A systematic review published on October 20 in JAMA Network Open set psilocybin’s clinical trial record beside its news coverage, condition by condition. Its sharpest contrast is PTSD: 74 of 172 news articles named the condition, and the review found no peer-reviewed psilocybin trial for it.
That reads as a story about media running ahead of science. The more useful finding is about timing. By the time the review’s trial search closed, Compass Pathways had already reported results from a psilocybin PTSD study, in an SEC filing. A literature search counted that evidence as zero. News coverage, meanwhile, made PTSD look further along than any controlled data showed. Neither view, on its own, described where the indication stood.
What the review covered
The review paired 70 psilocybin trial publications, from January 2014 to October 10, 2024, with 172 English-language online news articles from the Factiva database, 95 percent of them written for a general audience. Press releases, market reports, wire copy, editorials, broadcast and social media fell outside it. Its main text gives the news search an end date of April 22, 2025; its supplement gives October 10, 2024. The paper does not reconcile the two.
It describes a window that has closed. Several conditions in its media sample have produced trial results since.
Evidence does not arrive in one place
A drug’s evidence reaches the public through several channels, each answering a different question on its own clock. A peer-reviewed paper shows that reviewers accepted the analysis, often long after the data exist. A company filing shows what the sponsor has disclosed, usually as top-line numbers soon after a readout, with no independent review. A registry shows design and status, and sometimes results. A preprint shows the analysis before review. The JAMA review searched the first, a reasonable choice for a study of the literature, which means it measured an indication’s position in the literature, not everywhere it was public.
PTSD: visible, disclosed, published, still uncontrolled
News articles published January 2014 to April 22, 2025, as stated in the paper's main text; its supplement lists October 10, 2024. The trial search ended October 10, 2024. These are shares of one bounded sample of English-language online news, not of media coverage generally, and they do not describe the evidence available today. An article can name several conditions and make several claims.
Source: Madden K et al., "Clinical Evidence on Psilocybin and Its Representation in Media Reporting: A Systematic Review," JAMA Network Open, October 20, 2026, Tables 2 and 3.
Start with visibility. The review coded separately which conditions each article named and which efficacy claims it made, and for every condition more articles named it than claimed psilocybin worked for it. In proportion, the gap was widest for PTSD: 74 articles named the condition, and 12, about one in six, claimed psilocybin reduced its symptoms. The review does not say what the other 62 did with PTSD. It records that 62 articles in the full sample discussed MDMA, the drug most associated with psychedelic PTSD research, but does not cross-tabulate the two. Visibility, in other words, was not the same as an efficacy claim, and most of it was not one.
Then disclosure. Compass reported 24-hour safety data from a 22-person open-label PTSD study in December 2023. On May 8, 2024, it filed a Form 8-K with top-line efficacy results: on the CAPS-5, a clinician-rated PTSD scale, scores fell about 30 points from a baseline of 47.5 by week 4 and stayed there at week 12. There was no comparison group. That was inside the review’s trial window, and the review, which searched only the published literature, did not count it. Whether any of the 12 articles making efficacy claims drew on it, the review did not record.
Then publication. The peer-reviewed paper appeared online in August 2025, about 15 months after the filing. Its abstract said psilocybin “has not yet been investigated in PTSD,” a statement about the published literature, since the company had announced the study’s results the year before. An Ohio State study of 12 veterans with severe, treatment-resistant PTSD, published in July 2026, reported that 75 percent were in remission four weeks after their second dose, also with no control group.
Controlled evidence has not arrived. Compass’s randomized Phase 2b/3 PTSD trial, Redefine, is recruiting an estimated 300 participants, with primary completion estimated in the registry for November 2028.
Put together, the sequence gives three different pictures of the same indication. A literature search through October 2024 showed nothing. News coverage showed PTSD as one of the most discussed uses. The record across channels showed one company with open-label data in hand and a controlled trial still to be run. The third picture is the useful one, and it is only visible if the channels are read together.
Publication counts are not maturity
The review’s trial side needs the same care. Its 70 entries are publications, not 70 separate trials, and a number are secondary analyses or conference abstracts drawn from the same studies. Six of the 47 depression entries name in their titles the same comparison of psilocybin with escitalopram, one trial run at Imperial College London. The condition assignments are also loose. The four papers the review cites for its anxiety trials are the same four cluster headache and migraine studies it cites for headache, so its anxiety count appears to rest on a citation error. The one study it cites for obsessive-compulsive disorder is a Swiss analysis of 28 patients treated with LSD or psilocybin, and its outcome measure was depressive symptoms.
