Target trial emulation of randomised controlled trials: lessons from benchmarking and risk of bias assessment

Background Target trial emulations (TTE) offer a framework for assessing comparative effectiveness to complement randomised controlled trials (RCT). However, little attention has been given to challenges that arise in emulating a published RCT from routine data. Methods We aimed to evaluate the validity of a TTE by benchmarking it against the published RCT and assessing potential biases arising from limitations in the emulation. We followed methodological guidance in emulating the FIRST-ABC step-up RCT from PICANet data. This TTE compared two modes of non-invasive respiratory support in children admitted to paediatric intensive care according to the primary outcome, the time to liberation from respiratory support. The complete-case analysis included 5,405 children. To reduce the risk of confounding we applied doubly robust methods. A structured risk of bias assessment using the ROBINS-I framework was conducted to evaluate potential biases arising from departures of the TTE from the benchmark RCT, with the ROBINS-I domains applied across each component of the target trial framework. Results The estimated risk ratio for liberation from respiratory support at the median follow-up was 1.33 (95% CI 1.27–1.40) from the TTE versus 1.03 (95% CI 0.90–1.18) from the RCT. The risk of bias assessment identified important limitations in the emulation, including inability to fully replicate the RCT eligibility criteria and outcome measurement, and the potential for residual unmeasured confounding. Conclusions These findings highlight key challenges in replicating an ideal trial from routine data and offer practical insights to improve the conduct and interpretation of TTEs. Clinical trial number Not applicable