A Toxicokinetic–Toxicodynamic Modeling Workflow Assessing the Quality of Input Mortality Data

Abstract

Abstract:Toxicokinetic‐toxicodynamic (TKTD) models simulate organismal uptake and elimination of a substance (TK) and its effects on the organism (TD). The Reduced General Unified Threshold model of Survival (GUTS‐RED) is a TKTD modelling framework that is well established for aquatic risk assessment to simulate effects on survival. TKTD models are applied in three steps: parameterization based on experimental data (calibration), comparing predictions to independent data (validation) and prediction of endpoints under environmental scenarios. Despite a clear understanding of GUTS‐RED predictions’ sensitivity to the model parameters, the influence of the input data on the quality of GUTS‐RED calibration and validation is not systematically explored. We analyzed performance of GUTS‐RED calibration and validation based on a unique, comprehensive dataset, covering different types of substances, exposure patterns and aquatic animal species taxa that are regularly used for risk assessment of plant protection products. We developed a software code to automatically calibrate and validate GUTS‐RED against survival measurements from 59 toxicity tests and calculate selected model evaluation metrics. To assess whether specific survival data sets were better suited for calibration or validation, we applied a design where all possible combinations of studies for the same species‐substance combination are used for calibration and validation. We found that uncertainty of calibrated parameters was lower when the full range of effects (i.e. from high survival to high mortality) was covered by input data. Increasing the number of toxicity studies used for calibration, further decreased parameter uncertainty. Including data from both acute and chronic studies as well as studies under pulsed and constant exposure in model calibrations improved model predictions on different types of validation data. Using our results we derive a workflow, including recommendations for the sequence of modelling steps from the selection of input data to a final judgement on the suitability of GUTS‐RED for the dataset.

Identifying and Predicting Delayed Mortality with Toxicokinetic–Toxicodynamic Models

“…A general indication of the predictive power of short-term tests can be derived from the extensive study of Bauer et al (2024) on the calibration and validation performance of reduced GUTS models on various data sets for pesticides and aquatic invertebrates (14 species-chemical combinations, each with multiple toxicity tests). Figure 3 of Bauer et al (2024) shows that models that are "well calibrated" on acute toxicity studies with constant exposure will lead to a "successful validation" on chronic data sets (with both constant and pulsed exposure) in approximately half of the cases. Failure of shortterm data to predict longer term effects is therefore not uncommon, although the specifics of that failure would need to be more closely investigated (e.g., whether mortality in chronic tests is higher or lower than predicted from the acute test).…”

“…If we cannot expect to reliably extrapolate from the standard 2-day Daphnia test to longer time scales, what is the value of this test? To some extent, the information content from the 2-day test may be improved by including more observations over time, although this was not supported by the analysis of Bauer et al (2024). Perhaps improvement is possible by (additionally) performing tests with exposure pulses (as prescribed by EFSA, 2018) or by extending the test duration (e.g., until an incipient LC50 is reached).…”

Correspondence on “Mortality Pattern of Poecilus cupreus Beetles after Repeated Topical Exposure to Insecticide─Stochastic Death or Individual Tolerance?”

“…Fitting GUTS to a survival data set, and comparing the goodness of fit, is one way to test whether the hypotheses of SD or IT better explain the data. The body of literature on GUTS model applications shows that we can currently not select IT or SD as better performing (e.g., over 100 data sets ,,− ). Generally, both SD and IT explain the data well, and sometimes one or the other fits better, but not consistently across data sets.…”