Cut-point identification in biomarkers for a censored failure time outcome

Categorization is a crucial step for clinical decision making when dealing with diagnostic (or prognostic) biomarkers. We extend the use of three cut-point finding methods widely used in the absence of censoring, i.e., Youden index, concordance probability and point closest to-(0,1) corner in the ROC plane, to the case of censored failure time outcome. Further, we investigate the performance of the three methods by a simulation protocol. A motivating application of cut-point finding of a molecular biomarker in acute lymphoblastic leukemia is also presented