Progression of fibrosis in advanced chronic hepatitis C: Evaluation by morphometric image analysis

Authors


  • Potential conflict of interest: Dr. Goodman received grants from Intermune. Dr. Pockros is a consultant and received grants from Intermune. Dr. Afdhal is a consultant, advises, is on the speakers' burea of, and received grants from Intermune.

Abstract

Fibrosis progression in chronic liver disease has usually been evaluated by liver biopsy using insensitive semiquantitative numerical scores. An alternative to this is to measure fibrous tissue quantitatively using morphometric image analysis. The aim of this study was to quantify fibrosis progression in a cohort of patients with treatment-refractory chronic hepatitis C enrolled in a placebo-controlled clinical trial of interferon gamma-1b (IFN-γ 1b) for the treatment of advanced hepatic fibrosis. We used morphometry to quantify the amount of fibrous tissue in liver biopsies performed at baseline and after 48 weeks in 245 patients who had paired unfragmented, adequate-sized specimens and correlated the results with clinical and laboratory parameters. Eighty-seven patients were treated with placebo and 158 with IFN-γ 1b. No effect of the drug on fibrosis was found in the trial, and so data from all 245 patients were combined for analysis. At baseline, 78% had cirrhosis; 22%, bridging fibrosis. The mean morphometrically determined collagen content increased by 58% between baseline and 48 weeks. There were statistically significant but weak correlations of fibrosis with platelet count, albumin, bilirubin, INR, and hyaluronic acid; however, changes in these did not correlate with or predict changes in fibrosis in the liver biopsy. Conclusion: In advanced chronic hepatitis C, fibrosis increases at a rapid pace that can only be detected by morphometry. This technique can be used in future therapeutic trials of agents to inhibit fibrosis progression. (HEPATOLOGY 2007;45:886–894.)

The morbidity and mortality that accompany chronic liver disease are generally accepted as being attributable to the scarring that follows chronic hepatic injury.1–3 Hepatic fibrosis may eventually lead to cirrhosis, and although clinical decompensation and death seldom occur until after cirrhosis has developed, the progression of fibrosis is considered the key factor in the pathogenesis of cirrhosis. Consequently, a number of investigators have used liver biopsies to determine the rate of progression of fibrosis in chronic liver disease, especially chronic hepatitis C.4–14

Most studies have measured fibrosis with numerical systems of Scheuer,15 the Metavir group,16 or Ishak et al,17 using stages that range from 0 to 4 for Scheuer and Metavir or 0 to 6 for Ishak. The numbers, while intended to be semiquantitative, actually represent categories of increasing severity based on a combination of location and quantity of scarring, and whether the fibrous tissue forms septa, bridges, or nodules. Thus, a biopsy with a small amount of fibrous tissue in a particular location could have a higher stage than a biopsy with a greater amount of fibrous tissue but not in that location.

An alternative to numerical fibrosis scoring systems is direct measurement of the amount of fibrosis in the biopsy specimen by computer-assisted morphometric image analysis. Although the methodology is not standardized, a number of publications have described methods for quantifying hepatic fibrosis by image analysis, and all appear to yield similar results.5, 18–27 The methods are time consuming compared with simple histologic scoring, and the necessary equipment and expertise are not widely available, but in a study to evaluate progression or regression of fibrosis, measuring the amount of fibrosis as precisely as possible seems appropriate.

Two major limitations arise in using morphometric analysis of needle liver biopsies to study changes in hepatic fibrosis, but both can be overcome with appropriate study design. First, morphometry requires good quality, unfragmented specimens. In advanced fibrosis or cirrhosis, needle biopsies performed with suction techniques are often badly fragmented, consisting of parenchymal nodules that leave most of the fibrous tissue behind. Image analysis of such biopsies will markedly underestimate hepatic fibrosis. However, some biopsies performed with suction needles and most performed with cutting needles yield unfragmented specimens with both parenchyma and fibrous tissue suitable for morphometric analysis. Second, there is the problem of biopsy sampling error. In chronic liver diseases, an irregular distribution of fibrosis is seen throughout the liver tissue. Small needle biopsies may not be representative of the amount of fibrous tissue compared with large sections of tissue, and even large biopsies may not be entirely representative of the amount of hepatic fibrosis in an individual case.24, 25 However, even though a single needle biopsy may overestimate or underestimate the amount of fibrosis in that patient's liver, image analysis of the statistical sample provided by needle biopsies from many patients can be used to estimate the average amount of fibrosis in the cohort as a whole with a high degree of accuracy. Thus, a study using morphometry to measure fibrosis progression needs to be limited to unfragmented, preferably large biopsies from a sufficiently large number of patients for a statistical analysis to be valid.

