Quantitative histological-hemodynamic correlations in cirrhosis

Authors


  • Potential conflict of interest: Nothing to report.

  • This work was supported by the Clinical Core, Yale Liver Center (NIH P30 DK34989).

Abstract

We have previously shown, in a semiquantitative analysis of liver biopsies showing cirrhosis, that thickness of fibrous septa separating cirrhotic nodules and small size of cirrhotic nodules correlated independently with portal pressure (as determined by the hepatic venous pressure gradient; HVPG) and were independent predictors of the presence of clinically significant portal hypertension (PH). This study aimed to confirm these results using quantitative analysis of these biopsies using digital image analysis. Biopsies of 42 patients with cirrhosis and HVPG measurements within 6 months of the biopsy were included in the study. The following parameters were scored quantitatively and without knowledge of HVPG results: total fibrosis area, septal thickness, nodule size, and number of nodules per millimeter of length of liver biopsy. Fibrosis area was the only parameter that independently correlated with HVPG (r = 0.606; P < 0.0001). Correlation was significant, even among patients with clinically significant PH (r = 0.636; P < 0.005). Fibrosis area and nodule size were both independently predictive of the presence of clinically significant PH (r = 0.57; P = 0.003). Conclusions: On quantitative analysis, fibrosis area was the parameter that correlated best with HVPG and the presence of clinically significant PH. Beyond pathophysiological implications, this also has methodological implications that are discussed in this article. (HEPATOLOGY 2012)

Cirrhosis is the end stage of any chronic liver disease and is defined histologically by the presence of regenerative nodules surrounded by fibrous tissue. This architectural distortion leads to increased intrahepatic resistance, which, in turn, leads to portal hypertension (PH). PH is defined clinically as a portal pressure (as assessed by the hepatic venous pressure gradient; HVPG) greater than 5 mmHg. However, in patients with cirrhosis and PH, the complications of cirrhosis, including esophageal varices, clinical decompensation (i.e., development of ascites, encephalopathy, and variceal hemorrhage), and hepatocellular carcinoma (HCC), occur at a significantly higher rate in patients in whom the HVPG is 10 mmHg or greater.1-3 This threshold portal pressure level has been termed clinically significant portal hypertension (CSPH).4

Although histology is an important determinant of the severity of chronic liver disease, it is the degree of PH that is an important determinant of the severity of cirrhosis. In a study aimed at establishing the relationship between specific histological parameters in cirrhosis and HVPG, we demonstrated that thickness of fibrous septa separating cirrhotic nodules and small size of cirrhotic nodules were independent predictors of the presence of CSPH in patients with predominant hepatitis C.5 In this study, histological parameters were subjectively assessed by two independent observers in a semiquantitative fashion, in that septal thickness was classified as narrow, wide, or intermediate, and nodule size was classified as small, large, or mixed.

The need for validation of this semiquantitative assessment using a more objective, quantitative way of assessing these histological parameters was proposed in a previous study.5 Therefore, the aim of this study was to determine the relationship between HVPG, CSPH, and specific histological parameters in cirrhosis, which was assessed quantitatively by digital imaging analysis (DIA).

Abbreviations

CSPH, clinically significant portal hypertension; DIA, digital image analysis; HCC, hepatocellular carcinoma; HVPG, hepatic venous pressure gradient; PH, portal hypertension; QIA, quantitative image analysis.

Materials and Methods

Materials.

Biopsy slides of patients in whom the diagnosis of cirrhosis was established histologically and who had an HVPG measurement performed within 6 months of each other were included in the study. The same study population/biopsy material described in our previous study5 was used in the current study.

DIA.

