Integration of [F]FDG-PET radiomics with liquid biopsy improves outcome prediction in newly diagnosed diffuse large B-cell lymphoma.

Early identification of relapsing/refractory diffuse large B-cell lymphoma (DLBCL) represents an unmet clinical need. A real-life cohort of newly diagnosed DLBCL (n = 120) treated with R-CHOP was investigated. Using the standardized uptake value (SUV) threshold of 4.0, PET/CT radiomics variables (SUVmax, tMTV, tTLG and Dmax) were collected. Circulating tumor DNA (ctDNA) analysis by CAPP-seq yielded baseline ctDNA levels and LymphGen molecular clustering. The best cut-off for both PET/CT parameters and ctDNA levels were identified by max-stat statistics. tMTV, tTLG and Dmax retained independent prognostic value when adjusted for ctDNA levels and were grouped together in a variable named high-risk PET. By multivariate analysis, ctDNA-high and high-risk PET independently predicted PFS and were combined into a 2-factor prognostic model (C-indices: 0.712 for PFS and 0.696 for OS). Molecular clustering, by capturing high-risk biological features of DLBCL, further improved outcome prediction. Consistently, BN2/EZB/ST2 clusters maintained an independent association with better PFS when adjusted for the 2-factor model variables and were therefore included in a 3-factor prognostic score (C-indices: 0.745 for PFS and 0.746 for OS), that identified a very high-risk group of patients (n = 22, 40-month PFS 12.1%) which should be prioritized for early response evaluation and for access to novel agents.