Skip to contentSkip to filters

We use cookies for analytics.

How Should We Attempt to Emulate Target Trials Using Observational Data?

Oliver Dukes

Department of Mathematics, Computer Science, and Statistics, Ghent University

Abstract

Target trials are the randomized controlled trials one would like to run but cannot for ethical, financial or logistical reasons, so they must be emulated from observational data. In current target trial methodology each subject is enrolled for analysis in every target trial for which they are eligible rather than in a single trial, subjects are typically censored when they deviate from the protocol with inverse probability weighting to adjust for censoring, and results are pooled across the trials. Oliver Dukes situates target trial emulation within the framework of structural nested modelling, which generally yields more efficient estimators and more powerful hypothesis tests than those currently used, discusses issues that arise when pooling causal effects over time and tensions around model misspecification, and points toward a model-free approach to emulating target trials.

Topics