Liver biopsies are routinely formalin fixed and paraffin embedded at our institution, and slides are stained with hematoxylin and eosin and trichrome stain (i.e., Masson or Klatskin) for diagnostic purposes. Only trichrome-stained (i.e., Masson or Klatskin) slides were used for DIA. A liver biopsy size of at least 10 mm was required for inclusion in the study. In fragmented biopsies, the total length was estimated by adding maximum dimensions of each individual fragment. The equipment used for DIA consisted of a microscope (Olympus CX41; Olympus, Tokyo, Japan), with images recorded by a digital camera (Retiga 1300 color 12-bit; QImaging, Surrey, British Columbia, Canada) connected to a personal computer. Images were analyzed using Bioquant Nova Prime software (Bioquant Image Analysis Corporation, Nashville, TN), a Microsoft Windows application for automated and semiautomated quantitative analysis of fixed histological sections (Microsoft Corporation, Redmond, WA). All available liver tissue on each biopsy slide was entirely analyzed field by field for various histological parameters. As previously described,6 sections were placed on the x-y motorized tray (to avoid field overlap), and, after equalization of light intensity, each image was digitized in a 640 × 480 pixel picture. Images were obtained and evaluated at a 4× magnification. Optical field was defined as an area of 3.36 mm2 for a 640 × 480 pixel picture, resulting in an optical resolution of 0.09 μm2/pixel.

A single operator (S.S.) determined the following four histological parameters: (1) fibrosis area, expressed as a percentage of the total area of liver tissue in the biopsy occupied by fibrous tissue; (2) width of fibrous septa separating cirrhotic nodules; (3) nodule size; and (4) number of cirrhotic nodules per millimeter of length of the biopsy. The observer was blinded as to the HVPG results and the histological semiquantitative results.

To estimate fibrosis area, the color-editing function of the software was used to select the entire area of fibrous tissue (green pixels for Masson trichrome stain and blue pixels for Klatskin stain). Total biopsy area (expressed in μm2) was also selected using the software and consisted of the area occupied by the hepatic parenchyma, fibrous tissue, nodules, and vessels. Fibrosis area was calculated as the ratio between the area of fibrosis and the total biopsy area and is expressed as a percentage. In previous studies, this method has been shown to achieve high levels of accuracy, reliability, objectivity, and reproducibility.6-9 Septal width was determined as the distance (in μm) between adjacent cirrhotic nodules and was measured at multiple (at least three) points using a standard grid marker (Fig. 1). Median septal width was calculated for each septa and then for the entire biopsy. Nodule size was a measure of the maximum length (in μm) of each nodule along the long axis of the biopsy (Fig. 1). A minimum of five nodules were measured in each biopsy, and the median was calculated and expressed as the number of nodules per millimeter of biopsy. Incomplete or partial nodules at the end or at the edges of a fragment were counted as individual nodules. Small fragments that had only one or only partial nodules were not analyzed for septal width, nodule size, or number of nodules. In cases where multiple good cores of liver tissue were contained in the biopsy slide, the two to three most representative cores were used for image analysis. Cases that had extremely fragmented biopsy and contained less than two fragments that met the criteria for measuring the various histologic parameters were excluded from the study.

Figure 1.

Quantitative measurements. Septal width was assessed by measuring the distance between two cirrhotic nodules at multiple points (at least three, red arrows). Nodule size was assessed as the maximum length along the long axis of the biopsy (black arrows).

HVPG Measurements.

As described in our previous study,5 HVPG was measured at either the Hepatic Hemodynamic Laboratory at the VA-CT Healthcare System (West Haven, CT) or at an Interventional Radiology suite at Yale New Haven Hospital (New Haven, CT), following guidelines established in these centers.10 Clinically significant PH was defined as an HVPG ≥10 mmHg.

Statistical Analysis.

Analysis was performed on a database in which the results of DIA were entered before entering the HVPG data. Nonparametric statistics were used, and results are expressed as medians (ranges). The correlation between HVPG and the different DIA parameters was performed using Spearman's correlation. Variables that were significant at P values equal to or lower than 0.05 were entered in a stepwise linear regression analysis. Comparison of DIA parameters between patients with or without CSPH was performed using the Mann-Whitney U test. Variables that were significant at P values equal to or lower than 0.05 were entered into a backward logistic regression model. Statistical analysis was performed using the SPSS statistical package (SPSS, Inc., Chicago, IL).

Results

Characteristics of Study Population.

