New statistical methods for emulating target trials

The ‘target trial’ framework has recently gained prominence in medical research as a way of making causal inquiries when data from a randomised trial is not available. The guiding principle is that one should try to emulate the desired hypothetical trial of interest. This is relevant for questions where it may not be feasible to randomise patients, or a trial would take years to carry out but answers are urgently needed. 

In this project, we will construct a rigorous framework for the design and analysis of target trials. First, we will develop efficient estimators for treatment effects in these studies based on the statistical models currently used. Existing approaches often assume that the effects of interventions remain stable over time; in the second work package we will look at how to relax thes assumptions whilst also targeting a clearly defined causal estimated. Finally, we will develop new strategies to handle time-varying missing covariate data.