Artikel
How frequent are meta-analyses with “double-zero” studies in systematic reviews?
Suche in Medline nach
Autoren
Veröffentlicht: | 2. September 2009 |
---|
Gliederung
Text
Introduction: Meta-analyses for rare binary (e.g., adverse events or safety) outcomes often include studies that have no events in one (“single-zero”) or even both (“double-zero”) treatment arms. A variety of methods to deal with this problem statistically have been proposed. Most of them rely on adding pseudo-observations to the respective studies, or, in case of “double-zero” studies, simple deleting them. However, simply deleting “double-zero” studies might result in biased treatment effects as those studies (with having equal numbers of outcomes in both arms) point to a null effect. We recently gave an example of a meta-analysis where ignoring “double-zero” studies missed a clearly significant and clinically relevant advantage of the off-pump technique on post-operative stroke prevalence in bypass surgery [1]. While it has been reported that about 30 % of all meta-analyses contain at least one “single-zero” study [2], there is no evidence on the frequency of meta-analyses with “double-zero” studies.
Methods: We performed a systematic review on the sample of Cochrane reviews (n = 500) from the Vandermeer [2] study. Two independent reviewers searched the full text of papers for the number of meta-analyses containing at least one “double-zero” study. Disagreement was resolved by repeating the search in the respective review and finding consensus.
Results: In the 500 Cochrane reviews we found 1394 meta-analyses with at least one “double-zero” study. These 1394 meta-analyses originate from 172 of the 500 (= 34.4 %) reviews. The median number of meta-analyses with at least one “double-zero” study in those 172 reviews was 3.5 (Min: 1, Q1: 1, Q3: 9, Max: 102).
Discussion: Meta-analyses with “double-zero” studies occur frequently in Cochrane reviews. In future work we will check if statistical methods that adequately account for those studies (refraining from 1. adding pseudo-observations or 2. simply deleting them) would eventually give clinically different results.
References
- 1.
- Kuss O, Börgermann J. Meta-analyses with rare outcomes should use adequate methods – A case study from cardiac surgery. Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie. 53. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds). Stuttgart, 15.-19.09.2008. Düsseldorf: German Medical Science GMS Publishing House; 2008. Doc MBIO1-2. http://www.egms.de/en/meetings/gmds2008/08gmds049.shtml
- 2.
- Vandermeer B, Bialy L, Hooton N, Hartling L, Klassen TP, Johnston B, Wiebe N. Meta-analyses of safety data: a comparison of exact versus asymptotic methods. Stat Methods Med Res. 2008 Jun 18. [Epub ahead of print]