Bayesian multilevel discrete interval hazard analysis to predict dichlorodiphenyldichloroethylene mortality in Hyalella azteca based on body residues

Abstract

We exposed Hyalella azteca to p,p'-dichlorodiphenyldichloroethylene for intervals of 1 to 4 d and followed mortality out to 10 d. Mortality was determined as the cessation of heartbeat; dead organism body residue was determined daily. To model mortality probability, body residues of the living organisms were estimated using published kinetic data with concentration-dependent rate constants. The estimated residues compared favorably with measured residues in the dead organisms (predicted body residue = 1.302 ± 0.142 measured body residue + 10.351 ± 15.766, r² = 0.64, n = 50). The response data were collected at discrete intervals; thus, it was not possible to determine the exact time of death for organisms. Consequently, we analyzed the mortality data using discrete interval analysis, in a Bayesian hierarchical framework, with body residue as the dose metric. The predicted body residues to produce mortality were similar across the duration of exposure when postexposure mortality was considered. The concentration for 50% mortality was 0.47 μmol/g (148.6 tg/g, range 0.32-0.66 μmol/g), and predictions of response indicted 95% (range 73-99.9%) mortality at 0.79 μmol/g (250 μg/g) and 4% (range 1.2-9.6%) mortality at 0.16 μmol/g (50 μg/g). The lethal residue for 50% mortality based on interval analysis for short-term exposures with postexposure mortality resulted in values similar to long-term continuous exposures for exposure durations of more than 600 h.

General Unified Threshold Model of Survival - a Toxicokinetic-Toxicodynamic Framework for Ecotoxicology

“…The majority of the models in Table can be derived from GUTS by simply choosing different parameter values. However, several cases ,,, require a slight modification of the structure of one of the submodules. As they still follow the basic GUTS scheme of Figure , we consider them to be special cases.…”

“…To date, imprecise use of terminology has led to confusion, obscured the relationships between existing models and prevented the comparison of model parameter values from different studies. For example, TKTD models that include a damage state have used the assumption of an IT distribution , or the assumption of SD. ,, As both cases differ in their assumptions, also their model results and interpretation differ, but the use of the word “hazard” in both cases (wrongly suggesting the same meaning) has consequently led to confusion among readers. Further confusion may arise from the use of the same symbols for different model parameters or state variables (e.g., killing rate or recovery rate constants in refs and ) in different models, or the use of the term “elimination rate constant” for a fitted parameter in SD models for acute toxicity, which may also reflect a toxicodynamic recovery processes instead of toxicokinetics alone.…”

“…Even for an apparently simple endpoint as survival, a large number of very different TKTD approaches exist in the literature. ,− These approaches differ in their underlying hypotheses and assumptions, although the exact difference is often difficult to establish when the assumptions are not explicitly stated. This diversity of models and assumptions has led to confusion among scientists and has hindered the assimilation of TKTD models within ecotoxicological research and environmental risk assessment.…”