Gene expression profiles were studied by microarray analysis in 2 sets of archival breast cancer tissues from patients with distinct clinical outcome. Seventy-seven differentially expressed genes were identified when comparing 30 cases with relapse and 30 cases without relapse within 72 months from surgery. These genes had a specific ontological distribution and some of them have been linked to breast cancer in previous studies: AIB1, the two keratin genes KRT5 and KRT15, RAF1, WIF1 and MSH6. Seven out of 77 differentially expressed genes were selected and analyzed by qRT-PCR in 127 cases of breast cancer. The expression levels of 6 upregulated genes (CKMT1B, DDX21, PRKDC, PTPN1, SLPI, YWHAE) showed a significant association to both disease-free and overall survival. Multivariate analysis using the significant factors (i.e., estrogen receptor and lymph node status) as covariates confirmed the association with survival. There was no correlation between the expression level of these genes and other clinical parameters. In contrast, SERPINA3, the only downregulated gene examined, was not associated with survival, but correlated with steroid receptor status. An indirect validation of our genes was provided by calculating their association with survival in 3 publicly available microarray datasets. CKMT1B expression was an independent prognostic marker in all 3 datasets, whereas other genes confirmed their association with disease-free survival in at least 1 dataset. This work provides a novel set of genes that could be used as independent prognostic markers and potential drug targets for breast cancer. ' 2008 Wiley-Liss, Inc.Key words: breast cancer; gene expression; microarray analysis; prognosticThe heterogeneous nature of breast cancer reflects the complexity of the molecular alterations that underlie the development and progression of this disease and poses serious problems to clinical management, also due to the lack of reliable pathological or molecular markers.There are a number of major open questions, such as the evaluation of risk of distant metastasis in cases characterized by the presence of favorable indicators (negative axillary lymph nodes and/or positive estrogen receptors), the prediction of response to chemotherapy and/or antiestrogenic therapy and the prediction of metastases sites for high-risk cancers. Moreover, the principal molecular alterations leading to aggressive clinical behavior, representing potential therapeutic targets, still need to be identified in addition to the well-known factor ERBB2 and the estrogen receptor pathways.Molecular profiling using DNA microarrays have provided sound advancements in this field. The expression profile of gene clusters were useful in classifying breast tumors in biological subgroups with clinical relevance, 1 or in low-versus high-risk classes for relapse, 2-5 or to predict responsiveness to either hormonal-or chemo-therapy. 6,7 In a well-known study, van't Veer et al. addressed the risk of relapse in node-negative patients, 2 one of the most clinicall...