14.1. Introduction
We all make thousands of decisions every day. Decision-making is part of most areas of life. Consequently, decision analysis has applications in many areas of the society, business and, not least, medicine. Within the medical fields, decision analysis has, quite naturally, most often been used in order to find the optimal strategy for diagnosing and treating patients (e.g., Parmigiani, 2002, Chapters 5–6; Bertolli et al, 2003; Liu et al, 2003). The analysis is often made for a population although attempts have been made to also include the preferences of the individual patient (Protheroe et al, 2000; Elwyn et al, 2001). In health economics, the benefits of a treatment are weighed against its costs, and these components are often embedded in a decision analysis model (Cooper et al, 2002; Brown et al, 2003).
Whereas there are plenty of decision analytic applications to medical decision-making and cost-effectiveness analysis in the literature, less has been written about such applications to the core drug development process. One reason is probably the confidentiality of internal company decisions. However, as identified in the FDA's recent Critical Path Initiative (FDA, 2004, 2006), drug development faces large and growing problems, including increased development costs and fewer new drugs that reach the market. One part of the solution to the industry's productivity problem, we believe, is through model-based approaches (FDA, 2004, p. 24) coupled with a decision analysis (Poland and Wada, 2001; Burman et al, 2005). This chapter will give a few examples of decision analysis applications from different parts of drug development, often promoting but sometimes warning against their use. Special attention will be given to the use of decision analysis during clinical development since the use of clinical trials is not only one of the cornerstones of drug development, it is also something that adds a special structure to the problems and thereby differentiates the decision problems in clinical development from those in other parts of science and business.
It is important to recognise that there are many stakeholders in the process of developing a drug and distributing it to the patients in need. It is interesting to compare what is optimal for different stakeholders. Of course, we ought to have a system where industry investments, regulations, evaluations by health care providers, patent laws, guidelines, prescription habits, etc., all serve the good of the patients. The decisions of different agents obviously interact with each other. This chapter will mainly have an industry perspective. However, now and then we will touch upon some other perspectives and the interaction between the interests of different stakeholders.
14.1.1. Outline
We will begin with an introductory example considering a very simplified clinical program (Section 14.2). This example will introduce some of the ideas and notions. In Section 14.3, we will describe the fundamentals of classical statistical decision theory. We will then be ready to describe a number of different applications of decision analysis in drug development:
Go/no go problems (Section 14.4).
Sample size calculation from a perspective of maximising company profits (Section 14.5).
Sequential design of clinical trials (Section 14.6).
Finding the best dose in a dose-response study (Section 14.7).
Project prioritisation (Section 14.8).