Ninety-eight consecutive patients with myelodysplastic syndrome were randomized to a treated or a control group, both receiving conventional supportive therapy. The treated group were given 13-cis-retinoic acid 20 mg/d if marrow blasts were less than or equal to 5% or cytosine arabinoside 10 mg/d subcutaneously on 6 d/week if marrow blasts were 6-30%, to which retinoic acid was added after 12 weeks. Serum levels of the drugs in the treated group were similar to those that would produce inhibition of CFU-GM growth in vitro. In patients in the low blast group receiving retinoic acid, myeloid surface antigens reverted from an abnormal to a normal pattern. Log rank analysis carried out after 25 months showed no significant difference in survival between the treated and control group, either in the total patient population or in the high and low blast groups considered separately. However, analysis of 39 non-sideroblastic patients with less than or equal to 5% blasts showed an increase in survival in the treated group.