Thirty-five to 40% of patients with operable breast carcinoma develop metastases after primary therapy. There is a need for more specific prognostic parameters to identify patients who are most likely to benefit from adjuvant therapy. The success of such treatment stems from its ability to eradicate preclinical microscopic metastases. The bone marrow is an accessible and frequent site of breast carcinoma metastases. Following studies of Redding et al. (16), we used monoclonal antibodies that recognize membrane and cytoskeletal antigens expressed by epithelial cells (C26, T16, AE-1) in an immunohistochemical assay to find cancer cells in bone marrow aspirates. The assay can detect one cancer cell among 50,000-100,000 hematopoietic cells. None of the 44 control bone marrows (from normal individuals and patients with leukemias and lymphomas) contained antigen-positive (extrinsic) cells. We found extrinsic cells in the bone marrow of 35% (18 of 51) of patients with operable breast carcinoma; no extrinsic cells were identified by routine bone marrow cytology in these patients. Twenty-seven percent (six of 22) of patients with negative lymph nodes had antigen-positive cells, while 41% (12 of 29) of patients with lymph node metastases had such cells. Similarly, 23% (three of 13) of patients with TNM stage I disease, 38% (13 of 34) of patients with stage II disease, and 50% (two of four) of patients with stage III disease had extrinsic cells. In those cases where extrinsic cells were identified, stage II patients with negative lymph nodes and patients with stage I disease were found to have fewer such cells in their marrow than patients with lymph node metastases and patients with stage II disease. These trends did not reach the level of statistical significance in this small number of patients. The presence of extrinsic cells did not correlate with tumor size of lymphatic invasion around the tumor. We conclude that the epithelial cells detected in the bone marrow of the patients with breast carcinoma were carcinoma cells based on the following criteria: (a) they expressed both membrane and cytoplasmic epithelia-specific antigens, (b) they possessed the cytologic characteristics of malignant epithelial cells, and (c) these cells were not detected in the bone marrow from normal individuals or patients with nonepithelial neoplasms involving the bone marrow. We have shown that the technique described here can detect occult metastases in bone marrow and that the presence of extrinsic cells correlates with some established predictors of prognosis. Long-term clinical correlative follow-up studies are now underway.