Research Methodologies
Building on existing areas of expertise in the Unit
Bayesian methods: Bayesian methods are relevant in developing logical sequences of early phase trials, in sample size estimation in the presence of uncertainty about key parameters, in interpretation of results in the light external evidence, and in accounting for fuller uncertainty in the projections from health economic models for decision making.
Missing data, non-compliance and causal modeling: Missing data are a new-universal problem in clinical trials, and the Hub will work on applying modern statistical thinking in this area, including making full use of observed data and allowing for possible departures from the missing at random assumption.
Cluster randomized trials: Cluster randomized trials pose particular design and analysis problems, for which BSU has advocated Bayesian methods. These methods have not however been much used in practice, and the Hub will endeavour to apply these methods in the new collaborations.
Trial Design: Experimental design matter from pre-clinical and first-in-man studies, through demonstration of efficacy and safety, to the fourth hurdle of implementation, policy or cost-effectiveness studies.
Meta-analysis and evidence synthesis: Systematic reviews are needed to determine the appropriateness of any new trial, and meta-analysis is necessary to evaluate the added value of its results. The Hub aims to develop new methodology to support novel evidence synthesis to inform trial designs.
Cost-effectiveness analysis and health economic modelling: Pragmatic trials are an important basis for assessing cost-effectiveness, either directly or in estimating transition parameters in a long-term health economic model. The Hub is looking to link with other BSU programs to stimulate current methodological advances with application to collaborative studies.