Sixty-eight patients, followed in a prospective cohort study of 185 human immunodeficiency virus (HIV)-infected patients with severe immune thrombocytopenia (platelets < 50 x 10(9)/L), underwent splenectomy, 2 to 41 months (median: 10 months) after immune thrombocytopenic purpura (ITP) was diagnosed. The mean platelet count increased from 18 x 10(9)/L to 223 x 10(9)/L with a persistent increase in 56 (82%). It also led to a significant increase of the mean CD4 cell count from 475 x 10(6)/L to 725 x 10(6)/L within a mean delay of 10 months. In the whole cohort, with a mean follow-up of 63 months (range, 6 to 126), the 5-year estimated rate for progression to acquired immunodeficiency syndrome (AIDS) was 23% (95% confidence interval [CI], 15% to 31%) and the AIDS-free survival was 69% (95% CI, 61% to 77%). To investigate the potential impact of splenectomy, a Cox's multiple regression model was used; as splenectomy was not randomly assigned, it was incorporated as a time-dependent covariate. After adjustment on the CD4 cell count, no statistically significant differences were observed between the splenectomized and the nonsplenectomized patients: AIDS progression rate (P = 0.23), survival (P = 0.64) and AIDS-free survival (P = 0.72) were not influenced by splenectomy. Splenectomy is both effective and safe in the treatment of severe, refractory ITP associated with HIV infection.