Identification of a Prognostic Signature Based on the Expression of Insulin-Related Genes in Early Breast Cancer.

Abstract

Abstract Background: A substantial body of evidence indicate that the insulin pathway plays a key role in breast cancer development and progression and may represent a therapeutic target, especially in those patients exposed to high plasmatic level of insulin. However, the potential prognostic role of genes related to the insulin-pathway in breast cancer cells has not been explored. With these premises, we evaluated the prognostic role of the expression of genes related to the insulin pathway in early breast cancer. Methods: Candidate genes were selected from published literature, genomic databases, and gene expression profiling experiments performed in insulin resistant subjects, yielding 143 genes that were used to develop a molecular classifier. We used three publicly available breast cancer datasets, GSE1456, GSE3494 and GSE2990 that include gene expression data on a total of 502 cases with clinical follow up. The insulin gene signature was developed on GSE1456, containing microarray data from 159 early breast cancer patients. This dataset was split by a random procedure into a training set and a validation set. Univariate non-parametric Mann-Whitney U test was used to identify genes differentially expressed. Expression of genes significantly correlated with relapse was combined in a linear score. Patients were classified as low or high risk with respect to the median score. External validation was performed on GSE3494 and GSE2990. Results: On the training set, 15 genes resulted differentially expressed in relapsed and non relapsed patients: the 8-year disease free survival (DFS) was 91% (SE =4%) and 51% (SE = 8%) in the high and low risk group (p< 0.001); HR = 10.6 (95% CI 3.2-35.5, p<0.00001). In the validation set, the 8-year RFS was 97% (SE = 3%) and 54% (SE = 10%), respectively (p = 0.009); HR = 4.6 (95% CI 1.01 to 20.7, p 0.04). External validation was performed on two independent datasets, GSE3494 and GSE 2990 including 350 early breast cancer patients. In GSE3494 the 8-year DFS was 72% (SE = 5%) and 61% (SE = 4%) in the high and low risk group (p = 0.03), respectively. In GSE 2990 the 8-year DFS was 74% (SE = 7%) and 55% (SE = 8%), respectively, (p = 0.03). By multivariate analysis, the insulin signature resulted significantly associated with DFS, independently of age, tumor size, ER status, nodal status and grade. Conclusions: Our findings indicate that the insulin pathway is involved in breast cancer prognosis at a genomic level and might provide a better way to individualize therapeutic interventions targeting insulin signaling. Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 108.