Do trials quietly cut their enrollment target?

Whether trials that lower their planned enrollment after starting are less likely to report results. An agenda item: the registry’s change history is not programmatically reachable.

ImportantStatus: agenda. The data source is not reachable, and that is recorded rather than worked around.

On 2026-09-22 the ClinicalTrials.gov history endpoint returned HTTP 403 for two tested requests, with and without an identifying User-Agent. The AACT database, which offers archived copies of the registry, needs an account and a database restore; the URL tried for its snapshot listing returned 404. Neither was pursued further.

The question

A trial registration carries a planned enrollment that can be revised. A trial that starts, struggles to recruit, and lowers its target has not failed in any registry field: the final record shows the new target as if it had always been the plan. Whether trials that revise downward after starting are less likely to post results, and whether they report their shortfall when they do, needs the history of the record, not its current state.

Novelty

Log key P30.

corpus regex: (anticipated|target|planned|estimated) (enrollment|sample size).{0,80}
              (change|revis|reduc|modif)
PubMed:       ("sample size"[tiab] OR enrollment[tiab] OR enrolment[tiab])
              AND ("ClinicalTrials.gov"[tiab] OR registry[tiab]) AND ("history of changes"[tiab]
              OR "archived versions"[tiab] OR "registry changes"[tiab]
              OR "changes to registration"[tiab] OR "registration changes"[tiab])

One PubMed record: a 2011 study of completeness and changes in registered data in intensive care trials (PMID 21957485). The question is open. The data are the obstacle.

What would make it runnable

Programmatic access to registry change history at scale, by one of:

1. an AACT account and a restore of monthly archived snapshots, compared version against version;
2. written permission to retrieve the public History of Changes pages at a rate the registry accepts;
3. a data release of registry histories from the registry itself.

Any of the three turns this into a design. The hypothesis, sample and analysis would follow the structure of P28, which uses the same registry fields that are reachable.