The JMP® Passing-Bablok Add-In as a Tool for Robust Regression and Method Comparison in the Field of Diagnostics
Anja Wöerner, Project Manager, Roche Diagnostics GmbH
The method comparison (MC) approach is used as a statistical instrument in the performance evaluation of diagnostic assays. In this context, MC studies are conducted with the aim to demonstrate that two analytical methods (X and Y) are measuring the same, allowing both methods to be subject to experimental error. Simplified, it is a regression technique to determine the linear relationship and the bias at clinical decision points. The support for clinical decision making with in vitro diagnostic tests is a challenge in daily biostatistics service in a wide variety of indication fields, including oncology, virology, cardiovascular, inflammatory and infectious diseases. Within R&D, robust procedures for method comparison are strongly preferred to monitor changes in development and to show how close analytical methods do measure the same. The FDA mainly recommends Deming Regression, but the Passing Bablok (PB) procedure is accepted as well, and due to its robustness with respect to outliers and heteroscedasticity, PB is acknowledged by the Clinical and Laboratory Standards Institute (CLSI) Committee in the new guidance document CLSI EP09-A3 (Draft Guideline—Third Edition). Following the reference of the CLSI, PB is an appropriate method of choice to run method comparison studies. This talk shows the implementation of the PB regression procedure and the use of it within different typical experiments from assay development.