OBJECTIVES: To study mortality in African children born to HIV-1-infected mothers exposed peripartum to zidovudine. METHODS: A randomized placebo-controlled trial in Abidjan and Bobo-Dioulasso. Pregnant women received either 300 mg zidovudine twice daily from 36-38 weeks' gestation, 600 mg during labour, and 300 mg twice daily for 7 days post-partum or a matching placebo. Determinants of mortality were studied up to 18 months, overall and among the infected children: treatment, centre, timing of infection, mother and child HIV disease. RESULTS: There were 75 infant deaths among 407 live births. The risk of death at 18 months was 176/1000 in the zidovudine arm and 221 for placebo. Relative hazard (RH, zidovudine versus placebo) was 0.47 [95% confidence interval (CI) 0.2-1.0] up to 230 days of life. Maternal CD4 lymphocyte count < 200/mm3 (RH 2.92; CI 1.4-6.1) and child HIV-1 infection (RH 12.6; CI 6.6-24.3) increased mortality of all children born to HIV-1-infected mothers. There were 101 children infected (40 in the zidovudine group), and 51 died. Their 18 month probability of death was 590/1000 in the zidovudine group and 510 in the placebo group. Among infected children, maternal zidovudine reduced the risk of death on or before day 230 (RH 0.18; CI 0.1-0.5). Maternal CD4 lymphocyte count < 200/mm3 (RH 3.25; CI 1.3-8.4), maternal death (RH 9.65; CI 1.7-56.0), diagnosis of paediatric infection on or before day 12 (RH 18.1; CI 4.8-69.0) and between days 13 and 45 (RH 7.63; CI 2.0-29.5), clinical paediatric AIDS (RH 5.37; CI 2.3-12.7) were risk factors for death in HIV-1-infected children. CONCLUSION: Mother-to-child transmission reduction by zidovudine is safe and beneficial to African children. The mortality of HIV-1-infected children is high. Peripartum maternal zidovudine exerts a protective effect for at least 8 months.