Therapy prediction in a breast cancer primary systemic chemotherapy (PST) study applying docetaxel/epirubicin/cyclophosphamide (TEC) by gene expression profiling

Abstract

10049 Background: Currently there are no tests to assist in selecting the optimal PST regimen for breast cancer patients. Primary study goals of this prospective, single-armed multicentric investigation are pathologically confirmed tumor response and the rate of breast conserving therapy (BCT). Secondary goals are to find histopathologic and gene profiling patterns best correlating with tumor remission in a taxane- anthracycline based neoadjuvant setting as well as to evaluate cytostatic toxicity and quality of life. Methods: In this phase II study of totally 40 eligible patients with invasive breast cancer Human Genome Survey Microarray (HGSM) expression profiling is performed on jet-biopsy sample basis. The protocol was elaborated for the treatment of patients suffering from a primary tumor with 6 cycles of TEC (3-weekly) prior to the surgical treatment. The selection of predictor genes was done with BRB-ArrayTools Version 3.3 using a model based on the Compound Covariate Predictor, Diagonal Linear Discriminant Analysis, Nearest Neighbor Classification, and Support Vector Machines with linear kernel.We estimated the prediction error of each model using leave-one-out cross-validation (LOOCV) as described by Simon R. 2000 random permutations were used. Clustering was done using Cluster 3.0 and Java TreeView 1.0.12. Results: Tumor response (pCR, pPR) of more than 70% can be achieved using neoadjuvant TEC-regimen. 22% pCR (ypT0; ypN0) and 90% BCT in this study are comparable with data of other published PST trials. Preliminary expression profiling results reveal a subset of 148 genes that classifies all patients with a complete remission (pCR), in one cluster with a very closely related gene expression pattern (n=5; PPV = 100%). Furthermore 10 patients defined as responders due to selected MIB1-expression based criteria (expressing cells in the residual tumor ≤ 5% and a Δ MIB1-expression ≥ 20%) can be correctly classified in 9 of 10 cases. Comparable separation of the groups could not be achieved by established tumor factors. Conclusions: HGSM expression profiling is promising to have the potential to figure out genes that are related to chemotherapy response, especially in PST. No significant financial relationships to disclose.