Article type
Year
Abstract
Objectives: Traditional methods to identify evidence which are mentioned in the Cochrane Collaboration Handbook include using Collaborative Review Group trial registers, checking reference lists, personal communication, electronic databases and hand searching. The Internet represents a potential additional source for identifying evidence for systematic reviews. We aimed to determine whether and to which extent RCTs (randomized controlled trials) and other unpublished evidence can be identified by searching the World-Wide-Web, to identify websites which most likely contain hints to unpublished RCTs, to develop possible search strategies and tools for the Internet and to provide future directions for research in this area.
Methods: 8 completed, recently updated Cochrane Systematic Reviews (CSR) were randomly selected and their search strategies were retrospectively adapted for the World-Wide-Web. Different Internet search engines were tested for their suitability to handle complex queries. Using different methods, the Internet was searched and hints to potentially relevant unpublished or ongoing trials found on webpages were recorded.
Results: An advanced AltaVista search (www. altavista.com) using the Boolean expression "(intervention OR intervention-synonym) NEAR (condition OR condition-synonym)" led to optimal retrieval results. We failed to find any relevant unpublished trials that have been missed by the authors of the CSRs. However, in 4 of the 8 CSRs, hints to several ongoing and recently published trials could be found on the web, of which the CSR authors were partly not aware of.
Discussion: Searches on the Internet should be routinely included in the search strategies for CSRs. The development of specialised search tools for locating ongoing trials (a specialised Cochrane search engine) to facilitate Internet searches should be considered. A possible problem for reviewers could be the identification of non peer-reviewed, electronically published studies with Questionable reliability.
Methods: 8 completed, recently updated Cochrane Systematic Reviews (CSR) were randomly selected and their search strategies were retrospectively adapted for the World-Wide-Web. Different Internet search engines were tested for their suitability to handle complex queries. Using different methods, the Internet was searched and hints to potentially relevant unpublished or ongoing trials found on webpages were recorded.
Results: An advanced AltaVista search (www. altavista.com) using the Boolean expression "(intervention OR intervention-synonym) NEAR (condition OR condition-synonym)" led to optimal retrieval results. We failed to find any relevant unpublished trials that have been missed by the authors of the CSRs. However, in 4 of the 8 CSRs, hints to several ongoing and recently published trials could be found on the web, of which the CSR authors were partly not aware of.
Discussion: Searches on the Internet should be routinely included in the search strategies for CSRs. The development of specialised search tools for locating ongoing trials (a specialised Cochrane search engine) to facilitate Internet searches should be considered. A possible problem for reviewers could be the identification of non peer-reviewed, electronically published studies with Questionable reliability.