Detection of continuous biomarkers may be undermined by loss of statistical power when they are inappropriately dichotomized in statistical analyses. Similarly, assessment of companion diagnostics may be undermined when continuous disease outcomes are inappropriately dichotomized in the pursuit of 'sensitivity' and 'specificity' for ROC analyses. This presentation explores alternatives to dichotomization when evaluating continuous biomarkers, diagnostics and associated outcomes, with an illustrative example assessing the potential existence and diagnosis of a particular form of asthma which responds especially well to a unique class of drugs.