Predicting the liver histology in chronic hepatitis C: how good is the clinician?

THE AMERICAN JOURNAL OF GASTROENTEROLOGY © 2001 by Am. Coll. of Gastroenterology Published by Elsevier Science Inc.

Vol. 96, No. 11, 2001 ISSN 0002-9270/01/$20.00 PII S0002-9270(01)03837-0

Predicting the Liver Histology in Chronic Hepatitis C: How Good Is the Clinician? Joseph Romagnuolo, M.D., F.R.C.P.(C.), Gian S. Jhangri, M.Sc., Laurence D. Jewell, M.D., F.R.C.P.(C.), and Vincent G. Bain, M.D., F.R.C.P.(C.) Division of Gastroenterology, McGill University Health Centre, Montreal, Quebec; and Departments of Medicine, Public Health Sciences, and Pathology, University of Alberta, Edmonton, Alberta, Canada

OBJECTIVE: Liver biopsy is believed to be necessary before antiviral treatment in hepatitis C. Studies have found symptoms and biochemistry poorly predictive of grade and stage. In practice, a combination of factors is used to anticipate histology. The aim of this study is to evaluate the ability of global clinical assessment to predict histology in hepatitis C. METHODS: Fifty-four consecutive patients referred to a university center for consideration of antiviral therapy were enrolled. Clinical and laboratory data were recorded as was a prediction of the inflammatory grade (0 –3) and fibrotic stage (0 –3), with fibrotic stage 3 referring to cirrhosis. Liver biopsies were read by a blinded pathologist. The predictive value of the clinical assessment and individual parameters was assessed. RESULTS: All predictions were ⱕ1 point off the actual grade and stage. Thirty-six (66.7%) patients’ grades and 41 (75.9%) patients’ stages were exactly predicted. All four cirrhotic patients (sensitivity 100%, specificity 94%) and one case of hemochromatosis were correctly predicted. Spider nevi, organomegaly, white blood cell count ⱕ 4 ⫻ 109/L, ALT ⬎ 120 U/L, bilirubin ⬎ 20 ␮mol/L, albumin ⱕ 35 g/L, and ferritin ⬎ 200 ␮g/L predicted grade ⱖ2. Stage ⱖ2 was associated with age ⬎ 40 yr, previous decompensation, spider nevi, organomegaly, white blood cell count ⱕ 4 ⫻ 109/L, albumin ⱕ 35 g/L, platelets ⱕ 150 ⫻ 109/L, and international normalized ratio ⬎ 1.2. Grade correlated with stage (Spearman coefficient ⫽ 0.54, p ⬍ 0.001). By multivariate analysis, ferritin plus spider nevi or hypoalbuminemia was independently predictive of inflammation. Spider nevi and thrombocytopenia, with either splenomegaly or hypoalbuminemia, were useful three-variable models for predicting fibrosis. The corresponding scoring systems produced useful likelihood ratios. CONCLUSIONS: Global clinical assessment mirroring clinical practice in a tertiary liver transplant center is moderately accurate in predicting grade and stage in hepatitis C. Liver biopsy is the current gold standard; however, the amount of new information gleaned is less than was perceived. The need for routine biopsy before antiviral treatment in hepatitis C should be reevaluated in a multicenter study. (Am J

Gastroenterol 2001;96:3165–3174. © 2001 by Am. Coll. of Gastroenterology)

INTRODUCTION The treatment of chronic hepatitis C is rapidly evolving. Monotherapy with interferon achieves a sustained response in a minority of patients (1). Studies show that combination therapy with ribavirin is considerably more effective with sustained responses achieved in 31– 43% of previously treatment-naı¨ve patients (2, 3). Factors predictive of a good response include genotypes 2–3, female gender, age ⬍ 40 yr, a hepatitis C virus (HCV) RNA titer ⬍ 2 ⫻ 106 copies/ ml, and minimal fibrotic stage (2). However, antiviral therapy is expensive and can result in serious toxicity in a minority of patients. Unfortunately, the presentation in terms of symptoms, physical findings, and laboratory results is often misleading, such that patients with normal aminotransferases and no symptoms may have significant inflammation, and patients with normal physical findings and hepatic synthetic function may have considerable fibrosis. Therefore, a liver biopsy is usually recommended to better define a given individual’s risk-benefit ratio for antiviral treatment. The liver biopsy can give information that includes the grading of inflammation, the staging of fibrosis/cirrhosis, and the exclusion of other coexisting liver diseases. The stage of fibrosis, the hepatic iron grade, and the presence or absence of fat were found to be independent predictors of response by one Austrian group; in their study, not a single patient with a fibrosis stage ⱖ3 and an iron grade ⱖ3 responded to therapy with interferon monotherapy (4). Significant fibrosis was also a predictor of a poor response in the multicenter European-Canadian study (2), although certainly the sustained response in cirrhotics found in this study and others was more favorable than that seen with interferon monotherapy (2, 3). Not all physicians, however, agree that a liver biopsy is an essential step in the consideration of antiviral therapy for a patient with chronic hepatitis C (5). A survey among American gastroenterologists and hepatologists suggested that 91% would biopsy a patient with chronic hepatitis C with fatigue and raised aminotrans-

3166

Romagnuolo et al.

ferases (6). This proportion was considerably lower in nongastroenterologists. Although relatively safe, a liver biopsy is not without expense, discomfort, and risk (fatal and nonfatal hemorrhage rates of 0.04% and 0.16%, respectively, for nonmalignant disease) (7). The literature regarding the predictive value of liver biochemistry, serum HCV-RNA, and clinical parameters in the prediction of biopsy results is discouraging and at times conflicting. Based on the data available to them, the National Institutes of Health Hepatitis C Consensus Conference recently recommended that patients be biopsied before consideration of treatment (8). However, there are no prospective trials that mirror clinical practice, allowing the use of all the information normally available to the clinician to predict liver histology. The aim of this pilot study is to compare the accuracy of clinical prediction of the grade and stage of liver histology in patients with chronic hepatitis C with actual liver biopsy and microscopy. We also analyzed several clinical and laboratory parameters for correlation with the liver biopsy results to see how the clinical prediction could have been improved.

