In recent years, individualized genomic information that measures the underlying biology have proved to be clinically helpful tools in cancer research and patient management. Such biomarkers often are composite indexes based on multiple genes, predicting the disease progression, and some have been included in the guidelines for cancer staging or guiding personalized treatment. The clinical utility includes, but not limited to, electing conservative disease management to avoid overtreatment, choosing appropriate drug for improved efficacy, or expediting the enrollment of the investigational drug trials by providing non-invasive measurement of the biomarker. One of the most important aspects is that a novel biomarker needs to demonstrate that, adding the biomarker to the model based on important clinical and pathologic factors substantively improves the predictive accuracy of the model. This talk will discuss the statistical considerations to develop a validated, fit for purpose clinical diagnostic application in order to improve overall clinical outcome, using an example in prostate cancer to illustrate several abovementioned aspects in the tumor marker validation and utilization.