OBJECTIVE To determine whether zidovudine is effective in increasing the platelet count in patients with thrombocytopenic purpura related to human immunodeficiency virus (HIV) infection. METHODS Nonrandomized controlled trial with two consecutive regimens. METHODS Immunopathology and hematology clinics at two general hospitals. METHODS Consecutive sample of 34 patients infected with HIV who had thrombocytopenic purpura (platelets less than 50 x 10(9)/L) without visceral bleeding. Twenty-nine patients completed the study; one patient was removed because of drug toxicity. METHODS Zidovudine for 12 weeks, 250 mg every 6 hours orally in 10 patients; and 500 mg every 8 hours orally in 24 patients. RESULTS Three of ten patients receiving 250 mg every 6 hours and 12 of 24 patients receiving 500 mg every 8 hours had a persistent increase in their platelet counts. In both groups the mean value of the platelet count increased significantly by week 12; from 28 x 10(9)/L +/- 12 (SD) to 57 x 10(9)/L +/- 36 in the first group and from 20 x 10(9)/L +/- 13 to 77 x 10(9)/L +/- 42 in the second group (P less than 0.05 and P less than 0.01, respectively). CONCLUSIONS Zidovudine is effective on platelet counts in some patients with HIV-related thrombocytopenia. These results suggest that HIV itself may play a direct or indirect role in the pathogenesis of this disorder.