Comparison of Raw and Regression Approaches to Capturing Change on Patient-reported Outcome Measures

In April 2023, the US Food and Drug Administration (FDA) released “Incorporating Clinical Outcome Assessments Into Endpoints for Regulatory Decision-Making”, draft Guidance #4 in the FDA’s “Patient-Focused Drug Development Guidance Series.” The guidance document provides methods, standards, and technologies for collecting and analyzing clinical outcome assessment (COA) data to support regulatory decision-making, including methods for determining clinically meaningful change in a COA endpoint. The guidance re-emphasizes the need for sponsors to provide empirical support for defining a threshold of change in a COA that would be considered meaningful to patients as a critical first step to interpreting efficacy results on COA endpoints in a trial. The guidance offers some different thinking related to methodological considerations involved with generating this empirical data. Sponsors are going to continue to face pressure from the FDA to prove that changes in COA endpoints are meaningful to patients and to think differently about their methodological approach to addressing this.

Brandon Foster, Director of Psychometrics and Statistics in Lumanity’s Patient-Centered Outcomes team, recently collaborated with authors to publish “Comparison of raw and regression approaches to capturing change on patient-reported outcome measures” in Quality of Life Research. The publication is part of a special issue (published in May 2023) that brings together work from methodological leaders providing over a dozen papers covering the topic of “Methodologies and Considerations for Meaningful Change.”

In their work, Brandon and his co-authors argue that using raw change is limited within the commonly adopted framework for conducting anchor-based meaningful change analyses. They provide a new framework for calculating change scores on a COA using regression estimators, which also allows for information about measurement error and potential covariates, such as patient global impression measures, to be incorporated directly into calculating a COA change score. In addition to outlining the methodological approach, a proof of concept is shown using data from a real-world study of the PROMIS data.

Click the button below to access the article.