Does information from ClinicalTrials.gov increase transparency and reduce bias? Results from a five-report case series

Abstract

BACKGROUND: We investigated whether information in ClinicalTrials.gov would impact the conclusions of five ongoing systematic reviews. METHOD: We considered five reviews that included 495 studies total. Each review team conducted a search of ClinicalTrials.gov up to the date of the review's last literature search, screened the records using the review's eligibility criteria, extracted information, and assessed risk of bias and applicability. Each team then evaluated the impact of the evidence found in ClinicalTrials.gov on the conclusions in the review. RESULTS: Across the five reviews, the number of studies that had both a registry record and a publication varied widely, from none in one review to 43% of all studies identified in another. Among the studies with both a record and publication, there was also wide variability in the match between published outcomes and those listed in ClinicalTrials.gov. Of the 173 total ClinicalTrials.gov records identified across the five projects, between 11 and 43% did not have an associated publication. In the 14% of records that contained results, the new data provided in the ClinicalTrials.gov records did not change the results or conclusions of the reviews. Finally, a large number of published studies were not registered in ClinicalTrials.gov, but many of these were published before ClinicalTrials.gov's inception date of 2000. CONCLUSION: Improved prospective registration of trials and consistent reporting of results in ClinicalTrials.gov would help make ClinicalTrials.gov records more useful in finding unpublished information and identifying potential biases. In addition, consistent indexing in databases, such as MEDLINE, would allow for better matching of records and publications, leading to increased utility of these searches for systematic review projects.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 26 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.