Current Methodological Issues in the Economic Assessment of Personalized Medicine

Abstract

There is a need for methodological scrutiny in the economic assessment of personalized medicine. In this article, we present a list of 10 specific issues that we argue pose specific methodological challenges that require careful consideration when designing and conducting robust model-based economic evaluations in the context of personalized medicine. Key issues are related to the correct framing of the research question, interpretation of test results, data collection of medical management options after obtaining test results, and expressing the value of tests. The need to formulate the research question clearly and be explicit and specific about the technology being evaluated is essential because various test kits can have the same purpose and yet differ in predictive value, costs, and relevance to practice and patient populations. The correct reporting of sensitivity/specificity, and especially the false negatives and false positives (which are population dependent), of the investigated tests is also considered as a key element. This requires additional structural complexity to establish the relationship between the test result and the consecutive treatment changes and outcomes. This process involves translating the test characteristics into clinical utility, and therefore outlining the clinical and economic consequences of true and false positives and true and false negatives. Information on treatment patterns and on their costs and outcomes, however, is often lacking, especially for false-positive and false-negative test results. The analysis can even become very complex if different tests are combined or sequentially used. This potential complexity can be handled by explicitly showing how these tests are going to be used in practice and then working with the combined sensitivities and specificities of the tests. Each of these issues leads to a higher degree of uncertainty in economic models designed to assess the added value of personalized medicine compared with their simple pharmaceutical counterparts. To some extent, these problems can be overcome by performing early population-level simulations, which can lead to the identification and collection of data on critical input parameters. Finally, it is important to understand that a test strategy does not necessarily lead to more quality-adjusted life-years (QALYs). It is possible that the test will lead to not only fewer QALYs but also fewer costs, which can be defined as “decremental” cost per QALYs. Different decision criteria are needed to interpret such results.