Systematic Evaluation of Reference Authenticity in PubMed Central Papers

A report published in The Lancet has detailed the first systematic evaluation of the authenticity of references used in biomedical literature. The research, conducted by Topaz and colleagues, addresses the issue of fabricated references and the resulting threat to editorial integrity within scientific publishing.

The study utilized a massive dataset to examine the reliability of citations in the biomedical field. The researchers analyzed 2,471,758 papers sourced from the Open Access platform of PubMed Central that were published between January 1, 2023, and February 18, 2026.

During the initial phase of the study, the authors identified 125,615,773 references across the sampled papers. To conduct a precise evaluation, the team narrowed their focus to references that included a PubMed ID.

This subset represented 77% of the total references identified, totaling 97.1 million citations. These were the references that underwent further verification to determine if they were authentic or fabricated.

To ensure that the analysis did not misidentify legitimate papers as fake, the researchers implemented several filters. These filters were specifically designed to account for formatting discrepancies, such as incorrect titles or spelling errors in the names of authors.

Following the filtering process, the remaining references were verified against four separate databases to confirm their existence:

• PubMed
• Crossref
• OpenAlex
• Google Scholar

The use of multiple verification sources allowed the authors to perform a systematic evaluation of the authenticity of references on a scale previously unseen in biomedical research.

The integrity of references is fundamental to medical science, as new research typically relies on the validity of previously published data to build evidence-based conclusions. The identification of fabricated references suggests a vulnerability in the editorial process that could potentially mislead other researchers or clinicians.

By auditing millions of citations from a multi-year window, the work by Topaz and colleagues establishes a methodology for identifying non-existent references in the biomedical record, highlighting a need for more rigorous verification during the peer-review and editorial stages of publication.