Practical elements to consider when emulating a target trial

Target trial emulation, i.e., the analysis of observational data by explicitly emulating the design components of a randomized controlled trial (RCT), has gained attention in recent years as an approach to estimate causal effects of pharmacological or nonpharmacological interventions that are, on average, similar to those of RCTs.

This article addresses 2 objectives. First, we examine factors that may explain discrepancies in treatment effect estimates between RCTs and their target trial emulations, beyond issues related to bias or imperfect replication of the original trial. Second, we outline 11 practical considerations for the design and planning of a target trial emulation.

Treatment effect estimates from RCTs and their target trial emulations may differ because of differences in the underlying populations (beyond formal eligibility criteria), the definition of interventions, the measurement and definition of outcomes, and the causal estimand being targeted. To improve the validity of target trial emulations, several design elements should be considered. For example, control and intervention groups should be concomitant and geographically comparable, and temporal changes in clinical practice should be accounted for. Also, when grace periods are used, they should reflect logistical delays in treatment initiation while maintaining assignment at time zero. Finally, emulations necessitate adequate sample size and sufficiently rich data to address imbalances in prognostic factors.

This article provides key elements to consider when planning a target trial emulation.