Response to Letter by Zheng et al
We thank Dr Zheng and colleagues for their interest in our recent review article and the author’s important remarks considering search strategies in meta-analyses.
One main point of concern is the sole use of the Medline database for our literature search. In order to reach comprehensiveness of meta-analyses, Zheng et al strongly suggest the use of multiple databases, as well as the inclusion of “gray” literature such as dissertations. In this scenario, it would of course be of great interest to know which important articles have not been included in our review. Although it is possible that an additional database search, for instance using Embase, could have yielded a larger number of search results, it is questionable whether these would have fulfilled our stringent predefined inclusion criteria and thus have been of relevance to our review. In support of this notion, an additional Embase search only made a small contribution to the overall results of meta-analyses in a recent experimental study.1 Based on their results, Sampson et al concluded that searches in multiple databases are time-consuming and expensive without improving the overall findings in most instances.
Another point of criticism relates to the sole use of English articles in our review. We concur with the authors that all languages should principally be considered for a meta-analysis in order to avoid a potential language bias. On the other hand, publications and meta-analyses in particular should be comprehensible for other researchers and allow a reproducibility of the results. The use of a wide range of different language literature would definitely impose a limitation to other researchers in reproducing the results of meta-analyses. In general, the international research community should strive for publishing scientific data in a language understandable to the majority of their members. Currently, English is the most common language in scientific literature and was therefore chosen as the language of publication in our review.
Finally, it should be pointed out that we did not claim to fulfil the criteria for a Cochrane review, which are usually based on prospective, randomized multicenter studies. Hence, our work should be classified as a systematic descriptive review and will thus never reach the “highest level in hierarchy of evidence” irrespective of our search strategy.