Applying Causal Inference Techniques and Emulating Target Trials for Real-World Evidence Generation

The global burden of cardiometabolic diseases, such as heart attacks, stroke, and diabetes, has been rapidly increasing over the past few decades. This trend is projected to continue in the coming years, driven by factors such as ageing populations, unhealthy diets, smoking, excess alcohol consumption, as well as inadequate physical activity. There is, however, a lack of long-term evidence for long-term cardiometabolic interventions that span the entire life course, particularly using novel causal inference methods. Such evidence could better inform the development of long-term interventions for such conditions in the future.

Hernán and Robins first introduced the concept of target trial emulation (TTE) in 2016. TTE serves as a framework for using observational datasets to answer causal questions by emulating the design of an ideal randomised controlled trial (RCT). Although it has received valid criticism due to potential residual confounding in studies using this framework, it has been suggested that, with proper implementation, TTE can be a valuable tool to draw further insight from observational datasets, particularly in fields such as pharmacoepidemiology and health policy.

This PhD project would explore the implementation of the TTE framework, as well as additional causal inference techniques using observational data. Such techniques could include inverse probability weighting, the G-formula method, as well as integrating machine learning algorithms to balance treatment group characteristics. These include the longitudinal British birth cohort studies hosted at the UCL Centre for Longitudinal Studies, in addition to electronic health record data from the UK, and potentially from abroad. Particular exposures of interest include physical activity, diet and nutrition, as well as pharmacological interventions.