PD03-02: Prognostic and Predictive Predictors for Triple Negative Breast Cancer.

Abstract

Abstract Background: Both the prognosis and the therapeutic options in triple negative breast cancer (TNBC) are rather limited. Current prognostic gene expression profiles for breast cancer mainly reflect proliferation status and are most useful in ER-positive cancers. The identification of prognostic gene signatures from TNBC cohorts in previous studies was hindered due to relatively small sample sizes. Materials and Methods: All currently available TNBC gene expression datasets generated on Affymetrix U133 gene chips were assembled. To minimize inter-laboratory variation we analyzed only highly comparable arrays and data set-biased genes were filtered. Supervised analysis was applied to identify a prognostic signature from a finding cohort of 394 TNBC and validation was performed in an independent cohort of 261 TNBC. The genes from the prognostic predictor were analyzed for their correlation to known molecular phenotypes among TNBC. Results: Two supervised prognostic signatures consisting of 264 and 26 probesets, respectively, were obtained when applying different cutoffs for false discovery rates of 25% and < 3.5% in the finding cohort. In multivariate analysis in the independent validation cohort hazard ratios of 4.03 (95% CI 1.71−9.48; P=0.001) and 4.08 (95% CI 1.79−9.28; P=0.001), respectively, were obtained for the two signatures. When compared to 16 metagenes for previously described molecular phenotypes in TNBC the prognostic signatures displayed highest correlation to metagenes for IL-8/inflammation, VEGF/angiogenesis, and Histones. A subset of genes in the 264-probeset signature was inversely associated with a poor prognosis (29/264=11.0%). Most of these “good prognosis” genes are correlated with immune cell metagenes (21/29=72.4%). In contrast both identified supervised prognostic signatures did not correlate to previously published prognostic signatures (recurrence score, genomic grade index, Amsterdam signature, wound response signature, 7-gene immune response module, stroma derived prognostic predictor, and a medullary like signature). Regarding the response of TNBC to neoadjuvant chemotherapy the predictive value of the B-cell metagene was superior to the 264- and 26-probeset signatures. However combination of the B-cell metagene and the signatures increased the AUC in ROC-analysis from 0.606 to 0.656. Conclusions: The use datasets consisting only of TNBC allows identification supervised prognostic signatures for TNBC which are unrelated to previously known prognostic signatures. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr PD03-02.