We have previously used a chromium-release assay to demonstrate that the cocktail of monoclonal antibodies BA-1, BA-2, BA-3, and complement can effectively lyse human leukemic cells in the presence of excess bone marrow. Using a leukemic cell colony assay, we have reinvestigated the variables influencing lysis of human leukemic cells (KM-3, HPB-NULL, NALM-6) in bone marrow using BA-1, BA-2, BA-3, and complement. Specific variables addressed included the concentration of excess bone marrow cells, the number of treatments, the presence or absence of DNase during the treatment, the combination of antibodies, and the sensitivity of different leukemic cell lines to lysis. Using the colony assay, the BA-1,2,3 cocktail was shown to be more effective than any single antibody or combination of two antibodies. We also determined that the concentration of excess bone marrow cells and number of treatments had a direct bearing on leukemic cell lysis. Although two cycles of treatment were significantly superior to one cycle, three cycles were not significantly superior to two cycles. Inclusion of DNase (10 μg/mL) was a critical adjunct that eliminated clumping and facilitated plating cells in the colony assay. Finally, we could show that striking differences existed in the sensitivity of the leukemic cell lines to lysis with the BA-1,2,3 cocktail and complement. NALM-6 cells were the most sensitive (approximately four logs of kill), and KM-3 cells were the most resistant (less than two logs of kill). Our results strongly support the utility of sensitive leukemic cell colony assays in the analysis of marrow treatment variables in autologous bone marrow transplantation.