The opportunity to study fibrosis change by morphometric image analysis was made available by a clinical trial of interferon gamma-1b (IFN-γ 1b) for the treatment of advanced fibrosis attributable to chronic hepatitis C.28 IFN-γ 1b is a cytokine that has shown promise as an antifibrotic agent because of its effects on hepatic stellate cells in vitro, from animal studies showing regression of fibrosis, and from studies of human idiopathic pulmonary fibrosis. A placebo-controlled trial was designed to study the safety and efficacy of IFN-γ 1b in patients with chronic hepatitis C resistant to interferon alpha–based therapy and with marked fibrosis or cirrhosis on liver biopsy. The trial found no effect of treatment on hepatic inflammation or fibrosis in this cohort with advanced liver disease. However, patients in the trial had liver biopsies performed at a uniform time interval (48 weeks), and 245 patients had paired, unfragmented biopsies that were adequate for morphometric analysis of changes in hepatic fibrosis.

Abbreviation

IFN-γ 1b, interferon gamma-1b.

Patients and Methods

Liver biopsies were obtained in the course of a phase 2 randomized, double-blind, placebo-controlled trial of the antifibrotic efficacy of IFN-γ 1b in chronic hepatitis C patients with advanced fibrosis or cirrhosis.28 Patients recruited into the study had clinically compensated chronic hepatitis C refractory to standard IFN-α–based therapy, and they were required to have an adequate liver biopsy, evaluated by the central pathologist (Z.G.), showing marked bridging fibrosis or cirrhosis, stages 4 to 6 of the Ishak score.17 They were randomized 1:1:1 to receive placebo or IFN-γ 1b at 100 or 200 μg 3 times weekly monotherapy for 48 weeks.

Participation in the trial was approved by institutional review boards at each of 62 participating clinical sites and at the Armed Forces Institute of Pathology. Liver biopsies were performed before treatment and after 48 weeks of therapy. A baseline biopsy was required at screening for each patient unless an adequate one had been obtained within the previous 12 months. Only 48 patients (20%) had biopsies more than 60 days before randomization. The investigators at each site were permitted to choose the technique for performing the biopsy. The sites were encouraged to send the tissue in fixative to the central pathologist so that processing and sectioning could be as uniform as possible. Overall, 60% were submitted as formalin-fixed tissue, 26% with the tissue embedded in paraffin blocks and 14% as unstained slides prepared at the center. Of the 245 baseline biopsies used in the subsequent morphometric analyses, 144 (59%) were submitted as formalin-fixed tissue, 60 (24%) as paraffin blocks, and 41 (17%) as unstained slides prepared at the center. Of the 245 week 48 biopsies used for morphometry, 188 (77%) were submitted as formalin-fixed tissue, 38 (16%) as paraffin blocks, and 19 (8%) as unstained slides.

The primary efficacy endpoint of the trial was a reduction of one or more stage in the Ishak fibrosis score. Data collected for secondary endpoints included clinical findings and laboratory markers of hepatic function and injury, serum fibrosis markers, hepatitis C RNA levels, and quantitative assessment of changes in fibrosis by morphometry.

Histological Analysis.

