A Novel Pathological Scoring System for Hepatic Cirrhosis with Hepatocellular Carcinoma

Introduction

In more than 90% of cases, HCC develops within the background of established cirrhosis, and cirrhosis is the strongest predisposing factor for HCC.^1,2 Cirrhosis plays a vital role in determining and implementing an appropriate surgical strategy for HCC treatment; in addition, the severity of cirrhosis is closely associated with occurrence of liver-related events after surgery for HCC.³ Cirrhosis decreases the functional reserve of the remaining liver, as well as the recovery of liver function.⁴ Several portal-based staging systems have been used or proposed for grading of cirrhosis, including Histology Activity Index (HAI), Scheuer scoring system, histological semi-quantitative scoring system (SSS), Batts-Ludwig system, Metavir system, and Laennec system, etc.^5–12 Most of above staging systems have focused on grading hepatic fibrosis and inflammation, including only early cirrhosis, while in fact, a special histopathological sub-classification of cirrhosis is urgently needed in hepatic surgery based on the severity of cirrhosis.⁵ The Laennec scoring system mainly according to the thickness of the predominant type of septa graded cirrhosis into 3 stages, which were purely morphology-based. Obviously, it is not easy to evaluate the percentage of liver parenchymal loss and the thickness of septa on small biopsy tissues. In the meantime, grading of resection specimens from cirrhotic patients with HCC has not been routinely used and verified. Qualitative or semi-quantitative assessments have a large subjective error; therefore, morphometry and image analysis applied to liver sections have been considered to be the gold standard because the intra assay variability is very low.¹² Those scoring systems are subject to the valid criticism that the stages represented by numbers do not accurately reflect equal units in terms of severity.⁸ Therefore, based on morphometry and image analysis, we established a new quantitative scoring system, which is specifically applicable to patients of HCC with cirrhosis.

Inflammation and fibrosis can be used to assess the degree of cirrhosis. Portal inflammation is associated with liver injury and fibrogenic activity, and it is also strongly associated with regenerative activity at the portal-septal/lobular interface and with myofibroblast activity.¹³ The extent of fibrosis is considered to be indicative of the stage of cirrhosis. MRP14 is a small calcium-binding protein with several immunological functions mainly involved in chronic inflammation¹⁴ and has been proposed as one of the central inflammatory regulators capable of driving and responding to inflammation signals.¹⁵ On one hand, MRP14 protein levels are significantly increased in response to cytokine stimulation, while on the other hand, it enhances inflammatory cascades by inducing leukocyte chemoattractants and the expression of pro-inflammatory cytokines.¹⁵ MRP14 is also involved in fibrosis remodeling,¹⁴ induced in response to liver injury, and acts as a powerful leukocyte chemoattractant.¹⁶

In this study, the pathological indexes and protein markers of inflammation and fibrosis were systematically screened, 6 parameters were finally selected for analysis. Average number of inflammatory cells in the portal area of the liver (NICP), proportion of hepatic fibrosis area (RFA), the density of the  pseudolobules (DPL), MRP14 positive cell density(MRP14), the density of inflammatory cells in hepatic lobular area(DICH), average diameter of pseudolobules (ADP)  were selected. These 6 parameters were quantitatively measured by  the Image-Pro Plus v6.0 software. Diagnostic models were established by logistic regression analyses and externally validated in an independent testing group. In addition, we also evaluated the prognostic value of the new system and established the nomogram to predict overall survival in HCC patients with cirrhosis after liver resection.

