Clearly weighing the evidence in read-across can improve assessments of data-poor chemicals

Abstract

This paper provides a systematic weight-of-evidence method for read-across analyses of data-poor chemicals. The read-across technique extrapolates toxicity from analogous chemicals for which suitable test data are available to a target chemical. To determine that a candidate analogue is the 'best' and is sufficiently similar, the evidence for similarity of each candidate analogue to the target is weighed. We present a systematic weight of evidence method that provides transparency and imposes a consistent and rigorous inferential process. The method assembles relevant information concerning structure, physicochemical attributes, toxicokinetics, and toxicodynamics of the target and analogues. The information is then organized by evidence types and subtypes and weighted in terms of properties: relevance, strength, and reliability into weight levels, expressed as symbols. After evidence types are weighted, the bodies of evidence are weighted for collective properties: number, diversity, and coherence. Finally, the weights for the types and bodies of evidence are weighed for each analogue, and, if the overall weight of evidence is sufficient for one or more analogues, the analogue with the greatest weight is used to estimate the endpoint effect. We illustrate this WoE approach with a read-across analysis for screening the organochlorine contaminant, p,p'-dichlorodiphenyldichloroethane (DDD), for noncancer oral toxicity.

Keywords: Analogy; DDD; DDT; Evidence integration; Read-across; Weight of evidence.

Figure 1.

The process diagram for characterizing the toxicological endpoint for a chemical by read-across, has been adapted from Patlewicz et al. (2018). The weight of evidence (WOE) approach, elaborated on the right, is adapted from US EPA (2016).

Figure 2.

The hierarchy of properties and subproperties of evidence and of collective properties of bodies of evidence. The three properties of evidence and the collective properties are taken from the US EPA WoE guidelines (US EPA, 2016), but more detailed subproperties are specific to read-across.