Sections of each baseline and post-treatment liver biopsy stained with hematoxylin-eosin and Masson's trichrome stain were reviewed and scored for inflammation and fibrosis. The Ishak score17 was used for the primary endpoint, but data for other scoring systems were also collected. The central pathologist reviewed the paired biopsy slides from each patient while blinded to the order of the biopsies and the treatment group. Whether each biopsy was fragmented, and the length of the specimen, including all fragments, as represented on the slide, was measured. During the initial review, it was also noted whether each patient's baseline and post-treatment biopsies were of adequate size (at least 10 mm long) and unfragmented so as to be suitable for morphometric analysis (Fig. 1). Exceptions were allowed for 15 unfragmented wide (16-gauge) needle biopsies 8 to 9 mm long and one partly fragmented 22-mm biopsy in which the fragments were large and contained both parenchyma and fibrous tissue.

Figure 1.

Suitability of liver biopsies for morphometric image analysis (Masson trichrome stains). (A) A 25-mm-long biopsy from a cirrhotic liver that is too badly fragmented for morphometry. The fibrous tissue is underrepresented in the specimen. (B,C) Nine-millimeter unfragmented biopsies with cirrhosis. Because the fibrous tissue appears to be adequately represented, they are acceptable for morphometry.

Morphometry.

Sections of each biopsy stained with Sirius red were used for measurement of fibrosis. This stain binds to all connective tissue, but primarily to collagen, and the quantity of bound stain has been shown to correlate well with chemically determined collagen content and morphometrically determined hepatic fibrosis,29, 30 so the degree of red staining can be taken as proportional to the amount of collagen present. Two sets of images representing the entire biopsy were acquired using a 4× objective. The first set consisted of brightfield RGB images, and the second consisted of corresponding grayscale images using crossed polarization filters (Fig. 2). The automated image acquisition workstation included a digital Spot RT camera, Olympus BX-51 microscope, a motorized 4-slide Prior Stage, and desktop computer. The camera, microscope, and stage were driven by Image Pro Plus 4.0 imaging software (Media Cybernetics, Silver Spring, MD) and a Visual Basic 6.0 subprogram to automate the data acquisition. A manually edited image mask, created from a low-power overview of the specimen, was applied during image capture to ignore areas with technical artifacts or without tissue. A gray-valued overview image, obtained at one-fourth the resolution of the RGB images, was used to segment tissue of interest from background. First, all sufficiently dark pixels of the image were captured in foreground objects, automatically eliminating very small objects. The remaining objects were manually edited, eliminating technical artifacts and extraneous tissue such as muscle and skin to yield a mask corresponding to the tissue of interest. A custom-written combination of programs was used to calibrate the image sets against reference images and to mosaic the tiles, producing high-resolution polarized and nonpolarized images of each entire specimen. A dark image (obtained with the light source blocked) was subtracted from each brightfield image. Another difference image was made by subtracting the dark image from an averaged blank brightfield image. The ratio of the two difference images was rescaled to yield the calibrated image. Analogous calculations were applied to each polarized light image, using dark and blank images obtained with crossed and uncrossed polarizers, respectively, and a further step was added in dividing the result at each pixel by an estimate of the transmittance at the corresponding pixel of the brightfield image. The mosaicing program accounted for image and pixel sizes and corrected for minor rotation of the camera raster relative to imperfectly orthogonal stage movement axes. Collagen content at each pixel location was calculated as the square root of the product of the saturation of the red channel of the RGB image multiplied by the intensity of the corresponding pixel in the polarized image. Collagen stained with Sirius red has bright birefringence,31 whereas pixels that do not contain collagen are black in the polarized image, and so this permits measurement of collagen without interactive thresholding to subtract the background, which is a potential source of error.24 The collagen content per unit area of the specimen was calculated as the sum of the pixel-wise collagen measurements, divided by the number of summed pixels. This result is directly proportional to the thickness of the tissue section over the range of 2 to 10 μm thickness, as determined using serial sections from the same specimen (data not shown). Consequently, the thickness of each tissue section was determined using a confocal laser scanning microscope (Zeiss LSM 310), and the collagen content per unit area was divided by the section thickness to yield the final result, reflecting collagen content per unit volume of tissue. The values thus obtained are in arbitrary units providing a continuous variable that can be used to compare one specimen with another or in statistical calculations.

Figure 2.

Cirrhotic liver biopsy stained with Sirius red, viewed with transmitted light (left) and polarized light in grayscale (right).

