A Bayesian approach to analyzing ecotoxicological data

Abstract

Standardized chronic toxicity tests are usually analyzed using a NOEC (no observed effect concentration) or ECx (x% effect concentration) calculation. However,these methods provide very little information for the material cost they entail. It has been proposed that biology-based methods, such as the DEBtox approach, would make better use of the data available. DEBtox deals with the energy balance between physiological processes, and gives insight on how a compound disturbs it. We propose that data analysis can be further improved by estimating the DEBtox parameters using the considerable expertise available in laboratories and/or the literature. The Bayesian inference appears to be an appropriate estimation method for this purpose, as this technique takes expertise into account as prior probability distribution for each parameter, and provides the corresponding posterior distributions given the data. From these posterior distributions, point estimates can easily be deduced, but also credible intervals which are ideal for use in risk assessment. In this paper, we demonstrate this approach through the analysis of two 21-day Daphnia reproduction tests.