Reproducibility of the Banff classification of renal allograft pathology. Inter- and intraobserver variation.

The present study was undertaken to investigate the inter- and intraobserver variation in use of the scoring system for glomerulitis, vasculitis, interstitial inflammation, tubulitis and arteriolar hyalinosis that is an essential part of the recently proposed Banff classification of renal allograft biopsies. Seventy-seven biopsies done less than 90 days after transplantation were included. The scoring was done blindly by five pathologists on biopsies stained with H&E and PAS. The volume fraction of interstitial inflammation was estimated. Spearman rank correlation coefficient and kappa values were used for the evaluation of reproducibility. The results of both inter- and intraobserver variability showed a good correlation and reasonable kappa values for vasculitis, interstitial inflammatory infiltration, and tubulitis. Less-good correlation was found for glomerulitis and arteriolar hyalinosis. The interobserver kappa score for grading of the rejection severity was 0.40 overall but 0.56 when only presence or absence of acute rejection was considered and 0.66 for presence or absence of vasculitis. Weighted kappa values for interobserver vasculitis score and rejection grading were 0.58 and 0.55, respectively. A strong association existed between the volume fraction of interstitial inflammation and the semiquantitative scoring for interstitial inflammation. In conclusion, the good correlations for the key elements in the grading of the allograft biopsies in the present classification system, confirmed the utility of the defined criteria for grading rejection. More precisely defined criteria or simplification of the scoring system are needed for glomerulitis and arteriolar hyalinosis--parameters not used in the diagnosis of rejection.