Introduction

Despite a decreasing need to perform liver biopsy in patients with hepatitis C (HCV) in order to stage fibrosis prior to treatment with direct-acting antiviral agents (DAA)1, an adequate assessment of liver fibrosis remains important2. The amount of fibrosis present influences surveillance strategies for hepatocellular carcinoma (HCC)3,4, as well as in counseling patients as to their ultimate risk for liver decompensation and the need for liver transplantation5,6. There has been much interest in developing noninvasive modalities, such as MRI elastography (MRE) and transient elastography (Fibroscan), due to the invasiveness of needle liver biopsy and the sampling error that can be associated with it1,7. Questions of there being geographic variation in fibrosis within the same liver also remain8. Similar to HCV, in patients having nonalcoholic steatohepatitis (NASH), knowing the degree of fibrosis is important in recommending different interventions9,10, prioritizing the need to participate in clinical trials, and assessing the risk of developing HCC.

In HCV patients, achieving a sustained virologic response (SVR) and cure post DAA treatment, the hope is that there will be a regression in liver fibrosis. Cohort studies have shown that there is improvement of clinical portal hypertension in a subset of patients with cirrhosis post-SVR11, as well as a short-term decrease in the risk of HCC development12,13. However, in many patients with underlying cirrhosis, there is still a slow progression and decompensation of their liver disease despite having achieved SVR6. Higher trans-jugular hepatic venous-wedge pressure measurements and an advanced degree of fibrosis on transient elastography are noted in patients less likely to have a regression in their fibrosis14.

Cirrhosis is one of the key factors in determining the clinical course and management of most if not all chronic liver diseases, including chronic viral hepatitis, NASH, and chronic cholestatic liver disease5. Contradictory to the traditional view in which cirrhosis is regarded as homogeneous and static, recent evidence has suggested that cirrhosis is a spatially heterogeneous and temporally dynamic process due to a constant remodeling process15. This remodeling process is regulated by clinical and genetic factors and provides us promising therapeutic avenues for medical intervention16,17,18,19.

The Laennec system, a modification of the METAVIR system, was first developed in 2000 for cirrhosis substaging20. In this system, cirrhosis is subdivided into 4A, 4B, and 4C based on the thickness of the fibrous septa and the size of nodules. Although the Laennec system has been proven useful in predicting the risk of late liver-related events in cirrhotic patients, it cannot be used to assess the potential of fibrosis progression or regression after treatment. More recently, the Beijing classification was developed by analyzing the cirrhosis content and morphology in patients with hepatitis B (HBV) before and after antiviral treatment21. In this system, cirrhosis is sub-classified into predominantly progressive (P), indeterminate (I), and predominantly regressive (R) patterns. Clinical data show that P or I cases have increased risk of persistently active disease when compared with R cases. Although the Beijing classification is the first cirrhosis sub-classification system that takes the dynamics of remodeling into consideration, it has only been tested in HBV so far, with its usefulness in many other chronic liver diseases remaining largely unknown. In addition, it is still unclear how well it correlates with the Laennec system and more importantly, how these two schemes compare with each other in terms of clinicopathologic correlation.

Recent advancement of digital pathology has made quantitative measurement of the collagen-proportionate area (CPA) by using computerized quantitative morphometry more feasible in the daily practice of pathologists22. Studies have shown that liver biopsies with greater CPAs are less likely to have fibrosis regression23. Until now, there is little published data on how CPA correlates with the cirrhosis sub-staging by Laennec system and Beijing classification, and how these three methodologies are impacted by geographic heterogeneity within the liver. In this study, we aim to compare these three systems and determine how clinical parameters may be impacted by geographic heterogeneity within the same liver.

Methods

Patient samples

As part of a study initially designed to correlate the degree of fibrosis and HCV viral load within liver tissue, we performed five core biopsies in different geographic regions of cirrhotic livers, immediately upon explantation in the operating room at the time of liver transplantation. We examined these biopsy specimens in order to establish whether there is indeed geographic variability in the degree of fibrosis within the same cirrhotic liver, and then correlated the CPA with the Laennec stage and Beijing classification of cirrhosis, as well as with clinical parameters of the patients.

