OBJECTIVE: To assess the 9-month HIV-free survival of children with two strategies to prevent HIV mother-to-child transmission. DESIGN: Nonrandomized interventional cohort study. SETTING: Four public health centres in Rwanda. PARTICIPANTS: Between May 2005 and January 2007, all consenting HIV-infected pregnant women were included. INTERVENTION: Women could choose the mode of feeding for their infant: breastfeeding with maternal HAART for 6 months or formula feeding. All received HAART from 28 weeks of gestation. Nine-month cumulative probabilities of HIV transmission and HIV-free survival were determined using the Kaplan-Meier method and compared using the log-rank test. Determinants were analysed using a Cox model analysis. RESULTS: Of the 532 first-liveborn infants, 227 (43%) were breastfeeding and 305 (57%) were formula feeding. Overall, seven (1.3%) children were HIV-infected of whom six were infected in utero. Only one child in the breastfeeding group became infected between months 3 and 7, corresponding to a 9-month cumulative risk of postnatal infection of 0.5% [95% confidence interval (CI) 0.1-3.4%; P = 0.24] with breastfeeding. Nine-month cumulative mortality was 3.3% (95% CI 1.6-6.9%) in the breastfeeding arm group and 5.7% (95% CI 3.6-9.2%) for the formula feeding group (P = 0.20). HIV-free survival by 9 months was 95% (95% CI 91-97%) in the breastfeeding group and 94% (95% CI 91-96%) for the formula feeding group (P = 0.66), with no significant difference in the adjusted analysis (adjusted hazard ratio for breastfeeding: 1.2 (95% CI 0.5-2.9%). CONCLUSION:: Maternal HAART while breastfeeding could be a promising alternative strategy in resource-limited countries.