Approaches that assess comparative effectiveness by combining evidence from target trial emulations with RCTs

Background Randomised controlled trials (RCTs) can provide an unbiased estimate of the average treatment effect for the population represented by the study participants. For most RCTs, the sample size is insufficient to provide precise estimates of the treatment effect for patient subgroups. This has potentially serious implications, as erroneous identification of differential subgroup effects may lead to inappropriate provision or withholding of treatment. RCTs with insufficient participants for important outcomes or subgroups generate imprecise results of limited value for clinical decision-making and for patients. Non-randomised studies (NRS) that use routine data offer opportunities to supplement RCTs, but a major concern is confounding by indication. Target trial emulations (TTE) have the potential to address some of these challenges by implementing design principles similar to RCTs. Despite their ability to provide similar results to RCTs, their use as additional data has been disregarded. Bayesian approaches to combine evidence have been developed and could be suitable for this purpose, but their use in combining TTEs with RCTs has not been previously investigated. Aim The aim of this project is to define a Bayesian framework that combines information from an RCT with a TTE applied to routine data, and to exemplify this framework with two case studies. Methods I have undertaken preliminary work, including (i) a systematic review of published TTEs, and (ii) a literature review on Bayesian statistical methods for combining evidence. These literature reviews have exposed the absence of previous studies combining evidence from an RCT and a TTE. To help address this gap in the literature, I will (iii) emulate an RCT from routine data, (iv) conduct a simulation study to compare different Bayesian approaches, and apply the resulting methods to supplement the RCT with evidence from its emulation. (v) The Bayesian framework produced will be applied to a second case study, to critically examine the framework in different contexts. (vi) The literature review and simulation study will be extended to subgroup interactions, and the proposed methods applied to the patient subgroups within the case studies. The two case studies' heterogeneous populations, and their different characteristics, will enable the validation of the framework in different scenarios. The proposed methods will be used to estimate the treatment effect overall, on secondary outcomes, and by relevant patient subgroups. Impact and Dissemination The output will include a Bayesian framework that exemplifies how an RCT and a TTE can be combined to provide new evidence on treatment effectiveness for subgroups and relevant outcomes for clinical decision-making. Additionally, I will provide improved evidence from two RCTs of high relevance in crucial clinical areas. A group of patient representatives is involved in the project, providing feedback and insight on the subject, and recommendations on how the results are presented to the public. The proposed framework will be useful where additional evidence may enhance RCTs and improve healthcare decision-making, including underpowered outcomes, heterogeneous populations, and enhancing study external validity. Patients and the public will benefit from the use of extant data to help answer important research questions.