The MCP-Mod methodology - A statistical methodology for dose-response
Björn Bornkamp & Frank Bretz
Novartis
The ICH E9 guidance requires pre-specification of the data analysis technique for the analysis of the primary variable(s) in the trial protocol. For dose-response modelling this would mean that the used dose-response model would be pre-specified. This is in many situations difficult, because at the trial design stage it is not precisely known which model will be adequate. As a consequence of this model uncertainty, traditionally the assumptions made on the functional form for the dose-response relationship for the pre-specified analysis are kept at a minimum, in the extreme case leading to models that treat dose as a categorical variable (i.e. making no assumption on the dose-response relationship).
The Multiple Comparison Procedures and Modeling technique (MCP-Mod) ([1], [2]) is an alternative to such an approach. Its main idea is to deal with the resulting model uncertainty through the introduction of a candidate set of dose-response models (instead of one dose-response model) that is pre-specified at the design stage. Once the data are obtained, one then uses the whole candidate set of dose-response models to test for a dose-response signal (the MCP step) and to model the dose-response curve (the Mod step) either using model selection or model averaging techniques.
After a general tutorial overview of the MCP-Mod methodology, we will present an example data analysis, where we illustrate the usage of MCP-Mod in context of a longitudinal model. We will also provide an overview of the qualification opinion issued by the EMA in January this year [3].
References:
[1] Bretz, F., Pinheiro, J., and Branson, M. (2005). Combining multiple comparisons and modeling techniques in dose response studies. Biometrics 61, 738-748.
[2] Pinheiro, J., Bornkamp, B., Glimm, E., and Bretz F (2013) Multiple comparisons and modeling for dose finding using general parametric models, Statistics in Medicine online early, DOI: 10.1002/sim.6052
[3] European Medicines Agency (2014) Qualification Opinion of MCP-Mod as an efficient statistical methodology for model-based design and analysis of Phase II dose finding studies under model uncertainty, URL: http://goo.gl/imT7IT
