e23021 Background: Robust imaging data are critical to endpoint determination in oncology clinical trials and regulatory decision-making. Blinded independent central review (BICR) provides imaging assessments but is inherently dependent on site-acquired imaging data. A common limitation across oncology trials is the occurrence of non-evaluable (NE) timepoint responses (TPRs), which may delay response or progression responses, lead to censoring, and complicate endpoint analysis. NE TPRs occur when BICR radiologists are unable to assign an overall response category at a scheduled on-study timepoint (TP) due to incomplete or uninterpretable imaging data. Contributing factors include missing required imaging, incomplete anatomic coverage, inadequate imaging technique, motion or other image-degrading artifacts, or disease-related assessment constraints. The frequency and drivers of NE TPRs are expected to vary by response criteria and imaging modality; however, few studies have characterized NE TPR causes across oncology imaging frameworks. We hypothesized that NE TPRs in BICR are driven by a limited set of recurrent factors that differ by response criteria and imaging modality. Methods: We analyzed BICR-assessed imaging TPs from oncology clinical trials employing RECIST 1.1, PCWG3, and Lugano criteria. NE TPRs were identified at the TP level and categorized by contributing factors, including missing required imaging, incomplete acquisition or anatomic coverage, and image quality limitations. TP component data (target, non-target, bone, spleen, and metabolic elements) were evaluated to determine the cause of NE responses. Results: A total of 62,798 imaging TPs were analyzed (RECIST 1.1 n = 37,344; PCWG3 n = 21,002; Lugano n = 4,453). NE TPRs occurred in 3.5% (n = 1,317) of RECIST 1.1 TPs, driven predominantly by NE target assessments (93.3%). Missing required imaging accounted for 69.9% of RECIST NE TPRs. In PCWG3 trials, 3.6% (n = 761) of TPs were NE, commonly due to NE bone response (77.5%), with missing bone scans (65.0%) and missing cross-sectional imaging (28.5%) as primary drivers. In Lugano trials, metabolic NE TPRs (n = 738, 16.6%) were more frequent than anatomic NE TPRs (n = 76, 1.7%). PET imaging NE TPRs were largely attributable to missing PET imaging (n = 680, 92.1%), whereas CT imaging NE TPRs were uncommon. Conclusions: Across RECIST 1.1, PCWG3, and Lugano criteria, NE TPRs were uncommon but consistently driven by a limited set of recurrent factors, primarily missing or incomplete required imaging rather than image quality limitations or disease-related constraints. Differences in NE drivers reflected the modality dependencies inherent to each response criteria. Systematic classification of NE reasons may inform protocol design, imaging acquisition compliance, and proactive data checks, supporting consistent endpoint evaluation and methods to reduce NE imaging TPs in oncology clinical trials.
Phillippi et al. (Thu,) studied this question.