This pass attacks the review before outsiders do. It separates causal comparison from rare-event surveillance, promotes all-cause serious events over subjective attribution, and prevents “not reported” from becoming zero.
Comparative effect and monitored-exposure safety are no longer one analysis.
All-cause serious events lead; attribution-based events are secondary.
A focused depression search plus broad safety-capture search await PRESS review.
Information specialist, statistician, PPI, reviewers, and registration.
Randomized trials answer whether assignment changes risk. A wider set of prospectively monitored administrations can help detect rare events, but cannot inherit the same causal interpretation.
Can say: whether assignment to psilocybin changed a prespecified harm relative to the declared control.
Can say: what was observed after the studied monitored administrations and how uncertain or incomplete that observation is.
Investigator attribution can be affected by unblinding and reporting practice.
A study that never reports an outcome cannot contribute a zero.
Both are stored independently, preserving the source definition.
Reviewers may already know prominent public trials and results.
Harms are inconsistently indexed. The protocol now adds a broad psilocybin-only safety search rather than requiring an adverse-event word in the title or abstract.
People affected by the outcomes review prolonged distress, functional impairment, rescue procedures, and public misinterpretation before registration.
The sparse-data model and pooling rules must be signed before independent extraction results are unmasked.
AI helped draft code and methods. It cannot count as an independent reviewer, information specialist, statistician, or accountable investigator.
| Priority | Outcome | Why it is separate | Required evidence |
|---|---|---|---|
| Primary | All-cause serious adverse event | Does not depend on subjective drug attribution. | Explicit seriousness definition, participant count, denominator, window. |
| Primary | Death; transfer/hospitalization; ICU/organ support | Objective consequences should not be hidden inside a composite. | Verified event and exact window. |
| Primary | Severe psychiatric/behavioral event requiring intervention | Captures clinically important panic, mania, psychosis, dissociation, or agitation. | Symptom definition plus rescue, observation, restraint, transfer, or admission. |
| Primary | Self-harm or clinically important worsening | Ideation, behavior, and intervention are not interchangeable. | Instrument, threshold, event type, and intervention. |
| Secondary | Drug-related serious event | Useful, but attribution is a judgment and may be biased. | Who attributed, using what rule, with what source data. |
| Secondary | Rescue medication and extended observation | Process outcomes can reveal important distress without meeting SAE definitions. | Medication, indication, duration, and planned-vs-actual observation. |
PRESS-review the focused and broad searches, including database translations.
Freeze sparse-data, zero-event, pooling, and missing-data rules before results are exposed.
Review relevance, omitted harms, interpretation boundaries, and public-facing language.
Assign two reviewers and an adjudicator with completed independence and conflict forms.
Time-stamp the human-approved protocol, then run, archive, deduplicate, and hash the registered corpus.
It was built around randomized depression candidates. Product B requires a new broad safety-capture search. The project will not patch that gap by hand-picking a few extra favorable or alarming studies.
Cochrane Chapter 19: Adverse effects; PRISMA-Harms; CONSORT Harms 2022; PRESS 2015; and the official RoB 2 resources.