Methods & provenance

Trials are matched to a compound by their structured intervention name, not by a free-text search. We take accuracy over reach: a trial we can’t confirm from the intervention field is left out rather than shown under the wrong compound.

• Abbreviation collisions are filtered. “DMT” also means disease-modifying therapy (multiple-sclerosis trials); “LSD” also means lysosomal storage disease. Those are excluded.
• Ketamine is scoped to psychiatric indications (depression, PTSD, suicidality, substance use, bipolar). Unscoped, ketamine returns mostly anesthesia and pain trials, which aren’t what this resource is about.
• Ayahuasca is tracked separately from DMT. Although ayahuasca contains DMT, its oral route and MAOI combination give it a different safety profile, so its trials aren’t folded into DMT.

A trial that matches more than one compound is kept once and tagged with each.

• We store the raw counts each trial reported — participants affected and the number at risk — per study arm, for serious events, non-serious events, and deaths. We never store a bare percentage.
• A rate is computed at display time as affected / analysed. If a trial posted no usable denominator, the figure reads “Not reported” — never a fabricated 0%.
• Serious and non-serious events are kept separate, exactly as the registry records them.
• Every rate is shown next to the sample size it rests on, with a link back to the source trial so you can check it.

These two counts are different, and the difference is not an error. Enrolled is how many people entered a trial. Analysed for safety (shown as n) is how many were included in a given safety table — often fewer, because of dropouts, per-arm splits, or a safety population defined differently from the enrolled one.

Rates on this site always divide by the analysed count, because dividing events by the larger enrolled number would understate the rate. When a trial’s enrolled and analysed counts differ, both are shown so the gap is visible rather than hidden.

For each compound we show how many completed trials have posted a structured results section. This is neutral context — it tells you how much evidence a number rests on, which matters in a field that has drawn large conclusions from small studies. It is not a judgment, and we never single out an individual trial or sponsor for not reporting.

Every figure here is auto-extracted directly from the registry and has not been independently verified by a person. We make no “verified” claim. If this project later works with someone who can systematically check figures against their sources, a human-verification layer can be added — but only then, and labeled for exactly what it is.

We also don’t harmonize event terms into a controlled vocabulary yet, don’t normalize dose or setting across trials, and don’t pool rates across trials — raw rates from different trials aren’t a like-for-like comparison, and presenting them as one would mislead.

If a number here doesn’t match its source trial, or a trial looks miscategorised, please say so and it will be checked against the registry and fixed. Accuracy is the whole point of this resource, and corrections are welcome from anyone.