The aim of the two studies presented in this paper was to further improve the predictability of the original Multicentre Evaluation of In Vitro Cytotoxicity (MEIC) in vitro test battery for acute systemic toxicity. In the first study, whether a protein-free cytotoxicity assay could improve the prediction of human acute systemic toxicity was investigated. The cytotoxicity of 39 MEIC reference chemicals was measured by the neutral red uptake inhibition test after 30 minutes in phosphate-buffered saline (PBS), with hepatoma-derived Fa32 cells. The results were compared with the corresponding values obtained in complete culture medium, including 10% fetal calf serum. Mercuric chloride and hexachlorophene were much more cytotoxic in PBS, as was the case, to a lesser extent, for seven other chemicals. Potassium cyanide and eight other chemicals were less cytotoxic in PBS than in complete culture medium, probably because of poor physiological conditions. The correlation between the cytotoxicity measured in PBS and human acute toxicity was rather low, but became of the same order as for other assays, when mercuric chloride and hexachlorophene were withdrawn from the comparison. In the second study, modelling of human lethal blood concentrations by using the results of the three cell line tests of the original MEIC test battery were complemented by logP (octanol-water partition coefficient) values. The introduction of logP into the modelling did not improve the correlations, but some improvement of both R(2) and Q(2) was obtained by expanding the logP values with logP(2) values. The highest R(2) (0.84) and Q(2) (0.80) values were obtained for a model in which both experimental and calculated (ambiguous) logP values were used. When only experimental logP values were used, the corresponding values were 0.80 and 0.78. These two studies showed that including protein binding and the partition of chemicals in the MEIC in vitro test battery is important, in order to improve the predictability of the results obtained.