Statistical Considerations in Analysis and Interpretation of Biomarker studies

Biomarker development is a phased program and requires several years to develop for clinical use. For each phase of biomarker development, it is important to customize the study design, statistical analysis, and sample size calculation, in order to evaluate the clearly defined research objective. Biomarker classification performance should be quantified with appropriate metrics, such as TPR, FPR and ROC curves. Having a clear understanding of the research methodology and research goals can improve efficiency for successful biomarkers and prevent wastage of resources and effort on failed biomarkers. The development of biomarkers into diagnostic tests can be categorized into three broad phases: biomarker discovery, the evaluation of biomarker classification performance and the impact of using biomarkers in clinical care. Each phase requires unique statistical considerations and tailored study design to accurately evaluate research objectives. This chapter focuses on human studies evaluating the classification performance of diagnostic biomarkers.