Patients and Methods

Patients

We reviewed clinicopathological characteristics of 260 patients who underwent surgical resection and were diagnosed pathologically with HCC with cirrhosis at the Eastern Hepatobiliary Surgery Hospital from 2010 to 2015. Institutional review board approval and written informed consent from each patient were obtained. Hematoxylin-eosin (HE) stained slides, Masson-stained slides, and MRP14 immunohistochemically stained slides were prepared from formalin-fixed paraffin-embedded tissues of liver and reviewed by two experienced hepatopathologists (WM Cong and H Dong). Patients were randomly assigned to training group and testing group. We analyzed a training group consisting of 163 patients who underwent curative resection and classified them as Child–Pugh A (n=75), Child–Pugh B (n=40), or Child-Pugh C (n=48) according to Child–Pugh stage after 3 months of surgery. Additional 97 patients classified them as Child–Pugh A (n=47), Child–Pugh B (n=23), or Child–Pugh C (n=27) formed the testing group used to assess the validity of the model. Child–Pugh score 3 months after surgery was used as an indicator of liver function recovery. The morphology index and molecular markers of cirrhosis were chosen through systematic studies of histological characteristics of cirrhosis, cited references, and statistical analysis. We examined tissue area more than one centimeter away from the tumor, which would be more representative of the rest of the liver. The inclusion and exclusion criteria for patients and parameters in the current study were as follows: 1) preoperative liver function was Child–Pugh A and diagnosis of HCC with hepatitis B-induced cirrhosis; 2) there was no pre-surgical treatment for HCC including chemoembolization, radiofrequency ablation, or high-intensity focused ultrasound, which would confound the morphology of background liver; 3) cirrhosis was diagnosed according to the character of nodular regeneration surrounded by extensive septa, followed by parenchymal extinction and collapse of the liver structure;¹⁷ 4) the selected parameters should be related to the liver function of postoperative patients. Each parameter was established from morphometric measurements. Liver samples from various stages of cirrhosis were scored as an illustration in Figure S1.

Follow-Up

The patients were followed up once every 2–3 months during the first year after surgery and then every 3–6 months thereafter, tumor markers and liver function assessments were performed at each follow-up visit. The overall survival (OS) was defined as the length of time between surgery and death or the last follow-up examination.

H&E, Masson’s Trichrome, Immunohistochemistry, and Parameter Score

The resected specimens were embedded in paraffin, and 4-μm-thick sections were cut and stained with hematoxylin and eosin (HE) and Masson’s trichrome for routine light microscopy examination.^18,19 Immunohistochemical slices were stained with an automated immunostainer (Leica BOND-MAXTM, Darmstadt, Germany). The antibodies applied in our study were purchased from Changdao (Shanghai, China) or Maixin Biotech (Fuzhou, China) and were abbreviated and diluted, rabbit polyclonal antibody against calgranulin B (S100A9/MRP14) (sc-20173; Santa Cruz Biotechnology; 1/2500 dilution).²⁰ H&E staining was used to measure NICP and DICH; Masson staining was used to measure DPL, RFA, and ADP; and immunohistochemistry was used to measure the density of MRP14-positive cells.

The imaging system consisted of a Leica FC420 CCD camera connected to a Leica DM IRE2 microscope (Leica Microsystems Imaging Solutions, Cambridge, UK). Photographs of four fields were captured randomly from each slide under high-power magnification (100×) when measuring DICH and MRP14. All pseudolobules were captured under high-power magnification (100×) when measuring NICP. The entire area was captured from each slide under medium-power magnification (40×) when measuring RFA, and under low-power magnification (20×) when measuring DPL and ADP. All photographs were captured using Leica QW in Plus V3 software, the number and area for all photographs were counted and measured using Image-Pro Plus v6.0 software (Media Cybernetics, Bethesda, MD) for the three parameters of area sum, area, and sample. The number of inflammatory cells, MRP14 positive cells and fibrosis area were automatically measured by Iimage-Pro Plus's calibration measurement tool. The fibrotic region we selected was aniline blue staining region, the portal and inflammatory areas were extracted manually.

Statistical Analysis

Statistical analyses were carried out with IBM SPSS Statistics 22.0 software (IBM, Chicago, IL) and GraphPad Prism 5.01. Pearson’s correlation and Spearman correlation test were used to analyze the relationship among the pathological parameters and different system score and stage. Differences between the measured values of each group were analyzed by Mann–Whitney test for each pathological parameter. In the training group, the measured values were used to perform receiver operating characteristic (ROC) analysis. The ROC curve showed sensitivity plotted against 1–specificity for each cut-off value and was used to determine the best cut-off of each parameter based on optimal sensitivity and specificity.²¹ Multivariate associations with the diagnosis of cirrhosis were assessed by ordinal logistic regression models using generalized estimating equations with robust variance estimation and exchangeable correlation to account for correlations. The odds ratio (OR) was used as a coefficient of each parameter, the parameter grade minus one as the exponent of the coefficient, and the aggregate risk score was generated for each patient by summing the estimates of each factor.²² Additionally, separate two-way analysis of covariance (ANCOVA) were performed using intraoperative blood loss, total clamp time, and resected liver diameter as covariates and pathological parameters of the system as dependent variables.²³ Survival curves were depicted using the Kaplan–Meier method and compared using the Log-rank test. Cox regression analysis was used for multivariate analyses.

