The use of Mycobacterium bovis/Bacillus Calmette-Guérin (BCG) to vaccinate against tuberculosis remains controversial. The development of tuberculosis in human immunodeficiency virus (HIV)-infected children demands specific evaluation of the risk/benefit ratio of BCG vaccination in this situation. In our institution 9 of 68 HIV-infected children vaccinated with BCG before the diagnosis of HIV infection was suspected developed vaccine-related complications: 7 of these children had a large satellite adenopathy with or without skin fistulae, whereas the other 2 had disseminated BCG infection beyond the satellite ganglion (involvement of the spleen and mesenteric and mediastinal lymph nodes in one case and the liver and lungs in the other). The children were vaccinated soon after birth; no particular problems were observed at that time, but complications appeared 3 to 35 months later. All but one of these children had a rapidly progressive form of HIV disease. The possibility of delayed local or disseminated BCG infection must be considered in analysis of the risk/benefit ratio of vaccination of HIV-infected children. The prognosis of HIV infection must be taken into account, even if the child is asymptomatic when vaccination is being considered.