Statistical modelling of HER2-positivity in breast cancer: Final analyses from two large, multicentre, non-interventional studies in Germany. (February 2020)

• Record Type: Journal Article
• Authors: Rüschoff, Josef; Lebeau, Annette; Sinn, Peter; Schildhaus, Hans-Ulrich; Decker, Thomas; Ammann, Johannes; Künzel, Claudia; Koch, Winfried; Untch, Michael
• Is Part Of: Breast. Volume 49(2020)
• Page Start: 246
• Page End: 253
• Publication Date: 2020-02
• Subjects: Invasive breast cancer -- HER2 -- Immunohistochemistry -- In situ hybridisation -- Quality control -- Statistical modelling -- Breast: Pathology -- Breast: Diagnostic (and) imaging; Breast -- Diseases -- Periodicals; Breast -- Tumors -- Periodicals; Breast -- Periodicals; Electronic journals; Periodicals; 616
• DOI: 10.1016/j.breast.2019.12.005
• Languages: English
• ISSNs: 0960-9776

Abstract

Abstract: Background: The German NIU HER2 model was developed based on five variables found to have statistically significant influences on HER2-positivity, to allow exploration of deviations between model-predicted and actual HER2-positivity rates as a measure of testing quality. The prospective, non-interventional EPI HER2 BC study (NCT02666261) compared NIU and EPI data, aiming to validate the NIU model. Methods: HER2 status and patient-/tumour-related information were collected from eligible patients with invasive breast cancer. The influence of variables on HER2-positivity was compared between studies and the NIU model validated using EPI data with cut-off and variable coefficients from the NIU study. The influences of additional variables, centre effects and laboratory-specific parameters were also explored. Results: The study included 14, 729 EPI and 15, 281 NIU samples; HER2-positivity rates were comparable (13.5% versus 14.2%). The five covariates from NIU were shown to significantly affect HER2-positivity using EPI data. The Youden Index for the NIU model refitted to EPI data (0.3632) and the NIU model for prediction of HER2-positivity in EPI (0.3552) was close to that for the NIU model fitted to NIU data (0.3888), validating the NIU model. Replacing hormone receptor status with progesterone and oestrogen receptor expression, and adding method of sample extraction as a variable improved the model's predictive strength (ROC AUC 0.7402; Youden Index 0.3935). Conclusions: Reliable, high-quality HER2-testing methods are essential for selection of patients with HER2-positive breast cancer for HER2-tageted treatment. Integration of our model into a locally used software or website may improve its viability for use in clinical practice. Highlights: Five covariates significantly influenced HER2-positivity in the NIU study. The same five covariates also influenced HER2-positivity in the EPI study. The NIU HER2 model was successfully validated using the EPI study data. Integration of novel variables into the EPI model improved the predictive strength.