Revisiting the Banff Scores for Renal Allograft Pathology
M Mengel, J Reeve, B Sis, K Famulski, L Hidalgo, P Halloran. University of Alberta, Edmonton, Canada
Background: The Banff classification represents the international diagnostic consensus in renal allograft pathology. In the different morphological compartments of the kidney lesions are quantified and assigned a score of 0-3 according to arbitrary thresholds. Lesions scores are then again arbitrarily assembled into diagnosis with direct clinical implications. Consensus rules and thresholds for scoring lesions and assigning diagnosis were developed empirically in 1991 and later clinically validated by correlating them with response to treatment and outcome. However, significant advances in treatment and clinical management occurred since then. Thus we aimed to revisit the Banff scoring criteria and thresholds in terms of their current diagnostic and clinical utility.
Design: We assessed in 226 renal allograft biopsies for cause the extent and severity of lesions in interstitium, tubuli, vessels, capillaries and glomeruli as continuous variables. Using Principal Component and Regression Tree Analysis we compared Banff scores and continuous scores with each other and in terms of their predictive value for a molecular rejection classifier and allograft survival.
Results: Banff scores and continuous variables grouped closely together in Principal component analysis. The total i-score, i.e. the extent of all cortical inflammation was best predicting a high probability of rejection based on a recently described molecular rejection classifier. Interestingly the regression tree analysis suggested a total i-score threshold as the most predictive (>23%) similar to the current Banff cut-off for rejection (figure1). In terms of allograft survival the extent of tubular atrophy turned out to be the best predictor, at a threshold (>25%) identical to the current Banff ct2 score.
Conclusions: Continuous re-assessment of renal allograft lesions revealed total cortical inflammation as the best predictor of the molecular burden of inflammation and injury. The extent of tubular atrophy was a robust predictor of allograft survival. Surprisingly for both lesions the unsupervised regression analysis suggested thresholds virtually identical to the empirically derived Banff scores and thus independently confirms their clinical value.
Category: Kidney (does not include tumors)
Wednesday, March 24, 2010 1:00 PM
Poster Session VI # 221, Wednesday Afternoon