OBJECTIVES: The research objective was to develop a flexible Bayesian statistical framework for cost-effectiveness analysis using data from a clinical trial, and to illustrate the methodology in practical case studies. METHODS: The general methods of Bayesian statistical theory are used to develop the framework. For computation of results in case studies, simulation-based methods are used, including Markov chain Monte Carlo. RESULTS: A general framework is established, in which individual patient data arising in a clinical trial may be modelled using any appropriate probability models. Within the model, the true patient mean efficacy and true patient mean cost are represented as functions of the model parameters. Cost-effectiveness decisions are then based on inference about these true mean parameters for each of the two treatments under comparison in the trial. It is argued that appropriate decision indicators are whether the expected net benefit is positive, or the probability that net benefit is positive, with net benefit defined with reference to a specific threshold unit cost. When a range of unit costs must be considered, the relevant indicators are the break-even unit cost for expected net benefit, or the C/E acceptability curve (CEAC). Inference about these indicators is determined within the Bayesian statistical paradigm. Examples and case studies are presented illustrating the method with efficacy outcomes that are continuous, binary, ordinal or time to event, and with costs modelled as distributed normally, lognormally or nonparametrically. CONCLUSIONS: The Bayesian framework is demonstrated to be both a flexible and powerful tool for cost-effectiveness analysis from clinical trial data. STUDENT SESSION