Raison et al., 2023
50 psilocybin and 54 niacin participants in the safety population. No post-dose serious treatment-emergent AE. Four participants had severe related AEs through day 9; one had panic attack and paranoia.
Evidence pilot · framework v0.4
The first numbers survived only by staying narrow.
Two direct trials support one rare-event estimate. Most other prespecified outcomes remain unestimated because the reports use incompatible windows, event units, or causal definitions.
Direct2 trials
Raison 2023 and Yngwe 2026 closely match the fixed 25 mg MDD scenario.
Secondary0 / 67
Investigator-attributed drug-related serious adverse events after psilocybin. Secondary outcome under v0.8. The primary serious-event outcome is all-cause serious adverse events, which is not yet estimable.
Uncertain0–53.6
95% exact interval per 1,000 administrations—not evidence of zero risk.
Blocked3 outcomes
Exact severe-psychiatric, all-cause SAE, and rescue/observation estimates cannot yet be calculated.
01 · Releasable estimate — secondary outcome
No drug-related serious event was observed; precision is still poor.
Across the direct MDD trials, 0 of 67 psilocybin participants and 0 of 72 niacin participants had an investigator-attributed drug-related serious adverse event. Both studies followed participants longer than seven days, so zero over complete follow-up implies zero in the prespecified seven-day window.
Psilocybin · secondary outcome
0 / 1,000
95% exact binomial interval: 0 to 53.6 per 1,000. The upper limit is still compatible with about one event per 19 administrations.
0100 events / 1,000
Niacin comparator
0 / 1,000
95% exact binomial interval: 0 to 49.9 per 1,000. Risk-difference interval: approximately −50.7 to 54.2 per 1,000.
0100 events / 1,000
Interpretation boundary: this applies only to screened adults with MDD receiving a pharmaceutical 25 mg dose in a highly supported clinical setting. It is not a recreational-use estimate, a population-wide safety claim, or proof that risk is zero. The provisional certainty assessment is reported separately.
02 · Direct evidence
Two studies match; they still do not report safety the same way.
Yngwe et al., 2026
17 psilocybin and 18 niacin participants. No drug-related SAE across one year. Two psilocybin participants reported persistent severe anxiety requiring medical attention, but exact onset is not reported.
| Study | Directness | Useful safety quantity | Compatibility problem |
|---|---|---|---|
| Raison 2023 | Direct MDD / fixed 25 mg / niacin | Participant-level serious and severe AEs; day 9 and day 43 summaries | The severe-event window is day 9, not day 7; rescue/observation data are incomplete. |
| Yngwe 2026 | Direct MDD / fixed 25 mg / niacin | Dosing-day events, full-follow-up related SAEs, persistent anxiety signal | Some rows count events rather than participants; exact anxiety onset and seven-day counts are absent. |
| Goodwin 2022 | Indirect TRD / 1 mg comparator | Large safety dataset and suicidality signal | Different population and comparator; kept outside direct pooling. |
| Mertens 2026 | Indirect TRD / repeated-dose architecture | Systematic acute AE capture and serious reactions | Administration-level denominators and later second dose complicate participant-level synthesis. |
03 · Honest missingness
The protocol now proves when not to calculate.
All-cause SAE, 0–7 days
Procedure-related events and unrelated illnesses are not reported in an exact common seven-day participant window.
Severe psychiatric event
One study reports panic/paranoia through day 9; the other reports persistent anxiety without exact onset.
Rescue or extended observation
Complete arm-level participant counts are absent. The correct next action is author contact, not imputation.
04 · Preliminary bias review
Both direct studies are “some concerns” for this exact result.
Randomization and follow-up are strengths. The limitations are functional unblinding, subjective attribution of causality, small samples, and safety summaries that do not map cleanly onto the prespecified windows.
| Domain | Raison 2023 | Yngwe 2026 | Why it matters |
|---|---|---|---|
| Randomization | Low | Low | Supports arm comparisons. |
| Deviations / blinding | Some concerns | Some concerns | Participants and clinicians could often infer assignment, affecting reporting and rescue decisions. |
| Missing data | Low | Low for short term | Denominators are largely available. |
| Outcome measurement | Some concerns | Some concerns | Relatedness is investigator judged; event-vs-participant units differ. |
| Selective reporting | Some concerns | Some concerns | Exact 24-hour and seven-day outputs are not consistently published. |
These are author-pilot judgments only. A real review requires two independent result-level RoB 2 assessments and recorded adjudication.
05 · Next gate
The next work is replication, not another score.
- Run the full database and registry search with an information specialist.
- Have two reviewers independently screen and extract every record.
- Request exact-window counts for severe psychiatric events, rescue medication, and observation extensions.
- Adjudicate RoB 2 judgments before updating any synthesis.
- Freeze a public review dataset and publish every excluded record with its reason.
06 · Audit files