Prognostic value of combined baseline FDG PET/CT radiomic parameters in primary mediastinal B-cell lymphoma

ABSTRACT: Primary mediastinal B-cell lymphoma (PMBL) is an uncommon aggressive lymphoma with generally favorable outcomes; however, 15% to 20% of patients develop refractory disease. Reliable baseline prognostic biomarkers remain lacking. We evaluated the prognostic value of radiomic parameters derived from baseline 18F-fluorodeoxyglucose positron emission tomography (PET)/computed tomography in newly diagnosed PMBL. This post hoc analysis included 180 patients who were treatment naive from the multicenter Lymphoma Study Association (LYSA) cohort (2007-2017) treated with standard immunochemotherapy. Twelve 3-dimensional PET-derived radiomic features were extracted using Oncometer3D. Interparameter correlations were assessed using Spearman coefficients. Receiver operating characteristic analyses determined optimal cutoffs. Progression-free survival (PFS) and overall survival (OS) were analyzed using Kaplan-Meier estimates and Cox regression models. Four clusters of correlated features were identified: activity, tumor burden, massiveness/fragmentation, and dispersion. Among these, total metabolic tumor volume (TMTV), maximal interlesion distance (Dmax), and median peripheral centroid distance (medPCD) demonstrated the strongest prognostic associations. TMTV and Dmax were significantly associated with shorter PFS and OS, whereas medPCD predicted OS. In multivariable analysis adjusted for International Prognostic Index, bulky disease, B symptoms, and treatment regimen, a composite radiomic score (2-3 high-risk features vs 0-1) independently predicted OS (hazard ratio, 7.76 [95% confidence interval, 1.59-37.74]) and showed a strong trend for PFS. The composite score outperformed conventional clinical risk factors. Baseline PET-derived radiomic features capturing tumor burden, spatial dispersion, and compactness provide independent prognostic information in PMBL. A 3-parameter radiomic score outperformed conventional risk factors and identified a high-risk subgroup at diagnosis, warranting prospective and external validation before clinical implementation.