Evaluation of the performance of breast cancer cell line–derived multigene predictors of chemotherapy response in multiple clinical trials.

Abstract

64 Background: While various multi-gene predictors (MGPs) of chemotherapy response have been developed based on cancer patient primary tissues or cancer cell-lines, the accuracy and consistency of these predictors remain a concern in clinical validation studies. In this study we developed four unique MGPs for chemotherapy response from breast cancer cell lines and performed a systematic evaluation of the performance of these MGPs using data from five distinct clinical trials. Methods: Forty-six immortalized breast cancer cell-lines were exposed to various concentrations of drug combinations [paclitaxel, 5-fluorouracil, doxorubicin, cyclophosphamide (TFAC); 5-fluorouracil, doxorubicin, cyclophosphamide (FAC); 5-fluorouracil, epirubicin, cyclophosphamide (FEC) and epirubicin, cyclophosphamide (EC)] using an in vitro chemosensitivity assay. Utilizing publicly available breast cancer cell-line microarray data, genes highly associated with in vitro chemosensitivity were selected as candidate MGPs. Five independent and publicly available clinical trials were used for validation. In three of these clinical trials patients were treated by TFAC, while EC, FAC or FEC were used in the other two trials. All five studies involved neoadjuvant chemotherapy treatment, and pathologic complete response (pCR) was used as the endpoint. The association of MGPs with pCR was assessed using receiver-operator curve (ROC) analysis and area under the ROC (AUC) was used to evaluate the performance of prediction. Results: In five independent clinical trials, the MGPs predicted patient pCR to EC, FAC/FEC and three TFAC treatments with an AUC of, 0.671, 0.632, 0.735, 0.738 and 0.647 respectively. Conclusions: In the five independent clinical trials in which patients were treated by various chemotherapy agents, the performance of MGPs is promising. These results demonstrate the feasibility of using breast cancer cell-line derived MGPs to predict breast cancer patients’ chemotherapy responses.