The use of specific monoclonal antibodies (mAb) enables the detection of single malignant cells in the bone marrow of patients with breast cancer. Due to the current discrepancies regarding cell preparation methods and detection frequencies, which are very apparent when surveying the literature, we decided in this study to optimise and standardise these in order to determine the sensitivity and specificity of these immunocytochemical assays. For the immunocytochemical detection of tumour cells, we used the anti-MUG 1 mAb BM7, the anti cytokeratin (CK 8,18,19) mAb SD3, and the anti-GA733-2 mAb MOC31, all of which are expressed on breast cancer cells and which have been successfully used diagnostically. Cytospin preparations of bone marrow (BM) and peripheral blood (PB) cells obtained from normal healthy donors and patients with primary breast cancer were stained by the immune alkaline phosphatase anti-alkaline phosphatase (APAAP) technique using the Techmate 500 staining automate. In control experiments performed on BM and PB from healthy donors, BM7-positive cells were found in 4 out of 8 BM samples, in 6 out of 25 PB samples and MOC31-positive cells were detected in 2 out of 25 PB samples. No 5D3-positive cells were detected in these samples. However, 5D3-positive cells were detected by the APAAP technique in 35 out of 98 (35.7%) BM samples from patients with primary breast cancer. We were able to quantitate the number of tumour cells per slide on these cytospin preparations. The immunocytochemical technique using the 5D3 mAb against cytokeratins 8,18,19 for the detection of tumour cells in BM by this standardised procedure showed that it is an easy, quick, and reliable method with a sensitivity of one tumour cell per 10(6) normal cells.