The result exists. You cannot read it yet.
When a clinical trial finishes, its sponsor submits the results to ClinicalTrials.gov. They are not published on submission. They enter a quality-control review, and become readable when they clear it. The law is satisfied at submission. Medicine is only served at posting.
The primary outcome is withheld because estimator_recovers failing. PREREGISTRATION.md requires all three gates to pass before any queue estimate is published.
A trial under review is indistinguishable from one that has reported nothing at all. The registry publishes nothing about it while it waits, and discloses the submission date only once it posts. Measured on 2026-08-30: the counts of records carrying a submit date, a QC date and a post date are identical, and none of 120 sampled completed-but-unposted trials exposed a pending submission.
So the number above cannot be counted. It has to be estimated, and the estimate has to be marked against dates whose answer has since been disclosed. Ours does not pass that test yet, so nothing is published.
What is known
Once a wait is over it is measurable, and 79,892 of them are. From mature submission cohorts (2008–2016):
The gates
Three checks were fixed in advance, before any estimator existed. All three must pass before the primary outcome is computed at all.
| Gate | Verdict | Failed | Checked |
|---|---|---|---|
| capture faithful | pass | 0 | 200 |
| census agrees | pass | 0 | 18 |
| estimator recovers | fail | 6 | 9 |
The build is red on purpose. A green badge while a gate fails would mean a finding had been computed after its own validation failed, which is the one outcome the preregistration exists to make impossible.