Reliability and predictive value of the National Institute of Diabetes and Digestive and Kidney Diseases Liver Transplantation Database nomenclature and grading system for cellular rejection of liver allografts

Abstract

Pathologists participating in the National Institute of Diabetes and Digestive and Kidney Diseases Liver Transplant Database (LTD) created a histopathological grading system for acute liver allograft rejection, and tested it first for inter- and intra-rater reliability among a group of five pathologists experienced in liver and transplantation pathology. Specimens from post-transplantation biopsies from 48 patients with rejection, hepatitis, or other diagnosis(es) were reviewed. There was moderate to good (kappa = 0.40 to 0.55) inter-rater and good (kappa = 0.55 to 0.58) intrarater agreement for the diagnosis and exact grading of mild, moderate, or severe acute rejection, which improved when a short clinical history was provided. Thus, the scheme was reproducible, and few of the disagreements among the pathologists would have affected treatment decisions. Secondly, the ability of the grading system to predict an unfavorable short- or long-term outcome from the initial histopathological diagnosis of cellular rejection was tested on groups of 168 and 133 patients, respectively, from the three LTD clinical centers, who were followed up for at least 6 months after the first onset of rejection. This analysis showed that a higher histopathological grade of acute rejection on first biopsy diagnosis was significantly associated (P ≤ .006) with both an unfavorable short-term outcome, defined by failure of the episode to resolve within 21 days or the need for aggressive immunosuppressive treatment, and a long-term outcome defned by death or retransplantation from rejection within 6 months of onset. Lastly, an analysis was performed to determine whether subjective rejection grading by the pathologist or certain “objective” histopathological features identified by logistic regression modeling were more accurate in predicting an unfavorable outcome. The sensitivity (.86 vs. .71), specificity (.68 vs. .75), positive predictive value (.13 vs. .14), and negative predictive value (.99 vs. .98) for predicting an unfavorable long-term outcome (allograft loss from rejection within 6 months of onset) were similar for both prediction methods, although the comparison favored the logistic regression model. The low positive predictive value of both methods was attributed to the current immunosuppressive agents, which are highly effective in the prevention of liver allograft failure from acute rejection, and the difficulty in separating rejection grading from staging. To our knowledge, this study represents the first attempt to evaluate both the reproducibility and predictive value of a histopathological grading system for allograft rejection, using multiple pathologists and patients from more than one center. (Hepatology 1995;21:408-416.)