Evaluating response bias in RECIST 1.1: Impact on non-target lesion assessment in oncology.

Abstract

e13691 Background: In oncology, the RECIST 1.1 criteria serve as a cornerstone for assessing tumor response to therapeutic interventions. The criteria involve the evaluation of both target and non-target lesions to determine overall disease progression or response. Response bias, a concept well-documented in survey research and statistical analyses, refers to systematic errors that can skew data interpretations. In the context of RECIST 1.1, such bias may arise when reviewers' assessments of non-target lesions are subconsciously influenced by the observed changes in target lesions, which are analyzed before non-target lesions. This study explores the potential for response bias in evaluating non-target lesions, particularly when target lesions do not indicate progressive disease (PD). Methods: This retrospective analysis utilized data involving 321 subjects, with a total of 947 timepoints evaluated. The primary objective was to assess the frequency and implications of PD determinations based solely on non-target lesions compared to overall PD assessments. Each timepoint was examined for signs of PD as indicated by non-target lesion progression alone, independent of target lesion status. The percentage of PD timepoints attributed exclusively to non-target lesions was calculated to identify potential biases in clinical assessments. Results: Among the 947 total timepoints analyzed, 381 were identified as PD timepoints based on RECIST 1.1 criteria. Notably, only 57 of these (15%) were determined by progression in non-target lesions alone, without concurrent progression in target lesions. This relatively low percentage highlights a potential under-recognition or under-reporting of non-target lesion progression when target lesions do not show PD. The observed disparity underscores a possible response bias in radiological evaluations, where improvements or stability in target lesions might overshadow or influence the interpretation of changes in non-target lesions. This bias may potentially lead to an incomplete understanding of the true efficacy of new treatment options. Conclusions: The analysis highlights potential response bias in RECIST 1.1 evaluations, especially in non-target lesions when target lesions don't show PD. Only 15% of PD assessments are based on non-target lesion progression, indicating a need for greater awareness and rigorous methods, particularly in patients with high tumor burden. Addressing these biases is key to accurate treatment efficacy assessments and better patient care outcomes. Summary of PD assessments in RECIST 1.1 analysis. Metric Count/Percentage Total Number of Subjects 321 Total Number of Timepoints 947 Total Number of PD Timepoints 381 PD Timepoints Based on Non-Targets 57 Percentage of PD from Non-Targets 15%