Of the 43 patients/biopsy slides analyzed in our previous study,5 DIA could be performed in 42 liver biopsies. One biopsy had to be excluded because it was fragmented and contained less than two fragments. Of the 42 biopsies, 33 were obtained by the transjugular route and nine were obtained percutaneously. The 42 biopsies were obtained from 42 patients with a median age of 45 years (range, 23-79), and 31 (74%) were male. Etiology of cirrhosis was hepatitis C in 21 (50%), alcohol in 10 (24%) (2 with hepatitis C), 5 (12%) each with autoimmune and cryptogenic cirrhosis, and 1 case of hepatitis B. Median HVPG in this study population was 11.4 mmHg (range, of 5-28). Twenty-four (57%) had an HVPG ≥10 mmHg (i.e., they had CSPH).

Liver Biopsy Characteristics on DIA.

Median total fibrosis area was 21.1% (range, 4.3-46.0), median septal width was 186 μm (range, 32-1,151), median nodule size was 720 μm (range, 236 – 1,846), and the median number of nodules per millimeter of liver biopsy was 0.84 (range, 0.36-1.92).

Correlation Between HVPG and DIA Parameters.

As shown in Table 1, on univariate analysis, the two parameters that significantly correlated with HVPG were fibrosis area (r = 0.571; P < 0.0001) (Fig. 2A) and nodule size (r = −0.415; p = 0.006). These two parameters were entered in the multivariable model, and, on linear regression, the only parameter that independently correlated with the HVPG was fibrosis area, with a regression coefficient of 0.606 (P < 0.0001). Remarkably, the significant correlation between HVPG and fibrosis area was maintained if only patients with CSPH were analyzed (r = 0.636; P < 0.005) (Fig. 2B).

Table 1. Correlation Between Histological Parameters and HVPG
Histological FeatureSpearman's Correlation CoefficientP Value
Fibrosis area (%)0.571<0.001
Nodules/mm0.1430.367
Septal width (μm)0.0890.577
Nodule size (μm)0.4150.006
Figure 2.

Correlation between HVPG and total fibrosis area in all biopsy slides (A) and in biopsies of patients with CSPH (B).

Correlation Between CSPH and Histological Parameters.

As shown in Table 2, on univariate analysis, fibrosis area, nodule size, and number of nodules were significantly different between patients with and without CSPH. Septal width was not significantly different between groups. Given an (expected) significant correlation between nodule size and number of nodules (r = −0.482; P = 0.001), only fibrosis area and nodule size were entered into the multivariable model. By backward logistic regression, both parameters remained in the model, that is, both fibrosis area and nodule size were independently predictive of the presence (or absence) of CSPH (r = 0.57; P = 0.003). Of 11 patients with a fibrosis area <15%, only 1 had CSPH. Conversely, except for 1, all 13 patients with a nodule size <6 mm had CSPH.

Table 2. Histological Parameters and the Presence or Absence of CSPH
Histological FeatureHVPG <10 mmHg (n = 18)HVPG >10 mmHg (n = 24)P Value
  1. Results are expressed in medians (ranges).

Fibrosis area (%)12.6 (4.3-42.3)27.0 (14.2-46.0)0.01
Nodule size (μ)829.2 (596.0-1845.8)602.4 (235.8-1561.5)0.001
Number of nodules/mm0.72 (0.36-1.25)0.95 (0.521-1.92)0.05
Septal width (μ)212 (32.3-975.3)179.6 (71.6-1151.4)0.65

Discussion

Liver biopsy remains the gold standard for staging diffuse liver disease, with the highest degree of severity being the cirrhotic stage. In our previous study, we had proposed that the cirrhotic stage could be further subclassified based on histological parameters that correlated with HVPG and the presence of CSPH,5 which is defined by an HVPG ≥10 mmHg, because it is a strong predictor of variceal development,1 clinical decompensation (e.g., ascites, encephalopathy, and variceal hemorrhage),2 and HCC.11

In our previous study, in which liver biopsies were evaluated semiquantitatively, small nodule size and thick fibrous septa were identified as independent predictors of HVPG and the presence of CSPH, leading to a position article putting forward a pathophysiological classification of cirrhosis.12 Our results were subsequently confirmed in a similar study performed in patients with cirrhosis predominantly caused by hepatitis B.13 In the present study, we partially confirm these findings using a quantitative method previously validated for the assessment of fibrosis area in chronic hepatitis C.6, 14 Small nodule size predicted CSPH, and even though septal width did not, total fibrosis area was the most important predictor of CSPH and was the only parameter that correlated with HVPG. Examples of liver biopsies from 2 patients with and without CSPH are shown in Fig. 3A,B, respectively.

