Psilocybin Clinical Trial Results: How the Studies Worked and What They Found

How to use this page

This page explains how psilocybin trials are built and how to read what they report, using the best-known trials as worked examples. It is a reading guide, not a verdict on any one condition. For the current evidence summary, the strength of evidence and the lawful routes to treatment, use the condition pages: psilocybin for depression, psilocybin for smoking cessation and psychedelics for alcohol use disorder. For studies that are taking volunteers now, see the recruiting clinical trials.

As of October 2026, no psilocybin medicine is approved by the FDA, and psilocybin and psilocyn remain Schedule I controlled substances. (eCFR (Office of the Federal Register), current; U.S. Food and Drug Administration, Jul 2026) COMPASS Pathways has said it expects to complete a rolling application for its synthetic psilocybin in treatment-resistant depression in the fourth quarter of 2026; that is a company statement about a pending application, not an approval. (COMPASS Pathways plc, via SEC EDGAR, 5 Aug 2026)

What a psilocybin trial actually tests

A psilocybin trial almost never tests a pill on its own. It tests a package: screening, preparation sessions, a dosing day of several hours with trained staff present, and follow-up support. Published safety guidelines for this research call for careful preparation, a safe physical setting and support from at least two monitors during the session. (Journal of Psychopharmacology, 2008) When a trial reports that "psilocybin" reduced depression, it means that whole package did, compared with whatever the control group received.

Doses are reported as trial facts. The Johns Hopkins depression trial gave two sessions, 20 mg/70 kg and then 30 mg/70 kg, with about 11 hours of supportive psychotherapy. (JAMA Psychiatry, 2021) The COMPASS trial gave one dose of a synthetic formulation at 25 mg, 10 mg or 1 mg. (New England Journal of Medicine, 2022) The alcohol trial gave psilocybin at weeks 4 and 8 inside 12 weeks of manualised psychotherapy. (JAMA Psychiatry, 2022) These are research protocols with medical screening, not guidance for anyone else.

What the drug was compared with

The control condition decides what a result can mean. The landmark trials used at least five kinds:

• Waiting list. In the Hopkins depression trial, 27 people were randomised to start treatment immediately or after an 8-week delay. (JAMA Psychiatry, 2021) Waiting is not a placebo, so it cannot separate the drug from the attention and hope that come with treatment.
• A very low dose of psilocybin. The COMPASS trial used 1 mg as its comparator. (New England Journal of Medicine, 2022) The Hopkins cancer trial compared 1 or 3 mg/70 kg with 22 or 30 mg/70 kg, and gave instructions to participants and staff intended to minimise expectancy effects. (Journal of Psychopharmacology, 2016)
• An active placebo. Niacin causes flushing and diphenhydramine causes drowsiness, so both produce a noticeable effect. The NYU cancer trial and a 104-person depression trial used niacin; the alcohol trial used diphenhydramine. (Journal of Psychopharmacology, 2016; JAMA, 2023; JAMA Psychiatry, 2022)
• An existing medicine. Imperial College compared two 25 mg psilocybin doses plus daily placebo with two 1 mg doses plus six weeks of escitalopram. (New England Journal of Medicine, 2021)
• No control at all. The smoking pilot gave psilocybin to 15 smokers in an open-label design, so every participant knew what they received. (Journal of Psychopharmacology, 2014)

Crossover designs add another wrinkle. In the NYU cancer trial everyone eventually received psilocybin, and the comparison with niacin rests on the 7 weeks before the crossover. (Journal of Psychopharmacology, 2016) Later follow-up figures in a crossover trial have no untreated group to compare against.

The blinding problem and expectancy

A trial is "double-blind" when neither participants nor raters know who got what. With psilocybin that is hard to achieve, because a full dose is unmistakable. A 2021 review of psychedelic trials found they had generally not reported participants' expectations before the trial or whether blinding worked, and argued that effect sizes are likely overestimated as a result. (Expert Review of Clinical Pharmacology, 2021)

The same problem shaped the MDMA story. ICER, an independent drug-assessment body, concluded in June 2024 that "functional unblinding in the trials and additional concerns around trial design and conduct" left the public evidence on MDMA-assisted therapy insufficient to judge benefits against harms. (Institute for Clinical and Economic Review, 27 Jun 2024)

The evidence on expectancy is mixed rather than damning. In a reanalysis of the Imperial escitalopram trial, patients expected more from psilocybin, but their expectations predicted their response to escitalopram and not their response to psilocybin. (Psychological Medicine, 2024) One trial cannot settle it, and that is why newer trials are asked to measure expectancy and blinding directly. (Expert Review of Clinical Pharmacology, 2021)

Response, remission and the primary outcome

Read the primary outcome first: the one measure the trial was designed and powered to test. Everything else is secondary, and secondary results that are not corrected for multiple comparisons can mislead.

