What is GUTS?GUTS stands for "General Unified Threshold model for Survival". In other words, a modelling framework for the analysis of toxicity tests for the endpoint "survival". The method is expertly suited to analyse the results of standard acute toxicity tests (such as the 4day test with fish, and the 2day test with Daphnia). The "unified" relates to the fact that almost all models for the analysis of survival data can be seen as special cases within the GUTS framework. More background on GUTS. The simplest special case of GUTS was included into OECD and ISO guidances as early as 2006, under the denominator "biologybased methods". Download the OECD guidance. GUTS thus replaces the standard doseresponse analyses that are used to calculate an LC50. Advantages of a GUTS analysis:
A practical exampleThe advantages of the GUTSapproach are most easily demonstrated by an example analysis on a typical data set. The data used here are for the insecticide fenvalerate in fathead minnows (Pimephales promelas). The test design includes 6 treatments (including a control) with 20 individuals per treatment, and 4 days at which survival is scored. Normally, only the results on day 4 are used to calculate an LC50 (which is problematic here as there is only a single treatment with partial mortality). 
These data can be analysed with GUTS. Here, the fit is made with the special case of the "hazard model with scaled internal concentrations" (identical to the method as described in the ISO and OECD guidances). Left the fit to the data set, and right the 4 fitted parameters with their confidence intervals. 
The noeffect concentration (or NEC) should
be seen as the concentration that has no effect on
mortality, even after prolonged exposure (of course under
the assumption that the model is correct). The "elimination
rate" determines the rate at which the internal
concentration in the animal equilibrates with the
concentration in the medium (the rate is a combination of
the toxicokinetic elimination and possible transformation
processes and receptor and damage kinetics). The model parameters can now be used to generate a number of predictions. For example for the LC50 versus time (with 95% confidence interval), and even for the LC50 after 8 days (an extrapolation based on the estimated parameters). 
The LC50 depends on time, and after prolonged
exposure, the value will approach the noeffect
concentration (the "incipient LC50" thus equals the NEC). In
this case, the LC50 rapidly approaches the ultimate value,
caused by the relatively high value of the elimination rate.
The value for the standard 4dLC50 is determined with
greater precision than with a standard doseresponse
analysis as all data are used in the process. The parameters can also be used to predict mortality as a consequence of a specific exposure pattern (e.g., the results of a fate model). The prediction hinges on the assumption that the model is correct, and is only relevant for the specific other conditions in the test (temperature, lack of food, species and size of fish, etc.). As an example, below a prediction of the mortality (right, with 95% confidence interval) as a result of an exposure scenario with two pulses (left) of fenvalerate. 
Scientific literature, selection:
