This report describes the application of chemoinformatic methods to explore the applicability of the Threshold of Toxicological Concern (TTC) approach to cosmetic ingredients. For non-cancer endpoints, the most widely used TTC approach is the Cramer classification scheme, which categorises chemicals into three classes (I, II and III) depending on their expected level of concern for oral systemic toxicity (low, medium, high, respectively). The chemical space of the Munro non-cancer dataset was characterised to assess whether this underlying TTC dataset is representative of the "world" of cosmetic ingredients, as represented by the COSMOS Cosmetics Inventory. In addition, the commonly used Cramer-related Munro threshold values were applied to a toxicological dataset of cosmetic ingredients, the COSMOS TTC dataset, to assess the degree of protectiveness resulting from the application of the Cramer classification scheme. This analysis is considered preliminary, since the COSMOS TTC dataset and Cosmetics Inventory are subject to an ongoing process of extension and quality control within the COSMOS project. The results of this preliminary analysis show that the Munro dataset is broadly representative of the chemical space of cosmetics, although certain structural classes are missing, notably organometallics, silicon-containing compounds, and certain types of surfactants (non-ionic and cationic classes). Furthermore, compared with the Cosmetics Inventory, the Munro dataset has a higher prevalence of reactive chemicals and a lower prevalence of larger, long linear chain structures. The COSMOS TTC dataset, comprising repeat dose toxicity data for cosmetics ingredients, shows a good representation of the Cosmetics Inventory, both in terms of physicochemical property ranges, structural features and chemical use categories. Thus, this dataset is considered to be suitable for investigating the applicability of the TTC approach to cosmetics. The results of the toxicity data analysis revealed a number of cosmetic ingredients in Cramer Class I with No Observed Effect Level (NOEL) values lower than the Munro threshold of 3000 μg/kg bw/day. The prevalence of these "false negatives" was less than 5%, which is the percentage expected by chance resulting from the use of the 5th percentile of cumulative probability distribution of NOELs in the derivation of TTC values. Furthermore, the majority of these false negatives do not arise when structural alerts for DNA-binding are used to identify potential genotoxicants, to which a lower TTC value of 0.0025 μg/kg bw/day is typically applied. Based on these preliminary results, it is concluded that the current TTC approach is applicable to cosmetics, although a number of improvements can be made, through the quality control of the underlying TTC datasets, modest revisions / extensions of the Cramer classification scheme, and the development of explicit guidance on how to apply the TTC approach.
