The purpose of this study was to quantify the benefits of maternal-neonatal zidovudine (ZDV) administration for the prevention of vertical human immunodeficiency virus (HIV) transmission against the potential risks of drug-induced complications in uninfected children. A decision analysis model was created with use of a Markov cohort simulation, for evaluating both survival and quality of life for two hypothetical cohorts of HIV-exposed neonates: one with in utero and neonatal exposure to preventive ZDV therapy and the other not exposed. The model included the probability of congenital HIV infection with and without ZDV treatment (estimates derived from AIDS Clinical Trials Group study 076), the yearly probability of death with and without congenital HIV infection, a range of probabilities of adverse effects from ZDV use, and a range of ages in life when any adverse effect would manifest. In a series of scenarios, the impact of different estimates for the quality-of-life decrement from any adverse ZDV effect in HIV-uninfected children was assessed, and threshold values for this estimate were established, i.e., critical values below which withholding ZDV would be the preferred choice. Across a wide range of estimates for multiple contingencies, ZDV use was associated with a greater number of quality-adjusted life years than was non-use. Only in implausible, pessimistic scenarios (i.e., a high incidence of profound adverse effects beginning early in life) would withholding ZDV be the rational choice for an asymptomatic HIV-infected pregnant woman.(ABSTRACT TRUNCATED AT 250 WORDS)