Use of informative priors in model-informed drug development
Nicky Best
GlaxoSmithKline
Objectives:
- To introduce methods for constructing informative priors from historical data or by elicitation from experts
- To discuss different methods for weighting priors in relation to data and for assessing and handling conflicts between prior and data
- To share examples of how such priors are implemented in models to inform different stages of drug development
Overview/Description of presentation:
This presentation will provide a brief overview of some of approaches for combining prior information and data for model-based inference and decision-making for drug development. I will discuss the use of expert elicitation methods to “bridge the gap” between existing in vitro/pre-clinical data and expected treatment effects in patient populations, and show how the elicited priors can be used to calculate expected probability of success of clinical development plans. I will also discuss “dynamic borrowing” methods for constructing robust priors directly from historical data and using these in a Bayesian model which adaptively down-weighs the prior according to the observed conflict between prior and new data. This approach will be illustrated with case studies using historical control data to supplement the control arm in a new clinical trial, and extrapolation of clinical efficacy from adult to paediatric populations.
Conclusions/Take home message:
The methods and examples presented in this tutorial illustrate the possibility to enable robust inclusion of prior information into model-based drug development.
References:
[1] Dallow N, Best N, Montague TH. (2018). Better decision making in drug development through adoption of formal prior elicitation. Pharmaceutical Statistics. doi:10.1002/pst.1854
[2] Schmidli, H. , Gsteiger, S. , Roychoudhury, S. , O'Hagan, A. , Spiegelhalter, D. and Neuenschwander, B. (2014), Robust meta-analytic-predictive priors in clinical trials with historical control information. Biometrics, 70: 1023-1032. doi:10.1111/biom.12242
