Quantitative Integration of Dose-Response Data for Relative Potency Estimates of Dioxin-Like Chemicals
Author Block: C. Ring1, S. Fitch2, L. Haws1, M. Harris3, and D. Wikoff4. 1ToxStrategies Inc., Austin, TX; 2ToxStrategies Inc., Dallas, TX; 3ToxStrategies Inc., Katy, TX; and 4ToxStrategies Inc., Asheville, NC.
Toxic equivalency factors (TEFs) for dioxin-like chemicals (DLCs) are informed by an underlying database of relative potency (REP) estimates composed of heterogenous data from various study types, species, duration, dose, methods for deriving REPs, etc. The objective herein was to update the REP database and combine these heterogeneous data to infer quantitative dose-response (DR) distributions of REPs for each of 28 DLCs. Using a systematic search approach, > 50 studies were added to the database, doubling the size to >1200 datasets with DR and/or REP data. Individual DR curves were fitted to each dataset with DR data. Then, each DR curve was standardized by normalizing to the reference DR curve for each dataset. For each congener, from the set of study-specific standardized DR curves and/or study-specific REP values, an “average” standardized DR curve was inferred using a hierarchical Bayesian model. By comparing each congener-specific average standardized DR curve to the TCDD average standardized DR curve, a TEF value for each congener was inferred, along with uncertainty in the inferred value. Quantitative weighting based on study quality was incorporated by modeling increased heterogeneity of study-specific DR curves for lower-quality studies. For many data-rich congeners (e.g. 2,3,4.7,8-PeCDF), predicted TEF values are on the order of 2005 WHO TEF values; TEF 95% credible intervals are less than 1 order of magnitude wide. For congeners with little data (e.g. 1,2,3,7,8,9-HxCDF), TEF 95% credible intervals are 2 or more orders of magnitude wide. By modeling standardized DR curves, our approach harmonizes heterogeneous DR and REP data, and allows full consideration of shape and parallelism of DR curves — collectively demonstrating integration of dose-response data across study types and endpoints, including quantitative consideration of individual REP quality and relevance.