Agrochemical manufacturers must meet regulatory requirements to provide information about the acute, subchronic, and chronic effects potentially caused by their products and active ingredients. Studies run to obtain these results can require up to 100 animals per product and 7000 animals per active ingredient. In the United States alone, this results in the use of approximately 600 rabbits per year for eye irritation testing of agrochemical formulations and highlights the value of implementing non-animal approaches. While several in vitro1 methods have been found to be appropriate for specific applications, no single method has been identified as a complete replacement for the rabbit eye test for classification and labeling of agrochemical formulations. Development of a defined approach would leverage the strengths of different nonanimal eye irritation test methods to predict the complete range of ocular irritation potential of agrochemical formulations. However, development of such approaches requires data from representative formulations that have been tested in multiple in vitro methods. The National Toxicology Program Interagency Center for the Evaluation of Alternative Toxicological Methods worked to develop a defined approach for eye irritation testing. To support this effort, CropLife America member companies provided data on formulations tested in the in vivo rabbit test method and in one or more in vitro test methods. Results compiled for 232 agrochemical formulations were reviewed to determine if a combination of in vitro methods could accurately assign U.S. Environmental Protection Agency (EPA) eye irritation hazard classifications. However, there was insufficient overlap of formulations tested in multiple methods to conduct comparative analyses with which to justify a proposed testing strategy. Accordingly, new data for a short list of agrochemical formulations were generated in a variety of in vitro test methods, including the bovine corneal opacity and permeability assay (BCOP), neutral red release assay, isolated chicken eye assay, porcine cornea reversibility assay, and EpiOcular (EO) test method. In addition to the standard BCOP and EO testing protocols, protocols that adjusted incubation time, testing concentration, and/or analysis method were evaluated. Altogether, 16 donated formulations were evaluated in eight different test method protocols. In vitro test results were compared to regulatory hazard classifications (United Nations Globally Harmonized System and EPA) that were assigned based on retrospective rabbit test method data. Our analyses showed that no single method produced results that completely aligned with the rabbit test. While a combination of test methods may provide better information, development of such integrated strategies is still confounded by the variability and questionable human relevance of the reference animal data. We discuss issues associated with a reliance on animal data as a reference for the evaluation of new testing approaches and the potential advantages of a practical, more human-relevant strategy for hazard classification