A new study conducted by mathematical statisticians at the US Food and Drug Administration (FDA) bridges the gap between an objective endpoint and a patient-reported outcome (PRO).
The research, led by statisticians Chul Ahn and Xin Fang at FDA’s Center for Devices and Radiological Health (CDRH), seeks to arrive at an acceptable level of probability for a PRO measure to reveal the “truth” of a subject’s disease or health condition. It comes at time when the science of patient input has gained momentum at FDA.
With FDA pushing for greater use of PROs in product development, a major challenge has been the inherently biased nature of these measures in the context of clinical endpoints. This has remained a barrier not just from a clinical research standpoint, but also from that of healthcare providers’ varying interpretations of published statistical guidelines for tailored treatment plans.
A team of researchers from the University of Maryland School of Pharmacy, in an article published in August, questioned the listing of PROs in FDA’s pilot Clinical Outcome Assessment Compendium over whether entries are truly representative of measures that matter to patients.
PROs, however, have become “increasingly important in measuring the effectiveness of a drug or medical device,” the study researchers argued, pointing to a growing body of medical devices, new drug labels, new molecular entities and biologic license applications that have been granted or approved that include PRO claims over the past several decades.
CDRH reported an increase last year of more than 500% in the number of device premarket submissions that included PRO measures over a six-year period, though use in post-approval studies has fluctuated. The center committed to developing a “‘fit-for-purpose’ framework for assessing validation evidence” in order to “improve predictability by clarifying the methodology used to review PROs for various types of uses” in conducting pre and postmarket regulatory reviews.
In January, CDRH also committed to advancing the use of in silico tools to evaluate patient outcomes. Neither a framework on the process used by review staff to evaluate PROs nor a policy for use of such tools have been made public thus far in the device space, though new final and draft FDA guidances were issued this year for pharmaceutical companies.
“In order for a PRO [measure] to be claimed in labeling, the PRO has to be valid, reliable and able to detect a change if the targeted disease status changes,” the statisticians noted. They applied a simulation approach to hypothetical data for an ophthalmic device to arrive at Qiz, which is a subject-specific probability of a PRO measure revealing a disease status in a scale based on an objective endpoint. “Qiz can be also viewed as a new agreement statistic between a continuous endpoint and a binary endpoint with or without correlation among samples,” they argued. To apply the approach with multiple patients, the research suggests more PROs are needed to support what’s adequate enough for regulatory purposes.