Human epidermal growth factor receptor 2 (HER2, alias ERBB2)-targeted therapy in breast and gastric cancers depends on the reliable assessment of HER2 protein expression and (in equivocal cases) the quantitative evaluation of HER2 gene amplification. Typically, HER2 and centromere 17 gene copy numbers are evaluated using in situ hybridization (ISH) to calculate ratios for which cutoff values dividing nonamplified and amplified cases have been proposed. Although several studies have investigated how laboratory procedures affect diagnostics, a rigorous quantitative assessment of the diagnostic guidelines for data analysis is still missing. Here, we analyze the dependence of the diagnosed HER2/chromosome 17 ratios on i) sample size (evaluated cells), ii) gene/chromosome signal distributions, and iii) the approach used for quotient calculation using Monte Carlo simulations. Our data show that the current recommendation may lead to statistical HER2/CHR17 ratio variations of up to 0.94 and may therefore lead to incorrect HER2 status diagnoses, given the ratio threshold of 2.0 defined by the Food and Drug Administration. Moreover, borderline cases may receive different amplification diagnoses, depending on the ratio calculation approach: Brightfield-silver ISH with aggregated signal counts may underestimate the HER2/CHR17 ratio compared with two-color fluorescence ISH. Our results provide a basis for quantitative rationales behind HER2 diagnostic guidelines that call for increased numbers of evaluated cells and emphasize the importance of well-designed data analysis methods in diagnostic pathology, especially for predictive clinical application.
Copyright © 2012 American Society for Investigative Pathology and the Association for Molecular Pathology. Published by Elsevier Inc. All rights reserved.