As part of an IRB-approved protocol, 20 consecutive HCV patients undergoing liver transplantation were consented to participate. A core biopsy was taken from segments 8, 6, 4, 2, and 1 (five biopsies in total) of the explanted liver in the operating room immediately after explantation. Needle biopsies were performed (by TS) under direct visualization using a 14-gauge (width 0.2 cm) Jamshidi needle with specimens placed immediately into formalin. Biopsies were always performed as far away as possible from obvious HCC nodules. Clinical data and HCV treatment status were obtained from electronic medical record review.

Histology and morphometric analysis

After proper fixation, the biopsy specimens were routinely processed and embedded in paraffin (FFPE). Four-micron-thick sections were performed from FFPE with two sections stained with hematoxylin and eosin, one section stained with Masson trichrome. The slide stained with Masson trichrome was then digitalized by whole slide scanning at 20X using Panoramic 250 Flash III instrument (3D-Histech). Digital images were analyzed with Halo morphometry software (v.1.4, Indica labs) with Area Quantification Module to calculate CPA (Fig. 1). First, the entire biopsy tissue area in the image was precisely delineated by using a trackpad to measure the total biopsy tissue area. For each image, representative fibrotic areas, which were stained dark blue by trichrome stain, were selected and used to train the machine-learning algorithm to generate a fibrotic tissue classifier for each case. In order to avoid overestimation or underestimation, the generated fibrotic classifier was optimized by applying the real-time validation tool to different areas in the biopsy tissue in order to most accurately capture all fibrosis and eliminate artifacts at the same time. Next, the fibrotic tissue classifier was validated by running an entire biopsy specimen several times which was then compared with the original trichrome stained image. Last, the finalized fibrotic tissue classifier was used to analyze the entire image in a pixel-by-pixel manner in order to quantify the entire area of fibrosis for each individual biopsy. The CPA was calculated as the percentage of fibrosis area in the entire biopsy tissue area (Fig. 1). In order to avoid variability due to processing and staining intensity, each image was individually analyzed to generate its own specific fibrotic tissue classifier. Sub-capsular and perivascular fibrous tissue were manually corrected for.

Fig. 1: Collagen-proportionate area (CPA) measurement by computerized morphometrical quantification.
figure 1

Trichrome-stained slides are digitized by whole slide scanning at 20X magnification. Then, the entire biopsy tissue area in the image was precisely delineated by using a trackpad to measure the total biopsy tissue area using the Area Quantification function of HALO software (v1.4). Next, a fibrosis tissue classifier is generated in HALO v1.4 by analyzing the color spectrum of the dark-blue-stained fibrosis area on the trichrome-stained slide. After validation, a pixel-by-pixel analysis using this fibrosis tissue classifier is performed in order to label all of the fibrosis area (highlighted by pseudocolor yellow). The collagen proportionate area (CPA) is calculated as the percentage of fibrosis area in the total biopsy area for each slide.

Cirrhosis subclassification by Laennec system and Beijing classification

Cirrhosis substaging by either Laennec system or Beijing classification was assessed by an experienced liver pathologist (MIF) according to the previously published criteria20,21. Laennec 4A is defined as marked septation with rounded contours or visible nodules, but most septa are thin (one broad septum allowed). Laennec 4B is defined as at least two broad septa, but not very broad septa and that less than half of the biopsy length should be composed of minute nodules. Laennec cirrhosis 4C is defined as at least one very broad septum or more than half of the biopsy length being composed of minute nodules.

For cirrhosis substaging by Beijing classification, the P group is defined as wide and broad loosely aggregated collagen in >50% of the liver biopsy, with a mixture of light and dense dark-blue staining fibers on trichrome stain. The R group is defined as thin and densely compacted fibrous septa in >50% of the liver biopsy, trichrome stain showing predominantly densely stained dark blue. The I group is defined as a combination of P and R features. The current study is the first to apply the Beijing classification in assessing cirrhosis in HCV patients.

For the comparison and correlation among different methodologies for cirrhosis subclassifications, the overall CPA used for the entire explanted liver was determined by averaging the CPA score of all five biopsies, and then was correlated with the Laennec system and Beijing system cirrhosis subclassifications.

