What are RECIST 1.1 progressions made of? Variability in double-read oncology trials

Abstract Objective Blinded Independent Central Review (BICR) with double reads and adjudication is crucial in imaging-based clinical trials to ensure quality data. However, discrepancies in double reading can affect trial outcomes, particularly in assessing disease progression in phase 3 oncology studies using RECIST 1.1 criteria. This study examined discordance in the date of progressive disease (DoPD) across RECIST components: target lesion (TL), non-target lesion (nTL), new lesion (NL), exploring its impact on survival curves and whether discrepancies stem from timing differences or true/false PD detection. Materials and methods We retrospectively analyzed data from five clinical trials using BICR with double reads plus adjudication, involving 1932 lung cancer patients on immunotherapy or targeted therapy. RECIST components were examined to assess DoPD concordance, discrepancies, adjudicator acceptance, detection timing, and impact on survival curves. Results Readers showed a 39.3% discordance rate in DoPD assessments, with agreement on 17.3% of DoPD cases and 43.4% of non-PD cases. In 54.2% of concordant cases, multiple RECIST components contributed to PD. Discordance was primarily caused by NL (41.4%), sum of TL diameter increase (33.3%), nTL (11.8%), or multiple components (13.4%). In 49.2% of discrepant cases, PD was reported late, usually within one treatment cycle (79.8%). 62.5% of disputed PD cases were accepted by adjudication. Conclusion Different RECIST components vary in their likelihood of causing discordance and being accepted or refuted by adjudicators. NL detection is key for identifying progression but also the main source of disagreement. Using multiple RECIST components enhances the reliability of PD assessment. Key Points Question How does discordance across RECIST components—especially new lesion detection—influence progression assessment timing and potentially alter survival outcomes in oncology clinical trials? Findings Reader disagreement on progression dates is common, driven mainly by new lesion detection; incorporating multiple RECIST components increases alignment and strengthens PD determination reliability. Clinical relevance Imaging endpoints guide cancer treatment decisions. RECIST 1.1 is the standard, inter-reader variability can undermine PFS accuracy. By evaluating RECIST component reliability, this work aims to improve trial precision, enabling faster, more confident decisions that ultimately benefit patients. Graphical Abstract