Résumé : The knowledge of estrogen receptor (ER) status is important in the management of breast cancer patients. More precisely, analytical methods for ER determination have changed over the last two decades from ligand binding assay (LBA) dextran-coated charcoal (DCC) to enzyme immuno-assay (EIA) and more recently immunohistochemistry (IHC). We examined the respective clinical impact of ER determination according to these 3 methods over the period 1983-1999 within a group of 1940 patients, all operated and followed in the single institution Centre Antoine Lacassagne. Validated cut off values were 10 and 15 fmol/mg protein for both LBA-DCC and EIA, respectively and 10% of stained cells for IHC. During the years it was noted that the initial size of the tumor decreased and that the proportion of positive axillary nodes and negative ER tumors was different according to the ER method. ER negativity was 20, 13 and 10% in LBA-DCC, EIA, IHC, respectively. ER was a strong predictor of overall survival in the whole population (Mantel-Cox, p < 0.00001); however when stratifying the analysis on ER method groups, ER was still a prognostic indicator in the EIA, LBA-DCC group but not in the IHC group (the follow-up was too short). It is important to keep these data in mind when conducting large retrospective studies evaluating prognostic markers in breast cancer patients.