Retain & Refine and tiered approaches

A basic idea of Retain & Refine is that entities (e.g., substances) can be handled in different ways (more or less refined) while still being considered together in the same risk assessment (retain). We refer to such different ways as tiers.

In the modular design, a tier is defined here as a specific set of settings for a module or a group of modules. Tiers can differ in many respects, and there is no single dimension to rank tiers as low vs. high. In risk assessment, typical tiers contrast deterministic to probabilistic approaches, conservative to realistic approaches, approaches using restricted data to approaches using more extensive data, and approaches using different degrees of model complexity. For each of the modules of the toolbox, as many tiers are implemented as considered useful for the practice of risk assessment.

Each calculation in the modular design may involve multiple, nested, calculations of sub-modules. A risk (or health impact) assessment builds on an exposure assessment and a hazard assessment, the exposure assessment builds on a dietary and a non-dietary exposure assessment, the dietary exposure assessment builds on a consumption assessment and an occurrence assessment, etc. Tiers can be defined at each node of the assessment network. An example consists of the tiers ‘IESTI’, ‘EFSA basic optimistic’ and ‘EFSA basic pessimistic’ which are defined at the level of a dietary exposure assessment, but include the settings for the corresponding tiers at the level of the concentration model calculator.

Each calculator has as a main output entities that can be specified to have different tiers (tiered entities). For example, in a hazard assessment, some substances may be assessed using a tier ‘Hazard Dose from dose-response data’, other compounds may be assessed using a tier ‘TTCx100’ or ‘sample from general NOAEL distribution x100’ (which only requires knowledge of the Cramer class of the compound). As another example, in dietary exposure assessment some food-compound combinations may be recognised as risk drivers for which a more complex approach (e.g. probabilistic modelling) is required, whereas a simpler approach (e.g. deterministic modelling) may be sufficient for all other food-compound combinations. So in this case the tiered entity is ‘food-compound’. A typical risk assessment will start at a tier that is simple to perform for all tiered entities (potential risk drivers). Note that, based on data availability and ease of application, the initial assessment can already include more complex elements, such as probabilistic modelling. If the initial calculations produce risk estimates that do not exclude concern, refinement of the modelling for the perceived risk drivers is useful for checking whether this concern is real.