The analysis of a multiple-dose, combination-drug clinical trial using response surface methodology

We consider the situation where a multiple-dose, combination-drug clinical trial is conducted to identify one or more combinations that satisfy regulatory requirements. Generally, these requirements involve a compound hypothesis with multiple comparisons. The min test has been shown to be an optimal alpha-level test for testing a single combination drug. Analysis procedures in a multiple-dose, combination-drug study have typically involved classical ANOVA models or multiple regression models in a response surface methodology (RSM) framework. An inferential procedure based on an ANOVA model uses a screening test to address multiple comparison issues and multiple min tests to explicitly identify combinations satisfying regulatory requirements. An exploratory procedure based on RSM modeling is used to build a segmented linear model and a stairstep linear model to describe dose-response relationships. The two procedures are mutually supportive of one another in providing a broader assurance in the identification of effective combinations.