Statistics.

Differences between median baseline and week 48 collagen content were evaluated for the entire cohort and within subgroups with the Wilcoxon signed-rank test. Mann-Whitney and Kruskal-Wallis tests were used for differences in nonpaired data between groups. Correlation coefficients were Spearman rank correlations. P values were two-sided.

Results

The principal trial results have been reported.28 A total of 502 patients met entry criteria and were randomized, and 488 received at least one dose of the study drug. Of those, 162 received placebo, 169 received 100 μg IFN-γ 1b 3 times weekly, and 157 received 200 μg IFN-γ 1b 3 times weekly. At baseline 20% were stage 4 (marked bridging fibrosis) on liver biopsy, 29% stage 5 (incomplete cirrhosis), and 51% stage 6 (established cirrhosis). Post-treatment liver biopsies were available in 389 patients (80%). The post-treatment biopsy was inadequate for assessment in 16, with the other 83 patients either withdrawing from the study or declining biopsy.

In this cohort of patients with advanced hepatic fibrosis and cirrhosis, among those with adequate paired biopsies, no significant difference was seen between IFN-γ 1b–treated patients and placebo in the primary endpoint, change in Ishak fibrosis stage (Table 1), or in any other histologic parameters, including morphometrically determined collagen content. Scoring for inflammation using Ishak, Knodell, and Metavir systems found no change between baseline and week 48 and no differences between treatment groups in inflammatory scores, indicating no effect of IFN-γ 1b as monotherapy on hepatic inflammation in this cohort. Similarly, no significant differences were observed between treatment groups in laboratory measures related to hepatic synthetic function (albumin, INR), portal hypertension (platelet count), serum fibrosis markers (hyaluronic acid, YKL-40), or in the proportion of patients who died or suffered consequences of hepatic decompensation (11% overall).

Table 1. Changes in Ishak Fibrosis Stage and Inflammation After 48 Weeks of Therapy
Treatment Group
Change in Fibrosis stagePlacebo (n = 135)100 μg IFN-γ 1b (n = 140)200 μg IFN-γ 1b (n = 114)
−2157
−1231611
0919974
+1181820
+2222
Inflammation (HAI)   
 Baseline (median)9910
 Week 48 (median)9910

Biopsy Quality.

Of the 389 patients with both baseline and post-treatment biopsies, in 144 one or both biopsies were of insufficient quality to be informative for morphometric image analysis. In 79 patients, one or both of the biopsies was badly fragmented so that fibrous tissue was inadequately represented in the section (Fig. 1A), and many of these were also small. In another 37, one or both biopsies were less than 10 mm in length, whereas the remainder had various other technical problems. In 245 patients paired biopsies were suitable for morphometric image analysis. The mean length of these 490 needle biopsies was 21 mm with standard deviation of 8 mm. Fifteen of the 245 patients (6%) had one biopsy less than 10 mm in length, but these were unfragmented and of sufficient width to be adequate for morphometry (Fig. 1B, C). These 490 biopsies are the basis of the subsequent analyses. Thus, of the 502 randomized patients, 127 (25%) withdrew from the study or refused follow-up biopsy, and 144 (29%) had at least one biopsy that was technically unsuitable for morphometry.

Morphometric Assessment.

Of the 490 paired biopsies that were adequate for morphometry, 239 (49%) were Ishak stage 6, 141 (29%) stage 5, 93 (19%) stage 4, and 20 (4%) stage 3. A wide range of collagen values were seen within each stage and considerable overlap between stages, but there was greater mean collagen with advancing stages with a Spearman correlation coefficient of 0.49, P < 0.0001 (Table 2 and Fig. 3). Biopsy length did not correlate with collagen content (Spearman ρ = 0.03, P = 0.49).

Table 2. Morphometric Collagen Content for Each Histologic Fibrosis Stage
Fibrosis StageCollagen Content
MinimumMaximumMedianMeanSEM
  1. NOTE. The Spearman correlation coefficient is 0.49 (P < 0.0001). The medians of stages 3 and 4 did not differ significantly. There were highly significant differences between stages 4 and 5 (P < 0.0001) and stages 5 and 6 (P < 0.0001).

