Abstract
Objective: The aim of the present study was to establish noninvasive diagnostic models for liver fibrosis and assess their predictive accuracy (AC). Methods: A total of 349 patients with chronic hepatitis B virus infection were evaluated, who underwent liver biopsy and pathologic examination at Beijing Ditan Hospital affiliated to Capital Medical University. Patients were subdivided in disease-immune tolerant (n = 125) and immune reactive HBeAg positive (n = 224) groups. Diagnostic models were based on independent markers of liver fibrosis. Receiver operating characteristic (ROC) curves were used to set cutoff values and determine the diagnostic value of the models. Results: Wang I and Wang II models were constructed using independent disease markers. Wang I model cutoff values ≤ 1.75 and > 5.84 were used to identify patients in the immune tolerant phase with or without significant fibrosis. The area under the ROC curve (AUC) for this model was 0.866 (95% CI, 0.790, 0.942) and an AC of 92.0% was obtained. Wang II model cutoff values ≤ 3.79 and > 7.06 were used to identify immune reactive HBeAg-positive patients with or without significant fibrosis. AUC was 0.872 (95% CI, 0.824, 0.920), with an AC of 88.0%. Conclusions: Both Wang models enabled noninvasive liver fibrosis assessment with reliable predictive power and reproducibility for diagnosis of fibrosis in immune tolerant and immune reactive HBeAg-positive patients. With further development, these models may provide a clinical alternative to liver biopsy.
Keywords: Hepatitis B virus, liver fibrosis, noninvasive diagnosis, model, receiver operating characteristic (ROC) curve
Introduction
More than 400 million people worldwide are estimated to have contracted hepatitis B virus (HBV). Liver fibrosis is a serious consequence of chronic liver disease [1], and about 100 million people die because of chronic liver disease secondary to HBV infection and its complications each year [2,3].
Persistent HBV infection leads to inflammatory infiltration and necrosis of liver parenchyma, resulting in fibrous hyperplasia. The severity of liver fibrosis correlates with patient prognosis and determines the treatment strategy required. Therefore, accurate assessment of liver inflammation and fibrosis is an important step toward reducing or reversing liver fibrosis, and improving the prognosis of patients with chronic hepatitis B (CHB) infection.
Pathologic examination of a liver biopsy is currently the gold standard diagnostic technique for the assessment of liver fibrosis [4]. However, the procedure is invasive and not readily repeatable. It also carries the risk of clinical complications, sampling errors, interpretation errors, and poor patient compliance [5-8]. For these reasons, blood biochemical markers and serum fibrosis markers are commonly used as surrogates to detect liver fibrosis. However, these diagnostic techniques are limited in sensitivity (SN) and specificity (SP) [9]. In recent years, a number of clinical serologic markers, imaging, and noninvasive diagnostic models of liver fibrosis have emerged. These include the Forns’ score [10], APRI (aspartate aminotransferase-platelet count ratio index) [11-13], FibroTest [14], Fibrometer [15], and Hepascore [16]. However, most of these models have focused on chronic hepatitis C (CHC) and alcoholic liver disease, and are less accurate when used in patients with CHB [11,17].
Further studies have established more specific diagnostic models for the prediction of fibrosis in patients with CHB, including the Shanghai Liver Fibrosis Group model [18], Hui et al. model [19], S index [20], and FibroIndex [21]. These models were developed primarily for patients with CHB associated with elevated aminotransferases, but these techniques have not yet been sufficiently validated, and thus are not widely used.
Patients in the immune tolerant phase usually display normal liver function, and as a result, many remain inadequately treated, despite previous studies demonstrating that fibrosis occurs in 30-40% of patients during this phase [22]. Timely and accurate recognition of fibrosis is essential to improve the prognosis of liver disease. Antiviral and antifibrosis treatments can delay and reduce the incidence of liver cirrhosis and cancer [23]. However, models of the immune tolerant- and immune reactive HBeAg-positive phases have not yet been developed. We, therefore, aimed to establish two individual models to aid the early detection of liver fibrosis in the immune tolerant and immune reactive HBeAg-positive phases of HBV. These models were based on routine laboratory tests and liver biopsy results. The objective was to improve predictive accuracy, define a noninvasive method for screening therapeutic targets, and establish a basis for appropriate antiviral treatment and efficacy evaluation.
