = 0.70 |
| Finding This is the endpoint least dependent on subjective appraisal, and it shows no difference. The denominators here are those of randomization, 17 and 18, not those of the primary endpoint, 17 and 17. The mean delay only concerns patients who actually started an antidepressant, all these initiations having occurred after day 42. Here again, an undemonstrated difference does not amount to a demonstrated equivalence on a sample of this size. | |
| Blinding integrity | Participants 94.1% and 100% at day 365; rater 79.4% at day 8 |
| Finding At day 365, the rater still correctly identified the allocation in 82.4% of cases under psilocybin and 76.5% under niacin. Reported intensity of the psychedelic experience was 7.06 (SD 3.58) under psilocybin versus 1.22 (1.63) under niacin, on a 0-to-10 scale. Expected benefit before dosing, by contrast, was comparable, 56.7 (22.3) versus 56.8 (27.5) out of 100. The active placebo chosen therefore does not reproduce the subjective effects of the tested product, while both groups started from equivalent expectations. | |
| Safety | 2 of 17 patients with severe, prolonged anxiety under psilocybin, that is 11.8% |
| Finding These two cases persisted for several weeks after dosing and required medical care; no case of this kind is reported in the niacin arm in the body of the article. The most frequent adverse events under psilocybin were headache (9), anxiety (7), hallucination (5), agitation (3), hypertension (3) and paresthesia (3). Ten serious adverse events are reported across the entire follow-up, none judged related to psilocybin by the authors; their distribution by arm appears in the supplementary material, which we were not able to consult. | |
Critical appraisal
The assessment follows the RoB 2 tool. The notable feature of this trial is that its risk of bias is not diffuse: it is concentrated, identified, and partly quantified by the authors themselves.
| RoB 2 domain | Judgement |
|---|---|
| D1. Randomization process | Some concerns |
| Finding Adequate concealment and a key held externally, but deterministic allocation of the last six participants and marked prognostic imbalances, all tilted in the same direction. On 35 subjects, chance alone can produce such gaps, which does not make them harmless for interpretation. | |
| D2. Deviations from intended interventions | Some concerns |
| Finding The analysis follows the intention-to-treat principle and the co-intervention is identical in both arms, which is a strength. But the near-universal knowledge of allocation by participants may alter their care-seeking behavior during the twelve months of follow-up, in either direction. | |
| D3. Missing outcome data | Low risk |
| Finding One participant lost to follow-up out of 35 at twelve months, two sensitivity analyses announced in the statistical section. This domain is solid. | |
| D4. Measurement of the outcome | High risk |
| Finding This is the domain that drives the overall judgement. A fully subjective endpoint, raters who are also the study physicians, and a blind documented as broken at the very time of the primary measurement. The meta-epidemiological literature on the effect of unblinding for subjective endpoints, cited by the authors themselves, places the average inflation of the estimate at around 22%, with a ratio of odds ratios of 0.78 (95% credible interval 0.65 to 0.92). This benchmark concerns odds ratios drawn from trials across varied fields: it indicates a direction and an order of magnitude of bias, it does not provide a correction factor applicable to a 7.27-point difference in means on the MADRS. No corrected reanalysis is therefore possible here. | |
| D5. Selection of the reported result | Some concerns |
| Finding The primary endpoint is prespecified and reported without modification, which protects the core of the trial. However, the total absence of multiplicity correction on the secondary endpoints makes it impossible to interpret isolated point effects at certain late measurement times in isolation: a biological effect does not fade out and then reappear at two isolated dates. | |
| Overall judgement | High risk |
| Finding High risk through domain D4, in a trial whose conduct is otherwise of good quality. The consequence is not that the result is false, but that its size cannot be estimated from this trial. | |
Two points of interpretation deserve to be raised separately, as they concern the packaging of the work more than its conduct.
The title announces what the reference endpoint does not establish. The claim of late-term effects and of an effect duration beyond three months rests on an uncorrected secondary, self-reported endpoint, while the clinician-rated measure at twelve months is negative and no clinician-rated assessment exists between six weeks and one year. The authors themselves note that at early time points, effects were numerically larger in self-report than in clinician-report. This is exactly the expected signature of an expectation bias in an unblinded trial, and this observation is written without the conclusion being drawn from it. In fairness, the authors conclude that larger trials remain necessary to settle the question of long-term effects.
The interpretation of an abnormally weak placebo effect ignores a competing explanation. The low remission rate in the niacin arm at six weeks is presented as the sign of a modest placebo effect in this population. Yet underperformance of the control arm is precisely what one would expect from a fully unblinded placebo, in volunteers who came forward to receive a psychedelic. This competing explanation is not speculative: it has been documented in a meta-analysis by a team that includes this trial’s last author, a work the article cites, but only in the sense that reinforces its own reading.
Level of evidence
What is demonstrated: that the blind of a psilocybin trial in major depression does not hold, with figures to support it, and that the rater himself is not blinded at the time of the primary measurement; that a difference favoring psilocybin exists in the short term on the clinician-rated MADRS, in a design where psychotherapeutic support was identical in both arms; that at twelve months, on the reference instrument, this difference is no longer found, and that no difference could be shown in subsequent antidepressant use, which is not the same thing as an established equivalence.