3 (n = 20)0.0180.0980.0570.0580.004
4 (n = 93)0.0240.2050.0630.0730.004
5 (n = 141)0.0180.4260.0900.1060.005
6 (n = 239)0.0310.7750.1370.1500.006
Figure 3.

Distribution of measurements of collagen content at each Ishak fibrosis stage for the 490 liver biopsies in the study. The boxes show the median, 25th, and 75th percentiles, whereas the lines show the range of values. Three outliers in stages 5 and 6 are represented as separate points. The Spearman correlation coefficient is 0.49 (P < 0.0001).

Among the 245 patients with paired biopsies adequate for morphometry, 87 had been randomized to receive placebo, 87 had received 100 μg IFN-γ 1b 3 times weekly, and 71 received 200 μg IFN-γ 1b 3 times weekly. No difference was seen between the three treatment groups in morphometric collagen content, either at baseline (P = 0.85) or after 48 weeks of therapy (P = 0.59), and so the three groups were combined for subsequent analysis. Combining the three groups, the mean liver biopsy collagen content increased by 58% and the median by 82% (P < 0.0001) over the 48-week period of the study (Table 3). A decrease in the number of patients with low collagen content and a corresponding increase in the number with high collagen content (Fig. 4) were seen. A trend toward a greater percentage increase in fibrosis was seen in those with lower baseline Ishak fibrosis stage (Table 4), but the differences were not statistically significant (p=0.48). There were 146 patients in whom both liver biopsies were at least 15 mm in length. Mean collagen content in these patients increased by 56% between baseline and week 48, which was not significantly different from the entire cohort (Table 4).

Table 3. Morphometric Collagen Content of Entire Cohort of 245 Patients and Each of the Three Treatment Groups with Paired Biopsies Adequate for Morphometric Image Analysis
Entire CohortBaselineWeek 48Change
  1. NOTE. Differences between baseline and 48 weeks are highly significant by paired Wilcoxon test (P < 0.0001) for entire cohort and for each of the three treatment groups. Differences between treatment groups are not significant (P = 0.85 at baseline, P = 0.59 at week 48).

Mean0.09360.1475+58%
Minimum0.01800.0221+28%
10 %tile0.03800.0623+64%
25 %tile0.05190.0882+70%
50 %tile0.07540.1370+82%
75 %tile0.12020.1757+46%
90 %tile0.15620.2492+60%
Maximum0.38320.7746+102%
Treatment Group   
 Placebo (n = 87)   
  Mean0.09800.1479+51%
  Median0.07540.1327+76%
 100 μg IFN-γ 1b (n = 87)   
  Mean0.08920.1395+56%
  Median0.07500.1331+77%
 200 μg IFN-γ 1b (n = 71)   
  Mean0.09340.1568+68%
  Median0.08000.1475+84%
Figure 4.

Liver biopsy collagen content at baseline and after 48 weeks in 245 patients with advanced chronic hepatitis C. The group mean increased by 58%, and the median increased by 82% (P < 0.0001).

Table 4. Morphometric Collagen Content of 245 Patients with Paired Biopsies Adequate for Morphometry Analyzed by Baseline Ishak Fibrosis Stage and by Size of the Liver Biopsies
Baseline Ishak Fibrosis StageBaseline CollagenWeek 48 CollagenChange
  1. NOTE. Differences between baseline and 48 weeks are highly significant by paired Wilcoxon test (P < 0.0001) at each of the three baseline Ishak fibrosis stages (4-6). There is a trend toward a greater percentage increase in fibrosis in those with lower baseline fibrosis stage, but the differences are not statistically significant (P = 0.48). Size of the liver biopsies did not affect results, as the group of 146 patients with both biopsies 15 mm or longer did not differ significantly from the group of 99 with one or both biopsies less than 15 mm (P = 0.80 at baseline, P = 0.90 at week 48).

