All 74 renal transplant biopsies performed between 11/78 and 8/84 at Duke University and Durham VA Medical Centers having cellular infiltrates that could be phenotyped by immunoperoxidase labeling were examined to identify histologic and immunopathologic features which correlated with rejection. Monoclonal antibodies identifying hematopoietic cells, T cells, T helper cells, T cytotoxic-suppressor cells, macrophages, and B cells were used, and each phenotype population was graded separately based on pattern of infiltration: cortical-diffuse (CD), perivascular (PV), and cortical-aggregate (CA). Histologic and immunofluorescent studies were used to evaluate acute (humoral) and cellular types of vascular and interstitial inflammation. By univariate analysis, patients having irreversible rejection within 10 weeks postbiopsy (n = 23) had significantly higher grades of both acute vascular (humoral) inflammation and TC-S cell infiltrates in a CD pattern as compared with those who had good graft function. Multivariate (Cox regression) analysis was also performed considering biopsy changes, therapy before and after biopsy, and intervals between transplant, rejection onset, and biopsy--as well as other factors potentially affecting the biopsy or graft outcome. Of nine cell phenotype-pattern combinations, only TC-S infiltrates in a CD pattern were associated with a significant (P less than 0.03) relative risk (RR) of subsequent failure from rejection (RR = 8.2). When those cases with significant acute (humoral) inflammation (n = 10) were excluded, the relative risk of TC-S-CD infiltrates increased to 46.4 (P less than 0.04). These findings indicate that the location, as well as the number and type of cell infiltrates are critical in evaluating cellular forms of rejection, and that the extent of diffuse, cortical TC-S infiltration on biopsy provides the greatest predictor of subsequent irreversible graft failure.