What is suggested: that the durability reported by uncontrolled extensions in the earlier literature owed in part to the observational setup rather than to the product; that the divergence between self-report and clinician-report reflects an expectation bias; that the true short-term benefit, corrected for the broken blind, is smaller than the raw estimate.
What is opinion, including our own: the exact magnitude of the correction to apply, the respective share of the drug and of the therapeutic context, and the idea that a placebo able to reproduce the subjective effects would be the field’s only methodological way out.
The inference risks to keep in mind are those of a very small trial: power calculated on an effect hypothesis that was not attainable given the initial severity, hence a trial underpowered for the effect actually observed; very wide confidence intervals, particularly on binary endpoints; uncontrolled multiplicity, which makes any isolated secondary result uninterpretable on its own; a novelty effect specific to heavily publicized molecules, acting on both participants and raters. Confidence is high on the blinding measure and on the negative twelve-month result for the reference instrument. It is low on the quantification of the short-term effect, and very low on the six-week remission figure.
The colleague test
What an experienced colleague would say, presented with this study in two minutes, between two consultations.
“On the conduct of the trial, I have nothing to fault, and measuring one’s own blind when no one else did is an act of rigor worth saluting. My problem is elsewhere: the title promises me late-term effects while the one-year reference measure is negative, and the claim beyond three months rests on a self-questionnaire filled in by patients who all know what they took. Thirty-five subjects, one patient in two in remission at six weeks with a confidence interval running from one and a half to a hundred and seventy-five, and a placebo group that was sicker to begin with: I would not build anything on that. The body of the article is honest, it is its packaging that is not.”
Translation for practice: this trial is not to be cited to say that psilocybin works, nor to say that it does not. It is to be cited to say what can be measured and what cannot yet be measured in this field, and to bring the question of durability back to what the controlled data actually show.
What you can do with this
- When a patient arrives with a news article about a lasting cure from a single dose, this trial offers a factual, non-ideological answer: at six weeks, a difference exists; at twelve months, on the reference measure, it is no longer found, and no difference appears in subsequent antidepressant use.
- Keep the regulatory picture distinct from the scientific debate, and remember that authorization, marketing, access pathways and reimbursement are four separate questions: none of them being open for a product says anything about its efficacy, and a reimbursement decision never states that a treatment is ineffective. As a concrete, verifiable illustration, as of the verification date of August 12, 2026, no psilocybin-containing specialty holds a marketing authorization in France or at the European level, and psilocybin is not listed in any derogatory access pathway there, whether early access, compassionate access or compassionate prescribing. Readers practicing outside France should verify where psilocybin currently stands on each of these four questions in their own jurisdiction. Off-label prescribing is not a workaround either: in France, article L. 5121-12-1 of the Public Health Code makes it conditional on the absence of an appropriate authorized alternative, a condition not met in major depression. Outside an authorized research protocol, no legal pathway currently allows a patient to be exposed to psilocybin for depression, a point that generalizes well beyond the French case.
- Build the safety signal into the information you give: 2 of 17 participants experienced severe anxiety persisting for several weeks and requiring care, an incidence of roughly one participant in eight, within a highly controlled setting with strict selection and several hours of continuous supervision. Unsupervised use benefits from none of these protections.
- Use this trial as a methodological reference in case discussions or journal clubs: the measurement of blinding integrity, the divergence between self-report and clinician-report, and the effect of a three-point baseline imbalance on a fixed-threshold binary endpoint are all visible here in their purest form.
- Do not conclude that there is no long-term effect: with 35 subjects, the trial did not have the power to demonstrate a modest residual benefit. What is missing here is the demonstration, not necessarily the effect.
Frequently asked questions
A patient asks me if a single session can be enough for a year. What should I answer?
That this is exactly the question this trial tried to answer, with the first comparative follow-up maintained for twelve months, and that the answer obtained on the reference measure is negative. You can add, without contradiction, that the trial is too small to rule out a modest residual benefit, and that the time to starting an antidepressant was identical in both groups.
If everyone guesses their allocation, does the trial still have value?
Yes, but not the value usually credited to it. It loses its ability to quantify an effect on subjective endpoints. It keeps its value on behavioral endpoints, such as subsequent care-seeking, on safety, and above all on what it documents about the experimental setup itself. It is the first trial in this field to put a number on an objection that had only ever been stated before.
Isn’t remission in one patient out of two at six weeks a major result?
It is a spectacular and fragile result. It rests on 17 patients per arm, its confidence interval spans nearly two orders of magnitude, it is not adjusted for baseline value even though the psilocybin arm started three and a half points lower on a scale with a fixed remission threshold, and it is measured by raters who correctly identified the allocation in eight cases out of ten. Any one of these reservations taken alone would call for caution.
The placebo arm responded very little. Doesn’t that strengthen the case for the drug’s effect?
