Sensitivity in statistical evaluation of biomarkers

Abstract

In clinical research and practice, biomarkers help in understanding disease progression and are useful when monitoring patients and making treatment decisions. Correct quantification of biomarkers in predicting the clinical outcome is important for better treatment of individual patients. Despite the rich literature in statistical validation, application of these validation methods in real data is not well studied. Specifically, the question is whether the change in a biomarker or the actual assessed value of the biomarker should be used. Contrary to most published papers in which the actual value of the biomarker is used, we showed through theory, simulation, and an example that it is more appropriate to use the change or the actual value in the marker with adjustment for the baseline value in evaluating a marker. © 2013 John Wiley & Sons, Ltd.