In order to release correct biomarker results of a laboratory test, it is a regulatory requirement to apply quality control standards for controlling analytical errors. Releasing an incorrect test result might lead to wrong diagnosis or treatment of a patient in medical decision-making. In laboratory medicine, one of the means to control analytical errors is statistical process control procedures proposed by James O. Westgard and his coworkers nowadays known as "Westgard rules." To judge their performance for discriminating in-control from out-of-control processes, power curves are used. In this article, we describe functions for the power curves of the within-run Westgard rules. Based on these power curves, we use a benchmark approach for selecting a quality control procedure out of the set of Westgard rules. It is shown that two graphical procedures proposed by Westgard and his coworkers can be reduced to this benchmark approach. Besides, a commonly used measure in laboratory medicine for describing out-of-control processes is critically examined revealing the threat of selecting too optimistic quality control rules.
Keywords: companion diagnostics; individualized medicine; laboratory test; quality control; statistical process control.
© 2021 John Wiley & Sons Ltd.