From September 1963 to January 1968, 243 patients with operable breast cancer were included in a randomized trial designed to compare classical radical mastectomy (RM) alone to extended mastectomy (EM), i.e. RM plus internal mammary dissection. One hundred and seventeen patients underwent RM, and 126 EM. After a mean follow-up time of 20 years, no significant differences were observed between the two treatment groups for overall survival, for the relapse-free survival rates, nor for distant metastasis, or locoregional recurrence rates. From a regression model, significant interactions were found between risk of death, EM, and both nodal status and site of the tumour. When compared to RM, EM significantly decreased the risk of death for patients with internal or medial tumour and positive axillary nodes (P = 0.05). No beneficial effect of EM was observed for any of the other patients; on the contrary, EM seemed to increase the risk of death for the patients with external tumour and negative axillary nodes (P = 0.07).