Figure 3.

Liver biopsy of patients with (A) and without (B) CSPH (Masson stain). In (A), there are multiple small nodules (<1 mm) separated by wide fibrous septa, which are >1 mm thick. Fibrosis area was estimated at 46%, and HVPG was 28 mmHg. In (B), there are multiple large nodules (>2 mm) separated by thin fibrous septa, which are <1 mm thick. Fibrosis area was estimated at 10% and HVPG was 6 mmHg.

Semiquantitative analysis can be performed at the time of the histological evaluation of a liver biopsy, whereas quantitative fibrosis analysis requires additional resources and time and requires evaluating inter- and intraobserver reproducibility.

Based on both of our studies, we are able to compare and contrast the predictive value of each histological parameter by the method used and in the context of histological changes that occur in cirrhosis.

If the fibrotic process occurred uniformly, nodules would be perfectly rounded and distributed evenly throughout the liver. In this setting, increased fibrous tissue (and increased resistance in the liver) would be associated with increased total fibrosis area, increased septal width, smaller nodule size, and a larger number of nodules per unit volume of liver tissue (Fig. 4A). Conversely, with larger nodules, the area of fibrosis would be smaller, septae would be narrower, and there would be fewer nodules per unit of liver tissue (Fig. 4B). In these models, any of these parameters (i.e., fibrosis area, septal width, nodule size, and number of nodules) could be easily calculated by obtaining measurements of two other parameters. However, in reality, changes in liver histology are not uniform and nodules of different sizes and septa of different thicknesses are present in the same or different histological sections (Fig. 4C). Furthermore, because a needle biopsy specimen often represents <1/50,000 of the liver, this, in itself, could lead to a great potential for sampling error.15 Total fibrosis area is most reliably assessed by quantification using image analysis, whereas it cannot be evaluated reliably in liver biopsies by semiquantitative analysis. Conversely, assessment of septal width is most easily performed on a cursory low magnification examination (i.e., “bird's eye view”) of a liver biopsy, compared to measuring multiple points of the septae between two irregular nodules using image analysis. The situation is similar in a biopsy with very delicate thin fibrous septa, where measurement of septal thickness by DIA is difficult, but septal width is fairly obvious with a semiquantitative subjective analysis (Fig. 5A). This explains why, in our previous study, septal width was the parameter most predictive of HVPG and CSPH, whereas in the present study, it was fibrosis area that predicted these best.

Figure 4.

Three models of cirrhosis and theoretical biopsy slides from each model. (A) Small, perfectly rounded nodules are uniformly distributed throughout the liver, and fibrous septae are thick and fibrosis area is large. (B) Large nodules are uniformly distributed throughout the liver, and fibrous septae are thin and fibrous area is smaller. (C) Real model in which nodules are neither uniformly distributed nor are perfectly rounded. Consequently, septal width is also variable.

Figure 5.

Biopsies exemplifying problems confronted with quantitative analysis (trichrome stain). In (A), there are variably sized irregular nodules with thin, very delicate fibrous septa. The irregular outline of the nodules and the fibrosis pattern would be very difficult to assess quantitatively. The patient did not have CSPH. In (B), there are variably sized nodules with a very large nodule shown on the left edge and multiple small nodules with irregular outlines and thick fibrous septae. Though the semiquantitative analysis would conclude that nodules are small, size would be falsely overrepresented in a quantitative analysis by averaging the size of the large nodule. The patient had CSPH.