Materials and methods
Patient selection
A total of 349 patients with CHB treated at the Capital Medical University, Beijing Ditan Hospital from January 2009 to December 2011 were included in the study; they were diagnosed according to the American Association for the Study of Liver Diseases CHB prevention and treatment guidelines [24]. The included patients had a history of hepatitis B or had been hepatitis B surface antigen (HBsAg) positive for more than 6 months, and were HBeAg seropositive and/or had high HBV DNA levels. Exclusion criteria included coexisting viral infections and other forms of liver diseases or hepatic fibrosis-related diseases such as rheumatism, systemic lupus erythematosus, chronic obstructive pulmonary disease, renal failure, hematological diseases, or tumors. Patients receiving antiviral therapy in the year before hepatic biopsy, those receiving medical treatment within 3 months, or dialysis, blood transfusion or the administration of blood products within 1 week of hepatic biopsy were also excluded. The study protocol was approved by the ethics committee of the Capital Medical University, Beijing Ditan Hospital and written informed consent was obtained from each patient.
Patients were grouped according to disease phase: immune tolerant or immune reactive HBeAg positive. The immune tolerant phase of HBV infection is characterized by hepatitis B e antigen (HBeAg) positivity, high levels of HBV replication (reflected by high serum HBV DNA levels), normal-to-low levels of aminotransferases, mild or no liver necroinflammation, and no or slow fibrosis [25-28]. The immune reactive HBeAg-positive phase is characterized by HBeAg positivity, relatively low levels of viral replication (reflected by lower serum HBV DNA levels), increased or fluctuating levels of aminotransferases, moderate or severe liver necroinflammation, and more rapid fibrosis [25-29].
Assessments
Baseline patient characteristics including gender, age, height, and weight were recorded. Within 1 week of liver biopsy, blood samples were collected from all subjects, and white blood cells, red blood cells (RBC), platelets (PLT), hemoglobin, alanine aminotransferase (ALT), AST, total bilirubin (TBIL), direct bilirubin, albumin (ALB), globulin, A/G, gamma-glutamyltransferase (GGT), prothrombin activity (PTA) levels, and HBV marker levels were measured using a Sysmex automated blood cell analyzer (XE 5000; SYS-MEX, Kobe City, Japan) and a 7180 automatic biochemical analyzer (7180; Hitachi, Tokyo, Japan). Coagulation tests were performed using an automatic coagulation analyzer. The body mass index (BMI) and AST/ALT were calculated.
Histopathologic examination of liver biopsy
Biopsies were performed by Doppler ultrasound guided positioning of a 16G biopsy needle (16G-9-20T; Cook, USA). Each biopsy was completed within 1 s, measured at least 1.0 cm in length, and contained more than four portal tracts. The specimens were fixed in 10% formaldehyde solution, embedded in paraffin, and serially sectioned for hematoxylin and eosin, Masson trichrome, and reticular fiber staining. The stage of liver fibrosis was assessed according to the Scheuer (S) classification scale, where S0-S1 indicated no significant liver fibrosis, and S2-S4 indicated significant liver fibrosis [30]. The sections were independently examined by a pathologist.
Statistical analysis and Wang models establishment
Statistical analysis was carried out using SPSS version 18.0 software (SPSS Inc., Chicago, IL). In all tests, P < 0.05 was considered statistically significant. HBV DNA measurements were converted to natural logarithms to fit a normal distribution. Univariate analysis of continuous variables was carried out to determine significant differences between groups using t-tests and analysis of variance. The X 2-test was used to compare categorical variables between groups. Multivariate analysis was carried out to select predictors of CHB using stepwise logistic regression.
The regression equations were used to define Wang models for the prediction of fibrosis. The markers identified by univariate analysis were further analyzed using forward stepwise multivariate logistic regression, and were sequentially entered into the model to identify significant variables. Markers were considered predictive when P < 0.05, and insignificant when P > 0.1.
Receiver operating characteristic (ROC) curves were plotted and area under the curve (AUC) values calculated. Sensitivity and specificity were determined, and the maximum Youden index (SN + SP—1) corresponding to the best upper and lower limit cutoff values was combined with the positive predictive value (PPV), negative predictive value (NPV). Accuracy (AC) was defined as the degree to which the diagnostic results conformed to the actual results, and was applied to evaluate the model’s diagnostic value.
