MCP-Mod approach

Summary: The MCP-Mod approach affects both the design and the analysis of dose-finding studies. It is a hybrid approach that combines hypothesis testing and modeling to analyze phase II dose-ranging studies with the purpose of finding suitable dose(s) for confirmatory phase III trials. The MCP-Mod method was found adequate and appropriate for dose selection based on dose-response data.


The main goals in Phase II studies is to investigate the existence, nature and extend of dose effect (Ruberg, 1995). A well-known problem of failed Phase III programs is poor dose selection resulting from inappropriate knowledge of the dose-response relationship at the end of the learning phase of a drug development. Selection of a dose to carry into confirmatory Phase III of a clinical trial is one of the most difficult decision in drug development. The main difficulty in getting the right dose is the trade-off between wanted and unwanted effects, which highly depend on having chosen an appropriate model to represent the nature of dose-response relationship.

The MCP-Mod procedure is an approach for dealing model uncertainty by combining Multiple Comparisons Procedures (MCP) with Modeling Techniques (Mod). Recently, it has been qualified as an efficient statistical methodology for model based design and analysis of Phase II dose finding studies under model uncertainty. This approach provides the flexibility of modelling for dose estimation, while preserving the robustness to model misspecification associated with MCP.

CROS NT has experienced the application of this new methodology to a clinical trial.


The MCP-Mod procedure consists in five steps divided in two stages: the first one is the Trial Design Stage, while the second one is the Trial Analysis Stage.

Figure 1. shows all the five steps involved in this approach:blog image

  1. It starts at the trial design stage, by defining a set of candidate models (between a battery of commonly used dose-response models) to represent the underlying true dose-response shape.
  2. Then, optimum contrast coefficients are derived from each model based on the assessment of relevant metrics such as type I error rate, power to detect a significant dose-response shape, and the power to find the minimal effective dose.
  3. At the Trial Analysis Stage, significance of individual shapes is tested with a multiple contrast test based on the previously derived optimal contrasts coefficients, to assess whether a dose-response signal can be established. Proof of Concept (PoC) is settled when at least one of the model test is significant (MCP part of MCP-Mod).
  4. Later, a candidate model is selected using model selection criteria like Akaike Information Criterion (AIC) or Bayesian Information Criterion (BIC) or averaging.
  5. Finally, the selected model is used to produce estimates of target doses using modelling techniques (Mod part of MCP-Mod).


CROS NT has implemented MCP-Mod procedures by using two different statistical software: the completely free software R, with user-created packages (DoseFinding library) and SAS® 9.4 (S. Menon , R. C. Zink, Modern Approaches to Clinical Trials Using SAS®: Classical, Adaptive, and Bayesian Methods, 2015, Chapter 7) in order to check both software performances and we obtained the same results .

Final Considerations

On May 26, 2016, FDA determined that the Multiple Comparison Procedure – Modeling (MCP-Mod) statistical approach is fit-for-purpose (FFP). This new approach will be increasingly used in clinical trial.  CROS NT acquired further knowledge and expertise on this methodology and recently applied it on a study, where in addition covariates have been considered into the model (ANCOVA).