Dose Estimation Research Based On Uncertainty Parameter Model In Clinical Phase â…¡ Dose-response Studies

Objective:Finding the right dose is a critical step in drug development. In recent years, considerable effort has been spent on improving the efficiency of dose finding throughout drug development. Due to lack of sufficient dose-response knowledge for both efficacy and safety, right dose selection remains a key problem in drug development. So, we introduces a method MCP-Mod（Multiple Comparisons and Modeling）, which can estimate the dose of dose-response studies. Based on the application conditions and analysis method of MCP-Mod and generalized MCP-Mod method, we explore the power to detect a dose-response relationship and the accuracy of model selection and target dose estimation using simulation data. And the MCP-Mod method is applied to actual clinical drug data to estimate the MED of the new drug. In addition, MCP-Mod method can solve the problem of testing the presence of dose-response and selecting the adequate dose（s） in phase II clinical trials.Methods:Firstly, the research introduces the fundamental theory and application condition of the MCP-Mod method and generalized MCP-Mod method, respectively. Then, simulate data is set by three parameters: five dose levels d=0,0.05,0.2, 0.6,1,six kinds of sample size n = 10,25,50,75,100,150 for every dose; nine kinds of commonly used parameter models（constant model, linear model, logarithm linear model, Emax model, quadratic model, exponential model, logistic model, D- logistic model and convex model）.Different types of dose-response data was constructed based on the three parameters. We apply the R package MCPMod to analyze the data, and to evaluate the power to detect a dose-response relationship and the accuracy of model selection and target dose estimation. We applied the MCP-Mod method to the dose-response data in actual clinical trials to illustrate the procedure of estimating the MED of the new drug.Results:In the simulation, in the aspect of the power to detect a dose-response relationship: 6 groups of constant model have an approximately power 0f 0.05 to detect the dose response relationship, which was similar to a（28）0.05. For other parameter models, the larger the sample size, the higher power to detect a dose response relationship. In the aspect of accuracy of model selection: the larger the sample size, the higher degree to identify various models. In the candidate models, exponential model and quadratic model have a high degree of identification, but linear model and logarithmic function have a low degree of identification. The D- logistic model is approximated by quadratic model and the probability increases with the sample size increases, while the convex model is approximated by exponential model and the probability increases with the sample size increases. In terms of the accuracy of the dose estimation: MêD₁ underestimates the target dose,MêD₃ overestimates the target dose, and MêD₂ is close to the target dose, so we thinkMêD₂ has better consistency. As the sample size increases, the accuracy of the dose estimation increases. In addition, for different parameter model, the accuracy of the dose estimation is not the same. In actual clinical trials, different model selection criteria can lead to different estimation of MED.Conclusion:MCP-Mod method not only have a high power to detect a dose response relationship, but also have a high precision to select right model and estimate target dose, so MCP-Mod method is useful for phase II clinical studies, when one is interested in testing the presence of dose-response and in selecting the adequate dose（s） for the confirmatory phase of development.