MATERIALS AND METHODS Consecutive patients were recruited with chronic hepatitis C and raised liver enzymes, who were referred for consideration of antiviral therapy, and who consented to undergo a liver biopsy. The center to which they were referred was a tertiary care academic center with a liver transplant program. Outcome Measures The main outcome measure was accuracy of the clinician’s prediction for inflammatory grade and histological stage of fibrosis. The results of univariate and multivariate analyses of individual clinical and laboratory variables and their relative strengths in the prediction of the biopsy results were also studied. Patients The inclusion criteria were the following: 1) HCV antibody positive including confirmatory testing with recombinant immunoblot assay II and/or HCV-RNA by polymerase chain reaction, 2) age ⱖ 18 yr; and 3) elevated ALT in the past 3 months. Patients were excluded if they had any of the following: 1) prior liver biopsy; 2) coagulopathy or thrombocytopenia to a degree that would preclude percutaneous biopsy (fresh frozen plasma and platelet transfusion preprocedure were allowed); 3) ascites; 4) HIV positivity; 5) HBsAg positivity; 6) hemophilia; 7) chronic renal failure; or 8) previous liver transplant. Data Collection A thorough history and physical exam were performed on every patient by a single hepatologist (V.G.B.). A number of clinical features were specifically recorded including age,

AJG – Vol. 96, No. 11, 2001

gender, symptoms, risk factors for viral hepatitis, and the duration and amount (less than five drinks/wk, five to 10 drinks/wk, or more than 10 drinks/wk) of past and present alcohol use (an alcoholic “drink” was defined as 8 oz of beer, 4 oz of wine, or 1 oz of spirits). The following physical signs were specifically evaluated: presence or absence of spider nevi, palmar erythema, splenomegaly, and liver size and consistency. Decompensation was defined as ascites, jaundice, hepatic encephalopathy, or variceal bleeding. The following laboratory tests were performed in all cases: complete blood count, prothrombin time international normalized ratio, ALT, ALP, bilirubin, HBsAg, ferritin, fluorescent antinuclear antibody, and smooth muscle antibody. In selected cases, one or more of the following were also ordered: ceruloplasmin, ␣-1-antitrypsin, porphyrin screen, hemoglobin A1C, HIV, ␣-fetoprotein, and abdominal ultrasound. The clinician (V.G.B.) then made a prediction of the liver histology, employing the grading system of Hytiroglou et al. (9). Necroinflammatory activity was graded as none (grade 0), mild (lymphoid hyperplasia and lobular activity without significant interface necrosis, grade 1), moderate (lymphoid hyperplasia and lobular activity with marked interface necrosis, grade 2), or severe (severe interface and bridging necrosis, grade 3) (Fig. 1). The prediction of fibrosis was staged as none (stage 0), minimal fibrosis without architectural distortion (stage 1), significant fibrosis with septum formation and/or architectural distortion (stage 2), or established cirrhosis (stage 3) (Fig. 2). Fibrosis stage 2 of this classification system represents the collapsing of Hytiroglou et al.’s stages 2 and 3 into a single category (9). The term “significant fibrosis” is hereafter used to denote fibrotic stage ⱖ2, and again, stage 3 refers to cirrhosis. Finally, if coexisting pathology was suspected, this was indicated. A percutaneous liver biopsy was then performed without ultrasound guidance using a Menghini biopsy needle under sterile conditions. All biopsies were stained with hematoxylin and eosin, Massons trichrome, periodic-acid-Schiff (with and without diastase), rhodamine for copper, Perls’ Prussian blue for iron, modified aldehyde fuchsin (for HBsAg and metallothionine), and reticulin stains. The biopsies were read by a single gastrointestinal pathologist (L.D.J.) who was blinded to the above clinical details (except for the age, gender, and diagnosis of chronic hepatitis C infection). The pathologist was also blinded to the clinician’s prediction. If the biopsy was believed to be inadequate by the pathologist (less than three portal triads), it was rejected and that patient removed from the analysis. The inflammatory grade and stage were scored according to the same scale as was used for the prediction described above. Coexisting liver pathology was documented, and special stains were used at the pathologist’s discretion. Statistical Analysis The performance characteristics (accuracy, sensitivity, specificity, predictive values) of the clinician’s scores, using a

AJG – November, 2001

Global Assessment Predicts Histology in HCV

3167

Figure 1. This composite of hematoxylin and eosin high-power fields illustrates the inflammatory grading system used. Grades 0 to 3 are labeled G0 to G3, respectively.

