Detection of tumor cells in blood and bone marrow is increasingly used for the staging of patients with breast cancer and to evaluate the presence of tumor cells in peripheral blood progenitor cell collections to be used after high-dose therapy. We evaluated the sensitivity and specificity of three different methods for detection of tumor cells among non-tumor tissue. An immunocytochemical assay using antibodies directed against epitopes of the cytokeratin-19 (CK19) protein and two RNA-based methods: reverse transcriptase polymerase chain reaction (RT-PCR) and Nucleic Acid Sequence-Based Amplification (NASBA) for the same target gene were tested. With all the three methods, false-positive results were observed when peripheral blood mononuclear cells (PBMC) of healthy volunteers were tested. There was no concordance between the RNA-based assays and the immunocytochemical assay. The false-positive results in the RNA-based assays may be due to 'illegitimate expression' of epithelial genes in normal PBMC. The false-positive results in the immunocytochemical assay resulted from background staining of monocytes and granulocytes. This study demonstrates that CK19 is not a suitable target to detect the presence of breast tumor tells in PBMC. To reliably detect circulating tumor cells with RNA methods, the selection of suitable target genes is required, which are highly expressed in tumors but not at all in normal cells of blood and bone marrow. Genes with such characteristics may be identifiable with novel differential display techniques.