Integrated clinical and radiomic PET-FLIPI score for risk stratification in newly diagnosed follicular lymphoma

Abstract

Abstract Background: Follicular lymphoma (FL) has a heterogeneous clinical course, and despite advances in therapy, a significant proportion of patients experiences relapse. Existing prognostic models do not fully capture the biological heterogeneity of disease. PET/CT-derived biomarkers such as total metabolic tumor volume (TMTV) and the maximum lesion distance (Dmax) have emerged as prognostic tools, though their combined utility in FL remains underexplored. This study aimed to develop a composite prognostic score incorporating clinical and imaging metrics to improve baseline risk stratification in patients with newly diagnosed FL. Methods This is a post hoc analysis of patients enrolled in the FOLL12 trial which had avalaible baseline TMTV calculated. TMTV was calculated using the SUV4 threshold method, and the cut-off of 180 mL based on previously published study (Durmo et al., Am J Hematol) was used for analysis. Dmax was derived from lesion coordinates of TMTV and normalized by body surface area (SDmax). The optimal SDmax cut-point for progression-free survival (PFS) was determined using maximally selected log-rank test. A composite score was derived assigning weighted points to features that were indipendend in multivariable analysis. Model performance was assessed using Harrell's c-index. Internal validation for Sdmax and model through 1000 bootstrap resamples was used. Study primary endpoint was PFS and overall survival (OS) was a secondary endpoint. Results This analysis included 682 treatment-naïve patients with grade 1-3a FL. 336 (50%) was older than 60 years and 315 (46%) was male. The cohort was balanced across treatment arms (experimental arm n=351). A total 286 (42%) patients had R-Bendamustine as induction treatment. Median SDmax was 284 mm/m² (IQR: 247; range: 4–1037 mm/m²). An optimal SDmax cut-point of 400 mm/m² was identified, stratifying patients into low (SDmax ≤400 mm/m², n=538, 79%) and high (SDmax >400 mm/m², n=144, 21%) dissemination groups. SDmax >400 mm/m² and TMTV >180 mL were independently associated with inferior PFS (HR of 1.50, 95% CI: 1.12–2.02 and 1.32 95% CI: 1.01–1.73, respectively). FLIPI-2 and male sex also were independent prognostic factors. These variables were consistently selected in bootstrap resampling and used to derive a combined prognostic score ranging from 0 to 6 points, 1 point each for high TMTV, high SDmax, male sex and 3 points for high FLIPI2. The composite score effectively stratified our cohort into three risk groups: 307 (45%) patients into low-risk (score 0–1) with a 5-year PFS of 79% (95% CI, 73-83%), 140 (21%) of patients into intermediate-risk (score 2–3): 5-year PFS 65% (95% CI 56-72%, HR=1.86) and 235 patients (34%) into a high-risk (score 4–6) with a 5-year PFS 52% (95% CI 45-59% , HR=2.93). The prognostic role of the composite score was confirmed by treatment arm. Incorporating PET metrics improved model discrimination over FLIPI-2 alone (c-index: 0.636 vs. 0.600; p=0.015). Internal validation demonstrated excellent reproducibility of the model (bootstrap-corrected c index of 0.632) and a shrinkage factor of 0.995, indicating minimal overfitting and stable predictive performance. 5-year OS rates was 97% for score 0–1 (n=307), 94% for score 2–3 (n=140; HR = 2.44, 95% CI: 1.05–5.67), and 88% for score 4–6 (n=235; HR = 4.96, 95% CI: 2.45–10.0). Conclusion A composite prognostic score integrating FLIPI-2 with baseline TMTV, SDmax, and sex improves risk stratification in FL. This integrated model identifies patients at highest risk of progression and may support personalized therapeutic strategies in future trials.