global clinical assessment in predicting the degree of inflammation and fibrosis found on liver biopsy were calculated. These were reported in terms of the proportion of patients correctly scored in each category (i.e., inflammation or fibrosis) alone as well as in both categories together. Also, 95% CIs for these proportions were calculated. To determine the level of agreement of the clinical predictions with the histology, beyond chance, the ␬ statistic was calculated. A priori, the predictor variables were dichotomized according to clinically significant thresholds, decided upon by consensus of the study panel (Tables 1 and 2). Univariate analyses were then performed on the clinical and biochemical variables to see which factor(s) were most useful in the prediction of the abnormalities found on biopsy. Using the Fisher exact test, each predictor was assessed in terms of the strength of its association with each of the two clinically important outcomes. The point of dichotomization of the outcomes was determined a priori: 1) mild versus moderate/ severe inflammation; and 2) none/minimal versus significant fibrosis or cirrhosis. Predictors for the subset of patients with cirrhosis were also determined. The Spearman coeffi-

cient for the correlation between significant inflammation and advanced fibrosis was calculated. A significance level of 0.05 was used. A multivariate analysis was then performed by using LogXact for Windows (Cytel Software, Cambridge, MA), designed to handle smaller sample sizes, to determine which variables independently predicted inflammation and fibrosis.

RESULTS A total of 54 patients met the inclusion criteria and underwent a clinical prediction and subsequent liver biopsy. The patients’ historical and physical exam characteristics are summarized in the left-hand column of Table 1. Most patients were men (68.5%), symptomatic (57.4%), and without history of significant decompensation (94.4%). The most common route of transmission was injection drug use (74.1%). Thirty-five percent consumed less than five alcoholic drinks/wk in their past. Twenty-six percent of patients had spider nevi, 13.0% had hepatomegaly, and 9.3% had splenomegaly. Only one patient had had an abdominal ul-

3168

Romagnuolo et al.

AJG – Vol. 96, No. 11, 2001

Figure 2. This composite of Massons trichrome high-power fields illustrates the fibrotic staging system used. Stages 0 to 3 are labeled S0 to S3, respectively.

trasound, and it did not show any significant abnormalities. This ultrasound was not used to guide liver biopsy. The biochemical parameters are summarized in the lefthand column of Table 2. Of note, 20.4% of patients were hyperbilirubinemic, and 9.3% had a coagulopathy. In six patients, the ALT was elevated at the time of referral but had become normal before biopsy. None of the biopsy specimens were rejected because of inadequacy. In this series, the minimum number of portal tracts observed per biopsy was six. Thirty-eight (70.4%) patients were found to have mild inflammation, 15 moderate (27.8%), and one severe (1.9%). Thirty-seven (69.8%) patients had no fibrosis or minimal fibrosis, 13 (24.1%) patients had significant fibrosis, and four (7.4%) patients were cirrhotic. Performance Characteristics of Global Clinical Assessment Thirty-seven (68.5%; 95% CI ⫽ 56.1– 80.9%) patients were correctly identified in terms of inflammatory score (Table 3). In 11 (61.1%; 38.6 – 83.6%) of the 18 patients who had incorrect inflammation scores, the predicted grade was more

severe than the actual grade. The sensitivity for predicting moderate/severe inflammation (six of 16) was 37.5% (13.8 – 61.2%), the specificity (32 of 38) was 84.2% (72.6 –95.8%), the positive predictive value (PPV) was 50.0% (21.7–78.3%), and the negative predictive value (NPV) was 76.2% (63.3– 89.1%) (Table 4). In 41 patients (75.9%; 64.5– 87.3%), the degree of fibrosis was predicted exactly. Of the incorrectly staged biopsies, six of 13 (46.2%; 19.1–73.3%) were overstaged. Twenty-six patients (49.1%; 35.8 – 62.4%) had exactly matching predicted and actual scores for both inflammation and for fibrosis. For the detection of significant fibrosis or cirrhosis, the sensitivity (10 of 17) was 58.8% (35.4 – 82.2%), the specificity (34 of 37) was 91.9% (83.1–100%), the PPV was 76.9% (54.0 –99.8%), and the NPV was 82.9% (79.4 –98.6%) (Table 5). There were four patients who had cirrhosis on biopsy of which all four were correctly identified before biopsy (sensitivity 100%, NPV 100%). Only three of the 50 noncirrhotic patients, all of whom had extensive fibrosis, were predicted to be cirrhotic (specificity 94% [87.4 –100%],

AJG – November, 2001

Global Assessment Predicts Histology in HCV

3169

Table 1. Clinical Predictor Variables Percent of Patients Variable (Mean ⫾ SD) Age in yr (40.1 ⫾ 8.1) ⱖ40 (n ⫽ 35, 64.8%) ⬍40 (n ⫽ 19, 35.2%) Gender Male (n ⫽ 37, 68.5%) Female (n ⫽ 17, 31.5%) Symptoms Present (n ⫽ 31, 57.4%) Absent (n ⫽ 23, 42.6%) Fatigue Present (n ⫽ 21, 38.9%) Absent (n ⫽ 33, 61.1%) History of decompensation Present (n ⫽ 3, 5.6%) Absent (n ⫽ 51, 94.4%) Alcohol use (five or more drinks/wk) Past Yes (n ⫽ 35, 64.8%) No (n ⫽ 19, 35.2%) Present Yes (n ⫽ 10, 19.6%) No (n ⫽ 41, 80.4%) Injection drug use Yes (n ⫽ 40, 74.1%) No (n ⫽ 14, 25.9%) Spider nevi Present (n ⫽ 14, 25.9%) Absent (n ⫽ 40, 74.1%) Hepatomegaly Present (n ⫽ 7, 13.0%) Absent (n ⫽ 47, 87.0%) Firm or left-lobe enlargement Present (n ⫽ 8, 14.8%) Absent (n ⫽ 46, 85.2%) Splenomegaly Present (n ⫽ 5, 9.3%) Absent (n ⫽ 49, 90.7%)

