A multivariate analysis of prognostic factors for graft failure was performed on patients in the International Pancreas Transplant Registry. The analysis was restricted to the period January 1978 to June 1987 and included 764 patients. All patients had at least 1 year of follow-up. The following variables were studied: transplant year, continent (N. America, Europe, others), type of donor (cadaver, living related mismatched, living related HLA-identical), donor mismatch at the HLA A, B loci, donor mismatch at the DR loci, preservation time, kidney association (pancreas transplant alone, simultaneous pancreas and kidney transplant, pancreas after kidney transplant), whole versus segmental pancreatic transplant, graft duct management technique (polymer injection, enteric drainage, stomach drainage, bladder drainage), and immunosuppression. By stepwise, logistic regression analysis, we found that the following factors were predictive for 1-year graft function: donor mismatch at the DR loci (P = 0.0003), kidney association (P less than 0.0001), type of donor (P = 0.04), and immunosuppression (P = 0.0002). For donor mismatch at the DR loci, we found an odds ratio for success of 2.2 for 0 versus 2 mismatches. The odds for success were 2.9 for simultaneous pancreas and kidney transplant versus pancreas transplant alone. The best results--79% 1-year graft survival--were obtained for the combination of 0 mismatches at the DR loci, pancreas after kidney transplant, living related HLA-identical donor, and the immunosuppressive regimen consisting of cyclosporin, azathioprine, and prednisone. Patients receiving a pancreas transplant alone with 0 mismatches at the DR loci, living related HLA-identical donor, and triple immunosuppressive regimen had a predicted 1-year graft survival of 71%.