Results
Diagnostic model for the immune tolerant phase
The characteristics of patients with immune tolerant and immune reactive disease are shown in Tables 1 and 2, respectively [31]. Patients were further stratified according to significant fibrosis presence or absence (Table 3).
Table 1.
Univariate analysis of markers associated with significant fibrosis in the immune tolerant phase
| Variable | Modeling (n = 125) | No significant fibrosis (n = 87) | Significant fibrosis (n = 38) | P-value |
|---|---|---|---|---|
| Patients (n, %) | 125 (100.0%) | 87 (69.6%) | 38 (30.4%) | |
| Male (n, %) | 82 (65.6%) | 57 (65.5%) | 25 (65.8%) | 0.976 |
| Age (y) | 33.79 ± 8.86 | 32.17 ± 8.94 | 37.50 ± 7.54 | 0.002 |
| BMI (kg/m2) | 23.03 ± 3.59 | 22.51 ± 3.24 | 24.23 ± 4.08 | 0.013 |
| WBC (109/L) | 5.57 ± 1.33 | 5.56 ± 1.40 | 5.59 ± 1.18 | 0.894 |
| RBC (1012/L) | 4.84 ± 0.59 | 4.90 ± 0.59 | 4.70 ± 0.57 | 0.070 |
| Hb (g/L) | 145.63 ± 18.13 | 145.75 ± 18.16 | 145.34 ± 18.32 | 0.907 |
| Platelets (109/L) | 188.77 ± 49.00 | 199.87 ± 44.78 | 163.37 ± 49.33 | < 0.001 |
| ALT (IU/L) | 26.94 ± 9.94 | 26.05 ± 8.09 | 28.99 ± 7.27 | 0.057 |
| AST (IU/L) | 23.30 ± 5.11 | 22.18 ± 4.34 | 25.85 ± 5.83 | < 0.001 |
| AST/ALT | 0.93 ± 0.31 | 0. 93 ± 0.33 | 0.93 ± 0.24 | 0.953 |
| Total bilirubin (μmol/L) | 15.62 ± 10.61 | 15.33 ± 8.39 | 16.28 ± 14.59 | 0.645 |
| Direct bilirubin (μmol/L) | 4.09 ± 2.38 | 4.15 ± 2.54 | 3.96 ± 2.00 | 0.683 |
| Albumin (g/L) | 45.18 ± 3.86 | 45.75 ± 3.97 | 43.90 ± 3.32 | 0.013 |
| Globulin (g/L) | 30.25 ± 4.14 | 29.79 ± 4.29 | 31.28 ± 3.64 | 0.065 |
| Albumin/globulin | 1.52 ± 0.25 | 1.56 ± 0.26 | 1.42 ± 0.21 | 0.004 |
| GGT (IU/L) | 18.92 ± 13.73 | 15.38 ± 6.36 | 27.03 ± 21.01 | 0.002 |
| PTA (%) | 84.41 ± 11.27 | 86.07 ± 11.00 | 80.59 ± 11.09 | 0.012 |
| Ln (HBVDNA) | 15.16 ± 3.72 | 16.23 ± 3.26 | 12.69 ± 3.58 | < 0.001 |
ALT = alanine aminotransferase; AST = aspartate aminotransferase; BMI = body mass index; GGT = gamma-glutamyltransferase; Hb = hemoglobin; PTA = prothrombin activity; RBC = red blood cell; WBC = white blood cell.
Table 2.
