Abstract:
|
Problems in cancer research often require specialized but routine statistical methods, such as in the planning of clinical trials with commonly used endpoints. Consequently, applied statisticians in cancer centers may repeatedly confront similar research questions, the solutions to which are second-nature. Simultaneously, we trained to utilize and even create a wide variety of methodological tools when needed. Collaborations involving non-traditional problems provide opportunities for statisticians to contribute unique skills. This talk describes a long-term investigation of the relationship between Edmonton Symptom Assessment System (ESAS) and Common Terminology Criteria for Adverse Events (CTCAE) scores, i.e. patient reported outcomes and corresponding physician reported adverse events. Multiple methodological challenges are highlighted, such as the necessity and feasibility of fitting a joint multivariate, longitudinal model for a relatively small sample of zero-inflated scores measured over different and sparsely populated scales. The solution, interpretation and clinician reactions are presented, along with a discussion of the implications for applied statisticians.
|