To determine the optimal conditions for clonogenic assay, the antitumor activities of 5-fluorouracil (5-FU) in vitro and in vivo were compared from a pharmacodynamic viewpoint in a human carcinoma xenograft-nude mouse system. In the clonogenic assay, tumor cells were exposed to 5-FU continuously for 2 weeks, and the results of the assay were evaluated in terms of colony survival rates (T/C). Tumor-bearing BALB/c male nude mice were treated with 5-FU, and the results were evaluated in terms of the lowest values of relative mean tumor weights (T/C). To compute the area under the curve (AUC) after administration of 5-FU in vitro and in vivo, agar and serum levels of 5-FU were measured by bioassay. Antitumor activities in terms of T/Cs against a gastric carcinoma strain (H-111) depended highly on the AUCs in vitro and in vivo. The T/C in vitro (z) was correlated with the AUC in vitro (w), z = 58.4 x 0.99331w, and the T/C in vivo (y) was also correlated with the AUC in vivo (x), y = 52.1 x 0.88552x. On the assumption that the T/Cs of these two regression equations were equivalent (y = z), a correlation between w and x was derived. To predict the T/C at the maximum tolerated dose of 5-FU in mice, the optimal drug concentration in vitro was calculated to be 2.7 micrograms/ml, which also proved to be suitable for seven other carcinoma strains. Our experimental system was thought to be adequate for selecting the optimal drug concentration in the clonogenic assay.