Prognostic value of gene signatures and tumorbiological characteristics in breast cancer patients treated with anthracycline-containing chemotherapy

Background: Many prognostic and predictive multigene signatures have been established in breast cancer patients. For treatment decision the assessment of individual prognosis is essential. The choice of specific therapy is basically driven by empirical data although several predictive gene signatures already exist. In this context it would be valuable if specific signatures could be used for estimation of prognosis and prediction of therapy concurrently. Material and Methods: Microarray data (Affymetrix HG U133A) of a small samples set of n = 48 breast cancer patients who received an anthracycline-based adjuvant chemotherapy were analyzed. Tumor samples were classified according to four prognostic and two predictive previously described gene signatures and compared with standard parameters as histologic subtype, tumor size, nodal status, pathohistological grading as well as estrogen receptor and Her-2 status. Results: The gene expression values of ER, PR and Her-2 from microarray revealed a high concordance with protein expression assessed by means of immunohistochemistry. The determination of proliferative state of the tumors using gene expression of Ki67 showed a significant correlation with ER-status (p = 0.040, Mann-Whitney U-test) and pathohistological grading (p = 0.005, Kruskal-Wallis test). Neither of the six different signatures was able to predict event status of patients sufficiently. The main discriminatory power of the signatures was related to the ER status and to some extent to pathohistological grading. Conclusion: In a small cohort of uniformly treated patients prognostic and predictive gene signatures are incapable to predict disease outcome unambiguously. The main driving force of all signatures are ER-status and proliferation. The value of the individual signatures may be restricted to the specific setting from which they were derived.