Variability of grade and stage in simultaneous paired liver biopsies in patients with hepatitis C

Abstract

Background

Grading and staging of liver biopsies in patients with chronic hepatitis remains an inexact “gold standard” that is influenced by variabilities in scoring systems, sampling, observer agreement and expertise. Spatial disease variability relative to markers of the adequacy of biopsy has not been studied previously.

Methods

Paired liver biopsy specimens were obtained from the right and left hepatic lobes of 60 patients with chronic hepatitis C. Histological grade and disease stage were assessed according to the Ludwig scoring system, and scores were evaluated in relation to differences in size and number of portal tracts in all paired samples.

Results

The relative difference (%) in aggregate biopsy size and number of portal tracts was similar between paired samples with and without a difference in grade. Paired samples with a difference in stage showed a larger relative difference in biopsy size (p = 0.09) and in the number of portal tracts (p = 0.016).

Conclusions

Our study shows a difference of one grade or one stage in 30% of paired liver biopsies, due to a combination of sampling variability and observer variability. Acknowledgment of “built‐in” variability in grading and staging chronic hepatitis C by both clinicians and pathologists is essential for managing the individual patient with chronic hepatitis C.

Liver biopsy remains the gold standard for assessing disease severity and progression in chronic hepatitis C.¹,² Although virus‐related factors seem to be better predictors of response to treatment than host‐related factors, necroinflammatory activity (grade) and degree of fibrosis (stage) provide both a measure and a predictor of disease progression.¹ However, liver biopsy has several limitations, including interobserver and intraobserver variability in biopsy interpretation, sampling variation, and the morbidity and mortality associated with the procedure itself.³

Sampling variability is of considerable concern, as a standard liver biopsy specimen represents only about 0.0002% of the entire organ. This problem has been assessed in various ways through investigative studies.³,⁴,⁵,⁶,⁷ It is generally thought that size does matter,⁸ and that bigger is better,⁹ but universally accepted consensus criteria for the adequacy of biopsy regarding the minimum required biopsy size or number of sampled portal tracts do not exist.⁹,¹⁰ Moreover, several different grading/staging systems are used worldwide for the histological assessment of chronic hepatitis C.¹¹,¹²,¹³ The different sources of variability in interpretation of liver biopsy influence each other at least in part, but the degree of observer experience is probably more important than characteristics of the specimen.³

Two studies have recently dealt with the problem of spatial disease variability through sampling more than one area of the liver in patients with chronic hepatitis C.⁵,¹⁴ Both studies found differing histological grades and stages in a large fraction (24–45%) of paired biopsy specimens, but the relationship between observed disease variability and biopsy size has not been studied.

Patients and methods

In total, 60 patients with chronic hepatitis C and no clinical or histological evidence for coexisting liver diseases were enrolled into the study (2001–3), which had been approved by the Institutional Review Board at the University of Vermont (Burlington, Vermont, USA). Paired liver biopsy specimens were obtained from the right and left hepatic lobes under ultrasound guidance using 18‐gauge core biopsy needles. Specimens were fixed in formalin and routinely processed for histology including haematoxylin and eosin (H&E), iron and trichrome stains. Each biopsy sample was scored after examination of at least three histological H&E‐stained sections and corresponding trichrome stains. Biopsy samples from patients with coexisting steatohepatitis were excluded.

Histological grade (degree of necroinflammatory activity) and the disease stage (extent of fibrosis) were determined for each sample according to the four‐tiered Ludwig scoring system,¹⁵ which is similar to the grading tool originally published by Scheuer.¹⁶ Biopsy samples in which the liver capsule was identified were graded and staged using all but a 1‐mm subcapsular rim of tissue. The maximum aggregate length of each needle core biopsy specimen was measured and the number of portal tracts was recorded. All histological slides were coded to ensure blinded evaluation of the paired samples. The entire study set was then randomly evaluated by a hepatopathologist (HB) to obtain scores for both grade and stage, which were used for subsequent statistical analysis. Two weeks later, HB again blindly and randomly evaluated all paired biopsy specimens to assess intraobserver variability. A junior trainee (SS) with interest in liver pathology and experience with the Ludwig scoring system graded and staged the entire study set randomly to assess interobserver variability.

