Clinical trials design using a biomarker. In the first scenario, no biomarker is used in a randomized clinical trial. With enough patients, a 2 × 2 design may be used. In the second scenario, a biomarker that is being tested for specificity and sensitivity is being tested for treatment effects in experimental and control groups. In the third setting, patients are randomized to either test or not test a promising biomarker. Those with the biomarker are nonrandomly assigned to a specific treatment approach. Those without the biomarker receive conventional therapy. In the fourth scenario, a validated biomarker is being used to nonrandomly assign treatment. It is assumed in the fourth scenario that the biomarker has high sensitivity, specificity, and prevalence within the treatment population to merit its use to nonrandomly assign therapy. In the comparative, conventional therapy group, it is assumed that the therapy has not been significantly modified from that given to earlier patient subsets being used as historical controls. (Adapted from Young et al.⁴⁸ )