Generalizability and transportability in clinical trials

One of the principal aims of a clinical trial is to estimate the underlying causal effect of a treatment in the general population. The 'gold-standard' for a clinical trial is the randomized controlled trial (RCT), whereby causal effects are assessed by comparing outcomes between the control and the treated group. Often, however, the sample of RCT participants is not representative of the population in which the treatment is delivered. While the estimated causal effect will be internally unbiased for the RCT sample, it may not be externally valid for the target population. Observational studies, on the other hand, are typically more representative of the target population but can be subject to other internal biases due to confounding.

The aim of this PhD project is to assess and improve the generalizability and transportability of treatment effect estimates from clinical trials. To that end, the student will develop rigorous and principled statistical methodology to combine inferences from RCTs and observational studies. Registry-based RCTs, such as the ASCEND trial, offer particular promise for improved generalizability as participants are recruited directly from health and disease registries and are thus more representative of the target population. The student will assess the developed methodology through both extensive simulations and application to real clinical trial data.