Fig. 4

Evaluation of the nomogram and risk classification system for predicting 3- and 5-year survival. a Receiver operating characteristic (ROC) curve was used for discrimination of 3- or 5-year survival in the validation cohort (TCGA). The areas under the curves (AUCs) for the nomogram were 0.941 and 0.878 respectively, suggesting a very good predictive performance. Calibration curves for predicting patient survival at 3 years (b) and 5 years (c) in the validation cohort (TCGA). Nomogram-predicted probability of survival was plotted on the x-axis; actual survival was plotted on the y-axis. d The AUCs of the nomogram for predicting 3- and 5-years survival were 0.882 and 0.852, respectively, suggesting good predictive performances in the validation set (Gravendeel). e, f Calibration plots for the probability of survival at 3-year and 5-year showed an optimal agreement between the prediction and observation in the validation cohort (Gravendeel). Kaplan–Meier analysis of survival between the low-risk and high-risk groups according to the risk classification system in the primary cohort (g, CGGA) and the validation cohort (h, TCGA and i, Gravendeel)