This text is
based on p 32 of Hemilä
(2006)
These documents have up to date links to documents that are available
via
the net.
Harri Hemilä
Department of Public Health
University of Helsinki,
Helsinki, Finland
harri.hemila@helsinki.fi
One concern in meta-analyses of small studies is the possibility of
‘publication bias’ which indicates that studies with ‘negative’
findings tend to remain unpublished more often than those with
‘positive’ findings (Chalmers et al. 1990). In such cases,
meta-analytical conclusions from the published trials may be too
optimistic. However, in an analysis of 487 research projects approved
by the Oxford Research Ethics Committee, there was evidence of
publication bias in the case of observational and laboratory-based
experimental studies (odds ratio [OR] 3.8 for the comparison of
positive vs. negative results to be published), whereas there was no
evidence of publication bias in the case of randomized clinical trials
(OR 0.8) (Easterbrook et al. 1991). Thus, although publication bias is
probably of concern in certain conditions, and must be kept in mind in
considering the findings of a meta-analysis, its role should not be
exaggerated.
References
Chalmers TC, Frank CS, Reitman D (1990) Minimizing the three stages of
publication bias. JAMA
263:1392-5
Easterbrook PJ, Berlin JA, Gopalan R, Matthews DR (1991) Publication
bias in clinical research. Lancet
337:867-72