Systematic reviews and meta-analyses of the results of several studies are a common method for pooling data to gain further insight through access to larger numbers of patients. Researchers doing meta-analyses usually check for publication bias, which is the tendency for studies with negative results ("We found no evidence of ...") to be rejected for academic publication. Publication bias can skew the findings of meta-analyses in the direction of positive evidence, however, researchers doing meta-analyses can check for publication bias statistically, using asymmetry tests.
But new research by Drs. John Ioannidis and Thomas Trikalinos indicates that statistical conditions are often not met for employing asymmetry tests. The authors evaluated a large number of meta-analyses in the Cochrane Database of Systematic Reviews, and from print journals in 2005. They conclude that this erroneous use of asymmetry tests is a major threat to the validity of clinical research.
p. 1091 - The appropriateness of asymmetry tests for publication bias in meta-analyses: a large survey
- J.P.A. Ioannidis, T.A. Trikalinos
http://www.cmaj.ca/pressrelease/pg1091.pdf
Journal
Canadian Medical Association Journal