Résumé : The predictive value of the estrogen receptor (ER) assay with regard to the response to hormonal treatment was analyzed in women with advanced breast carcinoma. The significance of ten clinical variables of putative prognostic value was also investigated. A total of 49 courses of endocrine therapy were available for study. The respective merits of using the receptor information as a qualitative or a quantitative variable were compared. Linear logistic regression analysis showed that the quantitative information was significantly related to the therapeutic response (P <0.0001) and proved to be superior to the qualitative information. Compared with the clinical variables tested with the logistic model, receptor concentration was by far the most important single predictor of response. Nevertheless, introduction of two of these clinical variables (i.e., age and menopausal status) into the model in addition to receptor concentration improved its predictive value. Presented in graphic form, the improved model provides a simple means to estimate the probability that a given patient will respond to endocrine therapy. Successive ER assays were available in a series of patients who had received no systemic treatment. In ER+ cases, there was a significant correlation between receptor concentrations in the consecutive assays. There was no influence of the time interval between tissue samplings. Data were also consistent in ER- patients. These results give support to the practice of routine receptor determination in the primary tumor at the time of mastectomy. It is concluded that the distinction between hormone-responsive and hormone-resistant tumors appears artificial. The therapeutic implications of a continuous gradient of hormone-dependency among breast cancers are discussed.