Since the first reports demonstrating the ability of a total dose of 2 g/kg body weight (b.w.) of intravenous immunoglobulin (IVIg) to increase the platelet count in patients with autoimmune thrombocytopenic purpura (AITP), the optimal dose has remained controversial. We report the results of a randomized study which compared two low doses of IVIg (0.5 g/kg b.w., group A, n = 19, and 1 g/kg b.w., group B, n = 18) in 37 adults with AITP and platelet count <50 × 109/l, in preparation for surgery or in a situation with a risk of bleeding. On day 4 the proportion of responses, defined by a platelet count > 80 × 109/l and at least twice the initial platelet count, was significantly higher in the group receiving 1 g/kg b.w. (12/18 in group B versus 4/19 in group A, P = 0.005). All but one of the day 4 responders had already responded on day 3. The daily changes in the platelet count from the beginning of IVIg treatment were larger in group B, with a significant difference relative to group A on day 3 (92 × 109/l in group B versus 50 × 109/l in group A, P = 0.03) and on day 4 (106 × 109/l in group B versus 55 × 109/l in group A, P = 0.03). Patients who had not responded by day 4 subsequently received 1.5 g IVIg/kg b.w. (group A) or 1 g IVIg/kg b.w. (group B). A response was observed in 11/13 initial non-responders in group A, and in 2/6 initial non-responders in group B. Finally, on day 8, the proportion of responders was 78% (29/37) in the entire group and was similar in the two subgroups. In conclusion, (1) initial treatment with 1 g/kg b.w. of IVIg appeared to be more effective than 0.5 g/kg b.w. in adults with AITP; (2) infusion of a low dose of IVIg did not jeopardize the efficacy of IVIg reinfusion; (3) some adults who did not respond to 1 g IVIg/kg b.w. responded to a higher dose.