Baseline PET/CT metabolic parameters in the double-expressor subtype of diffuse large B-cell lymphoma: development of a clinical-radiologic-pathologic predictive model.

Background: Double-expressor lymphoma (DEL) is an aggressive diffuse large B-cell lymphoma (DLBCL) subtype (20%-30% of cases) exhibiting resistance to standard immunochemotherapy regimens [R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone)]. Multiple clinical studies have demonstrated that the combined use of new drugs such as Chidamide can significantly improve outcomes in DEL, underscoring the need for early identification of high-risk patients to guide therapy. In this context, baseline 18F-FDG positron emission tomography/computed tomography (PET/CT) metabolic parameters are poised to become a pivotal tool for optimizing risk stratification in DEL, owing to their unique capability to non-invasively quantify tumor metabolic heterogeneity. Methods: We retrospectively analyzed clinical and baseline 18F-FDG PET/CT imaging data from treatment-naive patients with DLBCL at the Northern Theater Command General Hospital from January 2020 to February 2025. Patients were classified into a DEL group and a non-DEL group. PET/CT parameters were correlated with clinical-pathologic features using Spearman analysis. Optimal thresholds for maximum standardized uptake value (SUVmax), total metabolic tumor volume (TMTV), and total lesion glycolysis (TLG) were determined by receiver operating characteristic (ROC) analysis. Univariate and multivariate analyses were performed to identify independent predictors, followed by the development and validation of a combined prediction model using ROC analysis, calibration curves, and decision curve analysis (DCA). Results: A total of 128 patients (71 men and 57 women, median age, 60 years, range, 16 to 87 years) were included in the non-DEL group (n = 82) and DEL group (n = 46). Spearman analysis revealed that PET/CT parameters significantly correlated with the International Prognostic Index (IPI), Ann Arbor stage, B symptoms, β2-MG, LDH, and non-GCB (r = 0.20-0.68, p < 0.05), but not Ki-67 (p > 0.05). TMTV demonstrated optimal DEL prediction (threshold = 510.22 cm³, AUC = 0.81, 95% CI, 0.73-0.88 p = 0.038). TMTV (>510.22 cm³, OR = 8.79, 95% CI, 3.20-24.11, p < 0.001), IPI (OR = 3.82, 95% CI, 1.44-10.11, p = 0.007), and Ki-67 (OR = 1.07, 95% CI, 1.03-1.11, p = 0.001) were identified as independent DEL predictors. The TMTV+IPI+Ki-67 combined model (AUC = 0.867, p < 0.05) showed significantly higher discriminative performance compared to dual-parameter models (TMTV+IPI, AUC = 0.798; TMTV+Ki-67, AUC = 0.844; IPI+Ki-67, AUC = 0.797, all p < 0.05). This superiority was further confirmed through calibration curves and DCA, indicating its reliable predictive accuracy and clinical utility. Conclusions: TMTV, IPI, and Ki-67 are robust independent predictors of DEL. The integrated clinical-imaging-pathological prediction model constructed from these three parameters synergistically combines multi-dimensional information, significantly enhancing early DEL identification capability and facilitating the implementation of risk-adapted therapeutic strategies.