OBJECTIVE To compare the safety and tolerance of stavudine (d4T) versus zidovudine (ZDV, AZT) in symptomatic human immunodeficiency virus-infected children 3 months to 6 years of age. METHODS In an initially double-blind trial, 212 evaluable human immunodeficiency virus-infected children who had received no more than 6 weeks of previous antiretroviral therapy were randomized to receive either d4T (1 mg/kg orally every 12 hours, maximum 40 mg orally every 12 hours) or zidovudine (180 mg/m2 orally every 6 hours, maximum 200 mg orally every 6 hours). The study was unblinded after a median follow-up period of 6.3 months; median follow-up at study closure was 17.3 months. Tolerance, safety, disease progression, and immunologic responses were evaluated. RESULTS The patient population was young (median age, 1.2 years; range, 0.3 to 6.4 years), with a median baseline CD4+ lymphocyte count of 965 cells/microL (range, 18 to 4238 cells/microL). Neutropenia < 400/microL occurred significantly more commonly among zidovudine recipients (1-year event rates of 20% both up to the time of unblinding and throughout the entire study) than among children receiving d4T (1-year event rates of 5% up to the time of unblinding and 6% throughout the entire study). In exploratory activity analyses using all data collected until study closure, children treated with d4T showed consistently greater positive changes from baseline in weight-for-age-and-gender z scores. As expected in this population of young children, median absolute CD4+ lymphocyte counts decreased in both treatment groups. Smaller changes from baseline were noted among d4T recipients. CONCLUSIONS In children between the ages of 3 months and 6 years, d4T and zidovudine are largely comparable in terms of safety and tolerance. Neutropenia occurs significantly less commonly among children treated with d4T. There was evidence that weight gain and absolute CD4+ lymphocyte counts were better maintained in children receiving d4T.