Moderate/Severe Inflammation

Significant Fibrosis/Cirrhosis

Cirrhosis

37.1 15.8

45.7† 5.3

11.4 0

32.4 23.5

32.4 29.4

8.1 5.9

35.5 21.7

35.5 26.1

9.7 4.3

38.1 24.2

38.1 27.3

4.8 9.1

66.7 27.5

100* 27.5

66.7* 3.9

37.1 15.8

37.1 21.1

8.6 5.3

20.0 29.3

20.0 31.7

0 7.3

27.5 35.7

30.0 35.7

7.5 7.1

64.3† 17.5

71.4‡ 17.5

21.4* 2.5

71.4* 23.4

71.4* 25.5

42.9† 2.1

50.0 26.1

62.5§ 26.1

25.0 4.3

80.0* 24.5

100† 24.5

60.0† 2.0

All p values by Fisher exact test. * p ⬍ 0.05. † p ⬍ 0.01. ‡ p ⬍ 0.001. § p ⬍ 0.1.

PPV 57.1% [20.4 –93.8%]). The predictive values are affected by the relatively low prevalence of cirrhosis in our cohort: four of 54 (7.4%). Fifteen patients had significant steatosis that was not predicted clinically; four patients were correctly identified as having fatty liver; and three patients were predicted to have a fatty liver but did not on biopsy (sensitivity 21.1% [2.8 –39.4%], specificity 91.4% [82.1–100%]). None of the patients with a fatty liver had other features of nonalcoholic steatohepatitis. One patient had iron overload, which was successfully predicted before biopsy. Kappa (␬) statistics were calculated to quantify the agreement of clinical impression and histology beyond chance. In general, a ␬ ⬍ 0.4 reflects poor agreement, 0.4 – 0.6 is fair, 0.6 – 0.8 is good, and ⬎0.8 is excellent (10). The agreement between global

assessment and histology for the classification of patients as moderate or severe inflammation versus minimal inflammation was poor and not statistically significant (␬ ⫽ 0.23; p ⫽ 0.08). Because one of the marginal totals in Table 4 is zero, a ␬ statistic considering all three grades was not able to be calculated for inflammation. For fibrosis, in contrast, the agreement was statistically significant and clinically fair (␬ ⫽ 0.46, p ⬍ 0.001). Dichotomizing fibrosis into the clinically relevant categories of minimal versus moderate to severe fibrosis, the agreement was even greater (␬ ⫽ 0.54, p ⬍ 0.001). Predictors of Inflammation Inflammation (moderate or severe) was associated with statistical significance with the following predictor variables:

3170

Romagnuolo et al.

AJG – Vol. 96, No. 11, 2001

Table 2. Laboratory Predictor Variables Percent of Patients Variable (Mean ⫾ SD)

Moderate/Severe Inflammation

Significant Fibrosis/Cirrhosis

Cirrhosis

40.0 28.6

20.0 32.7

0 8.2

100† 25.5

100* 27.5

33.3 5.9

Hemoglobin (152.7 ⫾ 17.5 g/L) ⱕ130 (n ⫽ 5, 9.3%) ⬎130 (n ⫽ 49, 90.7%) WBC (6.8 ⫾ 2.6 ⫻ 109/L) ⱕ4.0 (n ⫽ 3, 5.6%) ⬎4.0 (n ⫽ 51, 94.4%) Platelets (202.6 ⫾ 72.6 ⫻ 109/L) ⱕ150 (n ⫽ 13, 24.1%) ⬎150 (n ⫽ 41, 75.9%) INR (1.0 ⫾ 0.2) ⬎1.2 (n ⫽ 5, 9.3%) ⱕ1.2 (n ⫽ 49, 90.7%) ALT (132.1 ⫾ 102.6 U/L) ⬎120 (n ⫽ 23, 44.2%) ⱕ120 (n ⫽ 29, 55.8%) ⬎200 (n ⫽ 9, 17.3%) ⱕ200 (n ⫽ 43, 82.7%) Bilirubin (15.2 ⫾ 14.8 ␮mol/L) ⬎20 (n ⫽ 11, 20.4%) ⱕ20 (n ⫽ 43, 79.6%) Albumin (39.4 ⫾ 4.6 g/L) ⱕ35 (n ⫽ 7, 13.0%) ⬎35 (n ⫽ 47, 87.0%) Ferritin (239.6 ⫾ 227.5 ␮g/L) ⬎200 (n ⫽ 24, 47.1%) ⱕ200 (n ⫽ 27, 52.9%)

46.2 24.4

69.2† 19.5

30.8† 0

60.0 26.5

80.0* 26.5

40.0* 4.1

47.8* 13.8 33.3 27.9

39.1 24.1 33.3 30.2

17.4* 0 11.1 7.0

63.6† 20.9

54.5§ 25.6

27.3* 2.3

85.7† 21.3

100‡ 21.3

50.0† 11.1

41.7 22.2

42.9† 2.1 8.3 3.7

All p values by Fisher exact test. INR ⫽ International normalized ratio. * p ⬍ 0.05. † p ⬍ 0.01. ‡ p ⬍ 0.001. § p ⬍ 0.1.