A nomogram was formulated based on the results of multivariate analysis and by using the package of rms26 in R version 3.3.3 (http://www.r-project.org/). A final model selection was performed by a backward stepdown selection process with the Akaike information criterion. All statistical tests were two-sided and P values less than 0.05 were considered statistically significant.

Results

Results of Pathological Indicators Measurement

Six criteria compliant indicators were selected, and we analyzed the correlation between the pathological indicators and the Child–Pugh score. In the training group, Child–Pugh score showed a statistically significant association with NICP (r=0.7247, P<0.001), RFA (r=0.7182, P<0.001), DPL (r=0.5617, P<0.001), MRP14 (r=0.6013, P<0.001), DICH (r=0.6484, P<0.001), and ADP (r=−0.3044, P<0.001). NICP, RFA, DPL, MRP14, and DICH were higher in high Child–Pugh stage than in low Child–Pugh stage, ADP was lower in high Child–Pugh stage than in low Child–Pugh stage (Figure S1).

Diagnostic Value and Histological Score

The histological score of training group is detailed in Table S1. Elements of system were classified on a scale of 3 grades as follows: NICP (A≤1102, 1102>B≤2519, C>2519); RFA (A ≤ 0.1192, 0.1192>B≤0.2040, C>0.2040); DPL (A≤ 13.13, 13.13>B≤23.67, C>23.67); density of MRP14 (A ≤113.4, 113.4>B≤185.5, C> 185.5), DICH (A ≤719.9, 719.9>B≤1500, C >1500.0); and ADP (A≥1764, 1764>B≥1213, C<1213).

The histological score is detailed in Table S1. Predictive effect was represented by area under the curve (AUC) of ROC curves. For Child–Pugh A/B the AUC was 0.8877 for NICP, 0.8820 for RFA, 0.9272 for DPL, 0.9272 for MRP14, 0.9500 for DICH, and 0.6107 for ADP. For Child–Pugh B/C the AUC was 0.8396 for NICP, 0.8135 for RFA, 0.8396 for DPL, 0.7451 for MRP14, 0.7266 for DICH, and 0.6209 for ADP (Figure S1). In addition, the best cut-off value was determined by the ROC curve. The sensitivity and specificity of each parameter for detection of liver function are summarized in Table S2. A high sensitivity (86.05%) coupled to low specificity (41.33%) for discriminating Child–Pugh A from Child–Pugh B was seen for ADP. A low sensitivity (60.42%) coupled to high specificity (97.50%) for discriminating Child–Pugh B from Child–Pugh C was seen for DICH. Both low sensitivity (52.08%) and low specificity (69.77%) for discriminating Child–Pugh B from Child–Pugh C were seen for ADP.

Construction of a Diagnostic Model

In univariable analysis, all of 6 indicators showed a statistically significant association with recovery of liver function, but ordinal logistic regression analysis showed that only NICP (P=0.011, OR=3.27), RFA (P=0.019, OR=3.07), DPL (P=0.032, OR=2.31), and MRP14 (P=0.001, OR=3.32) were associated with recovery of liver function (Table S3). NICP, RFA, DPL, and MRP14 were independent prognostic factors and were therefore used as the components of the cirrhosis scoring system. Multicollinearity analyses showed all of 4 parameters NICP (P=0.002), RFA (P<0.001), DPL (P=0.002), and MRP14 (P<0.001) in the model have statistical significance (Table S7). Covariance analyses showed no significant differences between the coefficients of independent prognostic factors for the cirrhosis model for different levels of intraoperative blood loss, total clamp time, and resected liver diameter (Table S4). The sensitivity and specificity of the cirrhosis score system were 100.00% and 82.67% (cut-off value of 6.85) for Child-A/Child-B group, 81.25% and 85.00% (cut-off value of 24.15) for Child-B/Child-C group (Table S2).

