Risk factors for distant metastases following mastectomy and axillary node dissection for breast cancer were analyzed in a review of 1022 women. From diagnosis until the end of the adjuvant treatment, six stages were identified that corresponded well to patient data acquisition. At each stage, a prognosis study based on the Cox model was carried out using all acquired information from the first stage. The results demonstrated that tumor size, nuclear pleomorphism, mitotic index, and nodal status at the top of axilla were stable independent risk factors in predicting metastasis‐free survival (MFS). These analyses also revealed those factors that were significantly related to MFS at one or several stages and losing their significance at a subsequent stage. This was the case with clinical node status, age, and vascular tumor emboli. Other factors such as estrogen, progesterone, histologic grade, and clinical stage were never identified as independent factors at any stage. The four major stable risk factors were used to define a stratification of reference. The results demonstrated that the mere knowledge of clinical information such as tumor size, clinical node status, and age would enable 51% of the patients to be universally well classified according to that stratification. Knowledge of additional factors, such as nuclear pleomorphism and mitotic index, would bring the rate up to 61%, and then to 64% if supplementary information such as vascular tumor emboli were acquired. These percentages did not appear high enough to claim that the physician may make a reliable prognosis of operable breast cancer patients before acquiring information from the axillary node dissection. However, it was proven that there exist some subsets of patients with stable prognosis, i.e., subsets of patients who will belong permanently to the same risk group through the stages. Copyright © 1990 American Cancer Society