Article type
Year
Abstract
Background: Estimates of the benefits or harms of treatments pooled from randomized controlled trials could be distorted if only some outcomes are reported in published articles. This is known as outcome reporting bias. Originally planned outcomes can be studied in trial protocols submitted for ethical approval.
Objectives: To study the reporting of outcomes and the occurrence of outcome reporting bias in drug trials.
Methods: We identified protocols of drug trials submitted to the Research Ethics Committee of the Medical Faculty of the University of Berne, Switzerland from 1988 to 1998. We searched the Cochrane Central Register of Controlled Trials (CENTRAL) and contacted submitting investigators to identify subsequently published articles. Study outcomes were defined as variables intended for comparison between groups and measured at one or more time points. We extracted outcomes from study protocols and published articles for a random sample of studies.
Results: Of 518 eligible study protocols, 196 (38%) were randomly chosen for detailed analyses. One hundred and fifty (77%) were multi-centre studies and 102 (52%) of these named international collaborators. One hundred and fifty-nine (81%) received funding from for-profit organizations. Seventy-six of 196 (39%) protocols matched with at least one full publication. In these protocols, 649 outcomes were planned. Of those, 100 (15%) were labelled as primary outcomes. Of 909 outcomes reported in corresponding articles, 103 (11%) were labelled as primary outcomes. We are currently analysing the degree of concordance between pre-defined outcomes and those reported in articles. We will present our results about factors associated with non-publication of studies and outcomes, and time to publication.
Conclusions: Our preliminary analyses show evidence of selective reporting of the outcomes of drug trials. This suggests that systematic reviewers might need to ask authors of original studies for unpublished outcomes of already published studies as well as additional unpublished studies.
Objectives: To study the reporting of outcomes and the occurrence of outcome reporting bias in drug trials.
Methods: We identified protocols of drug trials submitted to the Research Ethics Committee of the Medical Faculty of the University of Berne, Switzerland from 1988 to 1998. We searched the Cochrane Central Register of Controlled Trials (CENTRAL) and contacted submitting investigators to identify subsequently published articles. Study outcomes were defined as variables intended for comparison between groups and measured at one or more time points. We extracted outcomes from study protocols and published articles for a random sample of studies.
Results: Of 518 eligible study protocols, 196 (38%) were randomly chosen for detailed analyses. One hundred and fifty (77%) were multi-centre studies and 102 (52%) of these named international collaborators. One hundred and fifty-nine (81%) received funding from for-profit organizations. Seventy-six of 196 (39%) protocols matched with at least one full publication. In these protocols, 649 outcomes were planned. Of those, 100 (15%) were labelled as primary outcomes. Of 909 outcomes reported in corresponding articles, 103 (11%) were labelled as primary outcomes. We are currently analysing the degree of concordance between pre-defined outcomes and those reported in articles. We will present our results about factors associated with non-publication of studies and outcomes, and time to publication.
Conclusions: Our preliminary analyses show evidence of selective reporting of the outcomes of drug trials. This suggests that systematic reviewers might need to ask authors of original studies for unpublished outcomes of already published studies as well as additional unpublished studies.