|
Activity Number:
|
259
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 10, 2004 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #301651 |
|
Title:
|
Statistical Methods for Analysis of Mixtures of Many Chemicals Using a Ray Design
|
|
Author(s):
|
Chris Gennings*+ and Hans Carter
|
|
Companies:
|
Virginia Commonwealth University and Virginia Commonwealth University
|
|
Address:
|
Dept. of Biostatistics, Richmond, VA, 23290-0032,
|
|
Keywords:
|
synergy ; dose-response ; interaction
|
|
Abstract:
|
The motivating strategy of our research program is to develop statistical techniques that are useful in detecting departure from additivity in mixtures with many chemicals while maintaining the fundamental definition of additivity as found in the toxicology literature. By focusing the inferential region to relevant mixing ratios of the chemicals through use of a ray design, it becomes experimentally practical for studies to be conducted of mixtures with many components. Single chemical dose-response data are used to estimate an additivity model for a fixed-ratio mixture. An "unconstrained" model is fit to observed mixture data along the ray. A test of the coincidence of these two dose-response curves is a test of additivity. A variation of this approach is demonstrated when the parametric form of the underlying response surface is assumed. Methods are described for testing hypotheses regarding the impact of subsets of chemicals in the mixture. A statistical model allowing for an interaction threshold parameter is presented. Finally, we describe a method of analysis where the additivity model is built implicitly. Each method is illustrated with various toxicology studies.
|
