The prognostic relevance of tumor cell detection in the bone marrow of metastasis-free primary breast cancer patients has been controversial ever since the first publications by the London group of R.C. Coombes almost 15 years ago. Up till now, studies comprise data of more than 1,100 patients with median follow-up periods of 24-76 months. In about 35% of primary breast cancer patients and about 25% of node-negative patients, tumor cells can be detected immunocytochemically at the time of primary therapy. The various detection methods used differ mostly in regard to their choice of primary detection antibody. Most authors find a significant prognostic impact of tumor cell detection on relapse-free survival in both univariate and – in comparison to established prognostic factors – multivariate analysis. The little data available also indicate significant prognostic impact in multivariate analysis for overall survival. About 50% of all relapses were predicted by tumor cell detection at time of primary therapy. Only a few studies present follow-up results for the clinically important subgroup of node-negative patients. Regardless of their nodal status, tumor-cell-positive patients seem to have about twice the relapse rate of tumor-cell-negative patients. In conclusion, methodogical improvements are necessary before tumor cell detection in bone marrow can be routinely performed in primary breast cancer patients. In addition, large retrospective as well as prospective studies, especially for node-negative patients, are required, in order to decide whether tumor cell detection identifies a subgroup of high-risk patients that would have been missed by sole consideration of established prognostic factors. However, published follow-up data as well as recent results about the malignant phenotype of tumor cells detected in the bone marrow at time of primary therapy support the notion that some of these detected cells indeed possess metastatic potential.