spider nevi, hepatomegaly, splenomegaly, white blood cell count (WBC) ⱕ 4.0 ⫻ 109/L, ALT ⬎ 120 U/L (three times upper limit of normal), bilirubin ⬎ 20 ␮mol/L, albumin ⱕ 35 g/L, and ferritin ⬎ 200 ␮g/L (Tables 1 and 2). The proportion of present drinkers (five or more drinks/wk) with moderate/severe inflammation was not significantly different from that of patients without significant alcohol intake (20.0% vs 29.3%). Elevation of aminotransferases greater than three times of normal was predictive for moderate to severe inflammation (47.8% vs 13.8%). Aminotransferases five times of normal had a weaker discriminating ability, but smaller numbers were involved in this analysis. In particular, mode of virus transmission (injection drug use vs transfusion, occupational exposure, etc) was not predictive of either inflammation or fibrosis. Predictors of Fibrosis The following clinical variables were significantly associated with the presence of advanced fibrosis or cirrhosis: age ⱖ 40 yr, spider nevi, history of decompensation, hepatomegaly, splenomegaly, WBC ⱕ 4.0 ⫻ 109/L, platelet count ⱕ 150 ⫻ 109/L, international normalized ratio ⬎ 1.2, and albumin ⱕ 35 g/L (see Table 1). There were trends for an association between bilirubin or either firm consistency or left lobe enlargement and fibrosis although these did not

reach statistical significance (p ⫽ 0.08 and 0.09, respectively). Predictors of Cirrhosis When analyzed for an association with cirrhosis itself, the same variables as for fibrosis were found to be significant with the following exceptions: ALT ⬎ 120 U/L and bilirubin were significantly associated with cirrhosis, whereas age of ⱖ40 yr and leukopenia were not. Twenty-five percent of those with a firm liver edge or an enlarged left lobe were found to be cirrhotic versus 4.3% without this finding. In addition to the high number of predictor variables found to be common to inflammation and fibrosis, there was also a significant association between fibrosis score and inflammation (Spearman coefficient ⫽ 0.54, p ⬍ 0.001). Sixty-eight percent of patients with moderate to severe inflammation were

Table 3. Proportion of Matching Between Predicted and Observed Biopsy Scores Fibrosis Score Exact matches 41 (75.9%) Scores off by ⱕ1 point 54 (100%)

Inflammation Concomitant Score Liver Disease 37 (68.5%) 35 (64.8%)* 54 (100%)

* Fatty liver was the only “concomitant disease” missed or overcalled.

AJG – November, 2001

Global Assessment Predicts Histology in HCV

Table 5. Biopsy Staging: Predicted vs Observed

Table 4. Biopsy Grading: Predicted vs Observed Observed Inflammation

Observed Fibrosis

Grade ⱕ1 Grade 2 Grade 3 Totals Predicted inflammation Grade ⱕ1 Grade 2 Grade 3 Totals

3171

32 6

10 5

1

38

15

1

42 12 0 54

cirrhotic versus 15.8% of patients with mild inflammation. Ferritin and present alcohol use were not able to be analyzed because of the small number of cirrhotic patients; however, past alcohol use was not associated with increased fibrosis. Multivariate Analysis Two multivariate models were found to significantly predict the presence of moderate to severe inflammation: 1) elevated ferritin (p ⫽ 0.01) and the presence of spider nevi (p ⫽ 0.03); or 2) elevated ferritin (p ⬍ 0.01) and low albumin (p ⫽ 0.02). Hypoalbuminemia and the presence of spider nevi were each significantly associated with one another (p ⫽ 0.02). None of the potential three-variable models were therefore found to be statistically significant. Of note, history of decompensation was not found to be independently predictive of advanced fibrosis in our model. This is partly explained by its low prevalence in this study (three of 54 patients) and partly because this information may have been redundant compared with other clinical and biochemical features. Based on these results, a predictive model including two criteria (high ferritin, and either spider nevi or low albumin) was constructed. Likelihood ratios (LR) were calculated, and the results are summarized in Table 6. LRs of ⬎1.0 raise the probability of inflammation whereas LRs of ⬍1.0 decrease that probability. The presence of two criteria is highly predictive of significant inflammation (LR ⫽ 5.54), whereas the absence of any criteria is highly suggestive of the absence of significant inflammation (LR ⫽ 0.11). The presence of only one criterion does not alter the pretest probability of inflammation significantly. The presence of spider nevi (ps ⫽ 0.004, 0.005, and 0.03, respectively) was independently predictive of fibrotic stage when compared with splenomegaly (p ⫽ 0.02), thrombocytopenia (p ⫽ 0.02), or hypoalbuminemia (p ⫽ 0.007). Hypoalbuminemia (p ⫽ 0.002) was also independent of thrombocytopenia (p ⫽ 0.03). Two additional models were identified that included three independent variables, although they did not quite reach statistical significance. Spider nevi (p ⫽ 0.006), thrombocytopenia (p ⫽ 0.09), and splenomegaly (p ⫽ 0.08) was one. The second model included spider nevi (p ⫽ 0.06), thrombocytopenia (p ⫽ 0.06), and hypoalbuminemia (p ⫽ 0.02). Splenomegaly and hypoalbuminemia were highly associated with one another (p ⬍ 0.001); thus, the presence of one did not offer additional infor-

Predicted fibrosis Stage ⱕ1 Stage 2 Stage 3 Totals

Stage ⱕ1

Stage 2

34 3

7 3 3 13

37

Stage 3

Totals

4 4

41 6 7 54

mation above the presence of the other for the prediction of fibrosis. Again, based on these results, a predictive model including three criteria (spider nevi, thrombocytopenia, and either low albumin or splenomegaly) was constructed. LRs were calculated, and the results are summarized in Table 7. Because of “zero cells,” the standard technique of adding 0.5 to each cell count in the table was used to avoid a LR of infinity for the presence of two criteria or three criteria (11). The presence of two or three criteria greatly increases the risk of significant fibrosis (LR ⫽ 18.5), whereas the absence of any criteria makes the presence of significant fibrosis unlikely (LR ⫽ 0.32). As with the inflammation model, the presence of one criterion alone does not alter the pretest probability of fibrosis significantly (LR ⫽ 1.19).

