Interobserver variation in interpretation of serial liver biopsies from patients with chronic hepatitis C

Abstract

Reliable and accurate assessment of liver histopathology in patients with chronic hepatitis C is important for decision regarding treatment and for evaluation of therapy. However, little data on interobserver variation have been published. In this study, five specialist histopathologists evaluated 46 liver biopsies from 20 patients treated with interferon-alpha. Knodell's and Ishak's scoring systems, De Groote's classification and a four level general necro-inflammatory activity score (GNAS) were applied. Besides kappa statistics, slide by slide analysis was performed. We defined an acceptable slide by slide agreement as eight of ten observer pairs agreed on 80% of the slides. The best agreement was seen for Knodell's and Ishak's fibrosis score, De Groote's classification and GNAS (mean weighted kappa (kappa(w)) = 0.49, 0.51, 0.50 and 0.44, respectively). By condensing data from Knodell's and Ishak's scores to presence or absence of cirrhosis and piecemeal necrosis respectively, concordance was substantial concerning cirrhosis (mean kappa = 0.69 and 0.72, respectively) but only moderate concerning piecemeal necrosis (mean kappa = 0.40 and 0.39, respectively). Slide by slide analysis showed the highest agreement on Knodell's fibrosis score and GNAS; only one point of difference in score was to be accepted to obtain 'eight of ten' agreement. In contrast, five points of difference were necessary to accept in order to reach the same agreement for Knodell's total activity score. Moreover, in serial biopsies the GNAS was sufficient to detect changes in disease activity following treatment. Thus, a simple scoring system with four category scales was reproducible and sufficient for detection of therapy induced changes.

Sources of variability in histological scoring of chronic viral hepatitis

“…This has already been observed for other scores. 9 The decrease in agreement as the number of scoring classes increases can be compared with the poor or fair agreement observed with the Ishak score including 7 classes, 7-9 despite the high correlation between Metavir and Ishak fibrosis scores, 16,17 whereas the necroinflammatory grades correlated moderately well. 17 Interestingly, when two classes were used, increased in relation to the fibrosis cutoff.…”

“…The relationship between agreement and specimen length is debated. 9,19 Bedossa et al 20 recently suggested that sampling variability of fibrosis was a significant limitation in the assessment of fibrosis with virtual liver biopsy using the Metavir scoring system. They suggested that a length of at least 25 mm was necessary to accurately evaluate fibrosis with a semiquantitative score.…”

“…5 Thus, inter-observer agreement has only been studied under selected conditions or in relatively few specimens. [4][5][6][7][8][9][10] The aim of this study was to evaluate inter-observer agreement for the histological interpretation of chronic viral hepatitis under different conditions, especially close to clinical practice, and to identify the sources of observer variability. Variability was evaluated with Metavir score according to elementary lesions, expertise in liver pathology, duration and location of practice, and length of specimens, mainly for inter-observer agreement but also for intra-observer agreement.…”

The Relationship between Treatment Response and Histological Scoring Systems Applied in Chronic Hepatitis B

“…The most important problem in evaluation of different histological activity scoring systems in CVH is intraobserver and interobserver variation [14][15][16][17] . Common feature of the studies in the literature is the low interobserver and intraobserver error rate in staging; but this rate declines when scoring systems become more difficult to apply.…”

E1 therapeutic vaccination for chronic hepatitis C: All that glitters is not gold

“…This problem is compounded by observer variation, which even in the hands of the best observer results in 10% to 30% of second reading values being either higher or lower than the first reading, and by sampling issues that arise from disease quantitation based on a needle biopsy 1/50,000 the size of the organ. [25][26][27] Nevertheless, it is possible to demonstrate changes in histology.…”