RESUMEN
Multiple comparison procedures and modeling (MCPMod) has established itself as a method for dose-finding under model uncertainty. A downside of MCPMod is that due to its frequentist nature in particular with respect to the multiple comparison part it is tough to incorporate historical information in a systematic fashion. A typical situation where such historical information is available is existing data for the placebo group from previous trials. There are multiple Bayesian concepts for integrating historical data in a systematic and even dynamic fashion like the meta-analytic prior approach. In this article, we define Bayesian MCPMod (BMCPMod) that is build upon these two aspects. BMCPMod is able to mimic the results of the classical MCPMod for non-informative priors. At the same time, it allows for the inclusion of historical data in a systematic fashion. After the definition of BMCPMod related characteristics for a Bayesian approach similar to the MCP-testing part are derived. The BMCPMod is compared to classical MCPMod/non-informative priors via simulations. Aspects of mixture priors, optimal contrast vectors, and impact of allocation ratios are discussed and an example for designing a BMPCMod trial is given.