Statistical analysis

CPA, Laennec stages, and Beijing classification were compared and then correlated with the clinical variables of the patients. Continuous variables were compared using the Student t-test. The significance threshold was set at 0.05. The coefficient of variation (CoV) was calculated as the ratio of the standard deviation (SD) to the mean of the five CPAs measured from the same explanted liver. The results are presented as means ± S.E., unless otherwise specified. All statistical tests were performed with GraphPad Version 8 (GraphPad Software, La Jolla, CA).

Results

The demographic and clinical data of the 20 patients in our study cohort are listed in Table 1. The average age at the time of liver transplantation was 61.8 ± 1.4 years. There were 16 male patients (average age 63.5 ± 1.5 years) and 4 female patients (average age 63.5 ± 4.4 years). The average total explanted liver weight was 1180.3 ± 83.6 g. At the time of transplantation, the average platelet count was 74 ± 7.3 × 109/L, the average total bilirubin level 5.5 ± 1.7 mg/dL, the average albumin level 3.3 ± 0.15 g/dL, and the average natural MELD score 22.2 ± 2.6. Three of these 20 patients received DAA treatment and achieved SVR, and they tended to have lower MELD scores than the non-SVR group (10.7 ± 0.9, n = 3; vs. 24.2 ± 2.7, n = 17; p = 0.056).

Table 1 Patient demographics, clinical and laboratory profile, liver explant Beijing and Laennec classification of cirrhosis and corresponding collagen proportionate area.

In our cohort, 15 patients (12 males and 3 females) had hepatocellular carcinoma in their explanted livers, including three patients which had already achieved SVR after receiving DAA treatment. The average age is similar to the patients without HCC (64.4 ± 2.2 vs. 60.9 ± 1.7 years, p = 0.28). When compared with patients without HCC, the patients with HCC have significantly lower MELD score (18.4 ± 2.3 vs. 33.4 ± 5.1, p = 0.006) and total bilirubin (18.4 ± 2.3 vs. 33.4 ± 5.1, p = 0.004) at the time of transplantation. There is no significant difference in the CPA (29.5 ± 7.5 vs. 25.4 ± 3.0, p = 0.54), platelet count (54.5 ± 11.7 vs. 80.5 ± 8.1, p = 0.12), albumin level (3.5 ± 0.4 vs. 3.3 ± 0.1, p = 0.57), and liver weight (1058 ± 155 vs. 1221 ± 961, p = 0.40).

Cirrhosis sub-stages by the Laennec system and Beijing classification correlate with CPA

The average CPA of biopsies with Laennec 4C cirrhosis was 37.2 ± 3.8% (n = 28), which was significantly higher than the CPA of 4B (21.5 ± 2.2%, n = 49, p = 0.0002) or 4A (16.5 ± 2.6%, n = 23, p = 0.00008) (Fig. 2). Similarly, when using the Beijing classification, the average CPA of biopsies showing the P pattern was 28.8 ± 2.3% (n = 68), significantly higher than those biopsies showing the R pattern (12.6% ± 1.6%, n = 14, p = 0.002), and I pattern (18.4% ± 4.2%, n = 18; p = 0.037) (Fig. 3).

Fig. 2: The CPA correlates significantly with Laennec stages.
figure 2

A–C Representative images of Trichrome stain of Laennec stage 4A, 4B, and 4C. D The CPA in Laennec 4C (37.2 ± 3.8%, n = 28) is significantly higher than both Laennec 4B (21.5 ± 2.2%, n = 49, p = 0.0002) and 4A (16.5 ± 2.6%, n = 23, p = 0.00008). (***p < 0.001; ****p < 0.0001).

Fig. 3: The CPA correlates with the fibrosis grade according to the Beijing classification.
figure 3

AC Representative images of Trichrome stain of predominantly regressive (R), indeterminate (I), and predominantly progressive (P). D The CPA in the predominantly progressive group (28.8% ± 2.3%, n = 68) is significantly higher than the predominantly regressive group (12.6% ± 1.6%, n = 14, p = 0.002) and indeterminate group (18.4% ± 4.2%, n = 18; p = 0.037) (*p < 0.05; **p < 0.01).