The relative difference (in %) in aggregate biopsy size and number of portal tracts between each pair of biopsies was calculated. For example, if the right lobe biopsy of a paired sample measured 7 mm, compared with a left lobe biopsy size of 10 mm, then the relative size difference was calculated as 30%.

Statistical analysis

Statistical analyses were performed using a two‐tailed t test, assuming unequal variance, and the χ² test, assuming a significance level of p<0.05. Cohen κ statistics assessed intraobserver and interobserver agreement.

Results

Paired bilateral needle biopsy samples from 60 patients (median age 44.5 years; range 20–55 years; 32 men, 28 women) with chronic hepatitis C and no other coexisting liver disease were evaluated. Biopsy characteristics were consistent with chronic hepatitis C in all cases. The percentage of hepatitis C‐associated steatosis varied from 0% to 25%, with a patchy distribution pattern, but features of steatohepatitis were not present in this study set. The degree of steatosis did not vary significantly between paired samples. The median aggregate biopsy size of both right and left‐sided biopsies was 14 mm, with a range of 4–30 mm for the right and 5–31 mm for the left. Similarly, the median number of portal tracts of all right‐sided samples was 13.5 (range 4–28), compared with a median of 12 (range 5–27) in left‐sided biopsies. The relative right versus left difference in number of portal tracts (mean (standard deviation) (SD)) for all paired samples was 27.4 (19.1), and that for aggregate biopsy size 21.4 (17). Table 1 shows the distribution of grade (degree of necroinflammatory activity) and stage (extent of fibrosis) according to the Ludwig scoring system for all samples.¹⁵

Table 1 Distribution of grade and stage in 60 paired right and left lobe liver biopsies.

	Right	Left
Grade
1	3 (5%)	8 (13%)
2	36 (60%)	27 (45%)
3	21 (35%)	25 (42%)
4	0	0
Stage
1	3 (5%)	3 (5%)
2	36 (60%)	40 (67%)
3	15 (25%)	11 (18%)
4	6 (10%)	6 (10%)

A difference of either one grade or one stage was seen in 18 (30%) paired samples. Of those, 10 showed a difference in both one grade and one stage. Paired biopsy samples with a difference in grade or stage were proportionally distributed (data not shown). The relative difference (in %) in aggregate biopsy size and number of portal tracts was similar between paired samples with and without a difference in grade (table 2).

Table 2 Relative differences (in %) in size and number of portal tracts in 60 paired right and left lobe liver biopsies.

	n	Relative difference (%) in size	Relative difference (%) in number of portal tracts
Biopsy pairs with difference in grade	18	20.5 (16.5)	30.5 (21.2)
Biopsy pairs without difference in grade	42	21.8 (17.4)	26.2 (18.3)
Biopsy pairs with difference in stage	18	28.4 (21.9)*	37.7 (21.7)†
Biopsy pairs without difference in stage	42	18.5 (13.7)	22.9 (16.2)

*p = 0.09.

†p = 0.016.

The relative difference in aggregate biopsy size was larger in paired samples that showed a difference in stage, but this difference failed to reach statistical significance (p = 0.09). The relative difference in the number of portal tracts was significantly (p = 0.016) greater in paired samples that showed a difference in stage (table 2). The smaller specimen of those paired samples consistently scored lower.

The absolute size (in mm) and the absolute number of portal tracts did not differ significantly between right‐sided or left‐sided biopsy specimens, whether or not there was a difference in grade or stage (data not shown). The fraction of paired samples showing a difference in grade or stage did not change significantly when restricting analyses to paired samples of a certain minimum size or number of portal tracts—that is, size at least 10, 15 or 20 mm, or number of portal tracts at least 10, 15 or 20 (table 3).

Table 3 Number of paired liver biopsy specimens showing a difference in grade or stage relative to biopsy size and number of portal tracts in 60 biopsies of the right and left liver.

