You are here:
Comparison of read-across in REACH registration dossiers with GenRA
Citation:
Patlewicz, G., P. Karamertzanis, M. Sannicola, K. Paul-Friedman, AND I. Shah. Comparison of read-across in REACH registration dossiers with GenRA. QSAR, Copenhagen, DENMARK, June 05 - 09, 2023. https://doi.org/10.23645/epacomptox.22939829
Impact/Purpose:
N/A
Description:
Read-across is a data-gap filling technique utilised to predict the toxicity of a target chemical using data from similar analogues. It is one of the most frequently used adaptations for higher tier endpoints under REACH, with approximately half of the non-intermediate substances produced or imported at >10 tpa using read-across and categories for repeated dose toxicity information requirements. Read-across acceptance remains an issue, mainly due to the difficulties of addressing residual uncertainties and because read-across is still often implemented as a subjective expert driven assessment. There have been many efforts to identify the sources of uncertainty in read-across, characterise them consistently and identify practical strategies to address and reduce those uncertainties. Notable of these efforts have been the creation of frameworks to develop, assess and document read-across. Efforts also include transitioning to data driven approaches such as Generalised Read-Across (GenRA) where uncertainties and performance can be quantified. GenRA affords opportunities for New Approach Method (NAM) data e.g. metabolic information, reactivity information, as well as biological data, to be incorporated to support the read-across hypothesis and strengthen its justification. A key issue that remains is how to reconcile an expert driven approach with data driven approaches in terms of establishing scientific confidence in the use of NAM data. This study aims to explore these issues by building a database of expert driven read-across assessments making use of REACH registration data as disseminated in the IUCLID REACH Study Results. The dataset contains hazard information for approximately 25,000 substances, out of which approximately 7500 are legally required to have at least a 28d repeated dose toxicity study. By focusing on repeated dose toxicity we extracted ~27 000 study documents with Klimisch score being reliable with/without restriction, that were either tests with the registered substances or a read-across source. By further filtering on the more recently reported read-across arguments leveraging the latest IUCLID structure, we produced a list of ~ 3000 read-across cases. This dataset is being used as a starting point to analyse the structural differences between the target and the source and compare their predicted toxicities. By applying the GenRA framework, combined with NAM data, we anticipate examining which expert analogue groupings appear insufficiently similar, which will in turn help evaluate to what extent these NAM data may refine or support target-analogue relationships. This study is a first attempt to systematically analyse the robustness of the most frequent adaptation under REACH. It charts a way towards analysing the received read-across and category arguments for all endpoints and assessing the potential of NAMs as bridging evidence for supporting the read-across hypothesis. Finally, it can be used to assess the bias of expert (rather than data) driven read-across and the extent to which it can result in unidentified hazards.
URLs/Downloads:
DOI: Comparison of read-across in REACH registration dossiers with GenRA
PRESENTATION.PDF (PDF, NA pp, 4487.795 KB, about PDF)