We compared the efficacy of continuous intraarterial versus intravenous 6-mercaptopurine (6-MP) infusion in a mongrel canine renal allograft model with regard to overall survival, incidence of systemic and renal toxicity, and systemic drug exposure. Arterial anastomoses were done end-to-end, and infusion catheters were placed in the iliac artery or vena cava and connected to a subcutaneously placed programmable pump. A dose of 0.5 mg/kg/day 6-MP did not prolong survival over heparin-treated or untreated controls (MST = 7 days for both groups) when administered either locally or systemically. However, 0.75 mg/kg/day 6-MP i.a. (MST = 20 days) significantly prolonged survival over both un treated (P= 0.007) and heparin-treated controls (P = 0.02), with all dogs eventually dying of rejection. In contrast, 0.75 mg/kg/day i.v. (MST = 7 days) failed to prolong survival over controls (P< 0.1) and produced death from systemic toxicity in 3 of 7 animals. Six of 7 dogs receiving 2.0 mg/kg/day 6-MP i.a. (MST = 12 days) developed azotemia secondary to drug-induced nephrotoxicity. Identical renal histologic changes occurred in the same time frame in autotransplants treated similarly. Of 7 animals receiving 2.0 mg/kg/day i.v. (MST = 12 days), 5 died from early, severe systemic drug toxicity and 2 from early rejection. During 6-MP infusion at 0.5 mg/kg/day, systemic exposure was significantly less in the locally treated than in the systemically treated dogs when Cr concentrations were normal or moderately elevated (P< 0.0005 andP= 0.01, respectively) but not when renal function became severely impaired (P= 0.34). In contrast to i.v. infusion, i.a. 6-MP delivery dissociated immunosuppressive efficacy from systemic toxicity, supporting previous work demonstrating high first-pass renal elimination of 6-MP. We conclude that tightly controlled local delivery of an immunosuppressive agent can effectively prolong graft survival with reduced systemic toxicity in a large animal model employing a pump/catheter system applicable to man.