Estimating total analytical error and its sources. Techniques to improve method evaluation.

Abstract

The process of method evaluation starts with identifying goals either to demonstrate the clinical validity of an assay or to identify assay error sources that require improvement. Taguchi's idea of continual quality improvement vs the notion of meeting or failing specification has been applied to clinical chemistry. In this article, I propose a model of assay performance that includes the terms random interferences and protocol-specific biases (a series of systematic errors). I explain these terms, as well as the consequences of failing to consider them. To validate an assay clinically, I recommend direct estimation of total analytical error from a method comparison. To identify assay error sources that require improvement, I recommend a multifactor protocol (in addition to a method comparison). Individual error sources are related to total analytical error with the use of an error propagation technique. Much of the proposed data analysis techniques are straightforward but not routinely practiced. I demonstrate principles with the use of a cholesterol assay.