Citation bias
In locating studies, researchers often supplement searches of computerised databases by contacting experts in the field and checking the reference lists of other studies and reviews. When reference lists are used, citation bias could have an important role.
In the field of cholesterol lowering, trials that are supportive of a beneficial effect are cited more frequently than unsupportive trials, regardless of the size and quality of the studies involved29 (box).
Thus the use of reference lists would be more likely to locate supportive studies, which could bias the findings of a metaanalysis. The journals in which papers are published could also influence the ease of their location and their inclusion in metaanalyses. One influential cholesterol lowering trial, for example, was originally planned as a
study with primary prevention and secondary prevention arms.34
The results of the primary prevention component were interpreted by the investigators as being favourable, and the results were published in the New England Journal of Medicine in 1987.35 The second ary prevention arm finished at the same time, but in this case the results were clearly unfavourable.36 The findings from this arm were not published until 1993,
in the Annals of Medicine,36 a journal with limited circulation. The paper in the New England Journal of Medicine received more than 450 citations in the three years
after publication, whereas the article in the Annals of Medicine received 17.
Multiple publication bias
Multiple publications from single studies can lead to bias in several ways.37 Studies with significant results are more likely to lead to multiple publications and presentations,13 which makes it more likely that they will be located and included in a metaanalysis. Furthermore, the inclusion of duplicated data may lead to overestimation of treatment effects, as recently shown for trials of ondansetron, a 5HT3 receptor antagonist that is used for prevention and treatment of postoperative nausea and vomiting.38
It is not always obvious that multiple publications come from a single study, and one set of study participants may thus be included in an analysis twice. This is a particular prob lem in multicentre trials. In one of the metaanalyses of low molecular weight heparin,8 for example, it seems that combined data from a multicentre trial were included, together with a subset of the same data that had also been reported separately from an individual centre which contributed to the trial.39
Indeed, it may be difficult, if not impossible, for metaanalysts to determine whether two papers represent duplicate publications of one trial or two separate trials, as exam
MARK HUDSON
Education and debate
BMJ VOLUME 316 3 JANUARY 1998 63
No comments:
Post a Comment