Optimal decision making in drug development
- Resource Type
- Electronic Thesis or Dissertation
- Authors
- Peck, Robbie
- Source
- Subject
- Language
- English
This thesis deals with quantitative decision making in a variety of different situations in drug development. In these different situations, optimal decision strategies may be derived and the value of different adaptive or model-based approaches may be quantified. We primarily use examples of Phase II/III programmes and portfolios of Phase III trials. Bayesian decision theory is a method that may be used to derive optimal decision rules given a gain function that aims to model the net present value of various assets to a trial sponsor. Deriving these optimal decision rules may be done using the method of dynamic programming with numerical integration routines. We investigate the benefit of adaptive methods such as group sequential designs or combination tests, or model-based approaches such as MCP-Mod to programmes and portfolios. Our results show that Bayesian decision theory coupled with the method of dynamic programming may find optimal decision rules in a variety of settings in drug development. These optimal decision rules may involve the choice of dose to take forward for a Phase III trial given Phase II data in a Phase II/III programme, or the choice of sample size for a trial in a portfolio of Phase III trials. It was found in simulation studies that group sequential designs may add value to a drug development programme or portfolio. Furthermore, the use of combination tests may add a smaller amount of value to a drug development programme. The problems discussed in this thesis are relevant to the running of clinical trials in industry. The methods we discuss may provide frameworks for the use of quantitative methods to help inform decision making in drug development.