Integrity of randomized controlled trials: challenges and solutions

Because the exacerbating situation regarding integrity of randomized controlled trials is undermining the authority of this top-level evidence-generating method, more attention is needed to this considerable challenge.
Integrity of randomized controlled trials: challenges and solutions

Volume 113, Issue 6, Pages 1113–1119


Wentao Li, M.D., Ph.D., Madelon van Wely, Ph.D., Lyle Gurrin, Ph.D., Ben W. Mol, M.D., Ph.D.


There has been increasing concern about compromised integrity in randomized controlled trials (RCTs) because it undermines the authority of this top-level evidence-generating method and puts patients at risk. Compromised RCTs also jeopardize the validity of systematic reviews and meta-analyses when they are included. The publication-driven research environment is responsible for the emergence of integrity issues in RCTs, which has been facilitated by insufficient efforts to prevent, identify, and disclose research misconduct. The ability to assess integrity in RCTs is important to deter research misconduct. Data integrity in RCTs can be assessed in three domains: general assessment, analysis with summary data, and individual participant data check. Statistical methods play an important role in finding evidence against data fabrication or falsification in RCTs. It is important to strictly implement the prospective registration policy and the Consolidated Standards of Reporting Trials (CONSORT) statement for RCT publication, but more effort needs to be devoted to developing and validating statistical methods that identify different patterns of data manipulation. Ideally the system of submission, review, and publication would be reformed in a way that allows the timely investigation and retraction of compromised RCTs. Authors of systematic reviews should be aware of the uncertainty posed by compromised RCTs to the quality of synthesized evidence, and they should participate in the campaign to fight compromised integrity.

Read the full text here.