Abstract
BACKGROUND: Follicular lymphoma (FL), as the most common indolent lymphoma, exhibits reduced survival rates in patients with early progression compared to those without early progression. We aimed to develop a scoring system integrating baseline (18)F-FDG PET/CT tumor metabolic parameters with clinical and other relevant factors to personalize the prediction of disease progression within 24 months (POD24) in FL patients. RESULTS: Among 123 patients with FL, 33 (26.8%) experienced disease progression within 24 months of follow-up. Among clinical indicators, Lactate dehydrogenase (LDH) [OR: 3.267 (1.055, 10.117), P = 0.040] was an independent risk factor for POD24. Based on the receiver operating characteristic (ROC) curve, the optimal cutoff values for Total lesion glycolysis (TLG), total metabolic tumor volume (TMTV), maximum distance of spread (Dmax), and voxel-based maximum interlesion distance (DmaxVox) in POD24 patients were 628.59 g, 94.08 cm³, 37.22 cm, and 77.92 cm, respectively. Additionally, TLG > 628.59 g [OR: 5.1430 (1.9360, 13.6580), P = 0.001], TMTV > 94.08 cm³ [OR: 11.8530 (2.6730, 52.5640), P = 0.001], Dmax > 37.22 cm [OR: 3.0000 (1.1880, 8.6878), P = 0.028], and DmaxVox > 77.92 cm [OR: 8.0620 (2.2830, 28.4720), P = 0.001] were all risk factors for POD24 patients. The correlation heatmap revealed significant positive correlations between TLG and TMTV (r = 0.95) and between Dmax and DmaxVox (r = 0.99). Patients were stratified into three risk groups based on clinical LDH and imaging parameters TMTV and DmaxVox, with POD24 incidence rates of 5.7%, 26.1%, and 57.1%, respectively (P < 0.001). The constructed integrated model demonstrated the highest predictive performance for POD24 (AUC = 0.759). Decision curve analysis (DCA) and calibration curves confirmed the model's robust predictive capability. CONCLUSIONS: TMTV and DmaxVox on baseline (18)F-FDG PET/CT imaging, along with the clinical parameter LDH, are independent predictors of early disease progression within 24 months (POD24) in patients with FL. A predictive model combining these imaging metrics with clinical parameters effectively stratifies patient risk.