OBJECTIVE To compare the effect of two dose regimens of zidovudine in the treatment of severe HIV-related thrombocytopenia (TP). METHODS Eighty-four patients with severe HIV-related TP and platelet counts < 50 x 10(9)/l were enrolled in an open study at six centres. Patients were randomized into two groups to receive zidovudine (group A, 500 mg per day; group B, 1000 mg per day) for 6 months. METHODS Platelet counts were determined monthly and patients categorized as complete responders (CR; platelets > 100 x 10(9)/l), partial responders (PR; platelets > 50 to < 100 x 10(9)/l), or failures (F; platelets to < 50 x 10(9)/l). CD4+ and CD8+ lymphocytes, HIV antigenaemia, beta 2-microglobulin, white blood cells, mean cell volume and haemoglobin were also determined. RESULTS Seventy-one patients (35 and 36 in groups A and B, respectively) completed the study; 11.4% of group A patients were CR and 45.7% PR; 38.9% of group B were CR and 33.3% PR. Increase in mean platelet counts was dose-related, more rapid in the higher dose group and remained significantly higher after 6 months of treatment (56.4 x 10(9)/l in group A versus 98.2 x 10(9)/l in group B; P < 0.01). CONCLUSIONS The results confirm the efficacy of zidovudine in the treatment of severe HIV-related TP. The average for CR and PR in the two groups was 64.8%; the higher dose of zidovudine was more effective at increasing platelet counts.