Eighteen patients with liver metastasis or locoregional recurrence of colon carcinoma received locoregional treatment by continuous 5-day infusions of 5-FU. 5-FU blood levels were measured by HPLC every day of the cycle at 8 am and 5 pm for a total of 87 cycles. Twelve patients were given the drug by an intra-arterial hepatic (i.a.h.) route, 3 by the portal vein (i.p.v.) and 3 by an intra-arterial pelvic (i.a.p.) route. These three routes were compared in respect of their relative pre-systemic drug uptake and the effect of dose escalation. Both the i.a.h. and i.p.v. routes, but not the i.a.p. route, resulted in a significant reduction in AUC 0-105 h compared to the i.v. route at the same dose range. Increasing the dose led to a modification in circulating 5-FU levels proportional to the dose for the i.v. and i.a.p. routes. By contrast, for the i.a.h. and i.p.v. routes, systemic drug delivery was significantly elevated, out of proportion with the dose, indicating a saturable process. For the i.a.h. route, increasing the 5-FU dose from 780 to 1000 mg m-2 day-1 caused a drop in hepatic extraction from 0.93 (0.90-0.95) to 0.44 (0.21-0.66). Liver saturation mechanisms were also evidenced by a mean increase of 2.6 times for the circulating drug level during the second part of the cycle as compared to the first part (P less than 0.001). The evolution of 5-FU AUC 0-105 h as a function of the dose was exponential (r = 0.75, P less than 0.001). Local extraction consecutive to i.a.p. was non-existent, implying that this route of drug administration has no potential advantage over classical i.v. infusion.