Both specimens of the paired biopsy sample feature a minimum of:	n	Paired samples with a difference in grade	Paired samples with a difference in stage
10 mm in length	50 (83.3%)	16 (32.0%)	15 (30.0%)
15 mm in length	18 (30.0%)	6 (33.3%)	6 (33.3%)
20 mm in length	5 (8.3%)	2 (40.0%)	1 (20.0%)
10 portal tracts	50 (83.3%)	14 (28.0%)	12 (24.0%)
15 portal tracts	23 (38.3%)	7 (30.4%)	8 (34.8%)
20 portal tracts	6 (10.0%)	2 (33.3%)	1 (16.7%)

Interobserver variability analysis (trainee v expert) indicated a moderate degree of agreement and showed differing grading scores in 41 of 120 (34.2%) individual samples (κ coefficient 0.47). Similarly, differing staging scores were seen in 18 of 120 (15%) individual samples (κ coefficient 0.59). Samples with disagreement showed a consistent trend toward higher grading and staging by the trainee. Intraobserver variability analysis (expert v expert on different days) indicated a substantial degree of agreement and showed differing grading scores in 30 of 120 (25%) individual samples (κ coefficient 0.72). Similarly, differing staging scores were seen in 14 of 120 (11.7%) individual samples (κ coefficient 0.78). Individual biopsy samples that showed interobserver or intraobserver variability in scores of grade or stage did not differ in size or number of portal tracts from those samples that scored consistently (data not shown).

Discussion

Liver biopsy continues to have a central role as the “gold standard” in evaluating patients with chronic hepatitis C,¹,⁹,¹¹,¹³ but limitations include cost, patient morbidity and mortality, observer variability, sampling variation and use of different histopathological scoring tools.

The Knodell histology activity index was the first system of its type to assess the severity of chronic hepatitis objectively, semiquantitatively and reproducibly. It was followed by the development of other grading/staging systems of variable complexity, such as the Ishak, Scheuer, Ludwig and METAVIR scoring systems.¹²,¹³ The resulting table with 552 possible numerals underscores the need to choose an appropriate system for the right setting. To be effective in everyday diagnostic practice, scoring systems must be simple to understand and to apply, be effectively communicated to the clinician, and be clinically relevant.¹³ Simpler systems have generally higher degrees of observer concordance,¹³ and agreement increases with the degree of observer experience.³ The Ludwig and Scheuer systems¹⁵,¹⁶ are less complex than the Knodell/Ishak or METAVIR systems, are associated with significant interobserver and intraobserver agreement,³,⁵ and have been adopted by many pathologists and hepatologists for the routine clinical evaluation of patients with chronic hepatitis C.⁵

Studies validating the most widely used systems have shown satisfactory degrees of intraobserver and interobserver variability.¹³ Our study was not designed to investigate observer variability of the Ludwig scoring system, but the data are consistent with previous studies. Agreement was generally greater for the assessment of fibrosis (stage), and the degree of expertise significantly influenced variability, using κ scores for assessment of intraobserver and interobserver agreement. A substantial degree of intraobserver agreement (κ scores of 72 for grade and 78 for stage in the current study) cannot minimise differing grading scores in 25% and differing staging scores in 12% of the 120 individual study samples. The generated numbers in hepatitis grading and staging do not represent measurements of a continuous variable, which presents a fundamental problem with histological scoring.¹¹ However, the relative simplicity of the Ludwig system used in this study would be expected to carry a low risk for substantial observer variation.³,⁵ We performed and compared the statistical analyses using both intraobserver and interobserver datasets without significant change in numbers, proportions and significance levels (data not shown). Thus, the observed intraobserver variation in grading and staging seems to be random, and did not change the interpretation of our data. It also indicates that grading and staging differences are more likely due to spatial disease variation.

