Estrogen receptor (ER) status is an essential determinant of clinical and biological behavior of human breast cancers. While ER-positive breast cancers respond well to adjuvant hormone therapy, ER-negative tumors are generally resistant. To date, no attempts have succeeded in finding molecular markers for classifying ER-negative breast cancers with respect to postoperative prognosis. To identify a set of prognostic markers for this type of cancer, we used a cDNA microarray consisting of 25,344 human genes to investigate expression profiles of ten primary breast cancers from patients who had died of breast cancer within 5 years after surgery (5y-D) and 10 from patients who had survived disease-free for more than 5 years (5y-S). Sets of genes characterizing each group were identified by Mann-Whitney and random-permutation tests. We documented 71 genes with higher expression in the 5y-D group than in the 5y-S group, and 15 with higher expression in the 5y-S group than in the 5y-D group. Semi-quantitative RT-PCR experiments were carried out to confirm the results of the microarray analysis. We established a scoring system for predicting postoperative prognosis of ER-negative breast cancers on the basis of aberrant gene expression. The list of genes reported here provides valuable information with regard to progression of breast cancer and is a source of possible target molecules for development of novel drugs to treat patients with ER-negative breast cancers.