Non-existent ClinicalTrials.gov identifiers in abstracts indexed by PubMed

Prospective registration plays an important role in ensuring the transparency and reliability of clinical trials. Preregistration of clinical trials has been required by the ICMJE since 2005 and mandated by law for most clinical trial types since 2007. It is one of the roles of peer reviewers of a clinical trial publication to confirm that there is concordance between the registry entry and the submitted publication. On October 22, 2019, abstracts for all articles indexed by PubMed with publication type "Clinical Trial" and a publication date after January 1, 2003 were downloaded. Clinical trial registry identifiers were automatically extracted and tested for the existence of a corresponding entry on ClinicalTrials.gov. Among 38,001 published clinical trial registry numbers, 215 (0.6%) do not correspond to a legitimate clinical trial registry entry. While there is a small proportion of non-existent NCT numbers in our sample, even a single non-existent NCT number in a publication represents a failure on the part of journals who publish clinical trials to systematically ensure that reviewers always check clinical trial registry entries for concordance with the text submitted for publication. These results cast doubt on how frequently editors and reviewers evaluate clinical trial reports in light of their corresponding registry entries.