Presently, there are no uniform criteria for assessing the adequacy of liver biopsies. On the basis of older reports, the recommended satisfactory length of liver biopsy in hepatitis C ranges from 10 to 40 mm, and a sample 15 mm long or containing 4–6 portal tracts was considered acceptable.⁶,⁹ The variability of the relative amount of fibrosis is negligible beyond a specimen length of 40 mm.⁴ When the criteria for minimum size (at least 10, 15 or 20 mm) or minimum number of portal tracts (at least 10, 15 or 20) were applied to our study set, the fraction of paired samples with a difference in grade or stage did not change. Earlier studies suggested that thin‐needle biopsy may provide information similar to large‐needle biopsy,⁷ but this notion has been generally refuted,⁸ and bigger seems to be better.⁹ Refined biopsy size criteria have been suggested to include a biopsy specimen at least 20 mm long and 1.4 mm wide, or the presence of 11 portal tracts.⁶,⁹ Schiano et al¹⁷ showed reliable grading and staging in specimens measuring at least 10 mm.

This study assessed the relationship between observed sampling variability and biopsy size. We showed that a size difference between paired samples did not influence the assigned inflammatory grade. By contrast, paired samples with a large relative size difference were more likely to show a difference in fibrosis stage. The smaller sample in each pair consistently scored lower, a finding supported by others.⁶,⁸ Two other groups found differing histological stages in a large fraction (24–45%) of paired biopsy specimens with hepatitis C obtained from different areas of the right lobe of the liver,¹⁴ or biopsies obtained from the right and left lobes of the liver chosen under direct visualisation during peritoneoscopy.⁵ This supports older studies showing confirmation of cirrhosis in only 50–80% when one biopsy specimen was examined, compared with the assessment of three spatially distinct samples.¹⁸,¹⁹ Interestingly, hepatic viral load does not seem to have the same degree of heterogeneity of sampling variability as does histology.²⁰ However, resolving the problem of disease heterogeneity by taking several biopsy specimens from the same patient must be balanced against the risk of increased morbidity and mortality.⁴,²⁰ Moreover, the relationship between observer agreement and specimen size is debated,³ suggesting that as the specimen increases in size the probability of various types of lesions increases, creating a paradoxical increase in observer variability. Sampling variability can never be completely eliminated, but may be overcome in large cohorts of patients and biopsies, as variation is likely to be random and multidirectional.⁹ This should minimise the role of sampling variability in the interpretation of clinical trials. However, sampling variability may become an issue in the management of the individual patient, particularly in the assessment of disease activity or progression, which may then influence treatment decisions. The relatively small number of biopsy specimens with minimum and maximum grade and stage in this study set is consistent with a random sample of patients with chronic hepatitis C seen at our institution. Patients at the extremes of inflammatory grade or histological stage may be particularly important in clinical decision making. However, paired biopsy samples with a difference in grade or stage (n = 18) were proportionally distributed along the different grades and stages, and did not cluster at potentially important clinical decision points. All paired biopsy samples with a difference in grade or stage (n = 18) showed only a one‐grade or one‐stage difference. Understaging, and to a lesser degree undergrading, would have occurred if only the smaller of the paired sample were to be examined. This could have influenced clinical decision making in some patients. Other clinical data such as viral load may help to balance the “built‐in” variability in grading and staging chronic hepatitis C, but the assessment of the adequacy of biopsy should accompany each pathology report.

Take‐home messages

• Histological grading and staging in chronic hepatitis is an inexact gold standard.

• A “built‐in” variability in histological grading/staging of chronic hepatitis has to be acknowledged by both the clinician and the pathologist.

• Variability in scoring systems, sampling, observer agreement and expertise necessitate a common ground approach in scoring biopsies.

• This common‐ground approach is best facilitated by (a) choosing the simplest scoring system to serve the intended purpose, (b) pursuing minimum sample criteria, including biopsy size of >10 mm or presence of >10 portal tracts, and (c) histological biopsy evaluation by an experienced pathologist.

In summary, grading and staging of liver biopsies in chronic hepatitis remains an inexact gold standard that is influenced by variabilities in scoring systems, sampling, observer agreement and expertise. Optimising one factor may confound the performance of other variables, necessitating a “common ground” approach. This would include (1) choice of the simplest scoring system that will serve the intended purpose; (2) pursuit of minimum sample criteria, including a size of at least 10 mm or presence of >10 portal tracts; and (3) assessment of biopsies by an experienced pathologist. Acknowledgment of a “built‐in” variability in grading and staging chronic hepatitis C by both the clinician and the pathologist is essential for managing the individual patient with chronic hepatitis C.