Prospective analysis of mRNA expression signatures as predictive tests in primary breast cancer

2531

Background: Preoperative chemotherapy is increasingly employed to treat primary breast cancer, allowing an ‘in vivo chemosensitivity test’. Markers which predict a pathological complete response are urgently needed to refine this strategy. This study was conducted to evaluate the use of gene expression profiling to predict response to neoadjuvant anthracycline- or taxane-based chemotherapy. Methods: Patients with operable or locally advanced HER2-negative breast cancer received preoperative chemotherapy: either dose- dense doxorubicin and cyclophosphamide (ddAC) or capecitabine and docetaxel (CD). Core needle biopsies were taken before treatment and gene expression profiling was performed using 35k oligo microarrays. Results: Gene expression profiles were obtained from pretreatment biopsies of 63 tumors. 27% of the patients achieved a (near) pathologic complete remission (pCR), 40% of the patients had a partial remission and 33% of the patients did not respond to chemotherapy. Based on the gene expression profiles, tumors were assigned to the previously identified “molecular subtypes” luminal, basal-like or ERBB2-like (Sorlie et al., PNAS 98: 10869, 2001). 13 out of 25 patients with a basal-like tumor (52%) achieved a complete remission, whereas for the luminal tumors a pCR was only obtained in 2 out of 29 patients. Using four published gene expression classifiers of response to chemotherapy, a reasonable separation between responders and non-responders could be observed for two of these. We also performed exploratory supervised classification analyses on our dataset to identify a novel classifier. This resulted in a classifier for response to therapy irrespective of the chemotherapy regimen used and a second classifier specifically associated with response to ddAC chemotherapy. We will perform validation of these classifiers in samples from patients that are currently being enrolled in the study. Conclusions: Basal-like tumors have a better response to neoadjuvant chemotherapy as compared to other tumor types. The identification of robust gene expression signatures for better response prediction may require larger patient groups and should probably be established separately for each of the molecular subtypes of breast cancer.

No significant financial relationships to disclose.

Prediction of response to neoadjuvant chemotherapy in breast cancer

20005

Background: Neoadjuvant chemotherapy is increasingly employed in operable breast cancer. Our initial studies on a cDNA array platform failed to identify gene expression patterns predicting response to neoadjuvant chemotherapy in breast cancer patients (J Clin Oncol 23:3331, 2005). Now we included more patients and used oligo microarrays. Methods: Patients with operable or locally advanced breast cancer were included in a randomized phase II study or received neoadjuvant chemotherapy off protocol. All except 7 patients began chemotherapy with 3 courses of dose-dense adriamycin and cyclophosphamide (ddAC) and response was evaluated by MRI. Patients with a response and a HER2-positive tumor were then randomized between either 3 additional courses of ddAC or six weekly courses of carboplatin, paclitaxel and trastuzumab (CPT). Patients without response were switched to CPT. Patients with HER2-negative tumors were randomized between 3 courses of either ddAC or capecitabine and docetaxel (CD). After evaluation, patients without response were switched to the alternative treatment arm. From all patients 14G core needle biopsies were taken before treatment and total RNA was isolated. Amplified mRNA was labeled and hybridized to 35k human oligo microarrays from our microarray facility. Results: So far, 77 patients have been included into the study. From 48 of these, good quality RNA from tissue with >50% tumor cells was isolated. 43 patients had received ddAC as initial chemotherapy; 32 of these had not been switched to another regimen. In a training set containing 11 pathological complete remissions (pCR) and 9 non-responders (NR) we could separate these groups by using 20 genes in a supervised classification and a 9-step cross validation. These results could be validated in an independent set of 11 samples (6 pCR, 5 NR). From 10 out of 11 samples, response status could be predicted correctly, independent from the treatment regimen. Although ER-positive tumors have a lower pCR rate than ER-negative ones, the steroid hormone receptors were not present in the classifier. Conclusions: We conclude that it should be possible to identify a reliable gene expression profile associated with response to adriamycin based neoadjuvant chemotherapy in breast cancer.

No significant financial relationships to disclose.