DISCUSSION This is the first study, which prospectively compares the clinician’s overall prediction of specific biopsy findings (grade, stage, and concomitant disease), using all usually available data, to the actual histology. In an attempt to replicate the practicality of clinical practice, we chose the biopsy scoring system described by Hytiroglou et al. (9), which has broad, clearly defined categories. Differences in the score in this system are more likely to be clinically significant. A minor change between scores in the more complex histologic activity index of Knodell (12) has never been shown to be clinically significant nor is it likely to be. A scoring system with fewer categories has the advantage of greater ease of application, and it is likely to have a high degree of reproducibility (9, 13). Furthermore, we thought it unlikely that clinicians would be unable to accurately predict biopsy scores using more complex classifications such as Knodell’s index. In this study, exact inflammatory grade could be predicted in 66.7%, histological stage in 75.9%, and both together in 49.1%. The sensitivity and specificity for identifying cirrhotic patients were 100% and 94%, respectively. The use of ␬ statistics revealed that the predictions were statistically significantly better than chance alone for predicting the stage, but not for the grade. Liver biopsy can give useful information in patients with hepatitis C infection. Many clinicians are more likely to treat patients with severe inflammation or early fibrosis but less likely to treat someone with advanced fibrosis or cirrhosis. Patients with advanced fibrosis or cirrhosis are at

3172

Romagnuolo et al.

AJG – Vol. 96, No. 11, 2001

Table 6. Prediction Model for Inflammation Significant Inflammation (Grades 2 or 3)

Lack of Significant Inflammation

Totals

Likelihood Ratios

7 8 1 16

3 13 22 38

10 21 23 54

5.54 1.46 0.11

Any two criteria* Any one criterion No criteria Totals

* Inflammation criterion 1: ferritin ⱖ 200 ␮g/L; inflammation criterion 2: presence of spider nevi and/or albumin ⱕ 35 g/L.

higher risk of hepatic decompensation (14). Screening for hepatocellular carcinoma may be recommended in these patients. As well, the presence of coexistant liver disease can be evaluated with liver biopsy. Lastly, it has been suggested that getting a baseline and follow-up liver biopsy may be the only way to identify a subset of patients who appear to be virological nonresponders to treatment but in whom histology improves (15). There are a number of studies looking at the utility of liver biopsy in different patient groups with suspected liver disease. Levin et al. (16) showed that 20% of patients suspected clinically of having alcoholic liver disease had other diagnoses on liver biopsy. In a second study, the positive predictive value of the prebiopsy diagnosis ranged from 88% for alcoholic liver disease to 56% for fatty liver in patients with chronically elevated liver enzymes (17). Yano et al. (18) studied the relationship of histological grade of inflammation to prognosis in patients with hepatitis C. They showed that 100% of patients with high-grade inflammation developed cirrhosis over 10 yr, compared with 96% of those with intermediate grade and 30.4% with low grade. The relevance of various laboratory factors in the prediction of the histological grade and stage of hepatitis C has also been studied, in an attempt to avoid the risk and discomfort of a liver biopsy in selected patients. In one epidemiological study, cirrhosis was found to be more common in patients ⬎ 40 yr old (34.4% vs 2.8%), and minimal inflammation was found more often in patients ⬍ 40 yr old (19). Mode of transmission did not correlate with histological grade or stage. Lebovics et al. (20) found chronic hepatitis C patients with severe, diffuse pruritus to have higher associated histological grades and stages, and to be almost 10 times as likely to have cirrhosis, compared with controls. There are recently published studies looking at the rela-

tionship between alcohol and histology and prognosis in hepatitis C. Pessione et al. (21) showed that self-reported alcohol consumption was highly correlated with serum HCV-RNA levels and that fibrosis was significantly correlated with age and alcohol consumption. In another study by Ostapowicz et al. (22), patients with cirrhosis had a greater total lifetime alcohol consumption on multivariate analysis. It was predicted in designing this study that past alcohol consumption would correlate best with fibrosis, and present consumption would correlate best with inflammatory grade. Because only 22 of the patients in our study could be classified as heavy drinkers (more than 10 drinks/wk), coupled with the small number of cirrhotic patients, it was not possible to exclude a significant relationship between alcohol use and histology. The predictive value of laboratory tests in chronic hepatitis C has been studied more extensively than has the value of clinical findings. Higher ALT levels have suggested more severe necroinflammatory activity, but there is such tremendous overlap and variability that its utility in the prediction of individual cases is poor (23–25). Furthermore, Vandelli et al. (26) found no correlation at all. HCV-RNA levels measured by various techniques have been found to show everything from a direct relationship (27–30) to even an inverse relationship (31) with liver histology. Bonacini et al. (32) found that a discriminant score using platelets, ALT/ AST ratio, and prothrombin time had a sensitivity of 46% but a specificity of 98% for the diagnosis of advanced fibrosis or cirrhosis. Finally, less standard tests such as serum procollagen III propeptide, type IV collagen, and IgG have been found to be of limited use (33). The studies to date have been helpful in evaluating the correlation of various laboratory tests with liver biopsy findings in hepatitis C patients. In clinical practice, however, all available data are normally used together by a clinician

Table 7. Prediction Model for Fibrosis

Any three criteria† Any two criteria Any one criterion No criteria Totals

Significant Fibrosis (Stages 2 or 3)

Lack of Significant Fibrosis

Totals

Likelihood Ratios*

4

0*

4

18.5

4 5 4 17

0* 9 28 37

4 14 32 54

18.5 1.19 0.32

* Calculated after adding 0.5 to each cell count to deal with the “zero cell.” † Fibrosis criterion 1: spider nevi; fibrosis criterion 2: platelets ⱕ 150 ⫻ 109/L; fibrosis criterion 3: splenomegaly and/or albumin ⱕ 35 g/L.

