A thousand finished trials. Four hundred and one reported back.
A structured sample of completed Phase 3 studies from the public registry, read into a grid space and written onto the board — who ran them, how large they really are, and how many posted results where the law says results belong.
This page is the exhibit. The space behind it is the product.
How big is a Phase 3 trial, actually?
Enrollment across the sample. The public image of a Phase 3 trial is tens of thousands of patients; the median in this sample is 230, and a quarter enroll fewer than a hundred. The long tail to the right is real but rare — and every bucket is countable, tap it.
Who runs them, and what they chase.
The ten largest in the sample.
How this was made.
The registry's own API, pulled on a daily schedule — a defined sample (completed, Phase 3), stated as a sample, never presented as the whole registry.
Each study lands as a record keeping its NCT ID — the registry's own receipt — at its own address.
The space's AI reads the records and writes the exhibits. A person chose the questions; the machine drew the answers.
Counts and structure only — nothing here is medical advice, efficacy analysis, or a judgment on any study.