Systematic review of the empirical evidence of the selective reporting of analyses

Article type
Authors
Dwan K1, Williamson PR1, Gamble C1, Higgins J2, Sterne J2, Altman DG3, Clarke M4, Kirkham JJ1
1The University of Liverpool, UK
2The University of Bristol, UK
3The University of Oxford, UK
4Queen’s University, Belfast
Abstract
Background: Selective reporting of information in trials may occur for many aspects of a trial. Examples include the selective reporting of outcomes, the selective reporting of analyses: e.g. subgroup analyses or per protocol rather than intention to treat analyses. Selective reporting bias occurs when the inclusion of analyses in the report is based on the results of those analyses.

Objectives: We review and summarise the evidence from studies that have assessed the selective reporting of analyses in randomised controlled trials.

Methods: Systematic Review of studies that have assessed the selective reporting of analyses in randomised controlled trials. Along with the collaboration with experts in this area, the review will feed into the development of guidelines to support the appropriate reporting of a clinical trial with respect to outcomes, outcome measures, subgroups and analyses.

Results: Fifteen studies have been included in this review and consider aspects of selective reporting such as statistical analyses (7); subgroup analyses (4); the use of different scales (1); composite outcomes (2) and quality of life measures (1). Guidelines will be discussed at a face to face meeting scheduled for July 2013 and the output presented at The Cochrane Colloquium.

Conclusions: This work highlights the evidence of selective reporting and demonstrates the importance of pre-specifying analysis and reporting strategies during the planning and design of a clinical trial, for the purposes of minimizing bias when the findings are reported.