Stage 4 (n = 53)   
 Mean0.05520.1015+84%
 Median0.05030.0978+94%
Stage 5 (n = 76)   
 Mean0.08560.1382+61%
 Median0.07240.1329+84%
Stage 6 (n = 116)   
 Mean0.11630.1746+50%
 Median0.10650.1528+43%
Biopsy Length   
 Both ≥ 15 mm (n = 146)   
  Mean0.09350.1456+56%
  Median0.07410.1388+87%
 One or both < 15 mm (n = 99)   
  Mean0.09360.1502+60%
  Median0.07940.1331+67%

The change in collagen content between baseline and week 48 biopsies in individual patients was subject to well-recognized biopsy sampling variability25 as well as to a true increase in fibrous tissue, but most showed worsening of fibrosis over the 48-week period. Only 53 patients (22%) had a biopsy with less collagen at 48 weeks than at baseline (Fig. 5). The median change from baseline was an increase of 65% over the 48-week period of the study (Fig. 6). A statistically significant (P < 0.0001) correlation between change in collagen by morphometry and change in Ishak fibrosis scores was seen, but because 64% of patients had no change in Ishak score, the correlation was relatively weak (Spearman r = 0.39). Paired biopsies from a representative patient are shown in Fig. 7, demonstrating a 55% increase in collagen but no change in Ishak score between baseline and week 48.

Figure 5.

Change in liver biopsy collagen content per patient, expressed in morphometry units. Solid bars represent patients whose biopsy collagen increased over 48 weeks, and the cross-hatched bars represent those whose collagen decreased.

Figure 6.

Change in liver biopsy collagen content per patient, expressed as percentage change from baseline. Solid bars represent patients whose biopsy collagen increased over 48 weeks, and the cross-hatched bars represent those whose collagen decreased.

Figure 7.

Paired biopsies from a single placebo-treated patient, stained with Sirius red. The biopsy on the left taken at baseline has 0.122 units of collagen, and that on the right at week 48 has 0.190, an increase of 55%. Both biopsies are cirrhotic, and it is only by quantitative image analysis that the degree of change can be determined.

Correlation with Liver Function and Laboratory Tests.

A slight worsening on average in tests related to hepatic function (albumin, INR, bilirubin), portal hypertension (platelet count), and fibrosis (hyaluronic acid) was seen between baseline and week 48, although the differences were not statistically significant (Table 5). Aminotransferase enzymes and hepatitis C RNA did not change. The baseline liver biopsies had statistically significant but weak correlations of collagen content determined by morphometry with serum bilirubin, INR, platelet count, and hyaluronic acid, whereas the week 48 biopsies had similar correlations with albumin, INR, platelet count, and hyaluronic acid. Change in collagen did not correlate significantly with changes in any of the laboratory tests except for INR, and the correlation was very weak (Table 6).

Table 5. Laboratory Values at Baseline and After 48 Weeks
Laboratory Values (mean)BaselineWeek 48
  1. NOTE. No significant differences were seen between baseline and week 48 in any laboratory test.

Albumin (g/dl)3.99(n = 241)3.88(n = 210)
Total bilirubin (mg/dl)0.90(n = 241)0.93(n = 210)
INR1.10(n = 235)1.16(n = 203)
Platelets151k(n = 225)147k(n = 192)
Hyaluronic acid252(n = 153)289(n = 159)
ALT (IU/l)107(n = 240)105(n = 208)
AST (IU/l)92(n = 238)99(n = 206)
HCV RNA (×106 IU/ml)1.66(n = 237)1.49(n = 219)
Table 6. Correlation of Laboratory Values at Baseline and Week 48 With Corresponding Liver Biopsy Collagen Content
 Correlation Coefficient (Spearman rank)P*
  • *

    P values are from analysis of variance of biopsies with collagen content ≤0.10, 0.10-0.15, 0.15-0.20, and 0.20.

  • Collagen levels as of week 48

Baseline  
 Bilirubin0.210.01
 INR0.23<0.001
 Platelets−0.260.001
 Hyaluronic acid0.230.045
Week 48  
 Albumin−0.180.017
 INR0.29 
 Platelets−0.200.004
 Hyaluronic acid0.230.026
Change from baseline  
 INR0.0810.027

Fifty-two 52 patients had clinical decompensation during the study with the development of ascites (26 patients), variceal hemorrhage (8), spontaneous bacterial peritonitis (2), encephalopathy (24), hepatocellular carcinoma (8), and/or death due to cirrhosis (5). Thirty-one of these had baseline biopsies that were adequate for morphometry, and their collagen content did not differ significantly from the baseline biopsies of patients who did not decompensate (P = 0.31).

