Additional literature search in Cochrane reviews: how, what and why?

Article type
Authors
J Mathew P, L Mathew J
Abstract
Background: Explicit and exhaustive literature search are hallmarks of Cochrane reviews; therefore authors commit to perform additional searches in and through conference proceedings, handsearching non-indexed journals, etc. However, the process, outcomes and therefore benefit (or otherwise) of these searches are often unclear. Objective: To assess whether Cochrane review authors perform additional literature searches; and to assess the process and outcome of these searches. Methods: New reviews published in The Cochrane Library Issue 1, 2009 were included in the study. One author masked identification details of reviews viz. Title, Authors, Cochrane Review Group and assigned a code number. The other examined the Methods section of each review to assess whether authors described additional literature searches in/through: (i) conference proceedings; (ii) handsearching of references/journals; (iii) contacting authors/experts; and (iv) contacting manufacturers. Thereafter the Results section was examined to determine: (i) details of how each of these four sources were searched/accessed; (ii) date of searching; (iii) outcome of the searches; and (iv) yield of additional trials. Results: There were 112 reviews published from 37 Review Groups. The Methods section revealed that 34 reviews (30%) included a search of conference proceedings, 96 (86%) included handsearching of journals, 58 (52%) contacted experts, and 19 (17%) contacted manufacturers. Most reviews did not describe details of how these sources were searched/accessed. The search date was specified in only six reviews. The outcome of these additional searches was also not described in most reviews. A small number of reviews reported finding additional trials through these searches. Table 1 summarizes the results. Conclusion: Although many Cochrane reviews suggest that additional literature sources have been searched, the process, output and outcome are poorly described in most; therefore the benefit (or otherwise) cannot be evaluated.