[Is the prediction of LD50 using cell cultures generally valid?].

In this article we use the linear regression model to compare the values of IC50 and LD50 p.o. (mice and rats) taken from the literature and estimated in our laboratory. By this analysis it was found that for different substance classes a similar relationship exists between IC50 and LD50, even if different cell types or cellular activities were used for estimating the IC50 on the one hand and different modes of substance application in animal toxicity tests on the other hand. The minimum and maximum values of the measured LD50 (LD50G) registered in animal experiments deviated from the theoretical LD50 values (LD50T) on the (log transformed) regression line in approximately 77 per cent of the cases by the factor FG less than or equal to log 5 only. The linear correlation between IC50 and LD50 is to be expected only for such substances which do not act specifically via higher levels of integration, cell type specific reactions or metabolites, respectively. These and previously reported results indicate a certain general validity not only for the positive linear correlation between IC50 and LD50 but also for the predicting the LD50 in a relatively narrow dosage range on the basis of in vitro estimated IC50 values. With these results new possibilities were opened for reducing animal experiments to estimate LD50.