Discussion

Most studies that have used liver biopsies to estimate changes in fibrosis in chronic hepatitis C have relied on the numerical scoring systems rather than direct measurement of the amount of fibrous tissue. Despite the fact that the scoring systems for fibrosis are categories rather than measurements, several studies have reported fibrosis progression rates in terms of “units per year” in whichever scoring system was used, calculated on the basis of sequential biopsies performed for clinical indications at varying intervals8–12 or on the basis of single biopsies with rates calculated from the presumed duration of infection.6, 13 This approach necessarily assumes that each patient arrived at the stage on the day of the biopsy, whereas it may actually have been years earlier, and it also assumes that the categories represented by the numbers of the scores are equal distance apart in terms of severity, amount of fibrous tissue, or time of development, none of which is supported by strong evidence. Indeed, our data, as well as those of O'Brien et al,22 have shown that in fact there is a great deal of overlap between the various stages of scoring systems in the amount of fibrous tissue measured in the biopsies. Thus, a progression through histologic stages is not necessarily the same as an increase in fibrosis.

A rate of increase in fibrosis can be determined when the connective tissue is measured directly and quantitatively by morphometry. All of the published methods provide results on a continuous scale, rather than merely four or six qualitative stages, permitting the use of a full range of statistical methods. The major drawback of using morphometry to assess disease in an individual patient is sampling variability.24, 25 There can be significant differences in the amount of fibrous tissue between closely adjacent parts of the liver, so this sort of data can only be used as a statistical sample to assess changes in a large cohort of patients, not to assess disease progression in an individual patient. Nevertheless, the current study has clearly shown that morphometry is a much more sensitive tool than histological staging to demonstrate fibrosis progression. By Ishak stages (Table 1), 68% of patients had no change in 48 weeks, 16% improved, and another 16% worsened, all of which could be due to sampling variability. By contrast, morphometry showed worsening of fibrosis in 68% and improvement in 22%. Some of this is due to sampling variability, which is why the average for the cohort is the most meaningful result, not the change per individual patient, but it is only by morphometry that the magnitude of change becomes apparent.

Four published studies have used morphometry to study changes in hepatic fibrosis after alpha interferon therapy,20, 23, 32, 33 and one has used morphometry to study disease progression in untreated patients with chronic hepatitis C.5 In an early interferon alpha-2b trial, Manabe et al.32 measured collagen content in 59 patients, 43 treated with interferon and 16 treated with placebo. Mean collagen content of the liver biopsies decreased in the interferon-treated patients, but mean collagen increased by 26% in those treated with placebo for 24 weeks. Similarly, three other studies20, 23, 33 found that collagen decreased after treatment with interferon alpha but included no placebo group. Finally, Kage et al.5 used morphometry to study 25 patients who progressed to cirrhosis caused by hepatitis C as well as 20 with hepatitis B. They found that over a mean period of 5.1 years, the amount of fibrous tissue increased an average of 2.8-fold in the hepatitis C patients or approximately 55% per year.

