Determining agreement between immunohistochemistry and RT-qPCR for standard biomarkers in breast cancer: Validation on GEICAM 9906 clinical trial.

611 Background: Common practice for comparing different technologies measuring the same biomarker is to establish cut-offs using Receiver Operator Characteristic (ROC) curves and maximizing sensitivity/specificity. We introduce a new method for establishing RT-qPCR cut-points using gene expression across biologic subtypes and compare to standard methods. Methods: A RT-qPCR training set was established to identify biologic subtypes of breast cancer using the PAM50 gene expression signature. 155 invasive breast cancers and 16 normal breast samples, procured in formalin-fixed paraffin-embedded tissue blocks, were used to identify prototypic samples for the common biologic subtypes (Luminal A, Luminal B, HER2-enriched, Basal-like, and Normal). Cut-points for ESR1, PGR, and ERBB2 were determined based on their expression across the training set subtypes and used to determine sensitivity/specificity in an independent test set of 793 invasive breast cancers from the GEICAM 9906 study (randomized phase III trial ...