We evaluated the best route of administration of TNP-470, an angiogenesis inhibitor, by comparing the anti-tumour effects and toxicity following injection via the hepatic artery, the portal vein, or the jugular vein in a rabbit model of liver metastases. Following the injections of 1 x 10(6) VX2 carcinoma cells into the portal vein of rabbits, 50 mg of TNP-470 was injected continuously into the hepatic artery, portal vein, or jugular vein for 7 days. The number of tumours on the surface of the liver was counted 14 days following the start of the infusion, and the serum glutamic-oxaloacetic transamine (GOT), glutamic-pyruvic transaminase (GPT) and total bilirubin concentrations were examined. In addition, a coloured silicon rubber was injected into the vessels of the liver to visualise the capillary networks around the tumours and assess the degree of suppression of angiogenesis by TNP-470. The mean number of tumours following intra-arterial injection (17.5 +/- 2.9) was significantly less than the control (237.0 +/- 34.0) (P < 0.05). The mean numbers of the tumours following intraportal (89.1 +/- 16.0) and intravenous (140.6 +/- 31.2) injection were both less than the controls (215.3 +/- 45.5, 284.8 +/- 55.4 respectively), but the differences were not significant. We conclude that intra-arterial injection of TNP-470 is the most effective method for preventing liver metastases in this model.