In this chapter, we have described modeling methods for analyzing dose-response studies. We introduced the framework of modeling approaches with their advantages and disadvantages as compared to alternative analysis methods. We described some commonly used dose-response models, while discussing and interpreting their parameters together with strategies for obtaining initial estimates. We then presented methods to determine suitable dose levels to meet clinical effect requirements. Particular emphasis was laid on the inherent problems related to model uncertainty, such as increased variability and bias of the parameters estimates of interest.
In light of these results a hybrid strategy for analyzing dose finding studies was investigated in detail. The proposed methodology combines the advantages of multiple comparison and modeling approaches, consisting of a multi-stage procedure. In the first stage, multiple comparison methods are used to test for PoA and to identify statistically significant contrasts corresponding to a set of candidate models. Once the PoA is established in the first stage, standard model selection criteria are used to chose the best model, which is then used for dose-fining in subsequent stages. The advantage of this new approach, in comparison to more traditional multiple comparison dose finding methods, is its added flexibility in searching for and identifying an adequate dose for future drug development while alleviating the aforementioned model selection problems.
The emphasis of this chapter has been on the motivation and the methodological foundation of the MCP-Mod approach. We did not present or discuss in greater detail the implementation aspects of the methodology, due to space limitations. A separate manuscript dedicated to the practical aspects of utilizing MCP-Mod, including software for implementing the various steps of the methodology, is currently under preparation.
An important research topic that can significantly enhance the performance of the methods described here is the re-evaluation of study designs for dose finding studies which take modeling of the dose-response relationship into consideration. Traditionally, dose-finding designs were developed with multiple comparison methods in mind. From the point of view of modeling, it would be better to have more doses spread out in the range of possible doses. Of course this raises other important issues, like the feasibility of manufacturing such doses and managing a trial with a larger number of doses. Whether or not these may be practical restrictions will be study-dependent. However, the need to re-think traditional dose finding designs in light of the need to model the dose-response relationship is evident and certainly deserving more research.
Was this article helpful?