Abstract P1-01-05: Agreement rates of pathologist-derived manual and digital read and image analysis quantitation for membrane and nuclear-based immunohistochemistry biomarkers in breast cancer clinical studies

Abstract

Abstract Introduction: Over the past 18 years, pathologists have had an increasing responsibility of quantitating immunohistochemistry (IHC) markers with the expectation of high intra- and inter-reader reproducibility. These markers are related to treatment prediction or patient prognosis and are predominately represented by estrogen receptor (ER), progesterone receptor (PR), human epithelial growth factor receptor 2 (HER2), and Ki-67. Many agencies have instituted quality programs to increase these metrics with dramatic effect; however, improvement is still needed. Digital imaging-based quantitation of IHC offers the potential for increasing intra- and inter-reader reproducibility. In this study, we culminate digital and imaging analysis data across multiple regulatory submissions to understand intra- and inter-reproducibility metrics in the context of five breast cancer biomarkers. Materials and Methods: For each IHC marker (ER, PR, HER2, Ki-67, and p53), 120 invasive whole tissue breast cancer cases were included from retrospective clinical archives and restained. Cases were represented by binning into three IHC staining categories (ER, PR, Ki-67, and p53: 0 to 0.99%, 1 to <10%, and >=10%; HER2: 0/1+, 2+, 3+) representing 33% of samples for each binned category. Three pathologists (9 total) scored each case per marker as positive/negative around a cut-point (ER and PR: 1%; Ki-67 and p53: 10%, and HER2: 2+) utilizing three scoring modes (manual read [MR], digital read [DR], and imaging analysis [IA]) with a 1-2 week washout. Primary end-point of DR and IA was non-inferiority to MR. Statistical analyses were performed using pair-wise analyses of MR vs DR or IA. Results: IA inter-reader reproducibility overall percent agreements (OPA) were at least similar to manual read OPAs for ER, PR, Ki-67, and p53. HER2 IA OPA showed an improvement over MR. DR inter-reader reproducibility OPAs were similar to manual reads for ER, PR, HER2, and p53. Ki-67 DR inter-reader reproducibility underperformed MR (Table 1). Intra-reader reproducibility OPAs across scoring modalities with respect to any single scorer were similar for ER, PR, HER2 and Ki-67 (>88% 95%CI 81.3-99.1%), while p53 was lower (82.9% 95%CI 75.1-91.6%). Table 1: Inter-reader reproducibility OPAs across MR, DR, and IA for each breast cancer marker MR (95% CI)DR (95% CI)IA (95% CI)ER94.9% (91.4-97.8%)92.0% (87.8-95.8%)95.3% (92.0-98.2%)PR94.4% (90.9-97.2%)94.0% (90.2-97.1%)94.1% (90.3-97.2%)HER285% (77-90%)92% (86-96%)95% (89-98%)Ki-6785.6% (80.4-90.4%)76.6% (70.9-82.2%)86.8% (82.1-91.4%)p5378.8% (72.2-83.3%)80.6% (75.0-86.0%)81.7% (76.4-86.8%) Conclusions: Overall, DR and IA IHC-based quantitation across these biomarkers showed at least similar results to MR. Some markers over or underperform from this MR baseline most likely related to analysis approach and/or marker heterogeneity. Citation Format: Barnes M, Srinivas C, Frederick J, Bai I, Little E, Sarkar A, Sabata B, Bamford P, Ranger-Moore J. Agreement rates of pathologist-derived manual and digital read and image analysis quantitation for membrane and nuclear-based immunohistochemistry biomarkers in breast cancer clinical studies. [abstract]. In: Proceedings of the Thirty-Eighth Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2015 Dec 8-12; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2016;76(4 Suppl):Abstract nr P1-01-05.