The psychometrics of clinical trial endpoints derived from patient-reported outcome measures: Causal inference validity
Edward Bein, PhD
Where: In-person | 2001 McGill College Avenue, room 1140; Zoom
Note: Social gathering (Open area outside Rm1140) from 3-3:30pm
The US Food and Drug Administration increasingly emphasizes the importance of assessing how patients “feel, function, and survive” when evaluating new treatments in clinical trials, often via the use of endpoints derived from patient-reported outcome measures (PROs). This presentation examines possible pitfalls in using PRO-derived endpoints to estimate and test hypotheses about treatment effects. It discusses the notion of whether endpoints possess “causal inference validity,” and presents inferential remedies when such validity is not achieved. The pitfalls and remedies are exemplified in a simulation study using the Rasch item response theory measurement model to analyze a PRO.
Ed Bein is on sabbatical from the US Food and Drug Administration, where he is a senior statistician at the Center for Drug Evaluation and Research. He has reviewed randomized trials evaluating treatments for infectious diseases, including COVID-19 and observational studies of the long-term safety of recently approved drugs. He has a Ph.D. in biostatistics from UC Berkeley and was a clinical psychologist in his previous career.