Feasibility Assessment Of Pharmacogenomic Predictors Developed From Breast Cancer Cell Lines To Predict Breast Cancer Patient Pathological Response In Neoadjuvant Chemotherapy

Background: Studies of developing pharmacogenomic predictors from cancer cell lines to predict cancer patient clinical response and outcome to chemotherapy have yielded both positive and negative results. The variability in these results may arise from the noise inherent of microarray technology as well as the small sample size of in cell line studies. Therefore, it is greatly needed to evaluate the feasibility of using cell lines to develop pharmacogenomic predictors of patient pathological response. Material and Methods: Thirty breast cancer cell lines were exposed to 10 concentrations of paclitaxel-5-fluorouracil-doxorubicin-cyclophosphamide (TFAC) to measure in vitro chemosensitivity. Two independent and publicly available microarray datasets (Hoeflich and Neve) on these breast cancer cell lines together with the chemosensitivity results were used to identify pharmacogenomic predictors from each dataset. Two independent clinical trials (Hess and Popovici) with publically available data, having 130 and 100 patients respectively, were used as test sets to validate the accuracy of the pharmacogenomic predictors. All patients received TFAC as neoadjuvant therapy and the gene expression profiles of patients were assessed before receiving chemotherapy treatment. The patient9s pathological complete response (pCR) was determined after treatment to evaluate the chemotherapy efficacy. The pharmacogenomic predictors developed from each of the cell line studies were evaluated for their ability to predict patient pCR in each of the clinical trials using the supervised principle component regression method. Results: The pharmacogenomic predictors identified from the Hoeflich and Neve cell line data (training sets) predicted pCR of the patients in the two clinical trials (test sets) with 64%-68% accuracy, 70%-87% sensitivity, and 60%-67% specificity when 100 genes were selected as pharmacogenomic predictors (Table 1). Conclusions: The four independent prediction results generated in this study demonstrate the feasibility of using breast cancer cell lines to identify pharmacogenomic predictors of response to chemotherapy treatment for breast cancer patients. Future studies will examine the use of drug responses from primary cultures of patient tumors to develop pharmacogenomic predictors of breast cancer patient responses to chemotherapy treatment. Table 1. Prediction of breast cancer patients’ pCR by pharmacogenomic predictors derived from breast cancer cell lines. Citation Information: Cancer Res 2010;70(24 Suppl):Abstract nr P2-09-10.