BACKGROUND: The AIDS Clinical Trials Group protocol 076 zidovudine prophylaxis regimen for HIV-1-infected pregnant women and their babies has been associated with a significant decrease in vertical HIV-1 transmission in non-breastfeeding women in developed countries. We compared the safety and efficacy of short-course nevirapine or zidovudine during labour and the first week of life. METHODS: From November, 1997, to April, 1999, we enrolled 626 HIV-1-infected pregnant women at Mulago Hospital in Kampala, Uganda. We randomly assigned mothers nevirapine 200 mg orally at onset of labour and 2 mg/kg to babies within 72 h of birth, or zidovudine 600 mg orally to the mother at onset of labour and 300 mg every 3 h until delivery, and 4 mg/kg orally twice daily to babies for 7 days after birth. We tested babies for HIV-1 infection at birth, 6-8 weeks, and 14-16 weeks by HIV-1 RNA PCR. We assessed HIV-1 transmission and HIV-1-free survival with Kaplan-Meier analysis. FINDINGS: Nearly all babies (98.8%) were breastfed, and 95.6% were still breastfeeding at age 14-16 weeks. The estimated risks of HIV-1 transmission in the zidovudine and nevirapine groups were: 10.4% and 8.2% at birth (p=0.354); 21.3% and 11.9% by age 6-8 weeks (p=0.0027); and 25.1% and 13.1% by age 14-16 weeks (p=0.0006). The efficacy of nevirapine compared with zidovudine was 47% (95% CI 20-64) up to age 14-16 weeks. The two regimens were well tolerated and adverse events were similar in the two groups. INTERPRETATION: Nevirapine lowered the risk of HIV-1 transmission during the first 14-16 weeks of life by nearly 50% in a breastfeeding population. This simple and inexpensive regimen could decrease mother-to-child HIV-1 transmission in less-developed countries.