Abstract P4-02-08: Statistical modeling of influential variables affecting HER2-positivity in breast cancer: Final analyses from two large, multicenter, noninterventional studies in Germany

Abstract

Abstract Background: While HER2 testing in breast cancer (BC) has been routine for over a decade, testing quality remains a challenge. Currently, HER2-positivity rate is the only recommended quality indicator. However, the large, observational, prospective NIU HER2 study in Germany quantified the impact of patient- and tumor-related characteristics such as histologic grade, hormone receptor (HR) status, histologic subtype, age, and nodal status on HER2-positivity, indicating that these factors need to be considered when evaluating HER2-positivity as a measure of testing quality (Rüschoff et al. Mod Pathol 2017). We now report the final analyses from the multicenter EPI HER2 BC study (ML29763, NCT02666261) in Germany where we compared NIU and EPI study data, and aimed to validate the NIU study model. Methods: Data from eligible patients with invasive BC were collected (HER2 test result; patient- and tumor-related factors) and variables influencing HER2-positivity identified and compared between studies. The NIU study model was validated and its predictive power determined using newly collected data from the EPI study, with cutoff and variable coefficients from the previous NIU analysis. Additional promising variables were explored, and their relative influence investigated, using multiple stepwise logistic regression. Results: In total, 14,729 (EPI) and 15,281 (NIU) BC samples were analyzed. Distributions of the main variables were comparable; overall HER2-positivity rates were 13.47% (EPI) and 14.24% (NIU). Fitting the NIU study model to EPI study data demonstrated that all five covariates from the NIU study analyses significantly affected HER2-positivity (p < 0.01); the influence for each covariate differed only slightly between studies (in EPI, histologic grade had most influence followed by histologic subtype, HR status, nodal status, and age). Prediction profiles were used to visualize the relationship between the model-predicted probability of HER2-positivity and the five identified covariates, which showed good comparability between studies. The receiver operating characteristics area under the curve (ROC AUC) of the NIU model used to predict HER2-positivity in the EPI study data was close to that of the model fitted to the NIU data, thus successfully validating the NIU model. To further improve the model, the categorical HR status was replaced by estrogen receptor (ER) and progesterone receptor (PgR) expression. Inclusion of ER and PgR as continuous variables improved the predictive strength of the model (ROC AUC = 0.74; sensitivity = 0.76; specificity = 0.63). Based on this improved model, PgR status had the highest influence on HER2-positivity, followed by histologic grade, histologic subtype, nodal status, ER status, and age. Conclusions: Results from our analyses confirm the statistically and clinically significant influence of patient- and tumor-related factors on HER2-positivity, and highlight the necessity to integrate these factors into the quality control assessment of HER2 testing. Implementation of this model in routine practice may assist in addressing issues with interlaboratory variation, and help to identify centers with HER2 testing problems more accurately. Citation Format: Rüschoff J, Lebeau A, Kreipe H, Sinn P, Schildhaus H-U, Decker T, Ammann J, Künzel C, Koch W, Untch M. Statistical modeling of influential variables affecting HER2-positivity in breast cancer: Final analyses from two large, multicenter, noninterventional studies in Germany [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P4-02-08.