Assessing site-central baseline imaging discordance in oncology clinical trials.

11028 Background: Eligibility determination is a critical and rate-limiting step in oncology clinical trials and directly influences endpoint interpretability. Across oncology indications, eligibility frequently depends on prespecified imaging criteria that function as binary gatekeepers to enrollment, with blinded independent central review (BICR) commonly used to standardize image interpretation. In many solid tumor trials using RECIST 1.1, participants must demonstrate measurable disease; in many lymphoma trials using Lugano criteria, FDG-avid disease is required; and in prostate cancer trials using PCWG3 criteria, the presence of a bone scan superscan is typically exclusionary. Often participant eligibility is established by site investigators, while BICR radiologists independently assess baseline and on-study imaging without determining enrollment, creating potential site–central discordance at baseline. We hypothesized that baseline imaging discordance exists between site and BICR and evaluated its frequency and potential relevance to clinical trial endpoint interpretation. Methods: We performed a preliminary analysis of 17,104 baseline BICR reviews among enrolled participants evaluated using RECIST 1.1 requiring measurable disease (n = 7,189), Lugano criteria requiring FDG-avid disease (n = 2,653), and PCWG3 (n = 7,262) criteria. Results: For the RECIST 1.1 subset, BICR determined that 11.0% (n = 791) of enrolled participants lacked measurable disease at baseline. Among these participants, subsequent responses included CR (n = 36), PD (n = 93), and SD (n = 207). Because partial response under RECIST 1.1 requires measurable target lesions, participants without measurable disease are inherently constrained in the spectrum of evaluable responses, potentially biasing response rate. For the Lugano subset, BICR determined that 1.5% (n = 41) of enrolled participants lacked FDG-avid disease at baseline, with subsequent anatomic-only responses of CR (n = 1), PR (n = 5), SD (n = 9), and PD (n = 2). Though infrequent, site-central differences under Lugano criteria may influence response assessment, as BICR evaluated anatomic change while site incorporates metabolic response. For the PCWG3 subset, 7.0% (n = 548) of enrolled participants demonstrated a baseline superscan by BICR, with subsequent responses of CR (n = 20), PR (n = 25), SD (n = 90), and PD (n = 91). In these cases, response assessment is largely limited to soft tissue disease. Conclusions: The observed site-central imaging discordance reflects differences between eligibility and imaging assessments rather than incorrect site enrollment decisions and is not uncommon across major oncology response criteria, with potential implications for efficacy assessment in trials where eligibility is site-determined but response is centrally adjudicated. This analysis supports further investigation incorporating additional criterion-specific analysis.