Abstract P5-03-01: An optimized 92-gene assay for the molecular diagnosis of triple-negative breast cancer

Abstract

Abstract Background: Triple negative breast cancer (TNBC) often presents as high grade, poorly differentiated tumors resulting in a more aggressive disease for which accurate and timely diagnosis is critical to treatment selection or clinical trial enrollment. Furthermore, the high rate of distant metastases and absence of breast-specific immunohistochemical markers that contribute to diagnostic uncertainty may delay or limit treatment modalities that can lead to poorer outcomes. The 92-gene assay is an RT-PCR-based cancer classifier that previously demonstrated 80% accuracy for the diagnosis of breast cancer. In this study, blinded validation of an optimized algorithm and assay specifically developed to improve performance in TNBC is described. Methods: To increase clinical scope for the diagnosis of TNBC, formalin fixed paraffin embedded specimens (N=103) representing a range of breast tumor histologies (e.g. TNBC, adenoid cystic, neuroendocrine, metaplastic, lobular, mucinous, DCIS) were added to the tumor reference database. A revised computational algorithm was constructed by the integration of machine learning techniques. For validation, tumor specimens (N=160) of TNBC (57%) and non-breast tumors (43%) were blindly tested using a 92-gene cancer classifier (CancerTYPE ID®, Biotheranostics, Inc). Tumor type predictions were reported as rank-order probabilities based on the degree of similarity to the tumor reference database. Assay sensitivity based on concordance of the main tumor type prediction with the reference diagnosis established by clinicopathologic review was analyzed. Results: Assay results included 85 breast carcinomas (TNBC) (53%), 23 Salivary gland carcinomas (14%), and 52 carcinomas (33%) representing 11 other tumor types. For performance in TNBC, the 92-gene assay demonstrated an overall sensitivity of 93% (CI, 86-98), and sensitivities of 96% [95% CI, 89-99] and 80% [95% CI, 52-96], in primary and metastatic tumors, respectively (P=0.085). Additional performance characteristics are shown in Table 1. Table 1Pathology subsetN, Validation setN, Correct 92-gene assay predictionsSensitivity (95% CI)All TNBC91850.93 (0.86-0.98)TNBC-primary76730.96 (0.89-0.99)TNBC-metastatic15120.80 (0.52-0.96)All Non-breast69550.80 (0.68-0.88)Salivary gland carcinoma25230.92 (0.74-0.99)Overall performance1601400.88 (0.81-0.92) Conclusions: An optimized 92-gene assay specifically modified to increase performance for the molecular diagnosis of TNBC showed strong accuracy in this blinded study. These findings support use of the 92-gene cancer classifier to aid in the diagnosis of primary or metastatic TNBC. With more refined tumor characterization, TNBC-specific chemotherapy regimens or clinical trial therapies may be pursued with the potential for improved patient outcomes. Citation Format: Sullivan PS, Soifer HS, Liu J, Zhang Y, Schnabel CA, Brachtel EF. An optimized 92-gene assay for the molecular diagnosis of triple-negative breast cancer [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P5-03-01.