Artikel
Metaanalysis for everyone?! A workshop
Suche in Medline nach
Autoren
Veröffentlicht: | 13. Februar 2006 |
---|
Gliederung
Abstract
As a metaanalysis of randomised controlled trials (RCT) currently is considered to represent the highest possible level of evidence, understanding the formal steps of a metaanalysis are vital to many HTA specialists and researchers. This workshop addresses metaanalytic techniques with special reference to the statistical principles and the assessment of publication bias.
In an fictitious example, two trials have been performed using the same intervention and the same study design. Both trials (named A and B) included 100 patients, but randomisation was unequal. In trial A the results were as follows: 18 of 80 patients died in the experimental group, whereas 8 of 20 control patients died. This gives a relative risk of 0.56. Trial B produced the following data: Two of the 20 patients in the experimental group died, but 12 of the 80 control patients experienced death. Therefore, the relative risk in trial B is 0.67. When both trials are combined by simple pooling, the contingency tables are added. This results in a total of 20 deaths in both experimental and control group. As the pooled group sizes are 100 patients per group, this (wrong) method of meta-analysis would produce an overall relative risk of 1. In the contrary, a true meta-analysis would yield a summary effect size of 0.59, which is much more in concordance with the primary studies.
The principle of the funnel plot is quite simple. It goes by the assumption that larger studies are more likely of being published, whereas smaller trials may get published only if they have a significant result. Thus, any difference between the results of larger versus smaller studies should cast a suspicion on the overall validity of the published evidence, because this finding indicates that some smaller studies have not been published due to their non-significance. Today, the funnel plot has become a standard procedure in meta-analysis, although sometimes constructing a funnel plot may be impossible due to the low number of available studies. Once the funnel plot shows clear asymmetry, the interpretation of the meta-analysis becomes complicated, since there is no opportunity to locate the apparently missing unpublished "phantom" studies.
Notes
The complete lecture can be found on the website of DIMDI: http://www.dimdi.de/static/de/hta/symposien/2005/index.htm