Compared with transplantation after chronic dialysis, the effect of PreKT on graft failure was modeled as the time from transplantation to kidney graft failure (including indication of return to dialysis), death, or retransplantation, using Cox proportional hazards regression. Patients were censored at the earliest of the following: last known follow-up date, the maximum date for which follow-up information was expected, or the end of study (June 30, 2003). The relative risk of death was modeled as the time from transplantation to death, using Cox regression, censored at the earliest of either the maximum date for which follow-up information was expected or June 30, 2003. These Cox models were developed to determine the adjusted and unadjusted relative risk of graft failure and death by type 1 or type 2 DM and transplant option (living donor, deceased donor, or SPK). Kidney graft failure and patient mortality models were adjusted for the following recipient variables: age, sex, race, ethnicity, panel reactive antibody (PRA) at time of transplantation, blood type, educational level, employment status at time of transplantation, and source of payment. The models also were adjusted for donor age, sex, race, and ethnicity. Transplantation-specific variables in the model included degree of HLA mismatch, year of transplantation, and whether the transplantation was preemptive. Models also included interactions by donor source and DM type to estimate subgroup differences in survival.