OBJECTIVE: To compare the accuracy of axillary dissection with that of axillary node biopsy for staging of lymph nodes in operable breast cancer. DESIGN: Randomised study. SETTING: District hospital, Oestersund, Sweden. SUBJECTS: 200 women with operable breast cancer who presented between 1985-87 and 1989-91. INTERVENTIONS: Women were randomised (n = 100 in each group) to have either axillary dissection (in which all fat tissue was removed from the axilla) or biopsy (in which the lower half of the axillary fat, together with obviously malignant nodes were removed for histopathological examination). MAIN OUTCOME MEASURES: Number of nodes harvested by each method, and number that contained metastases. RESULTS: Nodes that contained metastases were found in 43 of the patients in the dissection group and 46 of those who had undergone biopsy alone. The median (range) yield of nodes in the dissection group was 8.5 (0-16) and in the biopsy group 6 (0-14), p