Background: Clinical trial investigators and research ethics committees should ensure that clinical trials are methodologically sound, and that the results of those trials are made publicly available in an open and honest fashion that is not misleading and does not misrepresent the data.

Objectives: To investigate the selective reporting of the primary outcomes of clinical trials.

Methods: A follow-up study of all 103 published randomized controlled trials considered by Central Sydney Area Health Service Ethics Review Committee between 1 January 1992 and 31 December 1996. Selective reporting encompassed i) discrepancy in the identity of the primary outcome; ii) discrepancy in the definition of the primary outcome; iii) completeness of reporting of the primary outcome. Protocol related variables that may impact on the outcomes were explored using logistic regression.

Results: Discrepancies in the definition of the primary outcome were not quantifiable due to poor documentation. Seventeen percent of primary outcomes in the protocol were not reported as primary outcomes in the publication, and 15% of primary outcomes in the publication were not declared as primary outcomes in the protocol. Primary outcomes were more likely to be the same in the protocol and the publication in trials with a partially (OR 3.9, 95%CI 0.53-28.1) or completely (OR 12.9, 95%CI 1.9-86.1) documented sample size calculation. There was a significant relationship between completeness of the sample size calculation and completeness of reporting (global p=0.053). Trials with a partially (OR 7.7, 95%CI 1.3-46) or completely (OR 4.8, 95%CI 1.2-20) documented sample size calculation in the protocol were more likely than those without to report fully all of their comparisons. Trials where all of the comparisons were statistically significant were more likely to report fully all of their comparisons (p=0.06).

Conclusions: Selective reporting of the primary outcome existed in some form in a significant proportion of trial publications.