Defined approaches to predict GHS and EPA classifications for ocular irritation potential of agrochemical formulations

Regulations require that agrochemicals be labeled to indicate potential harmful effects caused by exposure. The in vivo Draize rabbit eye test has historically been the standard method used to assess the eye irritation or corrosion potential of chemical substances. However, as scientific confidence has been established for certain in chemico, in vitro, and ex vivo methods developed for this purpose, regulators are increasingly accepting data from such methods in lieu of the in vivo test. Defined approaches (DAs) may also be used to derive hazard and potency predictions by applying fixed data interpretation procedures to results from multiple methods, thereby leveraging strengths of different methods. Currently, the DAs accepted by regulators to predict eye irritation or corrosion potential do not specifically list agrochemical formulations within their applicability domains. To address this gap, we conducted testing to confirm the applicability of in vitro methods to agrochemical formulations and to develop DAs to predict eye irritation hazard labeling according to the Globally Harmonized System of Classification and Labeling (GHS) and the U.S. Environmental Protection Agency (EPA) classification system. Twenty-nine formulations were tested in up to four methods: bovine corneal opacity and permeability (BCOP; OECD TG 437) including histopathology, EpiOcular Eye Irritation Test (EO; OECD TG 492), SkinEthic time-to-toxicity for liquids (TTL; OECD TG 492B), and EyeIRR-IS. We propose four DAs comprising BCOP with histopathology alone, and combined with EO, TTL, or EyeIRR-IS. Instead of evaluating direct concordance of the four individual DAs with historical in vivo rabbit eye test data, for each formulation, we assessed orthogonal concordance of GHS and EPA classifications predicted across all five approaches. Predictions were considered orthogonally concordant when they aligned with the prediction of at least two other approaches (i.e. a majority, or at least 3 of the 5 approaches, achieved the same prediction), referred to as the ‘majority prediction.’ We also evaluated hazard labeling and PPE labeling associated with the GHS and EPA predictions, respectively. Relative to the hazard and PPE labeling associated with the majority predictions, each of the four DAs were as, or more, protective of human health than the rabbit test; hence, we conclude that these DAs can be used to predict the GHS and EPA classifications of agrochemical formulations.