OBJECTIVES: The present study presents a Bayesian cost-effectiveness analysis of HIV prevention in the instance when costs and effects cannot be measured directly. METHODS: A Bayesian approach to cost-effectiveness analysis was illustrated using empirical data from an HIV prevention randomized trial. We computed incremental net health benefit (INHB), and the analysis was conducted from the societal perspective. Intervention costs were estimated retrospectively. Clients were randomized into an intervention (advocacy training) (N=15) or comparison condition (N=15). Risk behavior data were collected at baseline and three months after the end of each intervention. In the Bayesian framework, we considered what could occur in a conceptual future study that is an identical replicate to the one actually conducted. Using posterior distribution of the behavior parameters, we sampled 5000 replicates. With the use of a Bernoulli model of HIV transmission, changes in the participants' HIV risk were combined with HIV transmission parameters (drawn from their respective prior distributions) and converted into an estimate of the number of HIV infections averted by the intervention. For each averted infection replicate, the corresponding savings in future HIV-related medical care costs and quality-adjusted life years (QALYs) were estimated. This sampling process was repeated 5000 times to yield a distribution of the INHB describing the full post-data uncertainty. RESULTS: We obtained a positive mean INHB, 0.0008 (close to zero), indicating that advocacy training is just slightly favored over the comparison condition for men, assuming a $50,000 per QALY threshold. We also displayed the acceptability curve: to be somewhat confident of the cost-effectiveness of advocacy training over the comparison condition, say at 0.7 probability, one should be willing to spend over $100,000 per QALY. CONCLUSIONS: The Bayesian framework provides a powerful tool to the economist who intends to advise policy makers by allowing the question -whether an intervention is cost-effective - to be addressed directly.