AUTHOR=Beaumont Hubert , Iannessi Antoine 

TITLE=Can we predict discordant RECIST 1.1 evaluations in double read clinical trials?

JOURNAL=Frontiers in Oncology

VOLUME=Volume 13 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2023.1239570

DOI=10.3389/fonc.2023.1239570

ISSN=2234-943X

ABSTRACT=Clinical trials utilizing imaging technology are extremely complex and expensive to manage. One of the main challenges associated with such trials is the management of inter-reader discrepancies. To reduce the burden of monitoring readers and anticipate discrepant evaluations as early as possible, analyzing data acquired at baseline can be beneficial. To understand how discrepancies manifest throughout trials, we studied the global evolution in time of the number of discrepancies.We identified three high-level factors that influence the occurrence of discrepancies: the rate of patient accrual, patient survival (or dropout), and the probability of discrepancy throughout patient visits. With this knowledge, we can model discrepancy occurrence before trials onset. We also tested the association between several aspects of the disease evaluations at baseline and the discrepancies that happened later in patient follow-up. We found several causes that had an association, although none of them had a strong association. Finally, we designed a predictive model for inter-reader discrepancy. The use of such a predictive model could help reduce the complexity and cost of clinical trials involving imaging technology.By anticipating discrepancies early during the trial, the monitoring of readers can be more efficient and effective.