Distant metastases are the major cause of morbidity and mortality in women with breast cancer. The ability to predict the metastatic proclivity is essential in choosing the optimal treatment. Tumor size and grade, which are frequently used markers in node-negative breast cancer patients, are inadequate markers for prognosis and individualized treatment design. The steps in metastatic progression include angiogenesis, invasion, and changes in adhesion characteristics. We developed a strategy for choosing biomarkers representing these steps in malignant progression to identify patients with occult metastases who will need chemotherapy and spare those women whose tumors have not developed the capacity to spread. To evaluate the added significance of E-cadherin to that of nm23-H1 and angiogenesis in determining metastatic proclivity, we used archival material from 168 node-negative breast cancer patients who were treated with mastectomy without any adjuvant chemotherapy or hormone therapy. Immunohistochemistry was used to detect E-cadherin and nm23-H1 expression, whereas angiogenesis was determined by microvessel count (MVC) after immunohistochemical staining. The median follow-up is 14 years. We found that E-cadherin is better in identifying the poor prognosis patients. The 14-year disease-free survival (DFS) is 84%, 80%, and 56% in patients with high, intermediate, and low E-cadherin. The worst prognosis group using nm23-H1 and MVC as biomarkers has a 14-year DFS of 62%. In this group, if E-cadherin is low, the 14-year DFS is further decreased to 44%. Nm23-H1 and MVC are better in identifying the good prognosis patients. The long-term DFS is >90% if MVC is low or if nm23-H1 is high. Multivariate analysis shows that E-cadherin, nm23-H1, and MVC are more significant prognostic biomarkers than tumor size or grade. Loss of E-cadherin appears to be a latter step in the metastatic progression compared to angiogenesis and the loss of nm23-H1 expression.