There is an urgent need for markers that can predict the efficacy of adjuvant chemotherapy in patients with solid tumors. This study was designed to evaluate whether monitoring of micrometastases in bone marrow can predict the response to systemic chemotherapy in breast cancer. Bone marrow aspirates of 59 newly diagnosed breast cancer patients with either inflammatory (n 23) or advanced (> four nodes involved) disease (n 36) were examined immunocytochemically with the monoclonal anticytokeratin (CK) antibody A45-B/B3 (murine immunoglobulin G1; Micromet, Munich, Germany) before and after chemotherapy with taxanes and anthracyclines. Of 59 patients, 29 (49.2%) and 26 (44.1%) presented with CK-positive tumor cells in bone marrow before and after chemotherapy, respectively. After chemotherapy, less than half of the previously CK-positive patients (14 of 29 patients; 48.3%) had a CK-negative bone marrow finding, and 11 (36.7%) of 30 previously CK-negative patients were CK-positive. At a median follow-up of 19 months (range, 6 to 39 months), Kaplan-Meier analysis of 55 assessable patients revealed a significantly reduced overall survival (P. 011; log-rank test) if CK-positive cells were detected after chemotherapy. In multivariate analysis, the presence of CK-positive tumor cells in bone marrow after chemotherapy was an independent predictor for reduced overall survival (relative risk 5.0; P. 016). The cytotoxic agents currently used for chemotherapy in high-risk breast cancer patients do not completely eliminate CK-positive tumor cells in bone marrow. The presence of these tumor cells after chemotherapy is associated with poor prognosis. Thus, bone marrow monitoring might help predict the response to systemic chemotherapy.