Background: randomized clinical trials (RCTs) play an important role in evidence-based medicine. Flaws in their design, conduct, publication and reporting may lead to erroneous conclusions with potentially serious consequences for patients. Responsible research practices are therefore essential to guide decision making in clinical practice.
Objectives: to assess trends over time in the design, conduct, publication, and reporting of RCTs.
Methods: we gathered full-text articles of all human RCTs published in PubMed until 2017 and additional relevant variables, using different machine learning and language processing tools. We retrieved information on authors, institutions, and clinical topics. In addition, we evaluated the following aspects related to responsible research: risk of bias (using RobotReviewer software); selective outcome reporting related to online trial registrations; quality of reporting in terms of the CONSORT statement; and statistical power and rigour.
Results: we collected almost 184,000 full-text RCT publications. Of all studies, a total of 18,871 were registered on ClinicalTrials.gov; only 24% uploaded their protocol before the start of the study, and 31% of studies changed the primary outcome in the registry during or after the study duration. We will present full data on all variables at the Colloquium.
Conclusions: this study provides insight into research and publication practices in more than 180,000 full-text RCT publications and illustrates changes in design, conduct, publication, and reporting over time. The results can be used to improve research practices in order to make them more responsible.
Patient or healthcare consumer involvement: there was no direct patient or healthcare consumer involvement.