The presence of circulating tumor cells in bone marrow and peripheral blood of cancer patients may reflect the aggressiveness of the disease. This also applies to cancers that rarely give rise to overt bone marrow metastases. The clinical validity of micrometastasis detection for staging and prognostication depends on the sensitivity and reliability of the detection method. In malignant melanoma, most studies have used reverse transcriptase polymerase chain reaction (RT-PCR) techniques, commonly with tyrosinase mRNA as the target molecule. Unfortunately, highly inconsistent results have been reported, raising doubts about this approach. In a study of 81 melanoma patients with metastatic disease, we used an immunobead rosetting method in which live melanoma cells are selected and identified by binding of paramagnetic beads coated with the 9.2.27 antibody against the high molecular weight melanoma-associated antigen. In bone marrow samples obtained from 60 patients, 14 (23.3%) were positive, compared to only two of 81 in blood. A highly significant correlation (p = 0.0001, log rank test) was found between micrometastasis positivity and overall survival from time of removal of the primary tumor. Moreover, in regression analysis it was found that the presence of micrometastatic cells was an independent and the most important indicator of poor prognosis, with a relative risk of 5.38. The immunomagnetic method is simple, rapid, and highly sensitive and will be used in further prospective clinical studies.