245O - A Prospective Randomized Trial Evaluating Gene Expression Arrays to Select Neoadjuvant Chemotherapy Regimen for Operable Breast Cancer: First Report of the Remagus04 Trial

Abstract

ABSTRACT Background DNA arrays allow quantifying gene expression at the whole genome level and could improve prediction of the benefit from a specific chemotherapy (CT). We have evaluated here whether DNA array could increase efficacy of neoadjuvant CT. Patients and methods This prospective randomized phase III trial compared the choice of a CT regimen based on DNA array to a standard CT. Pts with HER2 neg breast carcinoma not eligible for conserving surgery were included. DNA arrays (Affymetrix U133Av2) were performed within 15 days after frozen tumour biopsy. Only samples with >30% cancer cells, a RIN > 6 and a test array positive were eligible. DLD30 prediction score and topoisomerase 2A (TOP2A) level were calculated. Pts randomized to “genomic-driven” CT arm were allocated to 12 weekly paclitaxel then 4 FEC100 if DLD30+ , to 4*FEC → 4*docetaxel (D) if TOP2A+/DLD30- and to 6 D/capecitabine (DC) if DLD30-/TOP2A-. Pts randomized to standard arm were treated with 4*FEC*4 →4*D. The trial, designed to include 300 ps, was stopped after a preplanned interim analysis showing Results Overall 303 pts entered in the screening phase: 64 (21%) were not eligible for the main trial because of exclusion criteria, 60 (25%) presented less than 30% tumor cells in the sample and 37 (15%) presented a poor quality RNA. Overall, DNA array was feasible within 15 days in the context of daily practice in 59% of the pts. 142 pts entered in the clinical decision trial. 39% and 57% of the pts presented an ER-neg or high grade tumour. The overall pCR rate was 22%. No difference was observed between genomic driven arm and standard CT arm (pCR rates: 22 % and 21% respectively). DLD30+ score was associated with an increased likelihood of pCR (36% versus 3% for DLD30-). DLD30+ was associated with OR for pCR at 4.7 (0.87–35.9, p = 0.09) in a multivariate analysis. Additional unplanned retrospective analyses have shown that immune signature could improve prediction of pCR. Conclusion This is the first prospective trial showing that whole expression genome array is feasible in the context of daily practice, quantifies gene expression accurately and validates predictive value of DLD30 and immune signatures. Gene expression arrays could be a solution in the future to propose an all-in-one assay for personalized medicine. Disclosure All authors have declared no conflicts of interest.