The impact of multiple donor and recipient variables on functional survival of 307 cadaveric pancreas allografts transplanted in 253 recipients at the authors' institution between July 25, 1978 and September 4, 1990 was determined using the Cox proportional hazards regression model. Relative risk of graft loss was calculated for all cases as well as for technically successful (TS) ones. Factors with an impact in descending order of significance for TS cases were immunosuppression (RR = 3.9 for double-drug versus triple-drug maintenance, p < 0.0001); recipient category (RR = 2.4 for pancreas alone versus simultaneous pancreas/kidney, p = 0.009); retransplantation (RR = 1.8 for retransplants versus primary grafts, p = 0.007); donor hyperglycemia (RR = 1.7 for blood glucose ≥ 200 versus < 200mg/dL, p = 0.02); human leukocyte antigen (HLA) matching (RR = 2.1 for poor versus a good match, p = 0.04). A logistic regression analysis also was performed to determine which factors predisposed to technical failure; none were identified. To make the model as relevant as possible to their current program, the authors analyzed only the bladder-drained cases (n = 221; 1984 to 1990). All patients received triple therapy. Recipient category, retransplantation, donor hyperglycemia, and degree of HLA matching remained as significant risk factors. Construction of estimated survival curves showed that the results of retransplantation were significantly improved, and the penalty incurred by using hyperglycemic donors was partially ameliorated by using well-matched donors. Because preservation times up to 30 hours did not exert an adverse effect on outcome, an argument is made to share pancreata between centers to achieve good matches.