The prognostic value of oestrogen receptor (ER) and progesterone receptor (PR) was estimated through a multicentric study of 2257 operable breast cancer patients followed up for a median of 8.5 years. None of the patients had received adjuvant therapy. The series included 33.3% stage I patients, 57.1% stage II, 5.7% stage IIIa and 2.4% stage IIIb. At the end point of the study 589 metastases and 537 deaths from cancer were recorded. Receptor measurements were performed by radiolgand assay according to a uniform protocol. A total of 68.8% of the tumous were ER positive and 54.0% PR positive ( > or = 10 fmol mg-1 cytosol protein). In univariate analysis, ER and PR status (positive/negative) were of prognostic value (P < 0.001) for the disease-free interval (DFI), the metastases-free interval (MFI) and the overall survival (OS). The OS of the patients after a first metastasis was also significantly different between ER-positive and -negative tumours (P < 0.001). In multivariate analysis (Cox proportional hazard model, 1665 patients), only the ER status showed a significant difference (P < 0.01) between positive and negative groups regarding the DFI, MFI and OS. By using Cox non-proportional, time-dependent models, we show that the predictive value of ER status of the primary tumour decreases by approximately 20% per year, losing its significance after 8 years of follow-up. Overall, when compared with TNM and histological grading, ER and PR status have a low prognostic value, their major interest remaining solely in the domain of therapeutic decision.