The application of benchmark dose methodology to data from prenatal developmental toxicity studies

Abstract

The benchmark dose (BMD) concept was applied to 246 prenatal-developmental toxicity (DT) datasets from government, industry and commercial laboratories. Five modeling approaches were used, 2 generic and 3 specific to DT models. BMDs for both quantal and continuous data were compared with statistically derived no observed adverse effect levels (NOAELs) to determine similarities. Quantal (Q) endpoints included litter responses (e.g., one or more dead or malformed implants), and QBMDs were calculated using a Q Weibull (QW) model. Two types of continuous (C) data were modeled, the proportion of implants affected per litter, and the change in fetal weight (both mean and distribution); continuous power (CP) and DT models were used to calculate CBMDs. QBMDs for a 5% change in response (QBMD 05,) were 6-fold lower, on average, than the corresponding NOAEL. CBMD 05s on average were similar to the corresponding NOAELs, and CBMD 05s from different models were similar to each other. Including litter size but not threshold improved the fit of the DT models. For fetal weight data, specific cutoff values were used to calculate BMDs that were similar on average to the corresponding NOAELs: (1) changes from the control mean (5% of the mean, 25th percentile of the control distribution, or a decrease of 0.5 standard deviation), and (2) a 5 or 10% decrease in the proportion of fetuses below the 5th or 10th percentile, respectively, of the control distribution. These results support the use of BMDs as providing a more consistent basis for risk assessment than do NOAELs.