Tumor cell contamination of stem cell collections harvested from breast cancer patients is a common phenomenon described by several investigators but with findings that vary among reports. Although so-called co-mobilization of these cells has been hypothesized, the origin of tumor cell contamination in stem cells is still unknown. A total of 47 G-CSF mobilized stem cell grafts from patients with nodal-positive (n = 30), chemosensitive metastatic (n = 11), and 5 women with inflammatory breast cancer were evaluated for cancer cells by immunocytochemistry. Additionally, 40 bone marrow aspirations and 23 peripheral blood samples collected prior to apheresis and after one to two cycles of conventional chemotherapy were available for examination. Tumor cell contamination of leukapheresis correlated best with preharvest blood state. This was valid when the nominal (positive/negative) presence of tumor cells in blood was compared to the nominal presence of tumor cells in apheresis samples and when the it was correlated to the tumor cell load of apheresis samples (TCL = tumor cells per 10(6) nucleated cells investigated). The correlation between blood and stem cells was better (nominal and quantitative) than that between marrow and stem cells, despite the larger sample size of marrow aspirations. The presence or absence of cancer cells in apheresis samples could not be safely predicted by the presence or absence of tumor cells in marrow or blood alone. Diagnostic specificity seems to improve from a combination of results from marrow and blood analysis. No correlation was found in quantitative analysis of tumor cell contamination between marrow and blood. In conclusion, the results suggest that blood and bone marrow represent different compartments for epithelial cancer cells and that contaminating tumor cells in stem cell harvests may be derived from the blood and/or marrow compartment. The tumor cell contamination of a stem cell harvest cannot be safely predicted by a preceding blood or marrow analysis.