Reference bias
Gøtzsche, P.C.
Ugeskrift for Laeger 150(4): 228-229
1988
ISSN/ISBN: 0041-5782 PMID: 3376330 Document Number: 324525
Literature retrieval on review of reference lists may result in a skew selection of the literature if these lists are not neutral. Reference bias was demonstrated in a sample consisting of 244 articles about double blind trials of non-steroid antirheumatic drugs in rheumatoid arthritis. Among articles where bias was possbile, ten had a neutral selection of references whereas 44 out of the remaining 66 (67%, 95% confidence limits 54-78%) showed preponderance of investigations with positive results for the "new" preparation in the reference list (p < 0.01). No difference was present in the scientific standards of the periodicals quoted compared with those not quoted as assessed by the Science Citation Index. Forty-four articles were multiple publications. Reference bias is also of significance for the individual article as the conclusion may become less reliable.