Disease recurrence remains a major limitation to the use of marrow transplantation to treat leukemia. Previous transplant studies have demonstrated that higher doses of total-body irradiation result in less disease recurrence, but more toxicity. In this study, the possibility of delivering radiotherapy specifically to marrow using a radiolabeled anti-CD33 antibody (p67) was explored. Biodistribution studies were performed in nine patients using .05-.5 mg/kg p67 trace-labeled with 131I. In most patients initial specific uptake of 131I-p67 in the marrow was seen, but the half-life of the radiolabel in the marrow space was relatively brief, ranging from 9-41 hr, presumably due to modulation of the 131I-p67-CD33 complex with subsequent digestion and release of 131I from the marrow space. In four of nine patients these biodistribution studies demonstrated that with 131I-p67 marrow and spleen would receive more radiation than any normal nonhematopoietic organ, and therefore these four patients were treated with 110-330 mCi 131I conjugated to p67 followed by a standard transplant regimen of cyclophosphamide plus 12 Gy TBI. All four patients tolerated the procedure well and three of the four are alive in remission 195-477 days posttransplant. This study demonstrates the feasibility of using a radiolabeled antimyeloid antibody as part of a marrow transplant preparative regimen and also highlights a major limitation of using conventionally labeled anti-CD33--namely, the short residence time in marrow. Strategies to overcome this limitation include the use of alternative labeling techniques or the selection of cell surface stable antigens as targets.