Abstract: ObjectiveTo establish a tiered screening model for testing eye irritation of surfactants.MethodsTwenty surfactants were tested with FLT, RBCH, and NRU methods, then the test results were plugged into four screening models (FLT+NRU, FLT+RBCH, RBCH+NRU, NRU+RBCH) and compared the grading results with those of Draize test (based on the indices of Gamma coefficient，McNemar-Bowker coefficient, Kappa coefficient, and concordance rate).Finally, the predictive abilities of the four tiered screening models for eye irritation of surfactants were evaluated.ResultsCompared with the results of Draise test, the concordance rates of grading results of four screening models (FLT+NRU, FLT+RBCH, RBCH+NRU, NRU+RBCH) evaluated in this study were 90.0%, 80.0%, 80.0%, and 65.0%, respectively; the Kappa values were 0.84, 0.68, 0.69, and 0.47, respectively (P<0.01 for all); the Gamma values were 1.00, 1.00, 0.93, and 0.88, respectively (P<0.01 for all).ConclusionThe tiered screen model composed of FLT test (1st stage) and NRU test (2nd stage) has better detection performance than the other 3 models and is more suitable for the eye irritation detection of the surfactants.