The escitalopram trial is the standard example. Its primary measure, change on a self-rated depression scale at week 6, did not differ significantly between groups (P = 0.17). Remission was 57% with psilocybin against 28% with escitalopram, but that was a secondary outcome without correction for multiple comparisons. (New England Journal of Medicine, 2021) Headlines quoted the remission figure; the trial's own conclusion was that it showed no significant difference.

Response usually means at least a 50% fall in a depression score; remission means the score falls below a set threshold, such as 10 or less on the MADRS in the COMPASS trial. (New England Journal of Medicine, 2022) Response is easier to reach. In the 104-person niacin-controlled trial, more people on psilocybin had a sustained response, but not sustained remission. (JAMA, 2023)

How long the effects lasted

Durability is where the evidence is thinnest. In the COMPASS trial, sustained response at 12 weeks was not clearly supported for the 25 mg group. (New England Journal of Medicine, 2022) In the escitalopram trial's 6-month follow-up, both groups had kept their improvement and depression scores did not differ significantly between them; the authors called the results descriptive. (eClinicalMedicine, 2024)

Longer follow-ups exist, but they lose their control groups. The Hopkins depression participants were followed to 12 months, when response was 75% and remission 58% among the 24 who had completed treatment. (Journal of Psychopharmacology, 2022) At five years, 67% of the 18 people followed up were in remission. (Journal of Psychedelic Studies, 2025) In the smoking pilot, 10 of 15 were confirmed abstinent at 12 months. (The American Journal of Drug and Alcohol Abuse, 2017) These are encouraging, but people who stay in a study for years are not a random sample, and other treatments in the meantime are not controlled.

Who was left out

Trial results apply to the people who were allowed in. The 104-person depression trial excluded anyone with a history of psychosis or mania, an active substance use disorder or active suicidal ideation with intent, and people on psychiatric medicines could join only after tapering off them. (JAMA, 2023) The Hopkins trial took only people not using antidepressants and with no history of psychotic disorder, serious suicide attempt or hospitalisation. (JAMA Psychiatry, 2021) The alcohol trial excluded major psychiatric and drug use disorders and recent hallucinogen use. (JAMA Psychiatry, 2022) The smoking pilot recruited only psychiatrically healthy smokers. (Journal of Psychopharmacology, 2014)

These exclusions follow the research safety guidelines. They recommend excluding people with a personal or family history of psychotic or other severe psychiatric disorders. (Journal of Psychopharmacology, 2008) They also mean the results say little about people with bipolar disorder, for whom a survey found new or worsening symptoms after psilocybin use. (Journal of Psychopharmacology, 2023) If you take a psychiatric medicine, do not stop it to qualify for anything; talk to your prescriber, and see the interaction checker.

Adverse events

Controlled trials report side effects carefully, and they are common. In the 233-person COMPASS trial, 77% had an adverse event, mainly headache, nausea and dizziness, and suicidal ideation or behaviour or self-injury occurred in all dose groups. (New England Journal of Medicine, 2022) The niacin-controlled trial had no serious treatment-emergent adverse events but higher rates of overall and severe adverse events with psilocybin. (JAMA, 2023) Psilocybin also raises heart rate and blood pressure during the session. (National Institute on Drug Abuse (NIH); Psychiatry Research, 2024; EMCDDA (now EUDA), via Internet Archive snapshot 2024-06-18)

A 2024 meta-analysis counted nine serious adverse events across more than 1,000 administrations in controlled settings, and retreat settings do not necessarily reproduce that screening and monitoring. (Psychiatry Research, 2024)

Why phase 2 results often shrink

The trials above are mostly phase 2: small, often at one or a few specialist centres, with selected participants and highly trained staff. Several features push early estimates upwards. Weak blinding and high expectations are one. (Expert Review of Clinical Pharmacology, 2021) Small samples are another: the Hopkins trial's effect size of about 2.5 came from 24 completers against a waiting list, while the 233-person COMPASS trial found a 6.6-point difference on a 60-point scale against 1 mg, and no significant benefit at 10 mg. (JAMA Psychiatry, 2021; New England Journal of Medicine, 2022) The COMPASS authors themselves concluded that larger and longer trials, including comparison with existing treatments, are needed. (New England Journal of Medicine, 2022)

When the next result is announced, ask five questions: what was the control, was the primary outcome met, did the trial report how many participants guessed their group, how long was follow-up with a comparison group still in place, and who was excluded. The research library links each study with its design and limits.