survivalROC.C about Survival Model Predictive Accuracy and ROC curve

In Survival Model Predictive Accuracy and ROC curve (Heaherty and Zheng 2005), there are 2 examples, the first one is VA lung cancer data and the second one is mayo PBC data.

I want to know how they define markers in these two datasets and how they plot the AUC in the paper with R. A model score is derived using Cox regression with Karnofsky score, age, and cell type. ROC curves are estimated using a varying-coefficient Cox model with the derived model score as the single predictor.

I am not sure what it means. Can anyone help me to figure it out?