PROJECT GOAL
The goal of this project is to develop improved methods and approaches for utilizing developmental toxicity data to inform statistically based structure-activity relationship (SAR) models.
PROJECT RATIONALE
There are large numbers of chemicals in the environment that have not been fully characterized with regard to their potential to pose a health risk. Regulatory agencies in North America and Europe are increasingly called upon to review such substances and prioritize them for further study. There is special interest in identifying potential developmental toxicants. To make good use of limited resources and minimize animal testing, regulatory agencies need effective screening tools that can identify which particular substances are most deserving of further attention. Statistically based structure-activity relationship (SAR) models offer one approach. Such models rely on statistical analyses to discern relationships between chemical structure and activity/toxicity for a “training set” of substances, which is composed of substances that have been tested for the activity/toxicity. Using a trained SAR model, the activity/toxicity of an untested substance may be predicted, as long as the substance’s structural features are encompassed by chemicals of the training set.
In 2003, the International Life Sciences Institute Risk Science Institute (ILSI RSI) convened an expert working group to examine the potential of statistically based SAR models for use in screening environmental contaminants for possible developmental toxicants. The working group concluded that the reliability of such models could be enhanced with the development of improved methods for using toxicity data to inform the models. (See Birth Defects Research (Part A) 70: 902-911). The working group recommended that an interdisciplinary group of developmental toxicologists and SAR modelers be formed to consider: i) improved ways to identify specific endpoints, or groups of endpoints, that should be modeled; and ii) improved ways of characterizing the activity, or inactivity, of substances used to “train” the model.
PROJECT METHODS / APPROACH
A three phase project will address the recommendations described above.
Phase 1 Tasks
Design a prototype developmental toxicity database via an interdisciplinary effort involving modelers, developmental toxicologists and database experts.
• Determine scope of database.
Which species should be included? Which routes of administration? Which types of substances? Which types of studies (e.g., regulatory protocols? published literature?) should be incorporated?
• Determine specific nature of data to be gathered.
How refined should data collection be? Should specific types of abnormalities/effects be captured (e.g., bipartite ossification of thoracic centrum, decrease in male offspring anogenital distance)? Or is it adequate to capture more general information (e.g., decrease in survival, or increase in thoracic effects)? How can the database deal with the varying ways developmental toxicity data are presented (e.g., with litter as the experimental unit, or with fetus as the experimental unit)? What study design information is essential to collect?
• Determine how expert judgment will be incorporated into the database.
Should judgment be incorporated as a “summary call”, or as findings on specific endpoints? What if expert toxicologists differ with the judgment of the author of a published study?
Phase II Tasks
Assess the adequacy of the database design via review and testing by a panel of expert developmental toxicologists. Identify “lessons learned” during the database development phase, and revise the database design as necessary to improve its capacity to compile data in a transparent, objective and reproducible manner. The following questions will be addressed:
• Are the proposed database fields adequate to capture critical study design parameters and data on effects? If not, what fields should be added/dropped/changed?
• Within a given field, is it possible to enter data from diverse studies in a consistent and unambiguous manner? Are nomenclature/definitions clear and uniform?
Phase III Tasks
Explore implications for SAR model development. What caveats/considerations should be taken into account when selecting and utilizing data to train an SAR model? Given there are numerous and diverse endpoints, how should the “activity” of interest be defined? Is it appropriate to categorize certain types of effects together - or conversely, how finely should certain types of effects be divided for separate consideration? What are considerations for “scoring” or weighting the activity or effect?
PROJECT DELIVERABLES
- A prototype database, with representative data from published developmental toxicity studies in the rat.
- A public report that describes the “lessons learned” from this interdisciplinary process.
CURRENT STATUS OF THE PROJECT
Phase II of the project has been completed and the prototype database is being finalized and prepared for release.
PROJECT STEERING COMMITTEE
John M. DeSesso Mitretek Systems
Paul M. D. Foster National Institute of Environmental Health Sciences
Elizabeth Julien ILSI Research Foundation Risk Science Institute
Bette Meek Health Canada Existing Substances Division
Philip E. Mirkes Texas A & M University
Ann M. Richard USEPA National Center for Computational Toxicology
Jennifer Seed USEPA Office of Pollution Prevention and Toxics
PROJECT SPONSOR
Health Canada Existing Substances Division
FOR MORE INFORMATION:
Contact Dr. Beth Julien, 202-659-3306 ext 158.