Identifying and managing problematic trials: A research integrity assessment tool for randomized controlled trials in evidence synthesis

Abstract

Evidence synthesis findings depend on the assumption that the included studies follow good clinical practice and results are not fabricated or false. Studies which are problematic due to scientific misconduct, poor research practice, or honest error may distort evidence synthesis findings. Authors of evidence synthesis need transparent mechanisms to identify and manage problematic studies to avoid misleading findings. As evidence synthesis authors of the Cochrane COVID-19 review on ivermectin, we identified many problematic studies in terms of research integrity and regulatory compliance. Through iterative discussion, we developed a research integrity assessment (RIA) tool for randomized controlled trials for the update of this Cochrane review. In this paper, we explain the rationale and application of the RIA tool in this case study. RIA assesses six study criteria: study retraction, prospective trial registration, adequate ethics approval, author group, plausibility of methods (e.g., randomization), and plausibility of study results. RIA was used in the Cochrane review as part of the eligibility check during screening of potentially eligible studies. Problematic studies were excluded and studies with open questions were held in awaiting classification until clarified. RIA decisions were made independently by two authors and reported transparently. Using the RIA tool resulted in the exclusion of >40% of studies in the first update of the review. RIA is a complementary tool prior to assessing "Risk of Bias" aiming to establish the integrity and authenticity of studies. RIA provides a platform for urgent development of a standard approach to identifying and managing problematic studies.

Keywords: COVID-19 pandemic; evidence synthesis; good clinical practice; randomized controlled trial; research integrity; systematic review.

Figure 1. Hierarchical work flow and decision tree of the Research Integrity Assessment (RIA) tool

Potentially eligible RCTs identified during screening should be assessed for research integrity hierarchically considering domain 1 to 6. Retraction, lack of prospective registration, lack of adequate ethical approval with informed written consent, inconsistencies in the author group and the location of the study, lack of proper randomization, implausible study results should lead to exclusion of a RCT. Concerns with the RCT in any domain put the study in ‘awaiting classification’ and should lead to further investigations. If no concerns appear through all domains or could be clarified, e.g. in correspondence with study authors, the RCT meets criteria for inclusion in the review and can be processed further. In living systematic reviews, included RCTs and RCTs ‘awaiting classification’ must be reassessed for retraction notices.