The standard approach of RA includes four different phases, which are summarised below (EU TGD Part II, 2003; van Leeuwen and Vermeire, 2007).
- Hazard identification;
- Dose (concentration) – response (effect) assessment;
- Exposure assessment;
- Risk characterisation.
In the first step, the effects of concern are identified and the hazard classification of the substance according to GHS (Global Harmonised System) is established or reviewed.
In the dose (concentration) – response (effect) assessment, the predicted no-effect concentration (PNEC) is developed. For both steps, it is of high importance to evaluate the data with regard to their adequacy, relevance, and completeness. For the derivation of the PNECs, all available data should be taken into account. The adequacy of data is generally evaluated with a scoring system according to Klimisch et al (1997). This system has four scores (1: reliable without restrictions; 2: reliable with restrictions; 3: not reliable; 4: not assignable) and is used throughout the registration works for REACH. In principle, only data from categories 1 and 2 should be preferred. Data from category 4 might be used in a weight-of-evidence approach. From several studies a key study needs to be chosen which is then used for the derivation of the PNEC (Predicted No Effect Concentration) by applying appropriate assessment factors to account for the extrapolation from acute to chronic exposure, for the variability between experimental data, and for the uncertainty in extrapolating from single species data to a population. The PNEC is the threshold concentration which must not be exceeded in order to avoid deleterious effects on the environment.
In the subsequent exposure assessment, the environmental concentrations are determined either by using/gathering environmental monitoring data, or more commonly by modelling exposure in a hypothetical standard environment.
The starting point of the exposure assessment in predictive modelling is to define the use of a chemical and the rate Q at which the chemical is used (e.g. kg/day or tons/day). In order to define the emissions to the environment, release factors to water, air and soil are defined (Reihlen et al, 2016). A variety of sources is available to that end. (Reihlen et al, 2016). The emissions E to water, air, and soil are obtained according to equation ().
The emissions are an input to multimedia fate modelling. For REACH the EUSES model is employed (ECHA, 2016). It represents the environment as a nested set of boxes which are in exchange with each other (Brandes et al, 1996). It accounts for partitioning between compartments, degradation reactions and advective transport between compartments / boxes. The multimedia fate model yields so-called predicted environmental concentrations (PECs).
To ensure realistic PECs, all relevant exposure-related information on the substance is used. The PECs are derived for every single use of the compound and for every environmental compartment. Due to the lack of monitoring data, modelled PECs are usually used and the user has the option to refine his underlying modelling assumptions from conservative default parameters to more realistic conditions.
Calculation of RCR
The last step encompasses the risk characterisation in which the PEC and PNEC values for the different environmental compartments are used to develop a risk characterisation ratio (RCR = PEC/PNEC for a given compartment) which needs to be below 1.0. If this is not the case, further refinements must be done to ensure an RCR < 1, e.g. the use of risk management measures (RMMs) or the generation of additional experimental toxicity data to reduce the assessment factor.
As described for instance by van Leeuwen and Vermeire (2007), risk assessment is often a tiered approach. This is possible since the goal of the risk assessment is to assess whether PEC is below PNEC. In order to be resource efficient, the assessment often starts with conservative assumptions. If necessary, these conservative assumptions are replaced with less conservative assumptions, and if needed and possible with measured data. In this manner, the realism of the assessment is increased. There are multiple refinement options. For exposure modelling, these include replacing very generic emission estimation by measured release rates, replacing partition coefficients derived from quantitative structure-activity relationships with measured partition coefficients, using geo-referenced models (GREATER) rather than EUSES. Another option is to use measured environmental concentrations rather than predicted values. Finally, the effect threshold can be refined by using chronic instead of acute ecotoxicity data. Even further refinement is possible, e.g. the use of mesosome data or by species sensitivity distributions (SSDs).
For the sake of transparency, it is crucial that every deviation from default parameters needs to be documented and justified. At the end of the process, the entire assessment has to be communicated both to the authorities and to the downstream users handling the substance. The downstream users are then obliged to verify that they handle the compound in a way which is in compliance with the boundaries described in the assessment.