There is general agreement that bone marrow (BM) examination for staging in patients with retinoblastoma should be limited to cases with advanced disease. However, there are limited data about the yield of sampling multiple sites with aspirations and biopsies and immunocytology. Our policy for BM examination included: 2 aspirates and 2 biopsies at the posterior iliac crest scheduled only for cases with postlaminar optic nerve extension (n=56), scleral invasion (n=10) or orbital (n=5) or metastatic disease at diagnosis (n=7) or at extraocular relapse (n=18). Immunocytology with the antibodies 3A7 or 3F8 for the ganglioside GD2 was performed. From 1/1994 to 3/2005, 277 newly diagnosed patients and 5 at extraocular relapse were included. BM invasion was not found in any of the 66 patients enucleated with disease confined to the globe, but was found in 11/27 of those with overt extraocular disease. There were 2/11 cases with at least 1 negative aspirate with positive biopsy and/or immunocytology for GD2. GD2 positivity was found in 9/9 cases. A more aggressive BM evaluation has a low yield in enucleated patients with high-risk features but disease limited to the globe. However, in cases with overt extraocular dissemination, the use of BM biopsy and immunocytology for GD2 allowed for the detection of cases that would have been missed by aspirations alone. GD2 was intensively expressed and it may also be used to monitor disease response and the presence of minimal residual disease.