Dibromomannitol (DBM) is a new agent for the treatment of chronic granulocytic leukemia. A propsective evaluation of the drug was undertaken in a randomized comparison with busulfan. Forty previously untreated, Philadelphia chromosome-positive cases were treated, with 20 patients in each treatment group. The protocol provided for continuous maintenance therapy after remission induction, with a crossover to the opposite drug in patients who became refractory to the primary agent but are without evidence of blastic tranformation. There were 14 remissions in the DBM group and 15 in those treated with busulfan. The rate of decrease of the elevated leukocyte count was more rapid with DBM, but prolonged disease control off treatment occurred in only three of 14 cases as opposed to nine of fifteen busulfan-treated patients who required a median delay of 12 mo before maintenance could be initiated. Hypoplasia occurred in one DBM patient and two busulfan cases. Following recovery, crossover to the opposite drug in two cases again resulted in hypopllasia. Increased skin pigmentation, amenorrhea, pulmonary fibrosis, and cytologic dysplasia, commonly associated with busulfan adminstration, were also noted with DBM. The median duration of disease control with busulfan was 34 mo and 26 mo with DBM. There was no signigicant difference in the incidence of blastic transformation, and median survival for both groups was 44 mo. DBM appears to be as effective as busulfan in the treatment of the chronic phase of CGL but with a more predictable myelosuppressive action. The principal advantage of busulfan over DBM is the fact that more than half the busulfan-treated patients experienced prolonged disease control off treatment.