The cytotoxicity test of neutral red (NR) uptake in normal rabbit corneal epithelial cells (CornePack((R))) was validated as an alternative method to the Draize rabbit eye irritation test (Draize test). We tested 38 cosmetic ingredients as well as isotonic sodium chloride solution in three phases of the validation study. The test procedures were controlled among participating laboratories under a common standard operating procedure (SOP). The concentration of test substances that showed a 50% reduction in NR uptake relative compared with controls (median NR uptake concentration: NR(50), mug/ml) was determined and compared with in vivo Draize scores. Six laboratories participated in the first phase of the validation study, seven in the second, and five in the third. The average interlaboratory coefficient of variation (CV) was 32.9%. The correlation and rank correlation coefficients between the maximal average Draize total scores (MAS) and NR(50) were -0.583 and 0.587, respectively. When the anionic detergents were excluded from analysis, the correlation coefficient increased to -0.738. When the cut-off point for positive and negative irritation was set at MAS of 15 and the predictability of this method was assessed by liner regression line, six substances (two acids, two alkanolamines and two alcohols) were false negative. Through this project, it appeared that CornePack, supplied in kit form with frozen secondary cultured cell in serum-free medium, could provide an effective, highly sensitive and simple preliminary screen for determining the cytotoxicity of substances. These results suggested that CornePack might have the potential to predict the MAS if definite criteria can be established for the compounds to be applicable. However, it is important to understand the nature of CornePack responses since its NR(50) profile was quite different from other cytotoxicity assays.