<img alt="" src="https://secure.lote1otto.com/219869.png" style="display:none;">
Skip to content

MCP-Mod for the Modern Dose-Ranging Clinical Trial

| | |


MCP-Mod methodology for dose-ranging clinical trials has been gaining popularity since the 2013 publication of the qualification opinion by the European Medicines Agency Committee for Medical Products for Human Use. Since its development at Novartis, MCP-Mod promises to devise proof-of-concept and dose-ranging trials which generate superior statistical evidence for dose-selection, while providing safety and efficacy data that can prove critical data for Phase III clinical trial design. 

Although the general framework of the MCP-Mod method are becoming more familiar, its added complexity raises the question of whether it is a necessary supplement (or even substitute) for traditional dose-ranging trials. Here are a few shortfalls of the traditional approach that MCP-Mod is equipped to handle. 


The Traditional Approach:

Most people familiar with early phase trials imagine dose-selection as a series of three steps: proof-of-concept to establish dose-signal (i.e. the existence of a dose-response); dose-selection to determine which candidate doses should move forward; and dose-response modeling to determine the shape of an entire dose-response curve (e.g. for making predictions) [1]. 

The methodology employed at each of these stages may be described as follows.  

  • Proof-of-Concept: Conducted using (multiple) active arms and placebo
  • Selection of Target Dose: Determined by selecting the dose which (i) was statistically significant at the proof-of-concept stage; (ii) is the smallest of the statistically significant doses which is also clinically relevant
  • Dose-Response Modeling: Using data from proof-of-concept and earlier trials, find a statistical model capturing the effects of target dose on dose-response

Although the traditional approach is straightforward, it suffers from a number of operational flaws. Firstly, it tends to concentrate the range of doses studied to a narrow range where sponsors can have faith that they will establish a clear dose-signal, (i.e. to the “upper end of the dose response relationship” [2]). The opportunity to understand dose-response having considered a full range of doses is side-stepped in hopes of selecting a target dose with which to move forward. This is also partially a reflection of the fact that the sample size necessary to determine the existence of a dose-signal is in fact much smaller than that necessary for dose-selection or to model a dose-response curve.  

However, it is also clear that knowledge of dose-response across a broad range of doses can actually provide pivotal information for the selection of a good dose. This is due to the simple fact that once there is evidence of how a new medicine works across doses, study sponsors have more information on which to base dose-selection. Taking the time to study a wide dose-range before determining a target dose provides useful empirical evidence for both safety and efficacy considerations, and for choosing the right dose.  

A second, albeit related, concern is whether the dose-response model should itself play a greater role in choosing the right dose. According to an EMA’s Committee for Medical Products for Human Use, “It is rather obvious that a strategy based on a modelling approach that attempts to quantify a dose-response relationship may offer an improved basis for decision making…” [2] Although the information gathered at the proof-of-concept stage is sufficient to choose a clinically relevant dose, such information is typically sub-optimal for dose-selection. In fact, if sponsors had access to one or more dose-response models that captured the dose-response well enough, this information could supplement the information gathered while determining proof-of-concept. This would mean, however, that there is reason to reverse Steps 2 and 3 above, and to begin to consider how dosing decisions can be made under conditions of model uncertainty.

Third, the information needed for dose-response modeling should be an important consideration when designing the actual clinical trial, since it is important to ensure that the trial adequately examines the range of doses necessary to build a trustworthy statistical model. However, the standard approach focuses on modeling at the very end of the process. This means that trial design will likely generate sub-optimal evidence for the creation of a statistical model. A much better approach would be to design the trial with possible statistical models in mind.

The MCP-Mod Approach

The MCP-Mod method for dose-ranging studies is one strategy to tackle all three of the concerns raised with the traditional method. According to the European Medicines Agency Qualification Opinion, “The MCP-Mod approach is efficient in the sense that it uses the available data better than the commonly applied pairwise comparisons.”

The essence of the MCP-Mod approach begins with the idea of model uncertainty. We do not know the shape of the actual dose-response curve, but we can build a trial that is designed to test a proposed set of plausible dose-response models. We can then collect data from a clinical trial to determine the strength of evidence supporting the proposed models.  

Once we realize that the truth of a model can constitute the content of an alternative hypothesis, we can specify a null hypothesis based on several models considered together, (e.g. H0 = {M1, M2, M3…} where each Mi refers to a different statistical model). Maintaining Type I error while considering several models together requires the use of ‘multiple comparison procedures.’

In the MCP-Mod Approach:

  • Proof-of-concept: When multiple comparison procedures establish the statistical significance of the hypothesis (i.e. that one of the models is the shape of the actual dose-response curve), then this establishes proof-of-concept
  • Dose-Response Modeling: Given all of the models that demonstrate statistical significance, determine a dose-response model using data collected during the clinical trial
  • Selection of Target Dose: Once there is a statistical model of the dose-response use this to determine target dose levels

Liked this article? Click below to subscribe to the Cytel blog and get the latest biostatistics and clinical data trends, best practices and case studies direct to your inbox.


Related Items of Interest

[1] Bornkamp, Björn, et al. "Innovative approaches for designing and analyzing adaptive dose-ranging trials." Journal of biopharmaceutical statistics 17.6 (2007): 965-995.

[2] EMA Qualification for MCP-Mod

contact iconSubscribe back to top