When the page still works, whose policy does it carry?
When clinical-trial sponsors move data requests to a new portal, their older pages can remain in circulation. In three historical ClinicalStudyDataRequest (CSDR) pages, the transition notice identified the expected sponsor, but the policy body matched a different company. A reader could reach a working page and still encounter the wrong sponsor's access conditions.
The study grew out of a route audit following 14 trial records across five sponsors. Tracking where older listings and request instructions led brought a second question into view: whether the policy attached to a sponsor page came from that sponsor. Archived source comparisons became the central evidence for the paper.
Three historical policy-body mismatches
The comparison uses 15 dated Common Crawl WARC responses. The repository lists the archive location and cryptographic hash for each source. Readers can retrieve the original pages and repeat the checks.
| Page label | Policy text matched | Paragraph match |
|---|---|---|
| Lilly | Boehringer Ingelheim | 27/27 |
| Bayer | Sanofi | 25/25 |
| ONO | UCB | 21/21 |
Why attribution matters
A transition notice may point to the right destination while a policy underneath it describes someone else's eligibility rules, release timing, or data scope. Those details matter to researchers planning a request. The cases suggest checking whether the named sponsor, policy text, and request route agree, even when the link itself works.
A record others can verify
The reproduction package contains the public WARC coordinates, hashes, comparison scripts, tests, and expected outputs. Its frozen Zenodo v1.0.0 archive supplies a version-specific citation. Original third-party web pages are retrieved from Common Crawl when the checks run.
The earlier trial-route ledger and exploratory page comparisons remain available as context. The 15 preserved WARC coordinates support the exact historical replays. The work documents policy attribution on selected pages, rather than a platform-wide failure rate or the outcome of any individual data request.
Publication record
Accepted poster paper P283, IEEE BIBM 2026. The conference proceedings record is forthcoming. The Zenodo DOI above identifies the reproducibility software; the IEEE paper link will be added when published.