To determine the feasibility of using short-course zidovudine (ZDV) to prevent mother-to-child transmission of human immunodeficiency virus (HIV) in a breastfeeding population in a rural area in Kenya, pregnant mothers attending clinics in seven health centers in western Kenya between 1996 and 1998 were requested to volunteer for participation in this study. The HIV-infected mothers were given a daily dose of 400 mg of ZDV starting at 36 weeks of gestation and another 300 mg every three hours intrapartum. After delivery, mothers and their children were followed-up and clinically monitored every 3-4 months for two years, and child and mother mortality rates were analyzed. Of the 825 mothers who consented, 216 (26.2%) were infected with HIV. Of those infected, 51 (23.6%) took the full prescribed dose, 69 (31.9%) took only the prenatal dose, and the remaining 96 (44.4%) did not take any dose. Failure to take ZDV was attributed mainly to delivery occurring earlier than expected, while non-compliance to the intrapartum dose was due to mothers giving birth at home and fear of traditional birth attendants. By the end of the second year, 75 HIV-exposed children (34.7%) and 33 HIV-infected mothers (15.3%) had died. The HIV-free survival of children at 24 months was significantly associated with mother survival (P < 0.001) and prenatal ZDV compliance (P < 0.003). Our findings suggest that implementation of programs for prevention of mother-to-child transmission of HIV in rural areas of Africa need to consider the various socioeconomic and cultural barriers that may prevent successful uptake of antiretroviral prophylaxes. Similarly, the rapid disease progression in mothers may eliminate the increase in child survival due to ZDV prophylaxis.