TR 116 – Category approaches, read-across, (Q)SAR

Abstract

TR 116 : Category approaches, read-across, (Q)SAR | 21 November 2012

Background

An accepted practice for the assessment of human health and environmental safety of chemicals is the use of models and analogues to fill data gaps for specific endpoints either for single or multiple chemicals that share structural similarities, and/or comparable reactivity or similarities in metabolism in mammals, fish and other organisms.  For example, this approach is acceptable, with limitations, in preparing dossiers for REACH, and it supports efforts for reducing animal testing.  The OECD has published guidance on the formation and use of chemical categories for data gap filling.  In December 2010, an ECETOC Task Force produced TR 109: High information content technologies in support of read-across in chemical risk assessment; a project that has highlighted methods for read-across.

With the plethora of models and guidance growing for both human health and the environment, it would be prudent to identify recommended practices.  Additionally, the 2013 and 2018 REACH deadlines are pending; these deadlines require lower volume producers and importers to submit chemical safety assessments.  A report describing recommended practices in this area would be useful in supporting industry?s risk characterisation and prioritisation activities across all sectors.

To this end, an ECETOC Task Force of industry and regulatory experts on categorisation methods, read-across and the use of (Q)SAR in risk assessment has prepared Technical Report No. 116: Category approaches, read-across, (Q)SAR.

The PDF can be downloaded without charge from the download link on this page.

Summary

Considerable practical experience has been gained in applying non-testing approaches for regulatory purposes, most recently driven by the demands of the REACH legislation (EC, 2006). This ECETOC Task Force was convened to summarise guidance and tools available, to review their practical utility and to consider what technical recommendations and learnings could be shared more widely to refine and inform on the current use of read-across. A number of case studies were formulated and the generic insights of developing, evaluating, justifying and documenting read-across approaches were extracted as far as possible. Sharing this experience aims to inform users about the pitfalls and challenges associated with read-across approaches as well as about feasible practical strategies available to develop appropriate scientific justifications. The report is intended to lay down the foundations of robust yet practical read-across approaches whilst encouraging consistent application across industry.

Currently read-across strategies for REACH have tended to rely on the so-called ?analogue approach? as opposed to a ?category approach?, in addition to (quantitative) structure activity relationship [(Q)SAR] approaches. (Q)SAR approaches have been extensively relied upon to address data gaps for physicochemical properties such as log Kow, environmental fate parameters such as biodegradation, hydrolysis, bioaccumulation potential and ecotoxicity endpoints such as acute aquatic toxicity in the standard species (fish, daphnia and algae). For the aforementioned properties, such (Q)SAR approaches have been used as direct replacements to experimental testing. There are many expert systems available to facilitate these assessments including the Organisation for Economic Cooperation and Development (OECD) (Q)SAR Toolbox or the United States Environmental Protection Agency (US EPA) EPISUITE programme. For mammalian endpoints, (Q)SAR have been applied less frequently with the exception of endpoints such as Ames mutagenicity and, to a lesser extent, skin sensitisation where the mechanisms of action are relatively well understood and where the underlying data are more readily available. Read-across approaches have been attempted as a means to address data gaps for longer-term effects such as 90-day repeated-dose toxicity studies or reproductive/developmental effects. For such complex endpoints, (Q)SAR have been applied, but their role has been as supporting information as a means to highlight potential chemical modes of action or to offer indications for similarity in effect. For such endpoints in particular, information on likely transformation products and the rate of formation of these products as derived from experimental studies are strongly recommended to substantiate the overall read-across justification. Whilst toxicokinetic information is not a requirement under REACH per se, such information is viewed as key to help rationalise certain read-across approaches, particularly for endpoints like reproductive/developmental effects where current (Q)SAR approaches are still in early development.

Absence of toxicity is a particular challenge to justify. Despite provisions in REACH calling for the use of read-across and (Q)SAR for both the absence and presence of toxicity (see Annex XI in EC, 2006), the justification of ?absence? is not to be underestimated. In many cases, toxicokinetic information or physiologically-based pharmacokinetic (PBPK) modelling is considered desirable to provide valuable supporting evidence.

Read-across approaches for longer-term effects should ideally be structured to present an overall ?weight of evidence? (WoE) argument (SCENIHR, 2012). A justification needs to rely on several lines of corroborating evidence whether it be consistent metabolic profiles, similarity in effects at shorter exposures, (Q)SAR estimates or other supporting analogues with experimental data that are not necessarily part of the main category/analogue approach. In the latter case, tools such as Toxmatch, Leadscope or the OECD (Q)SAR Toolbox may prove helpful to identify related analogues (Patlewicz et al, 2011).

It is recommended that justifications are structured using a template such as the category/analogue reporting format (CRF/ARF) as outlined in OECD (OECD, 2007a) and REACH guidance documents (ECHAREACH TGD). These templates are an effective means of structuring the arguments on an endpoint per endpoint basis as well as presenting an overall data matrix for the analogue or category member under evaluation. A justification is strengthened by the presentation of an explicit data matrix which demonstrates a consistent profile for the members of the category or analogues under consideration. Presenting data for the source analogues for the endpoint that is proposed to be read across alone may not be sufficient.

By default, read-across is considered to be associated with additional uncertainty due to the fact that information on a target substance is being inferred from that available on a source substance(s). Whilst assessment factors can be a route by which uncertainty is addressed, these should be used on a case by case basis and driven by the confidence associated with the underlying similarity hypothesis as well as the quality of the study data forming part of the supporting WoE information.

In the future, less ?classical? toxicity data will be anticipated for each individual analogue member, and rather more ?omics data (van Ravenzwaay et al, 2012). Thus, there will likely be a commensurate shift towards deriving larger categories as contrasted with analogue approaches. This should facilitate analysis of trends, although the data gap-filling approaches will likely be contingent on the application of non-standard, alternative toxicity testing data including that from high throughput/high content technologies. Whilst this will be a challenge in interpretation, it does present a cost-efficient means of generating data in a relatively short time frame.

Guidance and experience will continually evolve as Tox21 (See Chapter 1 for further details) activities progress. US Environmental Protection Agency‘s ToxCast is one such example (Judson et al, 2010) and the OECD’s adverse outcome pathway (AOP) work programme is another (OECD, 2011). Both will have an impact on the development of read-across justifications. AOPs may in the future have the potential to provide the conceptual framework for how to utilise alternative data in the appropriate biological context as well as the chemical anchor by way of molecular initiating events (MIEs). Datasets such as those generated in ToxCast and related programmes may ultimately help to formulate practical strategies to quantify AOPs. OECD’s grouping guidance, which is currently under revision, discusses AOPs as a means towards developing new categories and read-across that are more mechanistically based (OECD, 2011). AOPs will also be implemented in some fashion in the OECD (Q)SAR Toolbox to extend the scope of its functionality. It is anticipated that regulatory agencies will start to consider these approaches. Indeed, the US EPA have alluded to a shift in the development of their chemical categories from those that are purely based on structure and physicochemical properties to ones that rely on the concepts of AOP information to inform their development and evaluation (Seed, 2012).

This report presents a snapshot of current practices and highlights possible future needs and opportunities. Read-across is clearly evolving and the challenge will be to keep pace and drive the scientific development and evaluation of these approaches including the critical assessment of its limitations.