Seventy-nine mongrel dogs underwent total pancreatectomy. Fifteen dogs served as apancreatic controls and died 7.0 +/- 4.2 days later (mean +/- SD). The pancreases of 44 dogs (group 1) were intraductally distended by manual injection of Hanks' balanced salt solution (HBSS). Thereafter each organ was mechanically disrupted and subjected to collagenase digestion as described by Mirkovitch et al. The pancreases of 20 dogs (group 2) were intraductally distended and subsequently perfused with collagenase by a roller pump. The organs were then mechanically disrupted and filtered through screens as described by Horaguchi et al. The resulting tissue suspensions were injected into the spleens of the dogs as autotransplants in both groups, by direct punction of the splenic capsule in group 1 and by retrograde infusion via a splenic vein tributary in group 2. The functional outcome was better in group 2 than in group 1, as assessed by the number of animals that became normoglycemic after transplantation [15/20 (75%) vs. 13/44 (30%); P = .0025]. The degree of islet purification, as measured by an increase in the tissue insulin/amylase ratio, was higher in group 2, and in both groups it was higher in normoglycemic than in hyperglycemic animals. The percent engraftment [i.e., amount of insulin recovered from spleen as percent of tissue transplanted (mean, 15.4% in group 1 and 14.5% in group 2) or as percent of original pancreas (mean, 4.9% in group 1 and 4.4% in group 2)] was low in both groups but again was higher in normoglycemic than in hyperglycemic animals within each group. In conclusion, both the degree of engraftment and purification and the route of implantation influenced the functional outcome after dispersed pancreatic islet autotransplantation to the spleen of totally pancreatectomized dogs, with purified tissue injected retrogradely functioning better than unpurified tissue injected directly.