Abstract P6-03-01: Variables influencing HER2-positivity in breast cancer: Assessment and validation of a statistical model based on two multicenter noninterventional studies in Germany

Abstract

Abstract Background: HER2 testing in breast cancer (BC), routine for >10 years, allows selection of patients (pts) for HER2-targeted therapy; however, testing quality remains a concern. While guidelines recommend assessment of HER2-positivity rates as a quality indicator, the influence of patient- or tumor-related factors on variability was unknown until we identified the effect (in order of influence) of histologic grade, hormone receptor (HR) status, histologic subtype, age, and nodal status in a large, multicenter, observational study in Germany (NIU HER2 study; Rüschoff et al., Mod Pathol 2017). Based on these variables and the statistical model developed, potential issues with HER2 testing quality in local practice may be identified. We now report interim analyses from a multicenter study in Germany (EPI HER2 BC study; NCT02666261), where data from the NIU and EPI studies were compared and the validity of the NIU study model assessed. Methods: Routine HER2 test results and patient- and tumor-related data were collected from eligible pts with BC. Factors influencing HER2-positivity rates in the EPI study were compared with those identified in the NIU study. The predictive power of the NIU study model, fitted to EPI data, was determined and assessments performed using the variable coefficients and cutoff resulting from the NIU study analysis. Attempts were also made to improve the model. Results: Analyses included 15281 (NIU) and 6019 (EPI) invasive BC samples. The distribution of relevant variables, including HER2-positivity rate (NIU: 14.4%; EPI: 13.5%), was comparable. When the NIU study model was fitted to EPI study data, all five covariates identified in the NIU analyses had a significant effect on HER2-positivity (p<0.001); the order of influence for covariates differed between studies (EPI [in order of influence]: histologic grading, histologic subtype, HR status, nodal status, and age). The relationship between HER2-positivity rate and the combined influence of covariates, visualized with the NIU study prediction profiler, was reproduced with EPI study data. The NIU study statistical model, with variable coefficients and cut-point determined in the NIU study, was used to predict the HER2-positivity of samples in EPI; if their NIU model-estimated probability of positivity was >0.1407, the resulting sensitivity, specificity, and receiver operating characteristic (ROC) area under the curve (AUC) were 0.7032, 0.6622, and 0.7259, respectively. Thus, initial validation of the NIU study model with EPI data was successful. Semiquantitative estrogen and progesterone receptor expression data were available from EPI only; their inclusion as independent continuous, rather than categorical, variables improved the model (ROC AUC = 0.7533). Conclusions: The statistical modeling approach used to analyze data from the NIU study showed that patient- or tumor-related characteristics should be considered when assessing HER2 testing quality. Our present analysis validates and improves upon this statistical model and further highlights the need to assess HER2 testing quality in BC. Comparison of calculated vs actual positivity rates may help identify centers with potential HER2 testing quality issues. Citation Format: Rüschoff J, Lebeau A, Kreipe H, Gerharz CD, Sinn P, Schildhaus H-U, Tennstedt-Schenk C, Ammann JU, Künzel C, Koch W, Untch M. Variables influencing HER2-positivity in breast cancer: Assessment and validation of a statistical model based on two multicenter noninterventional studies in Germany [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr P6-03-01.