Network meta-analyses could be improved by searching more sources and by involving a librarian

Objective: Network meta-analyses (NMAs) aim to rank the benefits (or harms) of interventions, based on all available randomized controlled trials. Thus, the identification of relevant data is critical. We assessed the conduct of the literature searches in NMAs. Study Design: Published NMAs were retrieved by searching electronic bibliographic databases and other sources. Two independent reviewers selected studies and five trained reviewers abstracted data regarding literature searches, in duplicate. Search method details were examined using descriptive statistics. Results: Two hundred forty-nine NMAs were included. Eight used previous systematic reviews to identify primary studies without further searching, and five did not report any literature searches. In the 236 studies that used electronic databases to identify primary studies, the median number of databases was 3 (interquartile range: 3-5). MEDL1NE, EMBASE, and Cochrane Central Register of Controlled Trials were the most commonly used databases. The most common supplemental search methods included reference lists of included studies (48%), reference lists of previous systematic reviews (40%), and clinical trial registries (32%). None of these supplemental methods was conducted in more than 50% of the NMAs. Conclusion: Literature searches in NMAs could be improved by searching more sources, and by involving a librarian or information specialist. (C) 2014 Elsevier Inc. All rights reserved.