The Banff scoring schema provides a common ground to analyze kidney transplant biopsies. Interstitial inflammation (i) and tubulitis (t) in areas of viable tissue are features in scoring acute rejection, but are excluded in areas of tubular atrophy (TA). We studied inflammation and tubulitis in a cohort of kidney transplant recipients undergoing allograft biopsy for new-onset late graft dysfunction (N = 337). We found inflammation ('iatr') and tubulitis ('tatr') in regions of fibrosis and atrophy to be strongly correlated with each other (p < 0.0001). Moreover, iatr was strongly associated with death-censored graft failure when compared to recipients whose biopsies had no inflammation, even after adjusting for the presence of interstitial fibrosis (Hazard Ratio = 2.31, [1.10-4.83]; p = 0.0262) or TA (hazard ratio = 2.42, [1.16-5.08]; p = 0.191), serum creatinine at the time of biopsy, time to biopsy and i score. Further, these results did not qualitatively change after additional adjustments for C4d staining or donor specific antibody. Stepwise regression identified the most significant markers of graft failure which include iatr score. We propose that a more global assessment of inflammation in kidney allograft biopsies to include inflammation in atrophic areas may provide better prognostic information. Phenotypic characterization of these inflammatory cells and appropriate treatment may ameliorate late allograft failure. © 2010 The American Society of Transplantation and the American Society of Transplant Surgeons.