Publication bias and diagnostic research: full publication of results initially presented in abstracts

Article type
Authors
Brazzelli M, Sandercock P, Deeks J
Abstract
Background: Whilst there is substantial literature on publication bias in systematic reviews of randomized controlled trials, little evidence is available on publication bias in systematic reviews of diagnostic studies. The results of systematic reviews of diagnostic studies may be biased by both failure to publish negative diagnostic results or delays in publication.

Objectives: To estimate the rate of full publication of results of diagnostic imaging studies initially presented as abstracts at international stroke meetings.

Methods: Relevant abstracts were identified by systematically reviewing international meeting proceedings for the International Stroke Conference and the European Stroke Conference published between 1994 and 2004 in special issues of the journals Stroke and Cerebrovascular Diseases. Abstracts were selected if they reported findings of a diagnostic study of accuracy. Full-test publications of diagnostic abstracts were identified through a MEDLINE search. Authors of diagnostic abstracts were contacted to ascertain the existence of a subsequent full publication only when a full publication could not initially be found in MEDLINE. The characteristics in terms of study design and results of abstracts published in full were assessed.

Results: Results will be available at the Cochrane Colloquium.

Conclusions: The findings of this study will contribute to the understanding of publication bias in systematic reviews of diagnostic accuracy.