AUC^a calculated by testing prediction models on SEER^b.

AUCa: area under the receiver operating characteristic curves. SEERb: National Cancer Institute's Surveillance, Epidemiology, and End Results. variable selectionc: the variable selection method which has the highest AUC. BSFSd: variable selection using backward stepwise feature selection. GAe: variable selection using genetic algorithms. Globalf: without variable selection. *: median AUC of 15 tests. **: comparing the AUC of prediction models with TNM staging system. &: comparing the AUC of prediction models with 20 variables to that with 14 variables.