All Times EDT
Keywords: Prioritized outcomes, non-parametric test, phase 2 trials, minimum effective dose, maximum safe dose, MCP-Mod
In recent years, the nonparametric tests based on U-statistics such as win-ratio or win-odds are increasingly adopted in assessing treatment effects in clinical trials due to complex distributions of endpoints and in order to increase efficiency. In this poster, we develop a novel win-odds based approach for multiple comparisons in dose finding studies where the selection of optimal dose depends on differentially prioritized efficacy and safety endpoints. The nonparametric method is harnessed with generalized multiple comparison procedure and can handle flexible distributions of correlated endpoints, where normality is undermined due to inflation or mixture of densities that are frequently encountered in several disease areas. The operating characteristics from numerical studies demonstrate the proposed method can yield elevated statistical power while controlling type 1 error for non-normally distributed endpoints with tied observations. Moreover, the method can hit the optimal dose more frequently than model-based approaches such as MCP-Mod when both efficacy and safety are jointly considered. The merits of the method are further elaborated through concrete phase 2 clinical trial examples. We also present an R package with RStudio shiny app for simple implementation of the proposed method.