Univariate analysis of markers associated with significant fibrosis in the immune reactive HBeAg-positive phase
| Variable | Modeling (n = 224) | No significant fibrosis (n = 95) | Significant fibrosis (n = 129) | P-value |
|---|---|---|---|---|
| Patients (n, %) | 224 (100.0%) | 95 (42.4 %) | 129 (57.6%) | |
| Male (n, %) | 163 (72.8%) | 70 (73.3%) | 93 (72.1%) | 0.791 |
| Age (y) | 34.26 ± 9.04 | 31.63 ± 7.66 | 36.20 ± 9.51 | < 0.001 |
| BMI | 24.41 ± 3.95 | 23.65 ± 3.59 | 24.96 ± 4.13 | 0.013 |
| WBC (109/L) | 5.57 ± 1.59 | 5.81 ± 1.59 | 5.39 ± 1.57 | 0.050 |
| RBC (1012/L) | 4.87 ± 0.54 | 5.01 ± 0.56 | 4.76 ± 0.50 | 0.001 |
| Hb (g/L) | 148.23 ± 17.66 | 149.90 ± 17.99 | 147.01 ± 17.38 | 0.226 |
| Platelets (109/L) | 184.99 ± 54.40 | 208.13 ± 52.26 | 167.95 ± 49.60 | < 0.001 |
| ALT (IU/L) | 173.03 ± 269.34 | 105.89 ± 175.16 | 222.48 ± 313.11 | < 0.001 |
| 1N < ALT ≤ 2N | 56.01 ± 11.68 | 56.07 ± 11.77 | 55.93 ± 11,67 | 0.947 |
| ALT > 2N | 340.07 ± 356.88 | 249.96 ± 300.5 | 373.20 ± 372.00 | 0.141 |
| AST (IU/L) | 105.57 ± 182.86 | 59.13 ± 93.31 | 139.77 ± 221.54 | < 0.001 |
| AST/ALT | 0.68 ± 0.42 | 0. 65 ± 0.50 | 0.70 ± 0.36 | 0.396 |
| Total bilirubin (μmol/L) | 20.93 ± 25.15 | 16.82 ± 16.93 | 23.96 ± 29.50 | 0.023 |
| Direct bilirubin (μmol/L) | 8.47 ± 19.51 | 5.95 ± 13.52 | 10.32 ± 22.82 | 0.075 |
| Albumin (g/L) | 44.30 ± 4.63 | 45.17 ± 4.03 | 43.65 ± 4.95 | 0.012 |
| Globulin (g/L) | 31.71 ± 4.30 | 31.34 ± 3.66 | 31.98 ± 4.72 | 0.256 |
| Albumin/globulin | 1.42 ± 0.24 | 1.46 ± 0.21 | 1.39 ± 0.25 | 0.033 |
| GGT (IU/L) | 58.87 ± 71.28 | 32.08 ± 34.83 | 78.61 ± 83.88 | < 0.001 |
| PTA (%) | 82.87 ± 12.89 | 87.45 ± 13.82 | 79.49 ± 11.06 | < 0.001 |
| Ln (HBVDNA) | 15.98 ± 2.81 | 17.02 ± 1.71 | 15.22 ± 3.20 | < 0.001 |
ALT = alanine aminotransferase; AST = aspartate aminotransferase; BMI = body mass index; GGT = gamma-glutamyltransferase; Hb = hemoglobin; PTA = prothrombin activity; RBC = red blood cell; WBC = white blood cell.
Table 3.
Inflammation grade and fibrosis stage in the immune tolerant and immune reactive HBeAg-positive phases
| Grade | Immune tolerant phase (n, %) | Immune reactive HBeAg-positive phase (n, %) |
|---|---|---|
| Inflammation grade | ||
| G0 | 0 (0.0%) | 0 (0.0%) |
| G1 | 79 (63.2%) | 75 (33.5%) |
| G2 | 36 (28.8%) | 79 (35.3%) |
| G3 | 10 (8.0%) | 64 (28.6%) |
| G4 | 0 (0.0%) | 6 (2.7%) |
| Fibrosis stage | ||
| S0 | 4 (3.2%) | 2 (0.9%) |
| S1 | 83 (66.4%) | 93 (41.5%) |
| S2 | 20 (16.0%) | 82 (36.6%) |
| S3 | 15 (12.0%) | 35 (15.6%) |
| S4 | 3 (2.4%) | 12 (5.4%) |
The 125 patients in the immune tolerant phase were diagnosed as having either no significant fibrosis (87 patients, 69.6%) or significant liver fibrosis (38 patients, 30.4%). Among the 224 patients in the immune reactive HBeAg-positive phase, 95 patients (42.4%) had no significant liver fibrosis, and significant liver fibrosis was present in 129 patients (57.6%).
Univariate analysis identified 9 markers, including age, BMI, PLT, AST, ALB, A/G, GGT, PTA, ln [HBVDNA] that differed significantly between these two groups (Table 1). Multiple analysis identified 4 independent predictors of fibrosis: PLT, AST, GGT, and ln [HBVDNA] (Table 4). The following multiple regression equation was constructed based on their relative regression coefficients:
Table 4.
