GARD Skin and GARD Potency: A Proof of Concept Study to Investigate the Applicability Domain for Agrochemical Formulations
Author Block: M. Corvaro1, J. Henriquez2, R. Settivari3, U. T. Mattson4, and S. Gehen2. 1Corteva Agriscience, Rome, Italy; 2Corteva Agriscience, Indianapolis, IN; 3Corteva Agriscience, Newark, DE; and 4SenzaGen AB, Lund, Sweden.
In vitro methods for detection of delayed dermal sensitization have been formally validated for regulatory use in the last two decades as an alternative to the animal use. Some methods have reached regulatory acceptance as OECD test guidelines. The Genomic Allergen Rapid Detection (GARD™) is a genomic based assay platform which is currently being assessed for inclusion in the OECD test guideline program. GARD is available in the two variants, GARDskin and GARDpotency, addresses Key Event 3 (dendritic cell activation) of the skin sensitization Adverse Outcome Pathway, and provides reliably potency information for several chemical classes. Understanding of the applicability domain of test methods is pivotal in providing confidence in assay outcomes, facilitating regulatory uptake in specific industry sectors. The purpose of this work is to verify the applicability domain of GARDskin and GARDpotency, for the product class of agrochemical formulations. For this proof of concept, 20 agrochemical formulations were tested using GARDskin. When GARDskin was positive, GARDpotency assay was used to determine the severity of sensitization potential. Tests were conducted according to the assay developer Standard Operating Procedures. The selected agrochemical formulations were liquid (11 water based; and 9 organic solvent based) with a balanced distribution (11 not classified; 7 GHS cat 1B; 2 GHS cat 1A, which is rare for agrochemical formulations). GARD results (available for 18 formulations at this time) were compared with in vivo data (mouse LLNA) already available for registration purpose, in order to verify concordance (GHS hazard and potency categories). For hazard, GARDskin was able to correctly identify 7/10 not classified (true negatives) and 7/8 GHS1B/1A (true positives), with 1 false negative and 3 false positives. The accuracy, sensitivity, and specificity for prediction of hazard were 77.8% (14/18), 87.5% (7/8) and 70.0% (7/10), when using available LLNA results as classification reference. Additionally, GARDpotency was able to correctly identify 5 GHS cat 1B and 1 GHS cat 1A out of 7 correctly predicted sensitizer (underprediction from 1A to 1B occurred in 1 case). In conclusion, GARDskin and GARDpotency, showed a satisfactory performance in this initial proof of concept.