What we mean when we say a drug "works"

Every few months a headline announces that a psychedelic compound has produced "unprecedented" results for depression, addiction or trauma. Some of those results are real and important. They are also frequently misunderstood — by readers, by reporters, and sometimes by the people running the studies.

This is not an argument that the research is weak. Much of it is careful, and the field has matured quickly. It is an argument that promising is a specific, limited claim, and that learning to hear it precisely is the most useful skill a curious person can bring to this subject.

A result is a sentence with hidden clauses

When a trial reports that a single dose "significantly reduced depression at six weeks", several quieter facts travel underneath that sentence: how many people were studied, who they were, what the comparison group received, how long anyone was followed, how the outcome was measured, and how large the difference actually was.

A number without its denominator is a rumour with a decimal point.

Take the largest programme in the field. Compass Pathways ran two Phase 3 trials of a synthetic psilocybin formulation for treatment-resistant depression. The first, with 258 participants, found that a single 25 mg dose beat placebo at six weeks by a mean of 3.6 points on the MADRS, a 60-point depression scale. The second, with 581 participants, found that two 25 mg doses beat a 1 mg comparison dose by 3.8 points. Both results were statistically strong. Both are also, in absolute terms, modest: a few points on a long scale, in the same neighbourhood as the average gap between a conventional antidepressant and placebo. That is not a criticism of the trials. It is what "significant" means, and it is a long way from "cured".

Smaller studies deserve even more caution. A 2021 trial comparing psilocybin to the antidepressant escitalopram enrolled 59 people and, on its main outcome, found no significant difference between the two. Early-phase studies exist to detect whether an effect is plausible, not to measure it accurately. The honest reading of an exciting early result is not "this works" but "this is worth the expense and risk of finding out properly".

The unblinding problem

Psychedelics present researchers with an awkward fact: people generally know whether they have taken one. That breaks the blind that ordinary drug trials rely on, and expectation is a powerful medicine in its own right. Good teams now design around this — low-dose active comparators, independent raters, careful measurement of what participants believed they received — but a study that ignores it should be read with that gap in mind.

This is not a theoretical worry. In 2024 the US Food and Drug Administration declined to approve MDMA-assisted therapy for post-traumatic stress disorder, despite two positive Phase 3 trials, and the difficulty of blinding was among the reasons its advisers gave. Regulators were not saying the treatment does not work. They were saying the trials could not yet prove how much of the effect was the drug.

"Durable" is a claim about a denominator

Durability is the whole question for a treatment given once or twice. So when a company reports that benefit "lasted six months", ask: among whom? In the Compass trials, the six-month durability data were reported for participants who had already responded at week six. That is a real and useful finding. It is also a different, narrower claim than "the treatment lasts six months", because it leaves out everyone for whom it did not work in the first place.

What good evidence will look like

Larger samples that include the people clinicians actually treat, not only the unusually well and unusually motivated.
Longer follow-up of everyone enrolled, not only responders.
Honest accounting of harms, including the difficult or destabilising experiences that averages tend to hide.
Results reported in absolute terms — points on a scale, proportion of people in remission — rather than only as "significant".

None of this is a reason for despair, and none of it is a reason for hype. The field is young and, for the first time, it has large trials to argue about. The most respectful thing anyone can do for it, and for the people waiting on it, is to describe it accurately while it grows up.

Sources and further reading

Compass Pathways, COMP005 topline results (June 2025), COMP006 topline results (February 2026) and COMP006 26-week results (July 2026), company announcements.
Goodwin GM et al. "Single-dose psilocybin for a treatment-resistant episode of major depression." New England Journal of Medicine, 2022.
Carhart-Harris R et al. "Trial of psilocybin versus escitalopram for depression." New England Journal of Medicine, 2021.
Raison CL et al. "Single-dose psilocybin treatment for major depressive disorder: a randomized clinical trial." JAMA, 2023.
Muthukumaraswamy SD, Forsyth A, Lumley T. "Blinding and expectancy confounds in psychedelic randomized controlled trials." Expert Review of Clinical Pharmacology, 2021.
Aday JS et al. "Great expectations: recommendations for improving the methodological rigor of psychedelic clinical trials." Psychopharmacology, 2022.
US FDA, Psychopharmacologic Drugs Advisory Committee meeting on midomafetamine (MDMA) for PTSD, June 2024; complete response letter reported August 2024.

This essay is general information, not medical advice. Trial results and regulatory status change; check current sources before relying on any figure here.