To characterize the nature, time course and dose dependency of zidovudine-related side effects, we undertook a multicenter, prospective, dose-range finding study. Our study group consisted of 74 HIV-positive homosexual men belonging to groups II B, III and IV C2 from the Centers for Disease Control (CDC) classification of HIV disease. Following a 3-week observation period, volunteers were treated with zidovudine 600 mg/day for 18 weeks, 900 mg/day for 9 weeks and 1200 mg/day for 9 weeks, followed by a washout period of 6 weeks after which they were re-started on 1200 mg/day or the highest tolerated dose at 8-hourly intervals. Subjects were randomly assigned to 4-hourly or 8-hourly regimens within CDC groups while taking 600 and 1200 mg/day. Clinical and laboratory evaluations were performed at 3-week intervals. Symptomatic adverse effects were present in 96% of subjects, most commonly nausea (64%), fatigue (55%) and headache (49%). These were generally self-limited, reappearing briefly at each dose increment. A decrease in hemoglobin occurred shortly after initiation of therapy. This was not dose dependent and reversed rapidly upon discontinuation of treatment. A red blood cell count decrease, a mean cell volume increase and a granulocyte count decrease developed early in a dose-independent fashion, reverting at least partially during the washout phase. The decrease in reticulocyte count was dose related between 600 and 900 mg/day with no further change when the dose was escalated to 1200 mg/day. Bone marrow changes occurred rapidly as demonstrated by megaloblastosis in 95% of 65 specimens at week 18.(ABSTRACT TRUNCATED AT 250 WORDS)