Statistical Methods for Assessing Biomarkers

According to the Dictionary of Epidemiology , a biomarker is “a cellular or molecular indicator of exposure, health effects, or susceptibility” (1 , p. 17). In this chapter, the primary focus is on markers of exposure, although the techniques described here can be applied to any type of biomarker. The process of assessing the quality of a biomarker consists of determining if the biomarker has adequate reliability and adequate validity. Reliability refers to “the degree to which the results obtained by a measurement procedure can be replicated” (1 , p. 145). (Reliability is often used interchangeably with the terms repeatability and reproducibility. ) The reliability of a measurement process is most often described in terms of intrarater and interrater reliability. Intrarater reliability (sometimes called intraobserver agreement ) refers to the agreement between two different determinations made by the same individual and interrater reliability (sometimes called interobserver agreement ) refers to the agreement between the determinations made by two different individuals. A reliable biomarker must exhibit adequate levels of both types of reliability. Also of concern in the assessment of the reliability of a biomarker are intersubject, intrasubject, and analytical measurement variability (2 ). The reliability of a biomarker must be established before validity can be examined; if the biomarker cannot be assumed to provide an equivalent result upon repeated determinations on the same biological material, it will not be useful for practical application.