C ontrolled clinical trials provide essential comparative data on safety and short-term surrogate end points such as rejection rates, but they are rarely powered to detect significant differences in graft survival or other “hard” end points. Registry analyses provide sufficient patient numbers to identify differences in graft outcomes between interventions, and they are conducted over the long term in a standard-risk patient population managed in routine clinical practice. We undertook a comparative intent-to-treat analysis of 3-year graft survival with use of either cyclosporine (CsA) or tacrolimus (TAC) in combination with mycophenolate mofetil (MMF) and steroids among renal transplant patients registered with the United Network for Organ Sharing Renal Transplant Registry. We selected a 2-year time frame to minimize the impact of concurrent improvements in patient management that were unrelated to immunosuppression. The most recent year for which full 3-year follow-up data are available is 1999, and, therefore, patients who underwent kidney transplants during 1998-1999 were selected for analysis. Univariate analysis was used to identify potential confounding factors including original disease, recipient age, body mass index, race, sex and panel reactive antibody status, previous transplant, previous pregnancy, waiting time on dialysis, and pretransplant blood transfusion; donor type, age, body mass index, race, sex, serum creatinine level, and mode of death; cold ischemia time, year of transplant, human leukocyte antigen compatibility, calcineurin inhibitor switch, cytomegalovirus status, and hepatitis B and C status. Cox regression analysis was utilized to adjust graft survival for known confounding factors that were found to differ significantly between treatment groups. A total of 21,267 patients were included in the analysis, 8038 of whom were treated with TAC-MMF and 13,229 of whom were given CsA-MMF. Approximately one third (7079) received a graft from a living donor. Three-year graft survival (KaplanMeier estimate) was significantly higher with CsAMMF than with TAC-MMF (86.6% vs 84.5%, respectively; P .0002) (Table 1). Twenty-one confounding factors varied significantly between treatment groups among the entire 1998-1999 cohort. After adjustment for confounding variables, the hazards ratio (HR) for graft failure with TAC-MMF versus CsA-MMF was 1.10, a difference that was statistically significant (95% confidence interval [CI], 1.03-1.20; P .007). When we repeated the analysis and excluded patients who died with a functioning graft, there was again a significantly higher risk of graft loss in the TACMMF treatment group when compared with the CsA-MMF treatment group (HR, 1.20, 95% CI 1.091.32, P .001). Fewer confounding factors (n 11) varied significantly between the CsA-MMF and TAC-MMF groups among living-donor patients, so we repeated the analysis using only living-donor recipients to further reduce risk of selection bias. Among living-donor recipients, 3-year graft survival was significantly higher with CsA-MMF than TACMMF (91.1% vs 88.1%, respectively; P .001), as was death-censored graft survival (94.5% vs 91.5%, respectively; P .0001). Estimated graft half-life was 20.8 1.2 years for the CsA-MMF group and 16.1 1.2 years for the TAC-MMF group (P .001). When graft survival was adjusted for confounding variables, the HR for all-cause graft failure with TAC-MMF compared with CsA-MMF was 1.27 (95% CI, 1.08-1.48; P .003); for death-censored graft failure, it was 1.48 (95% CI, 1.22-1.80; P .001). From the Multiorgan Transplant Program, Cedars-Sinai Medical Center, Los Angeles, CA; and Dumont Transplant Program & Immunogenetics Center, David Geffen School of Medicine, University of California Los Angeles, Los Angeles, CA. © 2003 Elsevier Inc. All rights reserved. 0955-470X/03/1704-0000$30.00/0 doi:10.1016/j.trre.2003.10.022