PRM9 - A LONGITUDINAL, ITEM-LEVEL ANALYSIS METHODOLOGY TO SUPPORT THE INTERPRETABILITY OF MULTI-ITEM PATIENT-REPORTED OUTCOMES

Abstract

There is increased interest in assessing the patient experience to measure treatment efficacy and tolerability in clinical trials. When included in clinical trials, patient-reported outcomes (PROs) provide this insight directly from the patient. However, the following complexities of PROs inhibit the interpretability at the score level: 1) PROs may be composite, multidimensional, multi-item measures which, at the domain level, are difficult to interpret without anchoring to a more direct concept; 2) assumptions of domain-level unidimensionality is often violated; and 3) scoring transformation algorithms conducted on domains with few items may inhibit the interpretation of achievable change thresholds. The objective of this study is to describe cross-sectional and longitudinal item level categorical analysis to enhance interpretability from the item-level to the domain-level. Using data from multiple clinical trials, we propose an approach to analyzing individual items on a PRO measure to fully understand how treatment affects the driving factors of disease symptoms and functional status. Important characteristics of PRO items can be summarized in the following four unblinded analyses: (1) proportions of participants endorsing each categorical response option at each time point, (2) responses in each level of categorical change from baseline, (3) likelihood of achieving meaningful levels of categorical change over the course of treatment using longitudinal analyses, and (4) time to meaningful levels of categorical change (worsening or improving). Results from multiple phase 3 clinical trials will be presented to demonstrate the above methodology with a focus on 1- and 2-grade categorical change from baseline. Endpoints derived from composite, multidimensional, multi-item measures should be described at the item and the domain level, specifically targeting proximal disease impacts (e.g., symptoms, related function) to adequately assess treatment effect and aid interpretability for stakeholders. This proposed approach of single item analysis provides a robust, interpretable framework for PRO-derived endpoints for all stakeholders.