Using Bayesian approaches to improve clinical trial efficiency

The frequentist statistical approach dominates clinical trial design and analysis. However, the problems with this approach are well described. Bayesian statistical methods provide an alternative framework for statistical modelling, using probability distributions to represent uncertainty of the model parameter estimates (e.g. treatment effects), and are increasingly being used in randomised controlled trials. Bayesian methods provide a formal approach for combining pre-existing information with data that are collected in the clinical trial into the analysis to update the current state of knowledge regarding the treatment effect(s). This is particularly appealing for clinical research given that healthcare advances usually occur through incremental gains in knowledge. Bayesian approaches may be used in clinical trials to incorporate information from previous studies (e.g. historical or external controls), to permit information-borrowing across patient subgroups (e.g. basket trials), or to sequentially update information on the parameters of interest when interim analyses are performed (adaptive trials). Bayesian approaches may offer a flexible and efficient framework for conducting clinical trials and may also provide results that are more useful and easier to interpret for clinicians, compared to frequentist approaches. This PhD project will focus on the application of Bayesian statistical methods to the design and analysis of clinical trials to determine where and how gains in efficiency may be obtained. The project will involve developing statistical methodology, or modifying existing methodology, and may explore the following types of novel designs/features (depending on the candidate’s interests): using external/historical data or other external information for the design and analysis of clinical trials in small populations; Bayesian adaptive designs; or master protocol trials such as basket, umbrella or platform trials. The PhD candidate should ideally have an undergraduate (Honours) degree or Masters degree in a quantitative field such as Statistics, Biostatistics or Mathematics, good skills in computer programming (e.g. using R), and preferably with some experience in Bayesian statistics.