Confronting Bias

For more than 20 years, steady progress has been made along the road to clinical trial transparency. What started with steps taken by medical journals, including disclosure of trial funding, disclosure of authors’ conflicts of interest, publication of trial protocols, and statements of authors’ contributions, has been strengthened by recent advances to require publicly accessible clinical trial registration and results reporting. Encouragingly, progress continues to be made, as there is now considerable momentum toward clinical trial data sharing. Each of these steps was taken with the intention to reduce the potential for bias in the peer-reviewed literature, mitigating problems such as selective publication and selective outcome reporting or guest authorship and ghostwriting and ensuring that research findings are fully disseminated, fairly and objectively, in a way that allows an informed assessment of the evidence.

Until this study by Emdin and colleagues,¹ there have been few evaluations of these interventions with the goal of determining whether they effectively reduce bias in the peer-reviewed literature. In their study of cardiovascular randomized clinical trials published in December 2012, they found that half of the 191 trials were reported as nonregistered and that these nonregistered trials were more likely than trials that were reported as registered, 70% vs 52%, to report statistically significant positive outcomes, even after accounting for other trial design differences. The higher rate of reporting positive outcomes among non-registered trials raises concerns about bias, suggesting, but not proving, that clinical trial registration at least partially mitigates selective outcome reporting. Bias may be mitigated by registration (1) because authors are more likely to present a balanced view of their research knowing that the prespecified protocol is available for review or (2) because editors and peer reviewers can access the registry and inquire about outcomes that were selectively reported.

More needs to be done to ensure compliance with clinical trial registration. The International Committee of Medical Journal Editors began requiring, as a condition of publication, registration in a public trials registry as of September 2005.² Moreover, the Declaration of Helsinki now states that “Every research study involving human subjects must be registered in a publicly accessible database before recruitment of the first subject.”^3(p2193) Yet half of the published trials that Emdin and colleagues¹ identified were not reported as registered. To more effectively confront bias, all medical journals should refuse to publish unregistered trials. In addition, journals should ensure publication of trial registration numbers so that editors, peer reviewers, and readers have access to summary information that was registered and prespecified as part of each study’s protocol.

Conflict of Interest Disclosures: None reported.