We are studying two cohorts of kidney transplant recipients, with the goal of defining specific clinicopathologic entities that cause late graft dysfunction: (1) prevalent patients with new onset late graft dysfunction (cross-sectional cohort); and (2) newly transplanted patients (prospective cohort). For the cross-sectional cohort (n = 440), mean time from transplant to biopsy was 7.5 ± 6.1 years. Local pathology diagnoses included CAN (48%), CNI toxicity (30%), and perhaps surprisingly, acute rejection (cellular- or Ab-mediated) (23%). Actuarial rate of death-censored graft loss at 1 year postbiopsy was 17.7%; at 2 years, 29.8%. There was no difference in postbiopsy graft survival for recipients with versus without CAN (p = 0.9). Prospective cohort patients (n = 2427) developing graft dysfunction >3 months posttransplant undergo 'index' biopsy. The rate of index biopsy was 8.8% between 3 and 12 months, and 18.2% by 2 years. Mean time from transplant to index biopsy was 1.0 ± 0.6 years. Local pathology diagnoses included CAN (27%), and acute rejection (39%). Intervention to halt late graft deterioration cannot be developed in the absence of meaningful diagnostic entities. We found CAN in late posttransplant biopsies to be of no prognostic value. The DeKAF study will provide broadly applicable diagnostic information to serve as the basis for future trials. © 2009 The American Society of Transplantation and the American Society of Transplant Surgeons.