There are no previous reports of trials that have directly compared the effects of adjuvant chemotherapy with oophorectomy in premenopausal women with node-positive breast cancer. During 10 years we recruited 332 such women who were randomised, after mastectomy or conservation therapy, to receive either ovarian ablation or cyclophosphamide/methotrexate/5-fluorouracil (CMF) chemotherapy, each with or without prednisolone 7.5 mg daily for 5 years. After a maximum follow-up of 12 years, we detected no significant overall differences in relapse rates, or in event-free or total survival for ovarian ablation compared with chemotherapy or for prednisolone versus no prednisolone, nor any suggestion of an interaction between these factors. Actuarial total survival at 8 years was 60% overall, irrespective of treatment, with a hazard ratio and 95% Cl of 1.12 (0.76-1.63) for the comparison of CMF with ovarian ablation and 1.26 (0.86-1.84) for prednisolone versus no prednisolone. Oestrogen receptor (ER) assays were done in 270 (81%) primary tumours but these results played no part in the randomisation procedure. When patient outcome was analysed in relation to the concentration of ER in the tumour, there was a statistically significant interaction between ER content and treatment, such that ovarian ablation was associated with improved survival in patients with ER concentrations 20 fmols/mg protein or more and CMF was more beneficial for patients with values less than 20 fmols/mg protein. No such interaction was seen for prednisolone therapy. Oestrogen receptor content has a role in decisions about treatment for primary breast cancer.