Background: Total Metabolic Tumor Volume (TMTV) and β2-microglobulin (B2M) are established prognostic biomarkers in follicular lymphoma (FL). B2M is included in FLIPI2 and PRIMA-PI scores, while TMTV has recently emerged as a strong independent predictor of outcome. Importantly, the use of a fixed SUV4 threshold for baseline TMTV determination (TMTV4) is now the consensus across lymphoma histologies. This study aims to assess a novel prognostic tool combining B2M and TMTV4, modeled as continuous variables using flexible functional forms.

Methods: The RELEVANCE study (NCT01650701) was a multicenter phase 3 trial in untreated grade 1-3A FL patients (pts), randomized to receive either rituximab (R) plus lenalidomide (18 cycles) or investigator's choice of R-based chemotherapy, both followed by R maintenance every 8 weeks (12 cycles). Efficacy outcomes were similar between arms. A post-hoc analysis was performed on PET-evaluable pts with available B2M and TMTV4 data. The B2M ratio (B2MR) was defined as patient B2M divided by the upper normal limit. Progression-free survival (PFS) was the primary endpoint. The prognostic value of a score combining TMTV4 and B2MR was assessed using Cox models with penalized splines for both covariates. Based on the fitted model, a composite score was derived from the estimated spline functions, and its prognostic performance was evaluated. Model fit, discrimination, and calibration were assessed using AIC, BIC, C-index, AUC, and Brier scores.

Results: From the initial RELEVANCE cohort of 1,030 pts, 42 were excluded due to unconfirmed FL diagnosis. An additional 601 pts were removed because of missing key variables required for model evaluation (TMTV4, B2M, or B2MR). This resulted in a final analysis population of 387 pts, whose baseline characteristics and treatment profiles remained representative of the original RELEVANCE cohort. Median age was 60 years (IQR 50–66) and 51% were female. Pts mostly had advanced-stage FL (92%), bulky disease (53%), bone marrow involvement (54%), and grade 1-2 FL (77%). Median TMTV4 was 332 cm³ (IQR 147 to 759), and the median B2MR was 1.04 (IQR 0.82 to 1.38). There was balanced randomization with 193/387 (49%) of pts assigned to the experimental arm.

Using a reverse Kaplan-Meier method, median follow-up was 121 months; 41% of pts experienced progression or relapse, and 14% died. Univariate Cox models confirmed the prognostic value of both B2MR and TMTV4. Each 1-unit increase in B2MR yielded a hazard ratio (HR) of 1.49 (95% CI: [1.142–1.939], p=0.003), while a doubling of log2-TMTV4 (log2-transformed) corresponded to HR=1.12 [1.035–1.22], p=0.006, for progression or death). Joint analysis using penalized splines in a Cox model showed near-linear associations for both variables, with no significant interaction, leading to a final combined model. This model outperformed FLIPI, FLIPI2, and PRIMA-PI in discrimination and calibration. It achieved a C-index of 0.573, an AUC of 60.1%, a Brier score of 20.3, and the lowest AIC (1735.3) and BIC (1741.5). In comparison, FLIPI, FLIPI2, and PRIMA-PI reached C-index values of 0.554, 0.549, and 0.544; AUCs of 57.6%, 57.0%, and 56.4%; Brier scores of 20.5, 20.6, and 20.7; and AICs of 1739.2, 1739.8, and 1739.6, respectively.

Calibration plots showed good agreement between predicted and observed outcomes, particularly beyond 24 months. The score combining TMTV4 and B2MR stratified pts into three risk groups using a quantile-based approach (Q33;Q66). Compared to the low-risk group, intermediate-risk pts had a HR of 1.58 [1.079–2.301], and high-risk pts had an HR of 1.83 [1.238–2.693] for progression or death. In the high-risk category, the score combining TMTV4 and B2MR matched 74.1% of PRIMA-PI classifications, 53.7% of FLIPI2, and 49.2% of FLIPI.

In addition, an equivalent model combining B2MR with TMTV calculated using a 41% SUVmax threshold showed similar performance, with a C-index of 0.601, AUC of 64%, and Brier score of 19.8%.

Conclusion: The score combining TMTV4 and B2MR is a robust prognostic tool in FL, with the continuous additive model showing the best predictive performance. These results confirm the added value of integrating PET-based tumor burden (TMTV) with biochemical parameters (B2M). This model could enhance risk stratification and guide therapeutic decisions in FL. The use of TMTV4, in line with international PET imaging consensus, supports its integration in future clinical trials and routine practice.

This content is only available as a PDF.
2025
Sign in via your Institution