2011 年 32 巻 1 号 p. 13-31
As clinical trials with “positive” results are more likely to be published, a meta-analysis of only published trials may be biased toward positive results (referred to as “publication bias”). A number of statistical tests have been proposed to detect publication bias. However, they have undesirable properties, particularly, the inflation of type I error and low power. A primordial countermeasure has been launched. In September 2004, the International Committee of Medical Journal Editors announced that they would no longer publish trials that were not registered in a public registry in advance. They embraced the WHO trial registration set consisting of 20 items including target sample size, which is related to the publication of results. The aim of this paper is to propose a new approach with a higher statistical power for detecting publication bias by using information on the sample sizes of all trials, including unpublished trials from the registry. We compared the proposed method to commonly used methods via simulations. The proposed method was found to have a higher power than the other methods in many situations. It will be useful for detecting publication bias because clinical trial registration will be more widespread in the near future.