Characterize and Dichotomize a Continuous Biomarker

Dichotomization of continuous biomarkers is both clinically relevant and desirable since clinical decisions are usually based on categorization of patients into groups (i.e., stratification or subgroup identification for tailoring). However, such categorization may result in significant loss of information. It is important that the analytical method evaluates and incorporates the continuum of biomarker values to guide the optimization of a clinically appropriate cutpoint. This chapter starts with an Sect. 1 that provides the key clinical objectives, motivating real-world examples and challenges in dichotomizing a continuous biomarker. A summary of current methods will be provided in the Sect. 2, including optimal cutpoint selection and continuum-assessment approaches. The methods discussed in this chapter not only include inferential-based approaches for dichotomizing a continuous biomarker, but also visualization methods that characterize the relationship between a continuous-scale biomarker and the clinical outcome. We also briefly discuss the use of biomarker signature as an extension to the single marker scenario. Finally, best practice and method extensions will be discussed in the Sect. 4.