Multiple regression of markers associated with significant fibrosis in the immune tolerant phase
| Variable | Immune tolerant Phase (n = 125) | ||
|---|---|---|---|
|
| |||
| β | OR§ (95% CI) | P | |
| PLT | -0.015 | 0.985 (0.974-0.996) | 0.010 |
| GGT | 0.071 | 1.073 (1.007-1.143) | 0.029 |
| Ln [HBV DNA] | -0.226 | 0.798 (0.701-0.909) | 0.001 |
| AST | 0.154 | 1.166 (1.052-1.293) | 0.003 |
OR is the risk of significant fibrosis in immune tolerant phase patients.
95% CI: 95%, Confidence Interval of the difference.
A = 0.153—0.015 × PLT + 0.154 × AST + 0.071 × GGT-0.226 × ln (HBVDNA).
‘A’ was converted to a score from 0-10 using the following formula:
Wang I = 10 × eA/(1 + eA).
Diagnostic model for the immune reactive HBeAg-positive phase
Univariate analysis identified 12 variables that differed significantly between patients with fibrosis from those without similarly. These included age, BMI, RBC, PLT, ALT, AST, TBIL, ALB, ALB/globulin, GGT, PTA, and ln [HBVDNA]. Multivariate forward stepwise logistic regression identified 7 independent predictors of fibrosis: RBC, PLT, TBIL, GGT, PTA, ln [HBVDNA], and BMI (Table 5). The following multiple regression equation was constructed to predict significant fibrosis:
Table 5.
Multiple regression of markers associated with significant fibrosis in the immune reactive HBeAg-positive phase
| Variable | Immune reactive HBeAg-positive phase (n = 224) | ||
|---|---|---|---|
|
| |||
| β | OR§ (95% CI) | P | |
| PLT | -0.011 | 0.989 (0.982-0.997) | 0.004 |
| GGT | 0.021 | 1.021 (1.010-1.033) | 0.000 |
| Ln [HBVDNA] | -0.258 | 0.773 (0.660-0.904) | 0.001 |
| TBIL | -0.019 | 0.981 (0.965-0.997) | 0.021 |
| RBC | -1.475 | 0.229 (0.106-0.496) | 0.000 |
| PTA | -0.052 | 0.950 (0.921-0.979) | 0.001 |
| BMI | 0.160 | 1.173 (1.054-1.305) | 0.003 |
OR is the risk of significant fibrosis of immune reactive HBeAg-positive phase patients.
95% CI: 95% Confidence Interval of the difference.
C = 13.657—1.475 × RBC—0.011 × PLT—0.019 × TBIL + 0.021 × GGT—0.052 × PTA—0.258 × ln (HBVDNA) + 0.160 × BMI.
‘C’ was converted to a score from 0-10 using the following formula:
Wang II = 10 × eC/(1 + eC).
Diagnostic value of the Wang I index in immune tolerant patients
The ROC curve demonstrated that the Wang I index predicted significant fibrosis among the 125 patients of the immune tolerant group. An area under the ROC curve (AUC) of 0.866 was obtained for Wang I index, which was significantly higher than the AUC values for individual markers (P < 0.05; Figure 1). At 89.5% SN and 96.6% SP, the Wang I index negative cutoff value was 1.75 and the positive cutoff value 5.84. With Wang I index ≤ 1.75, the NPV was 93.5%, indicating that 58 of 62 patients with a Wang I index ≤ 1.75 had no significant liver fibrosis (Table 6). A SN of 89.5% was obtained, indicating that 34 of 38 patients with significant fibrosis, confirmed by liver biopsy, had a Wang I index > 1.75. The diagnostic AC was 73.6%.
Figure 1.
ROC curve generated from the diagnosis of significant liver fibrosis by modeling patients in the immune tolerant phase using the Wang I index.
Table 6.