In our study, we found that in 245 patients with paired biopsies adequate for morphometric analysis, the amount of fibrous tissue increased on average by 58% over the 48-week period of the study. All patients had been previously treated with interferon alpha, which three studies already cited reported to decrease fibrosis even in patients without a sustained virologic response.20, 23, 33 Furthermore, 158 patients (65%) were treated with IFN-γ 1b in the course of this study, but because there was no difference between these and placebo-treated patients (Tables 3 and 4), we combined the data from all 245 patients for the analyses. Even though our patients had been treated with potentially antifibrotic agents, the findings in this cohort are quite similar to those in the small groups of untreated patients reported by Manabe et al.32 and Kage et al.5 Extrapolating the reported data, Manabe et al.32 found an average increase in fibrosis of approximately 52% per year in 16 patients, Kage et al.5 found 55% per year in 25 patients, and we found 58% per year in 245 patients, despite the differences in patients and in the baseline levels of fibrosis. This implies that in patients with chronic hepatitis C, the average amount of hepatic collagen doubles in approximately 2 years, and if this is true at all degrees of fibrosis, then the fibrous tissue may actually increase exponentially rather than linearly. Although there are clearly differences between individuals in the rate of progression of fibrosis, a doubling of the amount of collagen every 2 years could partly explain the apparent acceleration of fibrosis in older individuals who presumably have had the disease for longer time.5, 8, 13, 34 Only by direct measurement of fibrosis within a sufficiently large cohort does this become apparent, because histological stages are neither a continuous function nor a sensitive measure of this type of change. In the current study, we examined fibrosis progression in a cohort of patients who had already proved themselves rapid fibrosers. Furthermore, our conclusions are based on data from only 245 (49%) of the 502 patients randomized for the trial, because the rest either withdrew before the follow-up biopsy or had biopsies that were not technically adequate for morphometric analysis. Clearly, further studies are needed to confirm these observations, to see whether they hold true for patients with early fibrosis, and to determine the role of various cofactors that may accelerate fibrosis.

Progression of fibrosis leads to cirrhosis with its complications. In patients with advanced liver disease, one might expect a relationship between increase in fibrous tissue and increase in portal pressure or decrease in synthetic capacity of liver, but this was not found in the current study. Only weak correlations of laboratory tests that reflect portal hypertension and hepatic synthetic function were seen (Table 6), no correlation with clinical decompensation, and no significant change in mean test values (Table 5), despite the fact that the mean collagen content of the cohort increased by 58%. This suggests that factors other than the absolute amount of fibrous tissue play an important role in the clinical and laboratory changes in advanced liver disease or that the pathophysiologic effects of increased fibrosis lag the increase in fibrous tissue. The architectural distortion that accompanies nodular parenchymal regeneration as well as shunting of blood through vascularized fibrotic septa may be of equal or greater importance than the amount of scar tissue. Similarly, the mass of functioning hepatic parenchyma may be more important than the proportion of parenchyma replaced with scar tissue. Therefore, in the evaluation of an individual patient, the histologic diagnosis, based on the combination of architectural changes and amount of fibrosis, may well be more important than the amount of fibrosis alone.

Hepatic fibrosis is a dynamic process, characterized by a balance between fibrogenesis and fibrolysis. The current study shows the value of morphometric image analysis as well as the necessity of a placebo control group in testing new therapeutic agents designed to inhibit fibrogenesis or promote fibrolysis. If a new therapeutic agent is effective, significantly less increase in the amount of fibrous tissue compared with progression in a placebo group should be seen, and if it is truly effective in a clinically significant way, an actual decrease in the mean collagen of the treatment cohort should occur. This approach cannot be used to manage individual patients, but only to study cohorts with sufficient statistical power to show a meaningful change. Although liver biopsy is an invasive procedure, it is necessary for this type of study because standard laboratory tests do not appear to change enough to reflect changes in the quantity of fibrous tissue. Whether newer batteries of biochemical tests, serum fibrosis markers, imaging techniques, or measurement of liver stiffness by transient elastography will be able to detect changes in fibrous tissue with this degree of sensitivity remains to be determined.

In summary, our studies show that ongoing active fibrosis progression occurs in patients with advanced HCV-related liver disease that can only be identified by careful morphometric analysis. This increase in fibrosis is not initially reflected by clinical changes in outcomes or hepatic function over the short term, but future studies need to determine whether the true fibrosis progression rate by morphometry can predict subsequent clinical outcomes.

Acknowledgements

The authors thank Michelle Parks, who performed the morphometric analyses, Dr. Ruixia Zhou, who assisted with computer programming, and Phil Hormel, who performed statistical analyses.

Disclaimer: The opinions and assertions contained herein are the private views of the authors and are not to be construed as official or as reflecting the views of the Department of the Army or the Department of Defense.

Ancillary