BACKGROUND: Approximately half of the patients diagnosed with localized esophageal cancer die of metastatic disease within the first 2 years following tumor resection. The development of monoclonal antibodies (MAbs) directed against epithelial cell-associated and tumor antigens has enabled the detection of single disseminated tumor cells in secondary organs. PURPOSE: We used MAbs directed against epithelial cell antigens (i.e., cytokeratins) to determine the proportion of patients with esophageal cancer who display isolated tumor cells in their bone marrow. In addition, we evaluated the prognostic significance of a finding of bone marrow tumor cells in patients with esophageal cancer whose tumors were completely resected. METHODS: Prior to the initiation of treatment, bone marrow was aspirated from both sides of the upper iliac crests of 90 patients with squamous cell carcinoma of the esophagus. Bone marrow was also obtained from a population of 30 individuals who had not been diagnosed with cancer. Tumor cells in cytologic bone marrow preparations were detected by use of an assay that employed the MAbs CK2 (directed against cytokeratin 18), KL1 (directed against a 56,000-kd pan-cytokeratin component), and A45-B/B3 (directed against an epitope common to cytokeratins 8, 18, and 19) plus the alkaline phosphatase anti-alkaline phosphatasestaining method. Bone marrow biopsies, for conventional histologic examination with Giemsa staining, were performed on 62 patients. The Kaplan-Meier method and the logrank test were used to assess disease-free and overall survival according to the presence or absence of tumor cells in the bone marrow of 42 patients with completely resected tumors. Reported P values are two-sided. RESULTS: Cytokeratin-positive tumor cells were detected in the bone marrow of 37 (41.1%) of the 90 total patients. The number of tumor cells detected per 10(5) mononuclear bone marrow cells ranged from one to 82. No significant differences in the numbers of disseminated tumor cells were noted for patients diagnosed with tumors at different stages. Only two (3.2%) of 62 bone marrow specimens examined after Giemsa staining showed morphologically identifiable tumor cells. Tumor cells were not detected in the bone marrow of patients who had not been diagnosed with cancer. After a median follow-up of 15.5 months (range, 6-33 months), 15 (79.0%) of 19 patients with completely resected tumors and tumor cells in their bone marrow had relapses compared with three (13.0%) of 23 patients with completely resected tumors and no tumor cells in their bone marrow (P = .019, logrank test). Patients with completely resected tumors and tumor cells in their bone marrow had significantly shorter overall survival than corresponding patients without tumor cells in their bone marrow (P = .036, logrank test). CONCLUSIONS AND IMPLICATIONS: Dissemination of esophageal cancer cells to the bone marrow is more frequent than expected from the rare occurrence of overt skeletal metastases. In general, the presence of tumor cells in the bone marrow may be an indicator of the disseminatory potential of individual tumors.