부산대학교 도서관

Introduction: Randomised controlled trials (RCTs) provide data to help establish the harm-profile of drugs but evidence suggests that this data is underutilised and analysis practices are suboptimal. Aims To develop and assess methods for the analysis and presentation of harm outcomes in phase II/III drug trials that can facilitate the detection of adverse drug reactions (ADRs) and enable communication of informative harm-profiles. Methods: A systematic review looked at current practice for collection, analysis and reporting of harm outcomes and a scoping review to identify statistical methods proposed for their analysis was undertaken. A survey of clinical trial statisticians measured awareness of methods for the analysis of harm outcomes, barriers to their use and opinions on solutions to improve practice. Alternative strategies for analysis and presentation of harm outcomes were explored. Results: The review of current practice confirmed that data on harm outcomes is not being fully utilised, providing evidence of inappropriate and inconsistent practices. The scoping review revealed a broad range of methods for the analysis of both prespecified and emerging harms. The survey confirmed sub-optimal practices and while there was a moderate level of awareness of alternative approaches, use was limited. Guidance and training on more appropriate methods was unanimously supported. Recommendation were devised via consensus to encourage trialists to use visualisations for analysing and reporting harm outcomes. Of the evaluated methods for the analysis of emerging harms none were appropriate in trials ≤5000 participants with some utility in specific scenarios, recommendations for use are provided. Conclusion: Clinical trial statisticians agree that there is a need to improve how we analyse and report harm outcomes in RCTs. Efforts to date have focused on prespecified harm outcomes, with little thought given to emerging harms. Several solutions for immediate adoption are proposed but there remains the need for an easy to implement, objective, signal detection approach. Guidelines for best analysis practice that are endorsed by key stakeholders would also enable a more coherent and consistent path for change.