Article type
Year
Abstract
Background: Network meta-analyses (such as mixed treatment comparisons meta-analyses, multiple-treatments meta-analyses) could combine information from all randomized controlled trials (RCTs) when many treatment regimens already exist, to rank their benefits (and harms) to choose the best option for patients. Therefore finding all available RCTs is very important.
Objectives: To evaluate whether comprehensive searches were well conducted or reported in network meta-analyses.
Methods: We searched Pubmed, The Cochrane Library, Embase, ISI Web of Knowledge, and Google Scholar using 'Network meta analysis' OR 'mixed treatment comparisons meta analyses' OR 'multiple treatments meta analysis' in title/abstract without language, time, publication type restrictions. The first search was conducted on 2010 October 21, and updated on 2011 February 21. We included any network meta-analysis, regardless of the conditions or interventions. Letters or abstracts were excluded.
Results: We found 37 network meta-analyses, of which six included RCTs from relevant systematic reviews. For the remaining 31 network meta-analyses, the median number of databases searched was 3 (IQR 2-4, range 1-12) and 58.1% searched three databases or more. 37.8% did nothing beyond searching databases. Only 12.9% included the full electronic search strategy for at least one database. 41.9% did not present search terms, 12.9% did not report when they conducted their searches.
Conclusions: Our study results demonstrate that the searches in network meta-analyses are not well conducted, which may lead to biased results, and this needs to be improved in the future.
Objectives: To evaluate whether comprehensive searches were well conducted or reported in network meta-analyses.
Methods: We searched Pubmed, The Cochrane Library, Embase, ISI Web of Knowledge, and Google Scholar using 'Network meta analysis' OR 'mixed treatment comparisons meta analyses' OR 'multiple treatments meta analysis' in title/abstract without language, time, publication type restrictions. The first search was conducted on 2010 October 21, and updated on 2011 February 21. We included any network meta-analysis, regardless of the conditions or interventions. Letters or abstracts were excluded.
Results: We found 37 network meta-analyses, of which six included RCTs from relevant systematic reviews. For the remaining 31 network meta-analyses, the median number of databases searched was 3 (IQR 2-4, range 1-12) and 58.1% searched three databases or more. 37.8% did nothing beyond searching databases. Only 12.9% included the full electronic search strategy for at least one database. 41.9% did not present search terms, 12.9% did not report when they conducted their searches.
Conclusions: Our study results demonstrate that the searches in network meta-analyses are not well conducted, which may lead to biased results, and this needs to be improved in the future.