NIH project titles that name a registry ingredient are in the funding meter. FDA recalls and warning letters that say dietary supplement are on the enforcement log. Registry trial floors, with benefit kept apart from weak benefit, are the dose cards. No harm tag ≠ safe.

This check

Cited retractions0
Sources with identifiers540 / 65882.1% coverage
Unique PMIDs459
Unique DOIs498

Unique PMIDs and unique DOIs count identifiers cited on Supplement Brief. The positive control is not included. Sources with identifiers counts guide source lines that carry a PMID or a DOI.

Positive control

PMID 33612888

Known retracted publication. Used to confirm the matcher. Excluded from the counts above.

  • Retraction WatchRetractionRecord 74354
  • PubMedRetracted PublicationPublication type
  • Crossrefretractionupdate-to

This paper is not cited on Supplement Brief.

Coverage limitations

118 of 658 source lines have neither a PMID nor a DOI, so they are outside this check.

  • 74Regulators and registries
  • 6Standards and labs
  • 2PMC pages without an ID
  • 36Company and news pages

When a paper we cite is flagged

The same update does three things, and it ships before December 14, 2026.

  1. The retracted paper is removed or replaced with a source that still stands.
  2. The guide says the paper was retracted.
  3. The corrections log gets a dated entry for that guide.

verify-citations blocks a deploy while the retracted identifier is still cited.

Checks

This checkNovember 14, 2026
Next scheduled checkDecember 14, 2026

One check a month.

CheckRetractions we citedUnique PMIDsFile hash
November 10, 202604557b912811777e
November 4, 20260411e5351ee51ab3

Methodology and reproducibility

Matches in our citations

  • Retraction Watch0
  • PubMed, Retracted Publication0 of 459
  • Crossref retraction updates0

Crossref requests

  • DOIs in the bibliography498
  • Sent to Crossref, including the control499
  • Cited DOIs that resolved496
  • Control DOIResolved
  • Cited DOIs that did not resolve2
  • Corrections and other updates, not retractions8

PubMed types read from the snapshot of November 14, 2026.

Reproduce this check

SHA-256
4d0509f648ec6dc7f7de43a2260943ed15c56b239d882d73f468f4f40ef917dd
Identifiers we checked
Download JSON
Positive control
Record 74354. Not in the download.

How the join works

  • Retraction Watch

    68,628 of 73,822 records have an original PMID or DOI. A cited PMID matches OriginalPaperPubMedID. A cited DOI matches OriginalPaperDOI.

  • PubMed

    Publication type Retracted Publication, from the esummary field pubtype. The control has to carry that type or this check is discarded.

  • Crossref

    Works API field update-to. A retraction is type retraction or withdrawal.

Retraction Watch data is distributed by Crossref, and some records may overlap with Crossref's own update metadata. These checks provide multiple matching paths, not three fully independent confirmations.

We also check recorded expressions of concern and other publication updates where metadata is available. Their absence from this snapshot does not establish that no such notices exist.