Concordance between physician response assessment and a novel method per Lugano criteria in patients with lymphoma using real-world data.

Abstract

e19082 Background: Clinical response assessment criteria, such as RECIST v1.1 and Lugano 2014, provide a standardized approach to assess treatment effectiveness in oncology clinical trials. Corresponding methodologies are needed for real-world data (RWD) to improve interpretability alongside clinical trial data and facilitate replication of results across RWD studies. We developed a RWD novel methodology, real-world Lugano (rwLugano), derived from Lugano 2014 criteria to classify lymphoma treatment response among patients with diffuse large B-cell lymphoma (DLBCL). Methods: A retrospective multisite RWD study identified adult patients with DLBCL treated with first line chemoimmunotherapy in US clinical practice (01JAN2015-31DEC2022). Participating sites abstracted clinical data via medical chart review and positron emission tomography-computed tomography (PET-CT) reports. Deidentified digital PET-CTs taken at treatment initiation and first response assessment were uploaded for BICR. Three distinct methods were used to assess treatment response: 1) physician-reported obtained from unstructured clinical notes in medical records, 2) rwLugano calculated using data (i.e., components of Lugano 2014) from PET-CT reports, and 3) BICR adjudication by two lymphoma radiologists using Lugano 2014 criteria. We compared the proportion of patients by response category: complete (CR), partial (PR), stable (SD) and progressive (PD) for each method. Analyses included agreement (%), concordance (Cohen’s kappa [κ]), and odds ratios with 95% confidence interval (OR; 95%CI) estimated using multivariable generalized linear mixed models (GLMM) adjusted for baseline characteristics. Results: The study identified 174 patients with DLBCL (female n=71 [41%]; white n=135 [78%]; Hispanic n=13 [7%]; mean age 66 years). Fewer patients were classified as CR at initial response via physician-reported response (n=111, 64%) compared to rwLugano (n=145, 83%), and BICR (n=142, 82%). Overall and CR agreement with BICR (see Table) was higher for rwLugano (overall: 83%, κ=0.49; CR: 86%, κ=0.50) than physician-charted response (overall: 72%, κ=0.45; CR: 76%, κ=0.43). In GLMM models, physician-charted response underestimated CR compared to BICR (OR=0.22, 95%CI:0.12-0.43), whereas rwLugano was statistically similar to BICR (OR=1.2, 95%CI:0.6-2.4). Conclusions: This study found that physician-reported response may underestimate CR at initial response, and rwLugano may be more consistent with BICR for treatment response assessment. Use of rwLugano warrants further evaluation as a methodology to assess treatment response in RWD lymphoma research. [Table: see text]