Introduction
Are there fundamental differences in the statistical analysis of patient-reported outcomes and other non-biomedical endpoints, on the one hand, and so-called “harder” endpoints such as survival, tumor response, or patient blood pressure on the other?
In this chapter we emphasize that, while the application and interpretation of statistical methods in the outcomes research literature to date has been highly variable, this is not a signal that standard statistical approaches are not up to the task. Rather, they need to be applied with intelligence, completeness, and due consideration to the unique aspects of outcomes research. To that end, this chapter will provide specific examples of how the standard statistical methods have been applied skillfully, while also indicating where the use of novel or modern methods can and should be explored. The idea is not to address all relevant statistical topics and approaches de novo, or to produce yet another primer on statistical methods; there are already texts for this. Our aim, rather, is to ask what is, and what should be, the interplay between each statistical topic and the construction and selection of a cancer outcome measure. If there were little linkage between the statistical topic, on the one hand, and the choice of endpoint measures on the other, then that topic would receive only modest attention. Where the interplay is significant, the spotlight rises accordingly.