Geographic variation and fibrosis remodeling impact cirrhosis sub-stages by Laennec system and Beijing classification

In this cohort, the mean coefficient of variation (CoV) for the CPAs of the five biopsies taken from different lobes of the same explanted liver was 47.3 ± 4.5% (n = 20, range 14.2–97.5%). This suggests a significant degree of geographic heterogeneity.

When cirrhosis was substaged using the Laennec system, only five of twenty cases showed the same Laennec substages among all five biopsies taken from different segments (two cases all with Laennec 4C, one with Laennec 4B, and two all with Laennec 4A). The remaining 15 cases showed two to three different Laennec substages (Table 1). In the current study, when there were different Laennec substages in the different lobes, the highest Laennec substage was designated as the overall final Laennec substage for the case. By these criteria, there were 14 Laennec 4C cases, 4 Laennec 4B cases, and 2 Laennec 4A cases (Table 1).

Using the Beijing classification, seven of twenty cases showed the same fibrosis pattern among all five biopsies, all having P pattern. The remaining 13 cases demonstrated heterogeneous fibrosis patterns among the different segments. When there was heterogeneity present, the overall pattern was determined by the most predominant pattern (i.e., the pattern present in ≥3 out of the five cores). If there was no particular pattern present in ≥3 biopsies, the overall final Beijing classification pattern is considered as indeterminate (I). Using these criteria, there were 15 cases with P pattern, 2 cases with R pattern, and 3 cases with I pattern.

The two cases with the R pattern were both Laennec 4A cases, and both cases showed a uniform R pattern among all five biopsies. Among the three cases with the I pattern, two were Laennec 4B and one was Laennec 4A. Among the fifteen P-pattern cases, there were 12 Laennec 4C and 3 Laennec 4B. There was no case with a P pattern that showed Laennec 4A. Similarly, 12 of 14 Laennec 4C cases showed P pattern, with the remaining two cases showing I pattern. There was no case with Laennec 4C fibrosis, which showed an R pattern. The Laennec 4B group and the group with I pattern showed the least amount of overlap, with only one case overlapping (Fig. 4). These results suggest that the R and P patterns highly correspond with Laennec 4A and 4C, respectively.

Fig. 4: The effect of geographic variation and fibrosis remodeling on cirrhosis substaging using the Laennec grading system and the Beijing classification.
figure 4

A Representative images of the five biopsies from different segments of one same explanted liver showing different Laennec substages and Beijing classification patterns. B The violin plot of the distribution of the CPAs of five biopsies taken from the five different segments in all 20 explanted livers (central line: median; upper and lower line: interquartile range; outer shape: distribution), showing the significant degree of geographic variation of fibrosis content. Cases with the Laennec 4C cirrhosis and P pattern tend to have higher degrees of variation. The Laennec 4A and 4C cirrhosis correspond relatively well with the R and P patterns, but there is less overlap between Laennec 4B and I pattern.

The CoV in the Laennec 4A group was 34.3 ± 2.5%, lower than Laennec 4C (54.0 ± 15.8%, p = 0.36) and 4B (47.3 ± 5.0%, p = 0.45) but not statistically significant. Similarly, the CoV in the R pattern was 34.3 ± 2.5%, lower than in the P pattern (50.7 ± 5.8%, p = 0.33) and I pattern (39.1 ± 6.8%, p = 0.62) but not statistically significant (Fig. 4). There was no significant difference in the average CPA among the biopsies taken from the right lobe (segments 8 and 6) vs. the left lobe (segments 4 and 2) and even the caudate lobe (segment 1).

