OBJECTIVE This prospective study was performed in an attempt to evaluate (1) the rate of magnetic resonance imaging (MRI) demonstrating bone marrow (BM) abnormalities, (2) the correlation of these abnormalities with a pathologic malignant BM involvement, and (3) the possible modification of patients' usual disease staging in the light of these abnormalities. METHODS After extensive staging investigations, patients' diseases were classified as limited or extensive. Dorsal and lumbar spine MRI was performed for second staging. RESULTS Thirty-two patients were eligible for this study. In ten patients (31.2 percent), MRI showed abnormalities; in four of them, the BM sample from the posterior iliac crest was free from malignancy. Three of these four patients had extensive disease. The last patient, because he had limited disease, had biopsy at the site of MRI abnormalities; the biopsy specimen revealed a malignant involvement and therefore this patient, initially classified as having limited disease, was classified in the extensive disease group. In only 1 of 32 patients, BM-MRI data modified initial staging. CONCLUSIONS The metastases disclosure yield of the MRI in the detection of medullar involvement is higher than BM biopsy but especially in patients with extensive disease. Therefore, MRI should not be considered in routine practice, in particular in patients with extensive disease.