OBJECTIVE Up to 30% to 40% of metastases from hormone receptor-positive primary breast cancer do not respond to endocrine therapy. We studied how often hormone receptor status changes between primary and recurrent tumors and whether such a change might explain unresponsiveness to endocrine therapy. METHODS Primary breast cancer samples and matched asynchronous recurrences were studied from 50 patients who had not received any adjuvant therapy. Estrogen receptor (ER) and progesterone receptor (PR) status was determined immunohistochemically from histologically representative formalin-fixed paraffin-embedded tumor samples. ER status was ascertained by mRNA in situ hybridization. RESULTS Thirty-five (70%) of 50 primary tumors were positive for ER and 30 (60%) for PR. Hormone receptor status of the recurrent tumor differed from that of the primary tumor in 18 cases (36%). Discordant cases were due to the loss of ER (n = 6), loss of PR (n = 6), or loss of both receptors (n = 6). Receptor-negative primary tumors were always accompanied by receptor-negative recurrences. Among 27 patients with ER-positive primary tumors, loss of ER was a significant predictor (P = .0085) of poor response to subsequent endocrine therapy. Only one of eight patients (12.5%) with lost ER expression responded to tamoxifen therapy, whereas the response rate was 74% (14 of 19) for patients whose recurrent tumors retained ER expression. CONCLUSIONS Loss of ER expression in recurrent breast cancer should be considered as a cause for poor response to endocrine therapy in primarily ER-positive patients. We conclude that analysis of recurrent tumor samples may improve the predictive value of ER and PR assays.