An R package for simulation experiments evaluating clinical trial designs.

Abstract

This paper presents an open-source application for evaluating competing clinical trial (CT) designs using simulations. The S4 system of classes and methods is utilized. Using object-oriented programming provides extensibility through careful, clear interface specification; using R, an open-source widely-used statistical language, makes the application extendible by the people who design CTs: biostatisticians. Four key classes define the specifications of the population models, CT designs, outcome models and evaluation criteria. Five key methods define the interfaces for generating patient baseline characteristics, stopping rule, assigning treatment, generating patient outcomes and calculating the criteria. Documentation of their connections with the user input screens, with the central simulation loop, and with each other faciliates the extensibility. New subclasses and instances of existing classes meeting these interfaces can integrate immediately into the application. To illustrate the application, we evaluate the effect of patient pharmacokinetic heterogeneity on the performance of a common Phase I "3+3" design.