Diagnostic measures of patients in the immune tolerant phase
| Wang I index | Total (n = 125) | S0-S1 (n = 87) | S2-S4 (n = 38) | SN (%) | SP (%) | PPV (%) | NPV (%) | AC (%) | AUROC (95% CI) |
|---|---|---|---|---|---|---|---|---|---|
| 1.75 | 89.5 | 66.7 | 54.0 | 93.5 | 73.6 | 0.866 (0.790, 0.942) | |||
| > 1.75 | 63 | 29 | 34 | ||||||
| ≤ 1.75 | 62 | 58 | 4 | ||||||
| 5.84 | 57.9 | 96.6 | 88.0 | 84.0 | 84.8 | ||||
| > 5.84 | 25 | 3 | 22 | ||||||
| ≤ 5.84 | 100 | 84 | 16 |
AC = accuracy; AUROC = area under the receiver operating characteristic curve; NPV = negative predictive value; PPV = positive predictive value; SN = sensitivity; SP = specificity.
A patient with a Wang I index > 5.84 was suspected to have significant fibrosis. As shown in Table 6, the PPV using this cutoff value was 88.0% i.e. 22 of 25 patients with a Wang I index > 5.84 had significant fibrosis. The corresponding SP was 96.6%. Among the 87 patients in whom liver fibrosis absence was confirmed by liver biopsy, 84 individuals had a Wang I index ≤ 5.84, indicating a diagnostic AC of 84.8%.
Overall, 80 of 125 patients (64.0%) were diagnosed correctly, and 7 patients (5.6%) were diagnosed incorrectly. The diagnostic AC was 92.0%. The application of this model would have enabled 69.6% of patients (87 patients) to avoid liver biopsy. Among the remaining 38 (30.4%) patients with a Wang I index between 1.75 and 5.84, the presence of significant fibrosis ought to be determined by liver biopsy or other detection methods.
Diagnostic value of the Wang II index in immune reactive patients
The Wang II index predicted liver fibrosis in 224 patients in the immune reactive HBeAg-positive group. The AUC for Wang II (0.872) was significantly higher than that obtained for any of the individual markers (P < 0.05; Figure 2). With 90.7% SN and 91.6% SP, the negative and positive cutoff values for significant fibrosis were 3.79 and 7.06, respectively.
Figure 2.
ROC curve generated from the diagnosis of significant liver fibrosis by modeling patients in the immune reactive HBeAg-positive phase using the Wang II index.
With Wang II index ≤ 3.79, the SN and NPV were 90.7% and 84.4%, respectively (Table 7), and the absence of fibrosis was predicted correctly in 65 of the 77 patients (84.4%). Among the 129 patients with fibrosis confirmed by biopsy, 117 (90.7%) individuals had a Wang II index > 3.79; an AC of 81.2% was obtained.
Table 7.
Diagnostic measures of patients in the immune reactive HBeAg-positive phase
| Wang II index | Total (N = 224) | S0-S1 (N = 95) | S2-S4 (N = 129) | SN (%) | SP (%) | PPV (%) | NPV (%) | AC (%) | AUROC (95% CI) |
|---|---|---|---|---|---|---|---|---|---|
| 3.79 | 90.7 | 68.4 | 79.6 | 84.4 | 81.2 | 0.872 (0.824, 0.920) | |||
| > 3.79 | 147 | 30 | 117 | ||||||
| ≤ 3.79 | 77 | 65 | 12 | ||||||
| 7.06 | 62.8 | 91.6 | 91.0 | 64.4 | 75.0 | ||||
| > 7.06 | 89 | 8 | 81 | ||||||
| ≤ 7.06 | 135 | 87 | 48 |
AC = accuracy; AUROC = area under the receiver operating characteristic curve; NPV = negative predictive value; PPV = positive predictive value; SN = sensitivity; SP = specificity.
When the Wang II index cutoff value was > 7.06, the SP and PPV were 91.6% and 91.0%, respectively (Table 7). Fibrosis was predicted correctly in 91.0% (81/89) of patients. Among the 95 patients in which no fibrosis was detected by biopsy, 87 patients (91.6%) had a Wang II index ≤ 7.06. The AC was 75.0%.
Overall, 146 of 224 patients (65.2%) were diagnosed correctly. The diagnostic AC of the Wang II model was 88.0%. The application of this model would enable 166 (74.1%) patients to avoid liver biopsy. There were 58 patients (25.9%) with a Wang II index between 3.79 and 7.06 who would still require liver biopsy.
