Background: In systematic reviews, searching trial registries is important in detecting the risk of publication bias and indicating the breadth of ongoing trials. The Cochrane conduct and reporting standards recommend searching both ClinicalTrials.gov (CT.gov) and the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) and reporting of search terms used (1,2). Compliance with these recommendations within Cochrane reviews is unknown.
Objectives: To describe the current practices in searching trial registries in published Cochrane protocols.
Methods: We conducted a preliminary audit of Cochrane protocols published in Issues 1 to 3 (2016) in the Cochrane Library. We extracted information on whether or not 1) the Methods section described searching CT.gov and ICTRP, 2) the Methods or Appendices reported the search terms used in CT.gov or ICTRP, and 3) the Methods or Appendices reported the search strings on either the basic or advanced search functions or both. Findings were reported as frequencies and percentages.
Results: From 142 protocols identified, 44 were excluded due to being marked as withdrawn (36) or relating to either an overview (3) or a diagnostic test accuracy review (5). Of the 98 audited protocols, 88 (90%) mentioned searching both CT.gov and ICTRP in the methods section while six protocols intended to search either CT.gov or ICTRP. Of these 88 protocols, 30 protocols provided either search terms only (4) or search strings (26). The search strings provided in the protocols were in the form of: basic search only (8/26), advanced search only (12/26), basic or advanced search for each registry (4/26), or both basic and advanced searches (2/26).
Conclusions: The majority of audited protocols described both CT.gov and ICTRP as part of their searching resources but did not frequently provide search terms/strings. As ongoing trial research assists us in assessing the overall completeness of the evidence, further improvements in detailing search strings for trial registries is needed.