Two Procedures for the Prediction of Acute Toxicity (LD50) from Cytotoxicity Data

Abstract

Single linear regression analysis was used to characterise the relationship between cytotoxicity in a variety of mammalian cell culture systems and acute oral toxicity (LD50) in experimental animals. The following results were obtained. Firstly, in a cytotoxicity assay using the calf aortic endothelial cell line BKEz-7, IC50 values determined for 44 chemicals in culture showed significant correlation with the oral LD50 values for rat and mouse (computed correlation coefficient r=0.546). After eliminating three chemicals that were characterised by extreme lethality indices (LI = IC50/LD50), the correlation coefficient of the remaining 41 chemicals increased to a value of r=0.728. By using the linear regression model for these 41 chemicals, the oral LD50 for rat and mouse can be predicted correctly from the IC50 values for 83% of substances from a variety of chemical substance classes within a range of approximately one order of magnitude of dosage unit of LD50 for rat and mouse. Secondly, the mean IC50 values (IC50x¯) determined as the geometrical mean of two or more IC50 values per substance, which were generated in a wide spectrum of mammalian cell lines and collected in a “Registry of Cytotoxicity” (RC), gave similar results (r=0.644). Likewise, with the aid of this method, the oral LD50 for rat and mouse can be predicted for 74% of non-selected chemicals from structurally-different classes in the same dosage range, e.g., 1–25 millimoles per kg body weight. The prediction of LD50 values from in vitro cytotoxicity data may permit the calculation of a more precise dose range-finding and offers a new way for reducing the number of animals in acute toxicity testing.