Reliable and Representative Estimation of Extrapolation Model Application in Deriving Water Quality Criteria for Antibiotics

Deriving water quality benchmarks based on the species sensitivity distribution (SSD) is crucial for assessing the ecological risks of antibiotics. The application of extrapolation methods such as interspecies correlation estimation (ICE) and acute-to-chronic ratios (ACRs) can effectively supplement insufficient toxicity data for these emerging contaminants. Acute-to-chronic ratios can predict chronic toxicity from acute toxicity, and ICE can extrapolate an acute toxicity value from one species to another species. The present study explored the impact of two extrapolation methods on the reliability of SSDs by analyzing different scenarios. The results show that, compared with the normal and Weibull distributions, the logistic model was the best-fitting model. For most antibiotics, SSDs derived by extrapolation have high reliability, with 82.9% of R-2 values being higher than 0.9, and combining ICE and ACR methods can bring a maximum increase of 10% in R-2. Based on the results of Monte Carlo simulation, the statistical uncertainty brought by ICE in SSD is 10-40 times larger than that brought by ACR, and combining the two methods could reduce uncertainty. In addition, the sensitivity test showed that whether the toxicity data came from extrapolation or actual measurement, the lower the value of toxicity endpoints was, the greater the bias caused by the corresponding species in every scenario. Combining the two aforementioned extrapolation methods could effectively increase the stability of SSD, with their bias nearly equal to 1. Environ Toxicol Chem 2022;00:1-14. (c) 2022 SETAC