Clinicopathologic correlation with the Laennec system and the Beijing classification

The platelet count in the Laennec 4C group (68.0 ± 7.9, n = 14) was significantly lower than the Laennec 4A group (122.0 ± 24.0, n = 2, p = 0.036). There was no significant difference in platelet count when Laennec 4C was compared with the Laennec 4B group (71.3 ± 8.7, n = 4, p = 0.7). Statistically, significant differences were not observed among Laennec 4C vs. 4A and 4B when comparing other clinical parameters, such as MELD score at the time of transplantation (23.7 ± 3.1 in Laennec 4C vs. 20.5 ± 9.5 in Laennec 4A, and 17.5 ± 5.6 in Laennec 4B), total explanted liver weight (1198.6 ± 97.5 g in Laennec 4C vs. 1429.0 ± 171.0 in Laennec 4A, and 991.5 ± 194.4 in Laennec 4B), total bilirubin (6.2 ± 2.0 mg/mL in Laennec 4C vs. 1.2 ± 0.6 mg/mL in Laennec 4A, and 5.5 ± 3.8 mg/mL in Laennec 4B), and albumin (3.3 ± 0.2 g/dL in Laennec 4C vs. 3.6 ± 0.3 g/dL in Laennec 4A, and 3.4 ± 0.4 g/dL in Laennec 4B).

When using the Beijing classification, platelet count in P pattern group (62.4 ± 6.6, n = 15) was significantly lower than both the R pattern (122.0 ± 24.0, n = 2, p = 0.008) and I pattern groups (100.3 ± 6.0, n = 3, p = 0.024, Fig. 5). Statistically, significant differences were not observed among these three patterns when comparing other clinical parameters, such as MELD score at the time of transplantation (23.1 ± 3.0 in the P group, vs. 20.5 ± 9.5 in the R group, and 18.7 ± 7.3 in the I group), total explanted liver weight (1117.8 ± 304.9 g in the P group vs. 1429.0 ± 171.0 in the R group, and 1326.7 ± 304.9 g in the I group), total bilirubin (6.1 ± 1.9 mg/mL in the P group vs. 1.2 ± 0.6 mg/mL in the R group, and 5.5 ± 5.0 mg/mL in the I group), and albumin (3.2 ± 0.2 g/dL in the P group vs. 3.6 ± 0.3 g/dL in the R group, and 3.7 ± 0.3 g/dL in the I group, Fig. 5).

Fig. 5: The correlation between the Beijing classification and clinical parameters.
figure 5

A P pattern is associated with significantly decreased platelet count than both R and I patterns, B the trend of decreased total explanted liver weight. C R pattern is associated with the trend of lower total bilirubin levels. D There are no significant differences in MELD score. E albumin level among these three patterns (*p < 0.05; **p < 0.01).

Three of twenty cases in our cohort achieved SVR after receiving DAA treatment. There was a trend that SVR patients tend to have lower CPA than non-SVR patients (12.0 ± 3.6% vs. 27.0 ± 3.4%, p = 0.0866, Fig. 6). Among these SVR patients, there were two patients who had P pattern (one case shows uniform P pattern among all five biopsies, and the other case showed P pattern in four of five biopsies) and one with R pattern (R pattern in four of five biopsies). When substaged by the Laennec grading system, one SVR patient showed Laennec 4A cirrhosis in all five biopsies taken from different lobes (four of which were predominantly R pattern and one I pattern). The second SVR patient was substaged as Laennec 4C (two Laennec 4C biopsies, two Laennec 4B biopsies, and one Laennec 4A biopsy, all of which were predominantly P pattern). The third SVR patient was Laennec 4B (Laennec 4B in all five biopsies, among which three were P pattern, one R pattern, and one I pattern). Fig. 7

Fig. 6: The correlation between the Laennec grading system and clinical parameters.
figure 6

A The Laennec 4C cirrhosis group is associated with significantly decreased platelet count than Laennec 4A group. B There are no significant differences in the total explanted liver weight. C Total bilirubin level. D The MELD score at the time of explantation. E Albumin level among the Laennec 4A, 4B, and 4C cirrhosis.

Fig. 7: Sustained virologic response (SVR) tends to have lower collagen-proportionate area (CPA) values.
figure 7

A, B Representative trichrome stain images of a biopsy taken from a non-SVR patient and a SVR patient. C The SVR patient group shows the trend of having lower CPA than non-SVR patient group (12.0 ± 3.6% vs. 27.0 ± 3.4%, p = 0.0866).