AJG – November, 2001

to make predictions or to guide important clinical decisions such as the use of antiviral therapy. This prospective study is the first attempt to use an unstructured global clinical assessment to predict actual liver histology in chronic hepatitis C patients and clarify the added utility of a liver biopsy. The inherent advantage of this study proves, however, also to be a potential weakness. In attempting to be practical, one loses some reproducibility, in that the way that a clinician comes to a particular conclusion is often difficult to define precisely and thus difficult to replicate. Therefore, in an attempt to further evaluate this question, and to potentially increase the accuracy of the clinical prediction by making it more structured, multivariate analyses were performed to see which clinical and laboratory findings were important in this data set. The many common predictors support the notion that inflammation and fibrosis are intimately related. Patients with higher inflammatory activity over a period of time are more likely to suffer significant fibrotic change. Despite this overlap, those patients with cirrhosis could be accurately separated from the rest. The presence of spider nevi, an important clinical clue to advanced disease, as well as hypoalbuminemia, were found to be, not surprisingly, independent predictors of both inflammation and fibrosis in the multivariate analysis. Ferritin was an independent predictor of inflammatory grade. Ferritin is elevated in many types of liver disease, which may be explained by its role as an acute phase reactant, its release in hepatocyte destruction, or its role as a marker of the degree of iron loading, which may be a risk factor for more advanced inflammation via prooxidant effects. One curious finding was that all patients with WBCs of ⱕ4.0 had moderate to severe inflammation compared with 25.5% of patients with a higher WBC. This could be explained by some degree of bone marrow suppression as is seen in other viral illnesses, or by the high correlation between inflammation and significant fibrosis or portal hypertension (two of three of these patients had significant fibrosis; one of three was cirrhotic). Because of the small number of leukopenic patients here, this association remains to be confirmed. As in other studies, elevated aminotransferases and route of transmission were not associated with either inflammatory grade or stage of fibrosis. However, elevation of aminotransferases more than three times of normal was predictive of inflammation on biopsy. ALT/AST ratios were not evaluated in our study. Two models (Tables 6 and 7) were proposed based on the results of the multivariate analyses for the prediction of inflammation and fibrosis, and clinically useful likelihood ratios are presented; these remain to be validated prospectively. The generalizability of this study is limited by the use of a single clinician and a small sample size. As well, the center is a tertiary care liver transplant center with a large volume practice, and so one must be cautious in extrapolating these results to a general gastroenterologist’s practice.

Global Assessment Predicts Histology in HCV

3173

A multicenter trial is currently underway involving several clinicians making clinical predictions, to determine the generalizability of these results and to validate the models derived from the multivariate analysis. It is also unknown how significantly the under- or overestimations of pathology seen in this study would have affected the decision to treat or not to treat with antiviral therapy or would have altered the prognostic information given to the patient. This particular aspect, not addressed by this study design, certainly deserves further study, but would be highly dependent on the treatment philosophy of the practitioner and would be difficult to model in the form of a trial. In conclusion, clinical assessment is a powerful, undervalued tool in the evaluation of patients with hepatitis C before antiviral therapy. The performance of global assessment in predicting fibrosis is good, and the sensitivity and specificity for detecting cirrhosis are excellent. Accurate prediction of inflammatory grade is more difficult. The two models derived from the multivariate analysis may prove to be useful ways of improving clinical assessments in practice. Liver biopsy is still the gold standard in the assessment of patients with hepatitis C, but careful attention to clinical evaluation and simple laboratory tests can yield helpful information in the assessment of our patients with chronic hepatitis C. Reprint requests and correspondence: Vincent G. Bain, M.D., F.R.C.P.(C.), 2E1.22 Walter C. Mackenzie Health Sciences Centre, 8440 112th Street, University of Alberta, Edmonton, AB, T6G 2R7, Canada. Received Oct. 16, 2000; accepted Mar. 29, 2001.

REFERENCES 1. Poynard T, Leroy V, Cohard M, et al. Meta analysis of interferon randomized trial in the treatment of viral hepatitis: Effects of dose and duration. Hepatology 1996;24:778 – 89. 2. Poynard T, Marcellin P, Lee SS, et al. Randomized trial of interferon ␣-2b plus ribavirin for 48 weeks or for 24 weeks versus interferon ␣-2b plus placebo for 48 weeks for treatment of chronic infection with hepatitis C virus. Lancet 1998;352: 1426 –32. 3. McHutchison JG, Gordon SC, Schiff ER, et al. Interferon alfa-2b alone or in combination with ribavirin as initial treatment for chronic hepatitis C. N Engl J Med 1998;339:1485– 92. 4. Kaserer K, Fiedler R, Steindl P, et al. Liver biopsy is a useful predictor of response to interferon (IFN)-therapy in chronic hepatitis C (CHC). Hepatology 1996;24:154A (abstract). 5. Heintges T, Mohr L, Hensel F, et al. Value of liver biopsy prior to interferon therapy for chronic viral hepatitis. Dig Dis Sci 1998;43:1562–5. 6. Everhart JE, Stolar M, Hoofnagle JH. Management of Hepatitis C: A national survey of gastroenterologists and hepatologists. Hepatology 1997;26(suppl 1):78S– 82S. 7. McGill DB, Rakela J, Zinsmeister AR, et al. A 21-year experience with major hemorrhage after percutaneous liver biopsy. Gastroenterology 1990;99:1396 – 400. 8. Perillo RP. The role of liver biopsy in hepatitis C. Hepatology 1997;26(suppl 1):57S– 61S.