On the other hand, measurements of nodule size with image analysis, while more precise if the nodule evaluated is complete and rounded, is more difficult to evaluate when the nodule is irregular and/or incomplete, as often occurs (particularly in fragmented specimens). Subjective assessment of nodule size takes into account the predominant type of nodule (i.e., small or large) at a low magnification, whereas quantification of nodule size by image analysis would average different sizes and may overrepresent a nonpredominant type of nodule (Fig. 5B). This may explain why, in our previous study, nodule size correlated well with HVPG and was predictive of CSPH, whereas in the current study, nodule size was not an independent predictor of HVPG. Notably, in our previous study, we had arbitrarily defined small nodules as those smaller than 1 mm and this approximation was quite accurate, given that, in the present study, the median nodule size in patients with CSPH was 0.6 mm, whereas in those without CSPH, it was 0.8 mm. Therefore, though all these histological parameters in cirrhosis are conceptually important in assessing the structural remodeling of the liver, they may not be equally assessed by different methodologies. Depending on the method, different parameters may correlate better with HVPG. Though fibrosis area is most accurately assessed by quantitative image analysis (QIA), nodule size and septal width are better assessed by the subjective assessment of liver biopsies at low magnification. Thus, the use of QIA may not be currently practical in the substaging of cirrhotic biopsies, whereas this could be more easily achieved with some standardization of semiquantitative measurements of septal width and nodule size.

Beyond the clinical implications that this validation study puts forward (i.e., that specific histological findings in cirrhosis may be able to predict the likelihood of developing varices or clinical decompensation), this study shows that it is the total fibrosis area that is the most important determinant of PH. Our findings are not unexpected and are consistent with the pathophysiology of PH in cirrhosis, where an increased intrahepatic resistance plays a major role.16 The greater the fibrosis area, the greater the obstruction to portal flow and the higher the portal pressure, and this is shown by the positive correlation even among patients who already had clinically significant PH. Small nodule size is also indicative of greater architectural distortion (i.e., decreasing functional reserve of the liver) and would theoretically further increase intrahepatic resistance as well as indicate more severe disease. Although other factors, such as active constriction of intrahepatic vessels17 and an increase in portal venous inflow secondary to splanchnic vasodilatation,18 play an important role in the pathogenesis of PH, our findings support the important role that intrahepatic structural abnormalities play in the pathophysiology of PH. It may well be that the hyperdynamic splanchnic circulation becomes a more dominant factor in the maintenance of PH in cases of more severe cirrhosis and PH.19

A recent study also using DIA showed that the collagen proportionate area (similar to our fibrosis area) correlated with HVPG and was superior to the semiquantitative Ishak fibrosis system in patients with chronic hepatitis C.14 It is important to note that that study was different from ours because it included all stages of fibrosis (i.e., from no fibrosis to cirrhosis), whereas our study was restricted to only biopsies at the cirrhotic stage. In another study examining histomorphometry using an image analyzer in cirrhotic biopsies, the only significant (i.e., negative) correlation identified was between HVPG and the number of residual portal spaces (i.e., portal spaces not involved in the process of bridging fibrosis).20 However, evaluation of residual portal tracts is difficult, and our findings may be reflective of this process and more compatible with the pathophysiology of cirrhosis and PH. The issue of reversibility or regression of cirrhosis is an evolving and controversial concept,21 which has become more prominent with the development of new treatments for chronic liver disease and even antifibrotic therapy. It would be of great interest to identify histological patterns in cirrhosis that are more likely to reverse and to evaluate whether clinical and/or hemodynamic improvements in cirrhotic patients correlate with changes from a more severe to a less severe histological pattern.

Follow-up studies need to be undertaken in a larger population to validate the proposed subclassification of cirrhosis, particularly with regard to clinical outcome. Preliminary data in a separate patient population suggest that septal thickness in biopsies with cirrhosis predicts clinical decompensation.22 If confirmed, a modification of the staging system of biopsies with chronic liver disease could be proposed in which stage 4 (i.e., cirrhotic stage) could be further subclassified by severity based on area of fibrosis/septal width and nodule size.

In conclusion, we have validated our previous results performed using a semiquantitative analysis of histological cirrhosis with a quantitative method using DIA. The results indicate that fibrosis area is an important determinant of the degree of PH, and that both fibrosis area and nodule size are important histological predictors of clinically significant PH and are therefore predictive of the clinical severity of cirrhosis.

Ancillary