Development and validation of a prediction model based on the organ-based metabolic tumor volume on FDG-PET in patients with differentiated thyroid carcinoma.

BACKGROUND Although patients with differentiated thyroid cancer (DTC) generally have a good prognosis, patients with a large metabolic tumor volume (MTV) on FDG-PET may experience poor clinical courses. We measured organ-based MTVs and tested its prognostic performance in comparison to conventional MTV (cMTV). METHODS We retrospectively analyzed the cases of 280 patients who received their first I-131 therapy in 2003-2014 at our hospital and showed an FDG-avid metastatic lesion. We randomly divided the patients into training (n = 190) and validation (n = 90) datasets. We classified the MTVs as MTVneck-node, MTVdistant-node, MTVlung, MTVbone, and MTVother-organs and tested with/without dichotomization vis-a-vis overall survival (OS). Based on the estimated weighting coefficients of the organ-based MTVs, we propose a new index: the adjusted whole-body MTV (aMTV). Using the validation dataset, we compared the aMTV with cMTV for predicting OS. RESULTS In a univariate analysis, MTVdistant-node and MTVother-organs were more strongly correlated with the OS than the dichotomized forms, whereas the dichotomized forms of MTVneck-node, MTVlung, and MTVbone were more strongly correlated with OS than the continuous variables. The aMTV was thus expressed as 0.69 × dic(MTVneck-node) + 0.02 × MTVdistant-node + 1.05 × dic(MTVlung) + 1.58 × dic(MTVbone) + 0.01 × MTVother-organs, where dic(x) represents 0 or 1 based on the optimized cut-off. In the model evaluation using the validation group, aMTV was a significant predictor of OS with a higher c-index (0.7676) than cMTV (0.7218). CONCLUSION In DTC patients with FDG-avid metastasis before I-131 therapy, all organ-based MTVs were significant predictors of prognosis. As the aMTV outperformed the cMTV for predicting prognoses, we recommend measuring the MTV on an organ basis.