Does trial registration reduce research bias? A comparison of registered and unregistered trials in diabetes quality improvement interventions




Poster session 1 Wednesday: Evidence production and synthesis


Wednesday 13 September 2017 - 12:30 to 14:00


All authors in correct order:

Karunananthan S1, Danko K1, Sullivan K1, Ivers N2, Grimshaw J1
1 Ottawa Hospital Research Institute, Canada
2 University of Toronto, Canada
Presenting author and contact person

Presenting author:

Katrina Sullivan

Contact person:

Abstract text
Background: The purpose of trial registration is to increase transparency and quality in the conduct of trials. Despite the implementation of the ICMJE policy on trial registration in 2005, many trials are still conducted without registration. It is unclear whether unregistered trials are at greater risk of bias than registered trials. In this study, we compare the characteristics of registered and unregistered trials of diabetes quality improvement (QI) interventions.

Objectives: To compare the study characteristics of unregistered and registered trials of diabetes QI interventions.

Methods: In a systematic review of diabetes QI interventions, we identified 140 trials published between 2010 and 2014. We identified the proportion of trials that were unregistered and compared characteristics of unregistered and registered trials including country of study conduct, source of funding, ethics approval, sample size, number of study arms, cluster- versus patient-level randomisation, blinding, study duration, number of outcomes reported and the statistical significance of the reported associations.

Results: We identified 50 (36%) trials that were not registered in a clinical trials registry. Compared to registered trials, fewer unregistered trials involved multiple arms (4% vs. 16%, p<0.05) and blinding of assessors (8% vs. 30%, p<0.05). We found no significant difference between unregistered and registered trials in mean sample size, mean length of follow-up, or the use of cluster randomisation.

Conclusions: This study indicates that despite the ICMJE policy, a large proportion of diabetes quality improvement trials published after 2010 remain unregistered. Methodological differences between unregistered and registered trials suggest that registration does improve the quality of trials.