Serious recurrent bacterial infections are a major cause of morbidity and mortality in children infected with the human immunodeficiency virus (HIV). Because intravenous immune globulin has been shown to prevent bacterial infection in patients with primary immunodeficiency and in uncontrolled studies of HIV-infected children, we undertook a multicenter study of its safety and efficacy in children with symptomatic HIV infection. In a double-blind trial, 372 HIV-infected children (mean age, 40 months) with clinical or immunologic evidence of HIV disease were randomly assigned to receive either intravenous immune globulin (400 mg per kilogram of body weight) or placebo (0.1 percent albumin) every 28 days. The children were stratified into two groups according to CD4+ lymphocyte count at entry into the study and the clinical classification of the Centers for Disease Control. The median length of follow-up was 17 months. For children in either group with CD4+ counts greater than or equal to 0.2 x 10(9) per liter (greater than or equal to 200 per cubic millimeter) at entry, treatment with intravenous immune globulin significantly increased the time free from serious infection; estimated infection-free rates after 24 months were 67 percent for children receiving immune globulin as compared with 48 percent for those receiving placebo (P = 0.01). In addition, immune globulin was associated with an overall reduction in the number of both serious and minor bacterial infections (relative risk, 0.68; P = 0.01) and in the number of hospitalizations for acute care (relative risk, 0.65; P = 0.03). No such benefits were seen for children with CD4+ counts below 0.2 x 10(9) per liter at entry. For group 1 overall, there was a trend toward a difference in serious bacterial infection between immune globulin and placebo (24-month infection-free survival, 31 percent for intravenous immune globulin vs. 25 percent for placebo; P = 0.10). For group 2, the estimates of survival without serious infection were 73 percent with intravenous immune globulin as compared with 53 percent with placebo (P = 0.04). There was no effect of treatment on mortality for any group or CD4+ count at entry. Adverse reactions, noted for less than 1 percent of infusions, were minor. In symptomatic HIV-infected children the prophylactic use of intravenous immune globulin, is safe, and it significantly increases the time free from serious bacterial infections for those entering treatment with CD4+ lymphocyte counts greater than or equal to 0.2 x 10(9) per liter.