Modeling Effects of Agent Combinations Using Nonlinear Mixture Experiment Methods

• Activity Number: 259
• Type: Topic Contributed
• Date/Time: Tuesday, August 10, 2004 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Donald B. White*+ and William R. Greco and Leonid A. Khinkis
• Companies: University of Toledo and Roswell Park Cancer Institute and Canisius College
• Keywords: mixture amount model ; process variable ; nonlinear regression ; synergism ; antagonism ; response surface

Abstract

In pharmacology and toxicology, modeling the effects of chemical agents is important for developing optimal experimental designs, making effect predictions, optimizing treatments, and increasing scientific understanding. Extending such models to multiple agents is important for research in cancer therapy, antibiotic treatment, herbal remedies, toxic waste hazards, and many other applications. We introduce a new hierarchical approach to modeling the effects of multiple agents using a combination of nonlinear response surface modeling with mixture experiment methods. This approach allows us to model more than two agents, modulating agents having no effects singly but which enhance the effects of other agents, and complex patterns of synergism and antagonism. We are also able to produce effective descriptive graphics, well-grounded statistical inference, and methods for creating optimal experimental designs. These results are illustrated with three studies: a large in-vitro study of three cancer drugs, another large in-vitro study of two cancer drugs and a modulator, and a small in-vivo study of agents affecting cholesterol levels in mice. Supported by NIH RR10742 and CA16056.