Diagnostic ACs of noninvasive markers for predicting liver fibrosis
A comparison of different models, including the Wang I and Wang II indexes, S index [17], Zeng model [18], model proposed by Hui et al [19], APRI [12] and APAG is summarized in Table 8. Although the SN and SP of APAG were slightly higher than the values obtained for the Wang I and Wang II indexes, its PPV, NPV and AUC were significantly lower than in Wang models. The Wang I index had the best NPV (93.5%) in diagnosing no significant liver fibrosis, while the Wang II index had best PPV (91%) in diagnosing significant liver fibrosis. Interestingly, the AUCs of Wang I and Wang II indexes were the largest compared with all other models. These results indicated that the Wang I and Wang II indexes more accurately predicted liver fibrosis than other models.
Table 8.
Comparisons of noninvasive models for predicting liver fibrosis
| Noninvasive models | Cutoff value | Patients classified n (%) | Ishak Fibrosis | SN (%) | SP (%) | PPV (%) | NPV (%) | AUROC | |
|
| |||||||||
| Stage 0-1 (n) | Stage 2-4 (n) | ||||||||
|
| |||||||||
| Wang I index (N = 125) | ≤ 1.75 | 62 (46.0%) | 58 | 4 | 89.5 | 66.7 | 54.0 | 93.5 | 0.866 |
| > 5.84 | 25 (20.0%) | 3 | 22 | 57.9 | 96.6 | 88.0 | 84.0 | ||
| Wang II index (N = 224) | ≤ 3.79 | 77 (34.4%) | 65 | 12 | 90.7 | 68.4 | 79.6 | 84.4 | 0.872 |
| > 7.06 | 89 (39.7%) | 8 | 81 | 62.8 | 91.6 | 91.0 | 64.4 | ||
| S index (N = 386) [17] | < 0.1 | 61 (15.8%) | 40 | 21 | 90.4 | 24.0 | 60.9 | 65.6 | 0.686 |
| ≥ 0.5 | 103 (26.7%) | 23 | 80 | 36.5 | 86.2 | 77.7 | 50.9 | ||
| Zeng model (N = 200) [18] | < 3.0 | 43 (21.5%) | 37 | 6 | 94.8 | 44.1 | 70.1 | 86.1 | 0.840 |
| > 8.7 | 45 (22.5%) | 4 | 41 | 35.3 | 95.2 | 91.1 | 51.6 | ||
| APAG (N = 250) | < 0.27 | 39 (16%) | ND | ND | 95 | 29 | 62 | 82 | 0.792 |
| > 0.80 | 43 (17%) | ND | ND | 28 | 95 | 87 | 52 | ||
|
| |||||||||
| Stage 0-2 (n) | Stage 3-6 (n) | ||||||||
|
| |||||||||
| Model of Hui et al (N = 147) [19] | ≤ 0.15 | 55 (37.4%) | 52 | 3 | 93 | 49 | 41 | 95 | 0.803 |
| > 0.5 | 27 (18.4%) | 10 | 17 | 41 | 90 | 63 | 79 | ||
| APRI (N = 192) [12] | ≤ 0.5 | 55 (28.6%) | 47 | 8 | 91.0 | 47.0 | 61.0 | 86.0 | 0.80 |
| > 1.50 | 42 (21.9%) | 5 | 37 | 41.0 | 95.0 | 88.0 | 64.0 | ||
S index = 1000 × GGT/(PLT × ALB2); APRI = [AST level (ULN)/Platelet counts (109/L)] × 100. ND = not described.
Discussion
Serum HBV DNA levels, aminotransferase levels, and histologic findings have all been used as indicators of HBV presence and antiviral treatment efficacy. However, the extent of liver fibrosis is considered the key factor in determining the effectiveness of antiviral treatment in patients during the immune tolerant phase of HBV infection with normal serum ALT levels [24].
Liver biopsy is widely regarded as the gold standard diagnostic technique for assessing liver fibrosis, but the procedure is not conducive for monitoring histologic changes because of its invasiveness. In this study, the Scheuer S2-S4 histological classification was used to assess the severity of fibrosis in biopsies, and routine laboratory tests were used as screening markers. Based on their ALT levels, patients with CHB were divided into immune tolerant- and immune reactive HBeAg-positive groups. Two non-invasive diagnostic models, namely Wang I and Wang II, were developed to assess liver fibrosis in these two experimental groups.
