Abstract
3172
Introduction: Tumor fragmentation (TF) estimated by the ratio between tumor volume and tumor surface was previously shown to be an independent prognostic factor in diffuse large B-cell lymphoma (DLBCL) and to add prognostic value when combined with total metabolic tumor volume (TMTV) (Decazes et al., Eur J Nucl Med Mol Imaging, 2018) both being measured from whole-body [18F]FDG PET/CT scans. We evaluated this biomarker in an independent cohort of DLBCL patients and investigated its prognostic value when combined with TMTV and a second complementary prognostic factor, SDmax, which is the largest distance between two lesions (Cottereau et al., Annals of Oncology, 2021). An appropriate cutoff value for TF was also determined.
Methods: A database of baseline [18F]-FDG PET/CT scans of 286 DLBCL patients from the REMARC trial (NCT01122472) was used. TF, TMTV and SDmax were calculated for all patients based on semi-automated segmentation supervised by two nuclear medicine physicians.
The Spearman’s correlations between TF and TMTV and SDmax were measured. The univariate prognostic value of TF was determined using time-dependent Area under the ROC curve (ROC-AUC) on the entire cohort. A permutation test was performed to assess whether TF was significantly better than random. To test if TF adds prognostic information when combined with TMTV and SDmax, a multivariate permutation test was performed (1000 random reshufflings). At each permutation, a Cox-model with TMTV, SDmax and the shuffled TF values as covariates was trained and evaluated with a 100x5-fold cross-validation. The significance was computed by the proportion of Cox-models with a shuffled TF having a ROC-AUC above the ROC-AUC of the Cox-model with an unshuffled TF. The univariate log-rank test results obtained when using the cut-off published by Decazes (6.0 mm) or the median of TF on REMARC were compared.
Results: The Spearman’s correlations of TF with TMTV and SDmax were 0.64 (P < 1e-6) and -0.14 (P < 0.021) respectively. The ROC-AUCs for PFS of TMTV, SDmax and TF were 0.70 (P < 3e-5), 0.69 (P < 3e-5) and 0.58 (P < 0.034) respectively. For OS, the ROC-AUCs were 0.70 (P < 3e-5), 0.66 (P < 8e-4) and 0.57 (P < 0.084) respectively. Based on the multivariate permutation test, TF had significant prognostic value when used with TMTV and SDmax for PFS but not for OS (P < 0.037 and P < 0.241). The cut-off of TF from Decazes (6.0 mm) did not significantly separate the patients as a function of their survival on the REMARC cohort (P < 0.076 for PFS and P < 0.116 for OS). The median value of TF (3.5 mm) was found to be significant for PFS but not for OS (P < 0.009 and P < 0.078).
Conclusions: TF measured on the baseline [18F]-FDG PET scan was confirmed to be an independent prognostic factor for PFS on a different cohort of DLBCL patients, complementary to TMTV and SDmax, forming a trio of relatively simple and explainable biomarkers. The cut-off to be used to distinguish between short and long PFS still needs further investigation.
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.