To compare the response to intravenous immunoglobulin (IVIg) with the response to high-dose methylprednisolone (HDMP) in previously untreated adults with severe AITP and to determine if oral prednisone (PRDN) is required after IVIg or HDMP infusions. M ethods 116 adults, previously untreated, with platelet count ≤ 20 × 109/L were assigned by simultaneous double randomization to receive either IVIg (0·7 g/kg body weight per day) or HDMP (15 mg/kg body weight per day, total daily dose less than 1 g) from day 1 to day 3 (randomization A), and then to receive either PRDN, 1 mg/kg body weight per day, or a placebo (randomization B), from day 4 to day 21 in blinded manner. R esults Between day 1 and day 21, the number of days on which platelet counts were above 50 × 109/L and above 20 × 109/L was significantly higher among the 56 patients receiving IVIg than among the 60 patients receiving HDMP (P = 0·008 and P = 0·013, respectively), and the percentage of patients who had a platelet count exceeding 50 × 109/L on day 5 was also higher in the IVIg group. During the second treatment period, PRDN was more effective than the placebo on all short-term end points. Patients who received IVIg followed by PRDN had a significantly larger number of days with platelet counts > 50 × 109/L and > 20 × 109/L between day 1 and day 21 (P = 0·008 and P = 0·005, respectively) than patients treated with HDMP followed by PRDN. As regards the long-term response, the percentage of remitters was 40% overall and was not influenced by the treatment allocation in either randomization period (A or B). C onclusion IVIg followed by PRDN appears to be the more effective treatment in adults with severe AITP.