Probabilistic prediction of human skin sensitiser potency for use in next generation risk assessment

Abstract

Our aim is to develop and apply next generation approaches to skin allergy risk assessment that do not require new animal test data and better quantify uncertainties. We introduce the concept of the population threshold – a chemical-specific exposure level at which no individual in a population will experience induction of contact allergy to a chemical. A Bayesian multilevel (hierarchical) regression model is developed to estimate this population threshold under the conditions of a human-repeat-insult-patch-test. The approach is built on historical human (HRIPT) and murine (LLNA) data but, importantly, enables prediction based on in vitro (DPRA, Keratinosens™, hCLAT and U-SENS™) data. The Bayesian probabilistic framework allows us to explicitly quantify the uncertainty in the population threshold. Our skin allergy risk assessment defined approach (SARA DA) is used to estimate population thresholds for 30 chemicals using a weight-of-evidence across publicly available human, murine and in vitro data. Additionally, estimates for a further 43 chemicals are presented using chemical-specific data from in vitro assays only. Comparisons are made with current risk assessment metrics and across data types. We demonstrate that the approach can be used to derive a point-of-departure for next generation risk assessment based on in vitro data only.