Abstract
Personalized medicine is the notion that medical treatments can be adapted to individual patients based on a multitude of personal attributes. The set of personal characteristics that can together explain in part why patients respond differently to treatments is what we call patient heterogeneity. Economic evaluation traditionally uses a population-based approach; treatment recommendations and reimbursement decisions are based on the average outcome measured in an entire population sample. This can mask important sources of patient heterogeneity that could be used to improve decision-making. Instead, patients can be categorized in subgroups based on their personal characteristics and the cost-effectiveness analysis can be done exploring subgroup differences. However, in reality, this is rarely carried out. This is possibly because of the researchers' unfamiliarity with the methods and a lack of clear guidance in economic evaluation guidelines used by manufacturers and health technology assessment agencies. The guidelines published by the Norwegian agencies are vague and unclear on the topic of acknowledging patient heterogeneity and on how to conduct subgroup analyses in economic evaluation. Therefore, with the intention to ultimately make recommendations to improve the guidance in Norway, this thesis set out two objectives: (1) to describe and compare existing methodology to acknowledge patient heterogeneity and (2) to apply the methodology to the results of an RCT. These two exercises were carried out to allow for the identifcation of both theoretical and practical strengths and weaknesses of the methods. Three conceptual frameworks which in order are, Stratifed Analysis (SA), Expected Value of Individualized Care (EVIC) and Value of Heterogeneity (VoH) were selected for the exercises. Thoroughly discussing their theoretical foundation pinpointed that even though all three methods are very similar, each present important advantages/disadvantages. Applying the three methods to the results of an RCT showed that there are also several practical differences that needed to be considered before conclusively suggesting a best course of action. Some unexpected technical problems occurred when using RCT results rather than modelling results. However, some solutions were formulated to address these issues. Most importantly, the last exercise made it possible to identify future research questions that builds on the frameworks' concepts and could lead to important practical improvements. Ultimately, it was concluded that the use of either method alone is sub-optimal. Since the frameworks shared important similarities, it was possible to suggest an integrated approach that uses all three methodologies by playing to their strengths. This approach could serve as a rudimentary better course of action that may be recommended for HTA practices in Norway and from which to build on and improve with future research.