Model Evaluation in the Training Group

As shown in Figure S2, in the training group the Child–Pugh score correlated with new scoring system and other combinations of system factors, but the correlation coefficient of the system (r=0.8227, P<0.001) was higher than other combinations RFA+DPL+MRP14 (r=0.8167, P<0.001), NICP+DPL+MRP14 (r=0.8025, P<0.001), NICP+RFA+MRP14 (r=0.8096, P<0.001), NICP+RFA+DPL (r=0.8088, P<0.001); NICP+RFA (r=0.7922, P<0.001); NICP+DPL (r=0.7854, P<0.001), NICP+MRP14 (r=0.7705, P<0.001); RFA+DPL (r=0.7854, P<0.001), RFA+MRP14 (r=0.8001, P<0.001), DPL+MRP14 (r=0.7754, P<0.001). As shown in Figure S3, the AUC for Child–Pugh A/B and Child–Pugh B/C for cirrhosis scoring system (0.9500 and 0.8852) was higher than other combinations of system elements in the training group.

Cirrhosis Stage Scoring System

The standard cirrhotic scoring system was defined according to the following equation (Table 1):

Table 1 The New Cirrhosis Scoring System

Cirrhosis Score=3.27^G(NICP)-1+3.07^G(RFA)-1+2.31^G(DPL)-1+3.32^G(MRP1)-14

Cirrhosis Stage was classified as follows: I≤6.85, 6.85>II≤24.15, III>24.15. The pathological features of each cirrhotic are shown in Figure 1.

Figure 1 Four histological items for evaluation of cirrhosis system. H&E stain shows NICP of Cirrhosis Stage 1 (A), Cirrhosis Stage 2 (B), Cirrhosis Stage 3 (C) (F); Masson stain shows RFA of Cirrhosis Stage 1 (D), Cirrhosis Stage 2 (E), Cirrhosis Stage 3 (F), DPL of Cirrhosis Stage 1 (G), Cirrhosis Stage 2 (H), Cirrhosis Stage 3 (I); MRP14 stain shows of Cirrhosis Stage 1 (J), Cirrhosis Stage 2 (K), Cirrhosis Stage 3 (L). Black boxes, representative views of tissue; green boxes, enlarged representative views.

Parameters and Model Evaluation in the Testing Group

For the testing group, the clinicopathological features were summarized in Table S5 and the different systems’ scores were summarized in Table S6. As shown in Figure S6, the component factors of cirrhosis scoring system have good correlation with the serum indexes.

The correlation between the new cirrhosis system score and TIBL (P<0.001), ALT (P=0.008), AST (P<0.001), ALP (P<0.001), ALB (P=0.003), PT (P<0.001), Cre (P<0.001), ascitic score (P<0.001), and encephalopathy score (P<0.001) showed statistical significance. The Laennec score system was only correlated with PT (P=0.019) (Figure S4, Table S8).

There was a very close relationship between the stage of cirrhosis system and Child–Pugh stage (r= 0.8048, P<0.001), as well as correlation between the cirrhosis system score and Child–Pugh score (r=0.8058, P<0.001) and MELD score (r =0.666, P<0.001) 3 months after surgery, with higher correlation coefficient than any of the four independent variables. There was a low correlation between the Laennec score and Child–Pugh scores (r=0.2779, P=0.0059), MELD scores (r=0.3271, P=0.0011) and between the Laennec stage and Child–Pugh stage (r=0.1872, P=0.0663) (Figure S5, Figure S6). The AUC for cirrhosis I/II and cirrhosis II/IV for Child–Pugh scores were 0.840 (P<0.001) and 0.913 (P<0.001) in the testing group, suggesting that the model has good evaluation efficiency, and the efficiency was higher for cirrhosis II/III than cirrhosis I/II (Figure S7).

Prognostic Significance

At the time of last follow-up, among 260 patients in the training and testing group, 112 had died, the median OS was 40.5 months (range, 3.4 to 82.4 months) and overall survival of patients stratified by the Cirrhosis stage system was summarized in the Table S9. Kaplan–Meier curves demonstrated that the new cirrhosis system and Laennec system were significantly associated with OS in both training and testing group (Figure 2). The Laennec scoring system was statistically significant at univariate analysis, but it was not statistically significant at multivariate analysis. The results of the univariate and multivariate analysis are listed in Table 2. Multivariate analyses demonstrated that serum AFP, MVI, TNM, and new cirrhosis scores were independent risk factors for HCC with cirrhosis OS (Table 2).