That is the interpretation the authors propose, and it is not the only one possible. A control arm made up of volunteers who came to receive a psychedelic, and who all understand within a few hours that they did not receive it, has good reason to underperform relative to an ordinary placebo. A weak placebo response may therefore measure disappointment as much as the inertia of the illness. The same symmetry applies here as elsewhere: just as one cannot draw conclusions about the drug in an unblinded trial, one cannot draw conclusions about the comparator either.
Can I refer a patient to a protocol abroad?
This question goes beyond what this trial can answer and falls under regulatory and ethical frameworks rather than the data. Authorization, marketing, access pathways and reimbursement are four separate questions in every jurisdiction, and readers should verify where psilocybin currently stands under each of them where they practice. As a concrete illustration, in France, as of August 12, 2026, psilocybin is classified as a narcotic under Schedule III of the Order of 22 February 1990, holds no marketing authorization and is not listed in any derogatory access pathway; outside an authorized research protocol, no legal route currently allows a patient to be exposed to it there. On the data alone, this trial provides no durability argument that would justify such a referral, and it documents a risk of severe, prolonged anxiety in roughly one participant in eight even within a closely supervised setting.
Annotated bibliography
Source study. Yngwe H, Plavén-Sigray P, Ekman CJ, Henje E, Berglund A, Tiger M, Beckman M, Lundberg J. “Short-Term and Late-Term Effects of Psilocybin on Symptoms in Major Depression: A Randomized Clinical Trial.” JAMA Network Open 2026; 9(5): e2612589. DOI 10.1001/jamanetworkopen.2026.12589. PMID 42138922. Registration NCT04630964. Contribution: according to the authors, the first randomized controlled trial of psilocybin in major depression to formally measure the integrity of its blind and to maintain a comparative follow-up for twelve months. Limitation: 35 subjects, single-center, fully subjective primary endpoints and a blind documented as broken.
Context reference 1. Hieronymus F, López E, Werin Sjögren H, Lundberg J. “Control Group Outcomes in Trials of Psilocybin, SSRIs, or Esketamine for Depression: A Meta-Analysis.” JAMA Network Open 2025; 8(7): e2524119. DOI 10.1001/jamanetworkopen.2025.24119. PMID 40736734. Contribution: across 17 trials, the control arms of psilocybin trials markedly underperform, with a standardized mean change of 0.50 (SE 0.15) versus 1.00 (0.08) in SSRI trials and 1.12 (0.17) in esketamine trials, which provides a direct competing explanation for the weak placebo effect observed here. Limitation: a synthesis across heterogeneous trials, whose causal interpretation of the underperformance remains debated.
Context reference 2. Savović J, Jones H, Altman D, et al. “Influence of reported study design characteristics on intervention effect estimates from randomised controlled trials: combined analysis of meta-epidemiological studies.” Health Technology Assessment 2012; 16(35): 1-82. DOI 10.3310/hta16350. PMID 22989478. Contribution: quantifies the effect inflation associated with the absence or uncertainty of double blinding for subjective endpoints, with a ratio of odds ratios of 0.78 (95% credible interval 0.65 to 0.92), or about 22% average overestimation. Limitation: an average drawn from trials across varied fields, concerning odds ratios, not specific to psychiatry or to any given scale, to be used as a direction and order of magnitude of bias rather than as a correction factor.
Context reference 3. Gukasyan N, Davis AK, Barrett FS, et al. “Efficacy and safety of psilocybin-assisted treatment for major depressive disorder: prospective 12-month follow-up.” Journal of Psychopharmacology 2022; 36(2): 151-158. DOI 10.1177/02698811211073759. Contribution: represents the durability literature on which the narrative of a months-long persistent effect was built. Limitation: an uncontrolled extension, with no comparator beyond the randomized phase and no blinding, which makes it unfit to establish a duration of effect attributable to the drug.
Context reference 4. Carhart-Harris RL, Bolstridge M, Rucker J, et al. “Psilocybin with psychological support for treatment-resistant depression: an open-label feasibility study.” Lancet Psychiatry 2016; 3(7): 619-627. DOI 10.1016/S2215-0366(16)30065-7. And Carhart-Harris RL, Bolstridge M, Day CMJ, et al. “Psilocybin with psychological support for treatment-resistant depression: six-month follow-up.” Psychopharmacology (Berl) 2018; 235(2): 399-408. DOI 10.1007/s00213-017-4771-x. Contribution: the first of these works is the source of the effect hypothesis used for the power calculation of the trial analyzed here, and the field’s historical starting point. Limitation: very small samples, no randomized control arm, a population selected for treatment-resistant depression and participant expectations not neutralized.
Context reference 5. Sterne JAC, Savović J, Page MJ, et al. “RoB 2: a revised tool for assessing risk of bias in randomised trials.” The BMJ 2019; 366: l4898. DOI 10.1136/bmj.l4898. PMID 31462531. Contribution: the assessment framework used in this analysis, whose domain devoted to outcome measurement is decisive here. Limitation: a structured judgement instrument, which leaves room for appraisal between raters and does not produce a quantitative correction of the effect estimate.
What was consulted. References verified on August 12, 2026 against the published version. The supplementary material could not be consulted. Article submitted to an independent double reading.
Editorial collections
Tags
Verification
Verified on August 12, 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.