Sandrine Charles, Carlo Albert, Benoit Goussen, Tjalling Jager, Soeren Vogel, Roman Ashauer
SETAC Nantes 2016. Nantes, France
Publication year: 2016

Presenting author: Sandrine Charles


The General Unified Threshold model of Survival (GUTS) provides a theoretical framework for analysing stressor effects on survival over time through consistent model equations based on different assumptions about the stressor quantification, the compensatory processes and the nature of the death process. In ecotoxicology, stressors are toxicants characterised by a dose metric, e.g. the concentration in the medium surrounding an organism or inside the organism, or by the damage quantity they cause. The key GUTS feature is that mortality is estimated when the dose metric exceeds a certain threshold. Several GUTS flavours can be derived according to the assumption underlying the death process: (i) the threshold is distributed within a population, and when exceeded, the individual dies (individual tolerance, IT); (ii) there is one common threshold for all individuals, and when exceeded, the probability to die increases (stochastic death, SD); (iii) a unification of both previous assumptions (GUTS proper). While more realistic, GUTS proper requires the estimation of one additional parameter. Because environmental risk assessment of chemicals depends on robust estimates of GUTS parameters, we investigated parameter identification for GUTS proper, in relation to the experimental design of ‘short-term’ laboratory bioassays. In practice, standard survival datasets generally do not contain enough information to estimate all parameters of GUTS proper with sufficient precision. This is because a large number of individuals is required to provide strong information on probabilistic events. Hence, based on simulated datasets we identify appropriate experimental designs suitable to estimate all parameters of GUTS proper with the best possible precision. We show that datasets with a high number of animals per treatment allow for parameter estimation of GUTS proper with reasonable accuracy and precision. Moreover, increasing the number of animals or the duration of the experiment substantially reduce the uncertainty around the median value of the threshold. Nevertheless, general statements about optimisation for any chemical, any species, any test duration and/or any exposure concentration profile remain difficult. As take-home message, to the extent possible, we recommend not to use fixed experimental set-up for GUTS analyses, but rather tailor dedicated designs according to the chemical, the species and/or the research/regulatory question at hand.