Modern Approaches to Clinical Trials Using SAS: Classical, Adaptive, and Bayesian Methods

7.3 Considerations for MCP-Mod at the Design Stage

One main consideration when implementing MCP-Mod is the choice of the candidate set of models. As a rule of thumb, often 3-7 dose-response shapes, corresponding to 2-4 dose-response models are appropriate. Note that multiple dose-response shapes might be used for each dose-response model because the shapes define the optimal contrasts to be used in the MCP step and the models are those to be fitted in the Mod step. While in general the candidate set should be broad, when there is large uncertainty, one should keep in mind that using a very broad and flexible candidate set does not come for free. Generally the sample size will increase with a broader candidate set of shapes (i.e., contrasts) as ...