Abstract
Introduction: There has been significant progress in the development of quantitative biomarkers derived from FDG PET/CT in diffuse large B-cell lymphoma (DLBCL). Total metabolic tumor volume (MTV) has been the most extensively studied and has proven to be a strong independent prognostic factor in DLBCL. Here, we investigated the effect of body composition on treatment response and survival in patients with relapsed/refractory DLBCL treated with loncastuximab tesirine in a clinical trial cohort.
Methods: In this post-hoc analysis, we reviewed screening baseline FDG PET/CT images of patients enrolled in the LOTIS-2 trial (NCT03589469). Body composition analysis was performed using manual and deep-learning driven automated segmentation of three main tissue compartments, skeletal muscle (SM), subcutaneous fat (SF) and visceral fat (VF), obtained at the L3 level on the baseline CT scans. Manual segmentation was performed by a single experienced nuclear medicine radiologist using MIM Maestro (MIM Software, Cleveland, OH). The deep-learning driven automated method was constructed using the attention-guided U-Net architecture. We tested if these body composition compartments expressed as indices, SM/VF+SF, SM/VF, and SF/VF, could predict treatment response, progression-free survival (PFS) and overall survival (OS). For examining agreement between manual and automated segmentation, we performed Pearson’s correlation analysis. Logistic regression was used to assess the association between body composition indices and treatment response, and cutpoints were defined by maximizing Youden’s index. Cox regression was used to determine the effect of body composition indices on PFS and OS as continuous and binary variables defined by cutpoints obtained from Contal and O’Quigley method.
Results: 140 (96%) of the 145 patients enrolled in the LOTIS-2 trial were available for review. The median age was 65.5 years, the median LDH was 293, and most patients had advanced-stage disease (n=105; 76.1%). Manual (M) and automated (AI) segmentation were highly correlated (P<.0001) for each of the three body composition indices. Univariable analyses show that SM/VF index as dichotomized was a significant predictor of failure to achieve CMR (non-CMR) (M: OR=0.24; 95%CI=0.07-0.85; P=0.027) (AI: OR=0.24; 95%CI=0.08-0.78; P=0.017). Multivariable analysis shows that SM/VF index as dichotomized was a significant predictor of non-CMR (M: P=0.008, AI: P=0.042) after adjusting relevant clinical variables (age, stage, LDH, and MTV). Patients with higher SM/VF indices were more likely to achieve CMR; high SM/VF group (M: ≥3.690, AI: ≥4.045) had 54.5% and 53.8% of patients achieving CMR, respectively, while low SM/VF group (M: <3.690, AI: <4.045) had only 22.5% and 22.0% of patients achieving CMR, respectively. The optimal cutpoints of SM/VF indices determined for PFS were 1.073 (M) and 1.000 (AI), and the same cutpoints were used for OS. Univariable analyses show that SM/VF index as dichotomized was significantly associated with PFS (M: HR=0.56; 95%CI=0.34-0.93; P=0.025) (AI: HR=0.61; 95%CI=0.38-0.99; P=0.049), but not OS (M: HR=0.92; 95%CI=0.61-1.38; P=0.672) (AI: HR=0.87; 95%CI=0.58-1.31; P=0.508). Multivariable analysis shows that SM/VF index (manual only) as dichotomized was a significant predictor of PFS (M: P=0.033) after adjusting clinical variables (age, stage, LDH, and MTV), but not OS (P=0.988). The significant association of SM/VF index with non-CMR and PFS did not change when MTV was included or excluded from the multivariable model.
Conclusions: The present analysis demonstrates a higher SM/VF body composition index was associated with a greater likelihood of achieving CMR and improved PFS, but not a significant difference in OS, in patients with relapsed/refractory DLBCL treated with loncastuximab tesirine. Future studies are needed to establish predictive models incorporating body composition for individualized treatment planning.