Reading the Psilocybin for Depression Trials Without the Hype

The short version

The most cited modern trial of psilocybin for depression (Goodwin and colleagues, New England Journal of Medicine, 2022) was a phase 2 dose-finding study, not a verdict on whether the drug works. A single 25 mg dose lowered depression scores more than a 1 mg comparator at week 3 in people with treatment-resistant depression, and that difference was statistically real. But the trial could not cleanly separate the molecule from what people expected it to do, the comparator was not an inert placebo, and a better score at three weeks is not durable remission. Read the design before you read the headline.

What the trial actually tested

The study randomized 233 adults with treatment-resistant depression at 22 sites across 10 countries to a single dose of COMP360, a synthetic formulation of psilocybin developed by COMPASS Pathways, given alongside psychological support. Participants received one of three doses: 25 mg (79 people), 10 mg (75 people), or 1 mg (79 people). The 1 mg arm was built to serve as a control. The primary endpoint was the change from baseline in the Montgomery-Asberg Depression Rating Scale (MADRS), a clinician-administered depression score, measured at week 3.

The headline result is genuine. The least-squares mean drop in MADRS was 12.0 points in the 25 mg group, 7.9 in the 10 mg group, and 5.4 in the 1 mg group. The difference between the 25 mg and 1 mg groups was about 6.6 points, favoring the high dose (95% confidence interval roughly 2.9 to 10.2 points, P<0.001). Response, defined as at least a 50% reduction in MADRS at week 3, occurred in 37% of the 25 mg group versus 18% at 1 mg. Remission reached 29% versus 8%. The 10 mg dose did not separate meaningfully from 1 mg. So the signal is dose-dependent and clear. The harder question is what it means.

The blinding problem

The word that should follow you through every psychedelic trial is expectancy. A 25 mg dose of psilocybin produces unmistakable perceptual effects. A 1 mg dose usually does not. That means participants can frequently guess which group they are in, and so can the staff supporting them. This is called functional unblinding, and it erodes the core purpose of a placebo-controlled design.

Why does that matter here? Depression outcomes are sensitive to belief. If you are convinced you received an active psychedelic in a trial you volunteered for and traveled to attend, that conviction alone can lift mood and shape how you answer a rating scale. MADRS is administered by a clinician, but it is anchored in what the patient reports, and raters who are not fully blind can be swayed in either direction. None of this proves the drug did nothing. It means the gap between 25 mg and 1 mg mixes a pharmacological effect with an expectancy effect, and this trial cannot tell you the ratio between them.

The investigators have argued that the variability of the psychedelic experience, especially between the 10 mg and 25 mg doses, may have preserved the blind better than critics assume. That is a fair point and an unsettled one. It is exactly the kind of question a low-dose comparator cannot fully resolve.

What a response at week 3 means

A response rate is a snapshot. The primary endpoint captured MADRS at a single early timepoint, three weeks after one dose. Depression is a relapsing and remitting condition, so the useful clinical question is not whether a score improved by week 3 but whether it stayed improved. On that measure the trial is more sobering: the advantage of the 25 mg dose over the 1 mg dose seen at three weeks was not maintained through week 12. A meaningful share of early responders had lost that response by the end of the study.

This is not a knock on the molecule. It is a reminder that "37% responded" describes one moment, and single-timepoint response rates tend to read as more impressive than durable remission rates. When a future study or a press release leads with a response figure, the first questions to ask are simple: measured when, and did it last.

The safety line you should not skip

Adverse events occurred in 77% of participants (179 of 233), and most were the expected transient effects: headache, nausea, and dizziness around dosing day. The signal that deserves attention is that suicidal ideation, suicidal behavior, or self-injury was reported across all three dose groups, with instances of suicidal behavior among 25 mg participants, generally in people who had not responded and after the three-week mark. In a population defined by treatment resistance this is not automatically drug-caused, but it is why the field treats these compounds as serious interventions requiring structured monitoring, not wellness experiences.

How to read the next wave

Phase 3 trials are the ones that carry regulatory weight, and the questions to bring to them are the ones this phase 2 study raised. Is there a genuine active comparator rather than a low dose of the same drug? Did the investigators measure and report how well the blind actually held? Is the primary endpoint durable, assessed weeks or months out, rather than a single early response rate? Answer those three, and you can read past the headline whichever way it points. This article is educational and not medical advice; treatment-resistant depression is a serious condition that warrants individualized care.