Table 2 Univariate and Multivariate Analyses of Factors Associated with OS in Combined Training and Testing Cohorts

Figure 2 Kaplan–Meier survival curves of OS differences among patients in combination of training and testing group. Cirrhosis stage of training group (A) and testing group (B), Laennec stage of training group (C), and testing group (D).

Prognostic Nomogram for OS

The prognostic nomogram that integrated all significant independent factors for OS in the combination of training and testing group is shown in Figure 3. The C-index for OS prediction was 0.79 (95% CI, 0.73 to 0.85). The calibration plot for the probability of survival at 1, 3, or 5 years after surgery showed an optimal agreement between the prediction by nomogram and actual observation (Figure 4).

Figure 3 HCC with cirrhosis survival nomogram. (To use the nomogram, an individual patient’s value is located on each variable axis, and a line is drawn upward to determine the number of points received for each variable value. The sum of these numbers is located on the Total Points axis, and a line is drawn downward to the survival axes to determine the likelihood of 1-, 3-, or 5-year survival). MVI, microvascular invasion.

Figure 4 The calibration curve for predicting patient survival at (A) 1 year, (B) 3 years, and (C) 5 years in combined training and testing cohort. Nomogram-predicted probability of overall survival is plotted on the x-axis; actual overall survival is plotted on the y-axis.

Discussion

Diagnosing the severity of disease is an essential component of the baseline assessment of any disease. At the same time, cirrhosis plays a vital role in determining and implementing an appropriate surgical strategy for HCC treatment, and the severity of cirrhosis is closely associated with the occurrence of liver-related events after surgery for HCC.³ Although several staging systems for cirrhosis have been proposed, conventional systems combine all severity levels into a single category. However, there are variations in both the histopathology of cirrhosis and the severity of disease. Only the Laennec system subdivides cirrhosis to recognize the variability in severity.⁵ However, the Laennec system in our study was of limited value in the assessment of cirrhosis in HCC patients. Accordingly, the aim of this study was to propose an effective quantitative pathological scoring system to predict the post-operative liver function and to establish nomogram to predict postoperative survival of HCC with cirrhosis patients after hepatectomy.

Definition of the grades of severity of each lesion should be as precise as possible. In general, more complex systems have the capability to provide more information than simple ones.⁸ The new cirrhosis model consists of four quantitative parameters—NICP, RFA, DPL, and MRP14. All of the parameters are measured through the software. In order to get a more accurate assessment, the system was established based on the OR and the grade of each parameter (ie, cirrhosis system score=OR₁^G(NICP)-1+OR₂^G(RFA)-1+OR₃^G(DPL)-1+OR₄^G(MRP14)-1). The coefficients of the model in the scale are derived statistically so that appropriate weights are given to variables according to their relative risk value. Due to the large subjective error of qualitative or semi-quantitative evaluation, morphological and image analysis of liver slices is considered as the gold standard because of its low intrinsic variability.¹² Meanwhile, the establishment of quantitative grading system can promote the new software-assisted system in the pathological application to improve work efficiency and accuracy.

Liver function generally returns to a more stable level within 12 weeks after partial hepatectomy.²⁴ The main influencing factor of liver function after 3 months is pathological basis lesion, the liver function index after 3 months of surgery can better reflect the actual state of the liver. So we chose the liver function 3 months after surgery as the evaluation criteria. Therefore, this system can predict the actual state of the liver and the evolution of liver function. This pathology system provides new reference indicators for clinical evaluation of cirrhosis soon after surgery, this problem has not been solved in traditional systems. Ideally, there should also be a demonstrable correlation with clinical outcome.⁸ Thus, we built the model according to the postoperative liver function recovery to screen index.

