Improving the underlying disease or condition is a central goal of drug development. However, understanding patient experience while on therapy is increasingly of interest. The goal is accurate and interpretable patient-centric information that can inform providers and patients when making treatment decisions. Understanding the patient experience requires collecting data from patients. Patient-reported outcomes (PROs) such as symptoms and function are frequently collected on trials on a quantitative scale. These outcomes can provide valuable insight into the patient perspective.

However, like all trial data, PRO results may be biased. PRO data can present additional analytic challenges, and a better understanding of methods to analyze and interpret this data, while taking into account the potential for bias is needed. In this discussion paper, we consider two situations: 1) bias in responder analyses and 2) estimands for analyzing patient function in trials with severely ill patients.