3174

Romagnuolo et al.

9. Hytiroglou P, Thung SN, Gerber MA. Histological classification and quantification of the severity of chronic hepatitis: Keep it simple! Sem Liv Dis 1995;15:414 –21. 10. Schoenfeld P, Guyatt G, Hamiton G, et al. An evidence-based approach to gastroenterology diagnosis. Gastroenterology 1999;116:1230 –7. 11. Hosmer DW, Lemeshow S. Model building strategies. In: Hosmer DW, Lemeshow S, eds. Applied logistic regression. New York: John Wiley & Sons, 1989;82–134. 12. Knodell RG, Ishak KG, Bleck WC, et al. Formulation and application of a numerical scoring system for assessing histological activity in asymptomatic chronic active hepatitis. Hepatology 1981;1:431–5. 13. Brunt EM. Grading and staging the histopathological lesions of chronic hepatitis: The Knodell histology activity index and beyond. Hepatology 2000;31:241– 6. 14. Khan MH, Farrell GC, Byth K, et al. Which patients with hepatitis C develop liver complications? Hepatology 2000;31: 513–20. 15. Poynard T, Opolon P. Hepatitis C: Somber views of natural history and optimistic views of interferon treatment? Hepatology 1998;27:1443– 4. 16. Levin DM, Baker AL, Riddell RH, et al. Nonalcoholic liver disease. Overlooked causes of liver injury in patients with heavy alcohol consumption. Am J Med 1979;66:429 –34. 17. Van Ness MM, Mae Diehl A. Is liver biopsy useful in the evaluation of patients with chronically elevated liver enzymes? Ann Inter Med 1989;111:473– 8. 18. Yano M, Hiromitsu K, Masayoshi K, et al. The long-term pathological evolution of chronic hepatitis C. Hepatology 1996;23:1334 – 40. 19. Delladetsima JK, Rassidakis G, Tassopoulos NC, et al. Histopathology of chronic hepatitis C in relation to epidemiological factors. J Hepatol 1996;24:27–32. 20. Lebovics E, Seif F, Kim D, et al. Pruritus in chronic hepatitis C: Association with high serum bile acids, advanced pathology and bile duct abnormalities. Dig Dis Sci 1997;42:1094 –9. 21. Pessione F, Degos F, Marcellin P, et al. Effect of alcohol consumption on serum hepatitis C virus RNA and histological lesions in chronic hepatitis C. Hepatology 1998;27:1717–22. 22. Ostapowicz G, Watson KJR, Locarnini SA, et al. Role of

AJG – Vol. 96, No. 11, 2001

23. 24.

25.

26. 27.

28.

29. 30. 31. 32.

33.

alcohol in the progression of liver disease caused by hepatitis C virus infection. Hepatology 1998;27:1730 –5. Haber MM, West AB, Haber AD, et al. Relationship of aminotransferases to liver histological status in chronic hepatitis C. Am J Gastroenterol 1995;1:1250 –7. Healey CJ, Chapman RWG, Fleming KA. Liver histology in hepatitis C infection: A comparison between patients with persistently normal or abnormal transaminases. Gut 1995;37: 274 – 8. Papatheodoridis GV, Delladetsima J, Verghisi-Nikolakaki S, et al. Clinicopathological assessment of hepatitis C virus infection in parenteral drug abusers. Am J Gastroenterol 1995; 90:1843– 6. Vandelli C, Renzo F, Floridia A, et al. Do alanine aminotransferase reflect histological status in chronic HCV infection? J Hepatol 1997;26:209:C01/128 (abstract). Kato N, Yokosuka O, Hosoda K, et al. Quantification of hepatitis C virus by competitive reverse transcription-polymerase chain reaction: Increase of the virus in advanced liver disease. Hepatology 1993;18:16 –20. Naito M, Hayashi N, Hagiwara H, et al. Serum hepatitis C virus RNA quantity and histological features of hepatitis C virus carriers with persistently normal ALT levels. Hepatology 1994;19:871–5. Gordon SC, Kodali VP, Silverman AL, et al. Levels of hepatitis C virus RNA and liver histology in chronic type C hepatits. Am J Gastroenterol 1994;89:1458 – 61. Choi Y, Putti T, Win K, et al. Correlation of viral RNA, alanine aminotransferase, and histopathology in hepatitis C virus-associated hepatitis. Mol Diagn 1999;4:251– 4. Lau JYN, Davis GL, Kniffen J, et al. Significance of serum hepatitis C virus RNA levels in chronic hepatitis C. Lancet 1993;341:1501– 4. Bonacini M, Hadi G, Govindarajan S, et al. Utility of a discriminant score for diagnosing advanced fibrosis or cirrhosis in patients with chronic hepatitis C virus infection. Am J Gastroenterol 1997;92:1302– 4. Verbaan H, Bondeson L, Eriksson S. Non-invasive assessment of inflammatory activity and fibrosis (grade and stage) in chronic hepatitis C infection. Scand J Gastroenterol 1997;32: 494 –9.