Generation of novel in chemico in vitro skin sensitisation data to evaluate the human relevance of defined approaches
Skin sensitisation, leading to allergic contact dermatitis, is a common occupational health problem. Consequently, chemicals in products intended for human use must be assessed for their skin sensitisation potential. Traditionally this assessment has been carried out using in vivo assays like the murine local lymph node assay (LLNA), however, political and ethical pressure has led to increased use of non-animal alternatives such as in chemico and in vitro assays. Results from multiple information sources (non-animal assays, in silico models, physicochemical parameters) are often combined in what is known as a defined approach (DA).
It can be difficult to assess how well DAs predict human skin sensitisation potential as the amount of publicly available in chemico/in vitro data with corresponding human data is sparse. As such, a collaborative project was devised whereby chemicals lacking in chemico/in vitro data but with human potency data would be identified and new in chemico/in vitro data generated - ideally to test the relationship between in chemico/in vitro data and human potency.
Human data was taken from two publications which assigned a human potency category between 1 (extreme sensitiser) and 6 (non-sensitiser) to mainly fragrance-like chemicals based on human repeat insult patch test data, clinical data, and exposure. 80 chemicals were found with no or limited in chemico/in vitro data and of these, 34 were commercially available at a reasonable cost. These 34 were then prioritised into three groups based on the availability of LLNA data and (lack of) in chemico/in vitro data. The first group consisted of 8 chemicals and so far during this collaboration, DPRA, KeratinoSens™, and h-CLAT data have been generated using blind testing for this group. The predictivity of the individual assays when compared against LLNA data was between 50% - 88% and between 38% - 75% when compared against human data. When these novel data are used within 3 published DAs, the BASF 2/3, Kao ITS, and Lhasa DA, the predictivity against in vivo data ranges from 38% - 88%.