Avoiding Pitfalls in the Use of the Benchmark Dose Approach to Chemical Risk Assessments; Some Illustrative Case Studies
The USEPA's benchmark dose software (BMDS) version 1.2 has been available over the Internet since April, 2000 (epa.gov/ncea/bmds.htm), and has already been used in risk assessments of some significant environmental pollutants (e.g., diesel exhaust, dichloropropene, hexachlorocyclopentadiene, 2-butoxyethanol). From the many user inquiries addressed in 2000, the support staff for BMDS has become aware of several categories of potential "pitfalls" that can present themselves in a benchmark dose modeling assessment. These pitfalls can be categorized as (1) attempting to fit bad data sets, (2) attempting to fit good data sets with inappropriate models, (3) not trying enough appropriate models, (4) giving up too soon on an individual model and (5) relying on inappropriate goodness-of-fit statistics to determine or compare model fits, and (6) depending too much on goodness-of-fit statistics to determine or compare model fits. This presentation will focus on several case studies Examples that illustrate techniques and BMDS software features that can be used to help avoid such problems and improve your overall risk assessment.