The most serious complication of childhood acute immune thrombocytopenic purpura (ITP), intracranial haemorrhage, occurs in about 1% of children with platelet counts below 20 x 10(9)/L. We conducted a randomised study to explore three treatment options in this high-risk group. 146 children (> 6 months and < 18 years old) with typical acute ITP and platelet counts of 20 x 10(9)/L or lower were randomised to receive high-dose intravenous immunoglobulin G (IVIgG) 1 g/kg on 2 consecutive days (n = 34), 0.8 g/kg once (n = 35), intravenous anti-D 25 micrograms/kg on 2 consecutive days (n = 38), or oral prednisone 4 mg/kg per day with tapering and discontinuation of prednisone by day 21 (n = 39). The rate of response as reflected by the number of days with platelet counts at 20 x 10(9)/L or lower and the time taken to achieve a platelet count 50 x 10(9)/L or more was significantly faster for both IVIgG groups than for the anti-D group (p < 0.05); the difference between prednisone and IVIgG was significant (p < 0.05) only for the IVIgG 0.8 g/kg group, and responses to the two IgG groups were similar. These differences in response rates were reflected in the percentages of children with platelet counts of 20 x 10(9)/L or lower at 72 hours following the start of treatment: 3% (IVIgG 0.8 g/kg x 1), 6% (IVIgG 1 g/kg x 2), 18% (anti-D), and 21% (oral prednisone 4 mg/kg/day). Treatment-associated toxicities included a fall in haemoglobin with anti-D (to less than 100 g/L in 24% of cases); weight gain with oral prednisone; and fever, nausea, vomiting, and headache with IVIgG. On the basis of these results, intravenous anti-D cannot be recommended as initial therapy for children with acute ITP and platelet counts of 20 x 10(9)/L or lower. A single dose of 0.8 g/kg IVIgG offers the fastest recovery for the least treatment; additional IgG or oral prednisone can be reserved for the one-third of children who continue to have platelet counts of 20 x 10(9)/L or less at 48-72 hours after the start of treatment.