A systematic selection method for the development of cancer staging systems

摘要: The tumor–node–metastasis (TNM) staging system has been the anchor of cancer diagnosis, treatment, and prognosis for many years. For meaningful clinical use, an orderly, progressive condensation T N categories into overall needs to be defined, usually with respect a time-to-event outcome. This can considered as cutpoint selection problem censored response partitioned two ordered categorical covariates their interaction. aim is select best grouping TN categories. A novel bootstrap cutpoint/model method proposed this task by maximizing estimates chosen statistical criteria. criteria are based on prognostic ability including landmark measure explained variation, area under receiver operating characteristic (ROC) curve, concordance probability generalized from Harrell’s c-index. We illustrate utility our applying it colorectal cancer.