© 1994 Oxford University Press
research-article |
Dose-Response Assessment for Developmental Toxicity
I. Characterization of Database and Determination of No Observed Adverse Effect Levels1
*Department of Environmental Health, University of Washington, and Affiliate of the Child Development and Mental Retardation Center Seattle, Washington 98195
K. S. Crump Division, ICF Kaiser Ruston, Louistana 71270
Developmental Toxicology Division, Health Effects Research Laboratory, Office of Health Research. U.S. Environmental Protection Agency Research Triangle Park, North Carolina 27711
Reproductive and Developmental Toxicology Branch, Human Health Assessment Group, Office of Health and Environmental Assessment, U.S. Environmental Protection Agency Washington, DC 20460
Received April 19, 1993; accepted May 16, 1994
Developmental toxicity risk assessment currently relies on the estimation of reference doses (RfDDTS) or reference concentrations (RfDDTS) based on the use of no observed adverse effect levels (NOAELs) and uncertainty factors. The benchmark dose (BMD) has been proposed as an alternative basis for reference value calculations. A large database of 246 developmental toxicity experiments (Segment II-type studies) representing 1825 data subsets for various endpoints was compiled for use in comparing NOAEL and BMD approaches to developmental toxicity risk assessment. This paper describes the characteristics of the database used and the estimation of NOAELs using several approaches. For each endpoint evaluated, two NOAELs were calculated using the NOSTASOT procedure (Tukey et al., 1985). The first NOAEL calculation, the QNOAEL, was based on a quantal response where a litter was defined as "affected" if one or more fetuses or implants in the litter had the endpoint of interest. The second NOAEL calculation, the CNOAEL, was based on the proportion of fetuses or implants affected within each litter and was treated as a continuous response variable. Fifty-seven percent of the 246 experiments had at least one endpoint that showed a significant trend with dose. A total of 386 data sets were significant with respect to both the quantal and continuous test of trend. An additional 44 data sets were identified with significant trend only by the quantal approach whereas 177 additional data sets were identified with significant trend tests only by the continuous approach. Thus, the continuous approach appeared to be more powerful in detecting dose-related toxicity, but the patterns detected by the two approaches differed. QNOAELs and CNOAELs were compared by examining the degree to which they selected the same dose group. For 98% of the 386 data sets with both significant continuous and quantal trend tests, the CNOAEL was within one dose level of the QNOAEL. For data sets having significant continuous and/or quantal trend tests, 99% of the two NOAELs were within two dose levels. Twenty of the NTP studies were reviewed for comparison of "expert" versus statistically derived NOAELs. Most of the 360 NOAELs derived were identical for these two approaches; for the 8% that were different, approximately 80% were within one dose level. Thus, the statistically derived NOAELs that were used as the basis of our data comparisons were reflective of the typical methods used in deriving NOAELs. Compilation and characterization of such a large database of Segment II developmental toxicity experiments provide us with an opportunity to evaluate new quantitative approaches for developmental toxicity risk assessment and to make decisions about appropriate approaches for evaluation.