Figure 2

Predictive performance of prognostic models including CNTNAP2 or S100A8. (A) The lines represent the time-dependent area under the ROC curve (AUC (t)) for disease-specific survival using established clinical variables (lymph node status and tumor size) assessed individually and in conjunction with CNTNAP2 protein expression. The estimated performance of the model including CNTNAP2 was marginally better than the model containing the established clinical variables alone, increasing the C-index from 0.941 to 0.949. (B) The lines represent the time-dependent area under the ROC curve (AUC (t)) for overall survival using established clinical variables (lymph node status, tumor size, differentiation, age) assessed individually and in conjunction with S100A8 protein expression. Combining the established clinical variables with S100A8 protein expression increased the C-index significantly from 0.605 to 0.833. The x-axes depict Survival time in months and the y-axes depict AUC (t).