OBJECTIVES: Several interventions for reducing mother-to-child transmission of HIV (MTCT) have been implemented internationally. In some cities in China, HIV testing is required for all pregnant women and funding for MTCT prevention is paid by the government. However, a standard practice has not been determined. We evaluate the cost-effectiveness of five interventions for preventing MTCT of HIV in China. METHODS: We developed a decision analytic model to compare the costs, health benefits and cost-effectiveness of the five alternatives. We evaluated avoiding breastfeeding only, single-dose nevirapine, short-course zidovudine, long-course zidovudine, or long-course HAART. In the base case, we considered national program implementation in which approximately 15,000,000 women would receive HIV tests annually. We constructed additional scenarios to investigate regional differences in HIV prevalence cohort size. All costs were expressed in 2003 USD. RESULTS: HAART is the most cost-effective of the five interventions if implemented nation-wide. If HAART was offered to all women following a positive HIV test result, it would prevent 1,890 new infections annually and lead to a gain of 39,900 life years (LY). The total cost is $209,000,000, or $5,200/LY gained. HAART becomes more cost-effective as the baseline prevalence increases. In extensive sensitivity analysis, HAART remained the most cost-effective intervention. If avoiding breastfeeding is already the status quo, then single-dose nevirapine is the most cost-effective alternative and is cost saving. CONCLUSIONS: We are not aware of a cost-effectiveness threshold that is appropriate for China. However, HAART is the most cost-effective alternative in a number of different scenarios. In spite of being the most cost-effective option, the total drug cost of implementing HAART nationally may necessitate consideration of other alternatives.