Results publications are inadequately linked to trial registrations: An automated pipeline and evaluation of German university medical centers

Abstract

AbstractObjectiveTo evaluate links between registration and publication across clinical trials led by German university medical centers (UMCs) and registered in either ClinicalTrials.gov or the German Clinical Trials Registry (DRKS). Inadequate links make trial publications and registrations less findable and compromise evidence synthesis and health policy decision making. The World Health Organization (WHO) and others call for better adoption of this straightforward transparency practice.DesignCross-sectional bibliographic studySettingGerman UMC clinical trialsMethodsWe used automated strategies to download and extract data from trial registries, PubMed, and trial publications for a cohort of all registered, published clinical trials conducted across German UMCs and completed between 2009 and 2017. We implemented regular expressions to detect and classify publication identifiers (DOI and PMID) in registrations, and trial registrations numbers (TRNs) in publication metadata, abstract, and full-text.Main outcome measuresThe proportion of trial registrations that reference a known results publication. The proportion of results publications that report the known TRN in the metadata, abstract, and full-text.Secondary analysesWe constructed exploratory logistic regression models to investigate the relationship between trial completion date, registry, and registration-publication linking.ResultsOnly 20% (373/1,895) of trials have a linked publication (DOI or PMID) in the registration as well as the TRN in the publication full-text, abstract, and metadata, and only 25% (477) of trials met the CONSORT and ICMJE guidelines to include TRNs in both the full-text and the abstract. 17% (327) of trials had no links. The most common link was TRN reported in the full-text (60%, 1,137). ClinicalTrials.gov trials were overall better linked than DRKS trials, and this difference appears to be driven by PubMed and registry infrastructure, rather than by trialists. Of trials reporting a TRN in the abstract, trials in ClinicalTrials.gov were more likely than trials in DRKS to have the TRN captured in the PubMed metadata. Most (78%, 662/849) ClinicalTrials.gov registrations with a publication link were automatically indexed from PubMed metadata, which is not possible in DRKS.ConclusionGerman UMCs have not comprehensively linked trial registrations and publications by both including a structured reference to the publication in the registration, and reporting TRNs in results publications. In addition to improved linking by trialists, changes in the PubMed TRN capturing process (such as automated strategies like those developed in this study) and automated indexing of publications in DRKS would make trial registrations and results more findable.Open Data and CodeAll code and the final analysis data for this study are available at https://github.com/maia-sh/reg-pub-link. Raw data is available at https://github.com/maia-sh/intovalue-data.