Most Cochrane Reviews focus tightly on an intervention or class of interventions for a specific condition or in a particular population or setting. As a result, readers of the Cochrane Library may need to consult several related Cochrane Reviews to find the evidence they need. Cochrane Overviews were designed to address this problem by providing a 'friendly front end' to the Cochrane Library - with the expectation that Overview readers would use them to discover the presence of a set of related reviews in the library, and to get an idea of the contents of these reviews, with the option of going to the individual reviews if they wished additional detail.
The Overviews published to date have shown a wide variation in their scope, content and methods. Most have included only Cochrane Reviews, but a few have extended their search to systematic reviews published elsewhere. The number of reviews included in each Overview ranges from 3 to 54. Overview authors have also taken different approaches to the synthesis of data from included reviews.
In some cases, it has been possible to rank interventions by their effectiveness or adverse effects using network meta-analysis. This recently developed method allows Overview authors to make "indirect comparisons" between interventions for which no direct head-to-head trials have been found, and to include both the direct and indirect comparisons in their analyses. A key assumption underlying network meta-analysis is that the trials included in the analysis bear sufficient similarity to one another to make indirect comparisons appropriate. This has been labelled the "transitivity assumption" or "similarity assumption" and requires that the included trials should, at least in theory, be "jointly randomisable" - that is, that a subject in any one of the trials included in the network could have, in principle been assigned to any one of the other interventions being compared. More information about Network Meta-Analysis and its assumptions is available on the website of the Cochrane Comparing Multiple Interventions Methods Group (cmimg.cochrane.org). Four of the currently published Cochrane Overviews have taken this approach and present new analyses of the data from trials included in the Cochrane Reviews that they cover.
In the majority of cases, Overview authors have not been able to ensure that the transitivity assumption is satisfied, and have not attempted to rank interventions or to perform new analyses. Many of these Overviews include tables that juxtapose results of direct comparison meta-analyses performed in a number of related reviews to allow readers more easy access to this body of related evidence. While it may be tempting for the reader to try to rank interventions or do "informal" indirect comparisons using these tables, this temptation should be resisted. The transitivity assumption is equally important in this setting, and these informal indirect comparisons can lead to inaccurate conclusions if there were significant differences between the trials included in the different systematic reviews that contributed data to the table.
A minority of Overviews do not attempt to provide either direct or indirect comparisons. Overviews of this sort tend to deal with complex interventions where specific components may vary across trials. The authors use the results of the included reviews to provide a map of the existing evidence, often providing their own taxonomy of interventions or outcomes.
The Cochrane Overviews included in this special collection have been prepared by authors and editors from a diverse range of Cochrane Review Groups including the Cochrane Musculoskeletal Group, Cochrane Airways Group, Cochrane Tobacco Addiction Group, Cochrane Menstrual Disorders and Subfertility Group, Cochrane Hepato-Biliary Group, Cochrane Pain, Palliative and Supportive Care Group, Cochrane Drugs and Alcohol Group, Cochrane Effective Practice and Organisation of Care Group, Cochrane Consumers and Communication Group and Cochrane Pregnancy and Childbirth Group.