Trial registration was supposed to reduce publication bias, and also improve the quality of publications. Ensuring that what investigators say they did, is what they actually did.
Becker JE, Krumholz HM, Ben-Josef G, Ross JS. Reporting of results in ClinicalTrials.gov and high-impact journals. JAMA. 2014;311(10):1063-5. Results are also supposed to be reported on the registration databases, specifically on the database mentioned in the title of this article, (which was written by a medical student). The authors compared what was put in the database to what actually appeared in the publication, which they limited to major medical journals. They noted that nearly all of the 96 studies had at least 1 discrepancy in the results between the publication and the record on ClinicaTrials.gov. Many had differences in the primary outcome, and 6 of them changed the interpretation of the major outcome of the trial.
Kasenda B, von Elm E, You J, Blumle A, Tomonaga Y, Saccilotto R, et al. Prevalence, characteristics, and publication of discontinued randomized trials. JAMA. 2014;311(10):1045-51. This study looked at just over 1000 RCTs that were started in 3 different countries. They got the identification from ethics committees that had approved them, and then searched what had happened to them. 253 were discontinued, most often trials were discontinued because of poor enrollment (n=101). Other reasons included interim analysis showing no benefit, and interim analysis showing harm in one of the groups. Half of the discontinued trials were never published, but 21% of completed trials were never published either. Discontinuing a trial because of poor enrollment predicted a greater likelihood that no-one will ever see the results. It is very important to publish discontinued trials, other researchers, granting agencies, and patients can benefit from the data accumulated, even if you had to stop early. Once patients have volunteered to participate in a trial, it is an ethical failure to not publish the data that they provide.