Statistical Considerations in Analyzing Health-Related Quality of Life Data

Health-related quality of life in oncology studies is often measured using patient-reported outcome (PRO) instruments which quantify latent concepts, such as pain and fatigue, experienced by the patient. Certain statistical considerations are helpful when analyzing PRO data for results interpretable to a variety of stakeholders. For longitudinal studies, valid results depend on approaches that accommodate missing PRO data. While comparison of group means is a common way to evaluate interventions, other endpoints reporting proportions of people who have experienced meaningful improvement or the likelihood of improving, for example, are palatable to many end-users. This chapter begins by describing basic properties of PRO instruments. We discuss ways to develop PRO research questions to guide analytics and methods to analyze endpoints while accounting for missing data. We consider ways to analyze multiple PRO endpoints including the use of standardized effect sizes, effective visualization, and accounting for testing of multiple endpoints.