Abstract: SA-PO464

Mapping Potential Errors in Histology by Comparison with Molecular Phenotyping in 1208 Kidney Transplant Biopsies

Session Information

Category: Transplantation

  • 1702 Transplantation: Clinical and Translational


  • Halloran, Philip F., University of Alberta, Edmonton, Alberta, Canada
  • Reeve, Jeff, University of Alberta, Edmonton, Alberta, Canada

Group or Team Name

  • MMDx-Kidney study group

Histologic interpretation of biopsies is empirical because there has been no external standard for comparison. For kidney transplants, the emergence of a centralized molecular diagnostic system for kidney transplant biopsies, the Molecular Microscope™ system (MMDx) (Nature Reviews Nephrology 12:534, 2016) provides an external standard for assessing the accuracy of histologic diagnoses.


In 1208 prospective indication biopsies, we compared diagnoses of rejection (Banff criteria) assigned by local standard-of-care histology in 13 experienced international centers to the MMDx assessments by microarrays (Clinicaltrials.gov NCT01299168).


MMDx used archetype scores derived from 84 classifier equations to assign diagnoses of antibody-mediated rejection (ABMR) (subclassified early-stage, fully-developed and late-stage) and T cell-mediated rejection (TCMR) (J. Reeve et al. JCI Insight 2017 (In Press) . Disagreement with histology was 39%: 16% in biopsies with no histologic rejection, but 72% in biopsies with histologic rejection. Discrepancies were particularly common in TCMR, early-stage ABMR, and mixed rejection. Analyzing 161 selected discrepancies in detail (figure 1) revealed 3 classes of discrepancies: errors by pathologists applying complex Banff rules (N=59/161); previously flagged errors in Banff rules e.g. isolated v lesions (N=42/161); or inherent limitations (non-specificity) of histology lesions (N=60/161). For example, tubulitis occurs in many renal inflammatory diseases, including ABMR, and is not exclusive to TCMR.


There is a high rate of errors in standard–of-care histology assessments in experienced centers, particularly in rejection-related conditions. Potential solutions include simplifying and correcting rules; using probabilistic regression equations (AJT 16: 1183, 2016); and incorporating molecular assessments. Supported by Canada Foundation for Innovation and Transcriptome Sciences Inc.


  • Commercial Support