You are here:
20191103 - Variability in in vivo Toxicity Studies: Defining the upper limit of predictivity for models of systemic effect levels (SETAC NA)
Citation:
Friedman, K., L. Pham, R. Judson, AND Woodrow Setzer. 20191103 - Variability in in vivo Toxicity Studies: Defining the upper limit of predictivity for models of systemic effect levels (SETAC NA). Society of Environmental Toxicology and Chemistry NA 40th annual meeting, Toronto, Ontario, CANADA, November 03 - 07, 2019. https://doi.org/10.23645/epacomptox.10304855
Impact/Purpose:
Abstract and poster for Society of Environmental Toxicology and Chemistry (SETAC) 40th annual meeting in Nov 2019. The objective of this work was to quantify the variance within systemic LEL and LOAEL values, defined as potency values for effects in adult or parental animals only. Using the study descriptors and potency values from ToxRefDB, multiple linear regression (MLR) and augmented cell means (ACM) models were used to quantify the total variance, and the fraction of variance explained by the available study descriptors, for systemic LEL and LOAEL values.
Description:
New approach methodologies (NAMs) to predict hazard are often evaluated via comparison to results from animal studies; however, variability in these in vivo reference data limits NAM accuracy. The US EPA Toxicity Reference Database (ToxRefDB) enables consideration of in vivo mammalian point-of-departure (POD) variability from subacute, subchronic, chronic, multi-generation reproductive, and developmental toxicity studies. The lowest dose at which an effect was observed in a study, termed the lowest effect level (LEL), and the lowest observable adverse effect level (LOAEL), for in-life observations and pathology, along with associated study descriptors, were used to understand variability across studies. The objective of this work was to quantify the variance within systemic LEL and LOAEL values, defined as potency values for effects in adult or parental animals only. Using the study descriptors and potency values from ToxRefDB, multiple linear regression (MLR) and augmented cell means (ACM) models were used to quantify the total variance, and the fraction of variance explained by the available study descriptors, for systemic LEL and LOAEL values. The MLR approach considered each study descriptor as an independent contributor to variance, whereas the ACM approach combined all categorical study descriptors, i.e. chemical, study type, species, sex, and administration method, into cells to more stringently define replicates. Using these approaches, total variance in systemic LEL and LOAEL values (in log10-mg/kg/day units) ranged from 0.74 to 0.92, and the unexplained variance, approximated by the residual mean square error (MSE), ranged from 0.20-0.39. Restricting the datasets to subchronic, chronic, or developmental study designs reduced the dataset size, but resulted in similar values. Based on the relationship between MSE and R-squared for goodness-of-fit, the maximal R-squared for a systemic POD model using these data may approach 73 to 78%. The root mean square error (RMSE) ranged from 0.41 to 0.59 log10-mg/kg/day, and suggests that a prediction interval for systemic PODs may have a width of 40 to 200-fold. These findings suggest an upper bound on predictive performance of NAMs based on these data and may have important implications for the evaluation criteria used for NAM predictions of systemic POD values. This abstract does not necessarily reflect U.S. EPA policy.
URLs/Downloads:
DOI: 20191103 - Variability in in vivo Toxicity Studies: Defining the upper limit of predictivity for models of systemic effect levels (SETAC NA)
PAUL-FRIEDMAN_VARIABILITY_SYSTEMIC_TOX_POSTER_SETAC_NOV2019.PDF (PDF, NA pp, 1411.915 KB, about PDF)