Psilocybin: the 2022 depression trial.

What was studied?

This phase 2 double-blind trial randomly assigned screened adults with treatment-resistant depression to one dose of a proprietary synthetic formulation—25 mg, 10 mg or 1 mg as the control—with psychological support. The primary endpoint was change from baseline to week 3 in the Montgomery–Åsberg Depression Rating Scale (MADRS), a 0–60 clinician-rated scale in which higher scores indicate more severe depression. The control was a low dose, not an inert placebo, so subjective effects could still affect blinding. The support package and screening are part of the intervention as studied.13

The measured finding

There were 79 participants in the 25-mg group, 75 in the 10-mg group and 79 in the 1-mg group, with baseline mean MADRS scores around 32 or 33. The journal report gives least-squares mean MADRS changes at week 3 of −12.0, −7.9 and −5.4 respectively. The 25-mg versus 1-mg difference was −6.6 points (95% confidence interval −10.2 to −2.9; P<0.001). The 10-mg difference was −2.5 points (95% confidence interval −6.2 to 1.2; P=0.18).

The ClinicalTrials.gov results field labels a different set of figures (−12.0, −8.9 and −6.8) as means with standard deviations. The journal numbers are least-squares means from a statistical model. These are different descriptions and should not be interchanged. The registry field does not fully explain how its group values were calculated, so the remaining difference cannot be reconciled conclusively from these records alone. These are group estimates at one time point, not personal predictions or percentages cured.158

Why the timeline matters

Response and remission were assessed at week 3, while sustained response required response at week 3 and subsequent visits through week 12. The published report says that the 25-mg response and remission findings generally supported the primary result at week 3, but sustained response at 12 weeks was not similarly supportive. A hierarchical testing procedure stopped confirmatory significance testing after the nonsignificant 10-mg primary comparison, so the secondary findings should be described as directional rather than statistically established. An early rating-scale difference is therefore not the same claim as durable recovery. The study does not tell us how benefit and harm would look after longer follow-up or repeated treatment.1

What the registry adds

The ClinicalTrials.gov results record reports 233 participants: 79 in the 25-mg group, 75 in the 10-mg group and 79 in the 1-mg control, with 74/66/69 completing. Its primary outcome uses a full-analysis-set definition requiring study drug and at least one post-dose efficacy assessment. The registry reports week-3 group means with standard deviations of −12.0, −8.9 and −6.8 points. Separately, its mixed-model analysis reports a least-squares mean difference of −6.6 points for 25 mg versus 1 mg (95% CI −10.2 to −2.9); this modelled contrast should not be calculated by subtracting the displayed group means. A per-protocol sensitivity analysis uses smaller denominators, and the registry states that MAR and MNAR imputations were used for missing data. The sponsor’s final analysis plan defines the primary question hypothetically. It asks what the treatment comparison would be if participants could not start a new depression treatment after dosing. It prespecified imputation for missing outcomes and sensitivity analyses, so the result is a model-based estimate under those assumptions, not a direct observation of every participant under that hypothetical condition.58

Harms were observed

Adverse events occurred in 179 of 233 participants (77%) and included headache, nausea and dizziness. Suicidal ideation, suicidal behaviour or self-injury occurred in all dose groups. Reporting an event does not by itself establish causation, and this selected outpatient sample excluded people judged at clinically significant suicide risk. The findings show why a benefit headline must be read beside harms, screening and follow-up. A phase 2 sample cannot settle uncommon or delayed risks.15

What the design can and cannot answer

The trial did not compare psilocybin with an established antidepressant or a complete psychological-treatment package. The report did not establish that participants or staff correctly guessed assignment, and subjective effects make functional unblinding a material concern. The secondary analysis concerns the same trial population and is not an independent replication. The study also does not establish that mushroom products with variable composition have the same effects as the defined formulation, or that a result in screened adults with treatment-resistant depression applies to people excluded by the protocol. Because the lowest-dose control may have produced fewer noticeable subjective effects, the estimated contrast may include expectancy and treatment-contact differences alongside pharmacological effects.23

A result is not a treatment recommendation

The trial’s narrow conclusion is about a prespecified comparison at a prespecified time in a defined research population. It does not supply a dose recommendation, identify who should receive psilocybin, or establish that the same result would occur without the trial’s screening and psychological support. The study was funded by COMPASS Pathfinder; disclosure helps readers understand the evidence context without treating funding alone as proof that the findings are wrong.1

What would strengthen the evidence?

A stronger evidence base would include adequately powered comparisons with credible active controls or established care, prespecified handling of missing data, complete adverse-event reporting and follow-up long enough to examine persistence and delayed harms. Replication by independent groups and transparent access to protocols, analysis plans and supplements would also help readers judge whether the result is robust. Until then, this study supports a carefully bounded research conclusion rather than a settled clinical one.3

Use this study to ask better questions

What did treatment resistance mean in the protocol, and who was excluded? What evidence is there that blinding held despite subjective effects? Are the MADRS changes clinically meaningful? What do participant flow, withdrawals, missing-data methods and detailed adverse-event tables show? Was the 12-week analysis adequately powered to address durability? The official publisher report and registry results were inspected for this draft. The sponsor-submitted protocol and statistical analysis plan were inspected; the journal supplementary appendix still needs appraisal, and human scientific review remains pending. This is a historical research-literacy example, not a comprehensive evidence review.153678