RESUMO
Recognition that some lesions typical of T cell-mediated rejection (TCMR) also occur in antibody-mediated rejection requires revision of the histologic TCMR definition. To guide this process, we assessed the relative importance of various lesions and the performance of new histology diagnostic algorithms, using molecular TCMR scores as histology-independent estimates of true TCMR. In 703 indication biopsies, random forest analysis and logistic regression indicated that interstitial infiltrate (i-lesions) and tubulitis (t-lesions) were the key histologic predictors of molecular TCMR, with arteritis (v-lesions) having less importance. Histology predicted molecular TCMR more accurately when diagnoses were assigned by strictly applying the Banff rules to the lesion scores and redefining isolated v-lesion TCMR. This improved prediction from area under the curve (AUC) 0.70 with existing rules to AUC 0.80. Further improvements were achieved by introducing more categories to reflect inflammation (AUC 0.84), by summing the lesion scores (AUC 0.85) and by logistic regression (AUC 0.90). We concluded that histologic assessment of TCMR can be improved by placing more emphasis on i- and t-lesions and incorporating new algorithms for diagnosis. Nevertheless, some discrepancies between histologic and molecular diagnoses persist, partially due to the inherent nonspecificity of i- and t-lesions, and molecular methods will be required to help resolve these cases.