None of this reverses the review’s central contrast. Depression dominated the trial literature of the period, and no psilocybin PTSD trial had been published. But a count of publications measures how much has been written about a limited number of trials. An indication that generates many secondary analyses from one study looks more mature by that count than it is.
Safety shows the same unit problem. The review found 31 trial publications (44 percent) reporting safety outcomes and 4 (6 percent) assessing long-term safety, against news articles that mentioned safety 30 percent of the time and made efficacy claims 73 percent of the time. Those are publications, not trials, and some coded as silent on safety come from trials that reported adverse events elsewhere. What holds is narrower: long-term safety data were scarce in the period.
The label is not the population
Anxiety shows a different trap. The period’s strongest anxiety evidence came from two 2016 randomized trials in people with life-threatening cancer, which reported large, lasting falls in anxiety and depression. The review counts neither as an anxiety trial; by its citations, the Johns Hopkins trial sits under depression and the NYU trial under both depression and cancer-related distress. A 2025 phase 2b trial in Melbourne randomized 35 adults with life-threatening illness to psilocybin or niacin and reported greater falls in depression and state anxiety at six to seven weeks. A PubMed title search finds no published psilocybin trial in generalized anxiety disorder in the review’s window.
So the anxiety evidence is small in trial count, not in reported effect size, though small trials tend to report larger effects, and their active placebos face the same expectancy question as every psilocybin trial. It also comes almost entirely from people facing serious illness. Results in cancer-related anxiety do not establish psilocybin for anxiety disorders in people without a serious illness. That has not been tested with comparable rigor. An indication label and the population studied are separate questions, and evidence for the label has to come from trials in the population it describes.
Four indications, four different states
Treatment-resistant depression. The most-studied indication, named by 80 percent of sampled articles. Compass reports that its two Phase 3 trials met their primary endpoints: a 3.6-point difference from placebo on the 60-point MADRS with a single dose (June 2025), and 3.8 points between two 25 mg doses and two 1 mg doses (February 2026). Those results have come through company releases and SEC filings; no peer-reviewed paper from either trial has appeared so far. A rolling NDA submission is under way, and Compass expects to complete it in the fourth quarter of 2026, after which FDA review would change the picture.
PTSD. Named by 43 percent of sampled articles, a condition long associated with MDMA research. No controlled evidence; two open-label studies, of 22 and 12 people. The first data came through an SEC filing in May 2024, then journals in 2025 and 2026. What would change the picture is Redefine, the randomized trial of an estimated 300 participants with primary completion estimated for November 2028.
OCD. Named by 7 percent of sampled articles. Three small randomized trials, with at least 82 people randomized, none able to rule out that participants knew what they took. The results have appeared in a registry, in preprints and in one journal paper. Peer-reviewed versions of the Yale and Johns Hopkins results would change the picture, as would a Yale repeated-dose trial that has not posted results.
Anxiety in serious illness. Anxiety was named by 42 percent of sampled articles. Randomized trials exist in cancer (2016) and life-threatening illness (Melbourne, 2025), published in journals. What is missing is a controlled trial in people with an anxiety disorder alone.
Depression is the mature comparison. Its evidence moved through two randomized Phase 3 trials, so far reported by the company, to a rolling regulatory submission. PTSD was named about as often as anxiety and addiction in the review’s sample, and about half as often as depression, and has not yet produced a controlled result. OCD has randomized data but designs that cannot separate the drug from expectation. Anxiety has controlled evidence in a population narrower than the label suggests.
Reading an indication’s maturity
The review’s two samples point to what a serious reader should not infer maturity from: how often an indication appears in the news, how many publications mention it, how many secondary analyses one trial has produced, a single company’s top-line disclosure, or its presence in a systematic review.
What does carry weight is narrower. Has a controlled trial reported, with what design, comparator and blinding? Has the analysis been peer reviewed, or does it rest on a filing, a registry entry or a preprint? Who sponsors the program, and what regulatory stage has it reached? Where did the first data appear, and when? And which registered trial or filing would change the picture next?
The channels do not substitute for each other. A filing is not peer review, and a journal paper can arrive long after the data were public. For PTSD, the literature alone showed nothing and the news made it look advanced. Read together, the record showed early open-label data from one sponsor and a controlled test still years away.