When analyzing the correlation between CPA and clinical factors, we found that patients with MELD scores >15 showed a trend toward having higher CPA than those patients with MELD score <15 (27.4 ± 3.3%, n = 7 vs. 16.5 ± 4.0%, n = 13, p = 0.0617). There was no other significant association observed between CPA and other clinical parameters.

Discussion

Although most patients with HCV are now cured with DAA therapy, many questions remain to be answered regarding how to manage these patients. First and foremost, the question remains whether regression of fibrosis correlates with an overall decreased risk of liver decompensation and development of HCC1. For patients who had an improvement in their portal hypertensive complications after DAA treatment and SVR, it is also important to know if future liver transplantation can be abrogated. Current noninvasive modalities such as MRE and transient elastography have shown inconsistent results in evaluating the overall degree of fibrosis present and fall short in their ability to assess the potential for fibrosis regression vs. progression1. Being able to obtain additional information from baseline or future liver biopsies in such patients is extremely helpful in assisting clinicians to determine the best management and surveillance plan.

The current study demonstrates that on needle liver biopsy, the CPA correlates significantly with the histologic substages of cirrhosis using both the Laennec system and the Beijing classification. We found that there was a much greater amount of fibrosis measured by CPA in Laennec 4C cirrhosis and the Beijing P pattern with less interpatient variation, thus suggesting less potential for fibrosis regression.

The presence of more advanced Laennec stage, such as Laennec 4C or the Beijing classification’s P pattern, suggests that it is unlikely that there will be significant regression in the future and would thus highlight patients that need close follow-up for clinical decompensation and probable lifelong surveillance for HCC. We noted a trend in lower overall CPA in patients who had already achieved SVR, indicating that fibrosis regression may be possible after SVR. There was also a trend in higher CPA and Laennec stage in patients who had higher natural MELD scores, and thus in patients having more advanced liver disease.

The current study confirms that significant geographic heterogeneity in fibrosis is present throughout the liver and confirms that sampling error thus exists within the liver. This variation was present within and among the different segments of the liver.

The strength of our study is that we were able to obtain multiple liver biopsies from the same cirrhotic liver under direct visualization immediately after explantation, which negated any processing artifact. One limitation is that we did not have contemporaneous transjugular hepatic venous wedge-pressure measurements or transient elastography results to correlate with the histologic findings. However, in light of the geographic variation of fibrosis that we demonstrated, some inaccuracy in transient elastography results is to be expected, which potentially may underestimate the true degree of fibrosis. In addition, we were unable to account for the effects of locoregional therapies on the fibrosis present in the biopsy specimens, but we do not think that this would affect the correlations noted between the CPA, Laennec, and Beijing stages. We would anticipate that there is less inaccuracy with transjugular hepatic venous wedge-pressure measurements, although there could be sampling error with regard to the CPA, Laennec stage, and Beijing classification of fibrosis of the specimen obtained. Since all of the patients in the current study had HCV, our results may not be applicable to patients having NASH or other etiologies of liver disease. Current published data suggest that NASH patients having clinical portal hypertension and thus more advanced liver disease may not readily regress their fibrosis24, similar to the results shown in the current study. Thus, the performance of a similar study in patients having NASH might yield equally important prognostic data from liver histology. A recent study of 91 patients undergoing bariatric surgery who underwent needle liver biopsy of both the right and left lobes of the liver showed poor correlation in fibrosis stage, which is confirmed by the geographic variation noted in the current study25.

In conclusion, the current study demonstrates that CPA correlates significantly with the histologic subclassifications of cirrhosis using the Laennec system and the Beijing classification. Heterogeneous degrees of fibrosis exist diffusely throughout the liver, even within the same lobe. There is a much greater amount of fibrosis measured by CPA in Laennec 4 C cirrhosis and the P pattern of the Beijing classification with less variation, suggesting less potential for fibrosis regression. Performing CPA or assessing Laennec stage or applying the Beijing classification to previous or future liver biopsies in patients with HCV may yield important prognostic information with regard to their potential for fibrosis regression or risks of liver decompensation. Platelet count appears to be a good surrogate marker for the degree of underlying fibrosis.