In the present study, we found that MRP14 was associated with the liver function (with Child–Pugh, r=0.7206, P<0.0001), serum index (TBIL, ALT, AST, ALP, PT), and clinical parameters (ascites, encephalopathy) 3 months after surgery in cirrhosis with HCC. Inflammation is not only a factor of severity of disease but also reflects ongoing disease activity and is one of the most potentially responsive parameters to therapy. Therefore, including inflammatory assessment in the system and the system has the function of monitoring the treatment response. We found that NICP correlated with the recovery of liver function (with Child–Pugh, r=0.7143, P<0.0001), serum index (TBIL, ALT, AST, ALP, A/G, PT, Cre), and clinical parameters (ascites, encephalopathy) after surgery in cirrhosis with HCC. The extent of fibrosis was considered to be indicative of the stage of fibrosis and cirrhosis. In our study, RFA correlated with the recovery of liver function (with Child–Pugh, r=0.6627, P<0.0001), serum index (TBIL, ALT, AST, ALP, Cre), and DPL correlated with the recovery of liver function (with Child–Pugh, r=0.6167, P<0.0001), serum index (TBIL, ALT, AST, ALP, ALB, PT, Cre), and clinical parameters (ascites, encephalopathy) after surgery in cirrhosis with HCC (Table S8). These four parameters not only reflect the morphological changes but also reflect the liver damage and liver function. The correlation and diagnostic efficiency of the cirrhosis system were higher than combinations of part of elements and any individual element at the same time (Figure S3). The new cirrhosis system score value strongly correlated with Child–Pugh score value (r=0.8227, P<0.001) in training group, and in testing group the system score value strongly correlated the Child–Pugh score value (r=0.8058, P<0.0001), and serum index value of TBIL (r=0.6370, P<0.001), PT (r=0.6405, P<0.001) and a statistically significant correlation with value of ALT (r=0.2678, p=0.008), AST (r=0.4861, P<0.0001), ALP (r=0.3922, P=0.0002), ALB (r=−0.3012, P=0.0027), Cre (r=0.3903, P<0.0001), Ascitic (r=0.3772, P=0.0001), Encephalopathy (r=0.4537, P<0.0001) 3 month after surgery (Table S8), suggested that the new cirrhosis can reflect the liver recovery of liver function and may reflect the liver injury and blood coagulation and liver function compensation at 3 months after operation for HCC with cirrhosis. Higher cirrhosis system scores predicted poorer liver function and stronger liver damage 3 months after surgery.

The degree of liver damage is another important prognostic factor after HCC surgery.²⁵ Risk factors for morbidity after liver resection include blood loss volume, resected liver volume, and total clamp time.^26–28 ANCOVA is defined more purely as a continuous variable used to control for its effect when testing categorical independent variables.²⁹ Therefore, we used intraoperative blood loss, total clamp time, and resected liver diameter as covariates when performing covariance to detect statistical significance between the coefficients of independent prognostic factors for the new cirrhosis model when excluding covariates. Covariance analysis showed that liver damage had no efficacy among the four pathological parameters of the cirrhosis scoring model.

The relative diagnostic importance of each factor depends on the features of the patients in the training sample, as do the independent variables remaining after stepwise selection.³⁰ The Laennec scoring system grew from the recognition that systems in use for histologic evaluation of liver fibrosis and cirrhosis may not be appropriately used for the unique lesions of cirrhosis in HCC patients. The possible reason why the Laennec system has a poor correlation with liver function recovery was ignoring the effects of inflammation and the special environment of cirrhosis in HCC patients. Similarly, the applicability of the new cirrhosis scoring system is limited by the fact that it was established and validated on the basis of cirrhosis patients with HCC treated by surgical resection. However, it is increasingly recognized that the importance of grading cirrhosis in patients with HCC is due to its impact on prognosis.

In the present study, Kaplan–Meier analyses revealed that high cirrhosis stage was significantly linked to poor OS after surgical resection in HCC patients in both the training and testing groups. Meanwhile, the new cirrhosis system was a significant independent prognostic factor for OS. In our prognostic nomogram, the cirrhosis system was shown to be a poor prognostic factor that shortened OS. The nomogram also include important features of the tumor, including the serum AFP, MVI, and TNM. The C-indices of this nomogram was 0.79, and this nomogram also performed well in predicting long-term survival.

There are several limitations to this study. First, the nomogram was established based on data obtained from a single institution in China. Second, whether this cirrhosis scoring system and nomogram can be applied to liver biopsy remains to be determined.

In conclusion, we established a quantitative scoring system for cirrhosis in patients with HCC by measuring morphological and molecular markers associated with postoperative liver function. The prognostic nomogram was then developed to predict overall survival after hepatectomy. Taken together, these tools can provide a valuable reference for clinical assessment of the risks of postoperative liver function recovery and long-term survival.