OBJECTIVES: to demonstrate the use of bayesian decision theory to establish optimal commercial drug development programmes. METHODS: A probabilistic model of a drug in development was used to estimate the expected net return of alternative development programmes. Efficacy data were available from a phase II trial (n = 40). Based on the predicted distribution of treatment effect arising from a Bayesian analysis of these phase II data, sets of potential results arising from various phase III trial designs were simulated. The expected return for the various trial designs were then estimated considering both the fixed and variable costs associated with different trial designs and assuming that regulatory approval was conditional on a frequentist hypothesis test of efficacy, reimbursement was conditional on expected cost-effectiveness, and future sales (conditional on successful licensing and reimbursement) were also a function of expected cost-effectiveness. The optimal phase III design was established and the trade-off between more extensive phase III research and the likelihood of approval and re-imbursement was explored. RESULTS: Based on the particular phase II results considered, a phase III trial was found to have a positive expected return and further development with a trial of n = 420 was found to be optimal. If a lower cost-effectiveness threshold for re-imbursement was assumed, the optimal decision was to discontinue development. CONCLUSIONS: Where there are several potential candidates for further development and limited resources available the model could be extended to compare different portfolios of research. The model could also be used to compare different development strategies and to explore the relationship between the regulatory and re-imbursement environment and the resulting commercial incentives.