BACKGROUND AND OBJECTIVES: The diagnostic and therapeutic significance of axillary dissection has been questioned. We sought to define a subgroup of patients with early-stage breast cancer who are at low risk for positive axillary nodes. METHODS: Between 1970 and 1995, 1,598 women with stage I and II breast cancer underwent level I-II axillary dissection with a minimum of 10 nodes removed. The following factors were examined in univariate analysis for predicting positive nodes: race, method of detection, location of the primary tumor, age, menopausal status, obesity, ER status, PR status, pathologic tumor size, lymphatic vascular invasion, tumor grade, and histology. RESULTS: Four hundred and forty-five of the 1,598 patients (27.8%) had histologically positive axillary nodes. Significant factors in univariate analysis for positive nodes included: tumor size, lymphatic vascular invasion, grade, method of detection, primary tumor location, and age. The only group of women with a 0% risk of axillary nodes were those in whom the pathologic tumor size was < or = 5 mm and mammographically detected. A 5-10% risk of positive axillary nodes was identified in women with (1) pathologic tumor size 6-10 mm, mammographically detected, and age < or = 40 years, and (2) tubular carcinoma < or = 10 mm. Tumors detected on physical examination with or without mammography and women < or = 40 years had a significantly increased risk of nodes. In multivariate analysis lymphatic vascular invasion (P < 0.001), method of detection (P = 0.026), location (P = 0.01), and pathologic tumor size (P = 0.002) were significant predictors of positive axillary lymphadenopathy. CONCLUSIONS: The decision to forego an axillary dissection should be considered in (1) tumors mammographically detected and < or = 5 mm (2) mammographically detected, pathologic size 6-10 mm, age > 40 and (3) tubular carcinoma < or = 10 mm. All other groups had a > 10% risk of nodes and may benefit from axillary dissection.