Cyclophosphamide, methotrexate and 5-fluorouracile (CMF)-based chemotherapy for adjuvant treatment of breast cancer reduces the risk of relapse. In this exploratory study, we tested the feasibility of identifying molecular markers of recurrence in CMF-treated patients. Using Affymetrix U133A GeneChips, RNA samples from 19 patients with primary breast cancer who had been uniformly treated with adjuvant CMF chemotherapy were analyzed. Two supervised class prediction approaches were used to identify gene markers that can best discriminate between patients who would experience relapse and patients who would remain disease-free. An additional independent validation set of 51 patients and 21 genes were analyzed by quantitative RT-PCR. Applying different algorithms to evaluate our microarray data, we identified two gene expression signatures of 21 and 12 genes containing eight overlapping genes, that predict recurrence in 19 cases with high accuracy (94%). Quantitative RT-PCR demonstrated that six genes from the combined signatures (CXCL9, ITSN2, GNAI2, H2AFX, INDO, and MGC10986) were significantly differentially expressed in the recurrence versus the non-recurrence group of the 19 cases and the independent breast cancer patient cohort (n = 51) treated with CMF. High expression levels of CXCL9, ITSN2, and GNAI2 were associated with prolonged disease-free survival (DFS) (P = 0.029, 0.018 and 0.032, respectively). When patients were stratified by combined CXCL9/ITSN2 or CXCL9/FLJ22028 tumor levels, they exhibited significantly different disease-free survival curves (P = 0.0073 and P = 0.005, respectively). Finally, the CXCL9/ITSN2 and CXCL9/FLJ22028 ratio was an independent prognostic factor (P = 0.034 and P = 0.003, respectively) for DFS by multivariate Cox analysis in the 70-patient cohort. Our data highlight the feasibility of a prognostic assay that is applicable to therapeutic decision-making for breast cancer. Whether the biomarker profile is chemotherapy-specific or whether it is a more general indicator of bad prognosis of breast cancer patients remains to be explored.