Eighty bone marrow studies (each including 4 aspirates and 4 trephine biopsies) were performed in 37 children with stage IV neuroblastoma to assess the most accurate means for detection of invasion by neuroblastoma cells. Among 38 abnormal results, only trephine biopsy(ies) were found positive in 24 cases (63%), only aspirate(s) in 5 cases (13%), and both in 9 cases (24%). In 37% of abnormal results, only 1 of the 8 tests performed was found positive. No benefit was obtained from either associated touch imprints of iliac biopsies (67 investigations), or exploration of extra-iliac sites (66 sternums, 53 tibias). Two monoclonal antibodies, claimed to be specific for detection of neuroblastoma cells in bone marrow, were used in 56 investigations; they could detect minimal residual disease in some cases, but they were unreliable when no staining was obtained if initial phenotype of neuroblastoma had not been assessed, or when few isolated cells were observed. Prospective studies using immunocytology and immunohistology are thus warranted.