OBJECTIVES: Enhancements in renal transplant have led to reduced rates of acute rejection leading to a shift to kidney function as an accepted endpoint in efficacy trials.  Characterizing long-term benefits requires modeling to project long-term outcomes of treatment based on kidney function.  The goal of this study was to project rates of graft failure and quality adjusted life years over the lifetimes of hypothetical patients receiving belatacept (a recently-introduced selective costimulation blocker), cyclosporine or tacrolimus. METHODS: We developed a simulation with two phases that integrated trial-based information with a long-term four-state Markov model (functioning graft, graft failure on dialysis, functioning re-graft, and death).  In the first phase, three-year distributions of patients in four estimated glomerular filtration rate (eGFR) categories (>60, 45-59, 30-44 and 15-29, with graft failure assumed at eGFR <15 mLs/min/1.73m2) were estimated using a mixed treatment comparison of cyclosporine, belatacept, and tacrolimus.  The Markov phase was populated using transplant recipient data from the United States Renal Data System (n=34,130). Utilities for adjusting life-years were obtained from a study of US renal transplant patients. RESULTS: Over a 20-year modeled time-horizon for 1,000 hypothetical patients, belatacept was associated with 551 graft failures,and 9.6 quality-adjusted life years. Relative to cyclosporine, belatacept was associated with 0.9 additional quality-adjusted life years, and 56 fewer graft failures. Relative to tacrolimus, belatacept was associated with 0.9 additional quality-adjusted life years, and 18 fewer graft failures.   CONCLUSIONS:  This is the first long-term follow-up model of renal transplant patients to be based on trial-based graft function endpoints and include all relevant comparators.  This long-term extrapolation of differences in kidney function observed at three years shows clinically important differences between treatments.  While validation of these models will require long-term follow-up of patients, modeling based on renal function may prove to be a useful method of projecting long-term outcomes.