Background: The purpose of this study is to assess the predictive accuracy of a multi-gene predictor of response to docetaxel, 5-fluorouracil, epirubicin and cyclophosphamide combination chemotherapy on gene expression data from patients who received these drugs as neoadjuvant treatment.
Methods: Tumor samples were obtained from patients with stage II-III breast cancer before starting neoadjuvant chemotherapy with four cycles of 5-fluorouracil/epirubicin/cyclophosphamide (FEC) followed by four cycles of docetaxel/capecitabine (TX) on US Oncology clinical trial 02-103. Most patients with HER-2-positive cancer also received trastuzumab (H). The chemotherapy predictor (TFEC-MGP) was developed from publicly available gene expression data of 42 breast cancer cell-lines with corresponding in vitro chemotherapy sensitivity results for the four chemotherapy drugs. No predictor was developed for treatment with trastuzumab. The predictive performance of TFEC-MGP in distinguishing cases with pathologic complete response from those with residual disease was evaluated for the FEC/TX and FEC/TX plus H group separately. The area under the receiver-operating characteristic curve (AU-ROC) was used as the metric of predictive performance. Genomic predictions were performed blinded to clinical outcome.
Results: The AU-ROC was 0.70 (95% CI: 0.57-0.82) for the FEC/TX group (n=66) and 0.43 (95% CI: 0.20-0.66) for the FEC/TX plus H group (n=25). Among the patients treated with FEC/TX, the AU-ROC was 0.69 (95% CI: 0.52-0.86) for estrogen receptor (ER)-negative (n=28) and it was 0.59 (95% CI: 0.36-0.82) for ER-positive cancers (n=37). ER status was not reported for one patient.
Conclusions: Our results indicate that the cell line derived 291-probeset genomic predictor of response to FEC/TX combination chemotherapy shows good performance in a blinded validation study, particularly in ER-negative patients.