An in vitro pharmacokinetic simulation system that can simulate plasma pharmacokinetics was established to evaluate the cytotoxicity of two representative antileukemic agents, 1-beta-D-arabinofuranosylcytosine (ara-C) and daunorubicin. With this system, the survival rate of the cell line K562 treated with ara-C, relative to that of untreated control cells, was 7.1%, as determined by a clonogenic culture technique using a clinically intermediate dose of ara-C (1.0 g/m2; 2-h infusion). When the area under the serum concentration-time curve (AUC) was kept constant at infusion times of 2, 4, 8, and 16 h with a regular dose of ara-C (100-mg/m2 infusion), the relative cell survival rates were 75, 72, 34, and 14%, respectively. In contrast, with the use of a conventional culture system and a constant concentration-time product (C x T), no time-dependent inhibition effect by ara-C was observed, probably due to the ara-C inactivation in the cell-containing culture medium. For example, 93% of the ara-C in the cell-suspended medium in the conventional culture system was converted to its inactive form, 1-beta-D-arabinofuranosyluracil, within 16 h after addition of ara-C to the medium. For the simulations of the administration of 50 mg/m2 daunorubicin for 0.5-, 2-, 4-, and 8-h infusions, the relative survival rates [maximal concentrations (Cmaxs)] were 37.4% (0.24 microM), 49.7% (0.089 microM), 72.1% (0.055 microM), and 82.2% (0.032 microM), respectively. With the conventional culture system, the relative survival rates (Cmaxs) following daunorubicin treatment were 7.6% (0.48 microM), 18.6% (0.12 microM), 63.7% (0.06 microM), and 92.0% (0.03 microM) when the drug exposure times were 0.5, 2, 4, and 8 h, respectively, with a constant C x T. When the drug concentration for 90% cell killing by the conventional culture system was plotted against the exposure time on a logarithmic scale, the regression line for daunorubicin had a slope of -0.40, whereas the slope of cis-diamminedichloroplatinum (or cisplatin), a typical AUC-dependent drug, was -0.98. These results suggested that daunorubicin was Cmax dependent rather than AUC dependent. In the simulation system, this Cmax dependency was apparently reduced, probably because of the smaller difference of Cmaxs in the simulation system compared with the conventional system with the constant AUC. Thus, this simulation system can predict the effects of ara-C, daunorubicin, and other antineoplastic agents much more exactly than the conventional culture system in clinical use.