504 evaluable node positive oestrogen receptor (ER) positive breast cancer patients were randomly allocated to receive either 5 years tamoxifen treatment or chemotherapy [six courses of cyclophosphamide, methotrexate and 5-fluorouracil (CMF) followed by 4 courses of epirubicin] or a combination of both treatments. At a median follow-up of 5 years tamoxifen appeared to be more effective than chemotherapy, the difference being highly significant in postmenopausal women. The addition of chemotherapy to tamoxifen was not able to significantly improve the results achieved by tamoxifen alone, irrespective of menopausal status. Trends were similar even after stratification for the number of involved nodes. The protective effect of tamoxifen in terms of reduction of the odds of death increased with time and no rebound phenomena on recurrence or death has occurred so far after the completion of tamoxifen treatment. Overall, the prognostic value of number of involved nodes and of progesterone receptor (PgR) status was confirmed by multivariate analysis. However, the predictive value of PgR was lost in patients receiving tamoxifen alone. Similarly, the degree of ER positivity was not predictive of the response to tamoxifen. Tamoxifen treatment should still be regarded as the gold standard for postmenopausal ER positive patients. In younger women the antioestrogen proved to be safe and at least as effective as chemotherapy. However, the analysis of the annual risks suggests that the concurrent or the sequential use of chemotherapy and tamoxifen might represent a more appropriate treatment for this patient subset, particularly for those with four or more involved nodes. Different cut-offs of ER and PgR assays from those we have arbitrarily employed in the present analysis should probably be used to select more properly the patients who can benefit from endocrine therapy.