Reference-trial-informed design to explore treatment effects in trial-underrepresented subgroups

Randomized controlled trials (RCT) are often regarded as the ‘gold standard’ of clinical evidence. However, their strict eligibility criteria can impact cohort diversity and limit the inclusion of some subgroups, including patients with comorbidities, older individuals or those from minority ethnic groups. Observational data, including data from electronic health records, can be used to bridge the gap in evidence but are subject to bias and confounding owing to the lack of randomization.

The ‘target trial’ emulation framework, which emulates the design of a hypothetical ‘target trial’ using observational data, has increasingly been used for causal inference^1,2. Despite this approach implementing design decisions to limit the effects of bias and confounding, uncertainty remains as to whether the observed result aligns with those which would have been obtained in RCT settings. Using a specified existing RCT (the ‘reference trial’) offers an opportunity to add further validity to inferences. This goal is achieved by basing the target trial design on the reference trial and benchmarking findings from the emulated study against the reference trial results. This approach adds confidence to the results obtained from the observational study before extending analysis to trial-underrepresented groups.