Conducting observational analyses with the target trial emulation approach: a methodological systematic review.

Abstract

Objectives: Target trial emulation is an approach that is increasingly used to improve transparency in observational studies and help mitigate biases. For studies declaring that they emulated a target trial, we aimed to evaluate the specification of the target trial, examine its consistency with the observational emulation and assess the risk of bias in the observational analysis.

Design: Methodological systematic review reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement.

Data sources: The database MEDLINE (Medical Literature Analysis and Retrieval System Online) was interrogated for all studies published from 1 January 2021 to 3 July 2022. We performed an additional manual search of 20 general medical and specialised journals that spanned the same period.

Eligibility criteria: All studies that declared emulating a hypothetical or real randomised trial were eligible.

Data extraction and synthesis: Two independent reviewers performed the whole systematic review process (screening and selection of studies, data extraction and risk of bias assessment). The main outcomes were the definition of the key protocol components of the target trial and its emulation, consistency between the target trial and its emulation and risk of bias according to the ROBINS-I (Risk Of Bias In Non-randomised Studies - of Interventions) tool.

Results: Among the selected sample of 100 studies, 24 (24%) did not specify the target trial. Only 40 studies (40%) provided detailed information on all components of the target trial protocol. Eligibility criteria, intervention strategies and outcomes were consistent between the target trial and its emulation in 35 studies (46% of those specifying the target trial). Overall, 28 studies (28%) exhibited serious risk of bias and 41 (41%) had misalignments in the timing of eligibility assessment, treatment assignment and the start of follow-up (time-zero). As compared with studies that did not specify the target trial, those that did specify the trial less frequently seemed to have both time-zero issues (39% vs 52%) and serious risk of bias (26% vs 33%).

Conclusions: One-quarter of studies declaring that they emulated a target trial did not specify the trial. Target trials and their emulations were particularly inconsistent for studies emulating a real randomised trial. Risk of methodological issues seemed lower in observational analyses that specified versus did not specify the target trial.