Canine pancreata obtained at total pancreatectomy were cannulated via the ducts and perfused with collagenase to prepare a tissue suspension that was isografted into the spleen (preparation ⋍2 h, mean graft vol = 10 ± 1 ml containing 24% of the B-cell mass/pancreas). In 13 dogs the tissue was implanted by reflux into terminal splenic veins: two died postoperatively, and in two the intrasplenic vein wall was inadvertently punctured during cannulation. In the remaining nine, mean fasting blood glucose (BG) was ≤150 mg/dl initially; one was killed at 2 wk (distemper) and one at 6 wk (sepsis, diabetes), and one died at 9 wk (intestinal obstruction). Mean BG was 94 ± 4 mg/dl at 1 mo and remained in this range until the dogs were killed at 5 mo (91 ± 1 3 mg/dl). During glucose-tolerance testing 1 wk preimplantation and 1 mo and 2-3 mo postimplant, mean values were: K (decline in glucose concentration, %/min), 3.4 ± 0.2,1.4 ± 0.1, and 1.5 ± 0.1; peak insulin μU/ml), 50 ± 5,12 ± 1, and 11 ± 2; fasting serum glucagon (mg/ml), 33 ± 3, 59 ± 12, and 53 ± 9, with no change in the glucagon response. Histologically, the spleens contained prominent islets. In five other dogs, the tissue was injected into the splenic pulp: mean BG rose to ≥250 mg/dl at 2 wk (compared with initial series, P < 0.001) and remained elevated until death at 6 wk, when histologic examination of the spleens showed severe fibrosis and no islets. Apancreatic controls (N = 4) survived 10 ± 3 days; BG was 343 ± 11 mg/dl terminally. We conclude that this modified method for collagenase perfusion of a single large-mammal pancreas via the ducts provides sufficient viable islets to induce prolonged normoglycemia (5 mo) and preserve the response to glucose challenge. Reflux of pancreatic fragments into splenic veins appears more efficient than intrapulp implantation.