We found that the application of different diagnostic models to both the immune tolerant and immune reactive HBeAg-positive phases of CHB improved predictive AC. For patients in the immune tolerant phase, the Wang I model would have enabled 69.6% of patients to avoid liver biopsy. In the immune reactive HBeAg-positive phase of CHB, using the Wang II model, 74.1% of patients were assessed as having liver fibrosis at an AC of 88.0%.
The previously reported noninvasive diagnostic model for liver fibrosis in HBV infection did not distinguish between the immune tolerant and immune reactive HBeAg-positive phases [18-21]. We found that independent predictors of liver fibrosis in the immune tolerant phase differed from those in the immune reactive HBeAg-positive phase. The Wang models improved diagnostic efficiency, SN, and SP, especially in the immune tolerant phase of CHB, suggesting that they may be suitable for assessing long-term disease progression and guiding antiviral and anti-fibrosis treatment decisions.
A model termed PPT derived from three independent predictors of liver fibrosis, including platelet count (PLT), prothrombin time (PT) and total bile acid (TBA) was recently developed and showed an AUC of 0.83, indicating its superiority to other models such as APRI, FIB-4, age-AST model, AP index and APGA model [33]; however, the PPT model is at best comparable to the Wang methods described herein. In addition, the Wang models demonstrated a higher diagnostic AC than the S index [11,17], Hui et al. prediction model [19], APAG index (including age, prothrombin time, albumin and gamma-glutamyltransferase), and Wong et al. model [32]. These models all have strengths and weaknesses. The S index model is relatively easy to use, but relies on a single parameter and is unable to distinguish between S1 and S2 fibrosis. Currently, this can only be achieved by pathologic examination. However, a combination of the S index with other markers may improve its predictive AC.
The Hui et al. model [19] applies strict requirements for > 105 copies/mL of HBV DNA to determine fibrosis. However, significant risk factors for cirrhosis and cancer can also occur at HBV DNA levels > 104 copies/mL [24]. The diagnostic model described by Wong et al. involves Fibroscan, a device that significantly improves the diagnostic AC with high SN and SP (> 90%) [32]. However, most hospitals do not have access to this equipment, limiting its widespread use in clinical practice. The APAG index model evaluates the influence of age, PLT, ALB and GGT on the development of fibrosis and has proved to be a convenient and accurate method. We attempted to improve the AC of this technique by adding several new parameters.
The Wang models in our study used conventional biochemical markers, such as PLTs, AST, GGT, PTA and BMI, all which have been previously validated as independent indicators of significant liver fibrosis [16,24,33]. PLTs are a useful measure of liver fibrosis and used in the FibroIndex, S index, APRI index, and the Fibrosis 4 model. HBV infection causes inflammation in the liver that continuously stimulates fibrosis and destroys the liver structure, resulting in reduced hepatic production of thrombopoietin I [34,35]. The associated splenomegaly further reduces platelet production [34,35].
We also observed that RBC counts were reduced in the immune reactive HBeAg-positive group. The hypersplenism secondary to liver fibrosis causes RBC damage via the nuclear-macrophage system [36]. Because RBCs supply oxygen to liver, their destruction causes liver tissue hypoxia, increasing the rate of liver cell injury, activating hepatic stellate cells, and promoting fibrosis. These findings suggest that RBCs constitute a reliable marker of liver fibrosis during the immune reactive HBeAg-positive phase of CHB.
We identified PTA and GGT as important parameters in the models, suggesting that these markers are closely related to the degree of liver fibrosis. The synthesis of clotting factors including prothrombin has been found to be significantly decreased in the presence of liver cell degeneration, necrosis and fibrosis [37]. Gamma-glutamyl transferase is a membrane-binding enzyme mainly distributed in the hepatocyte microsomal and canalicular bile. Inflammatory stimuli increase membrane permeability, resulting in increased GGT activity in the serum [38,39]. Previous studies have found that GGT is a reliable predictor of CHC and CHB fibrosis [38,40].
BMI is an important predictor of fibrosis in the Hui et al. model. In patients with CHC, high BMI, which is closely related to obesity and nonalcoholic fatty liver disease, reduces the antiviral response and increases the risk of fibrosis progression [41-43]. However, this marker has not been used in the CHC diagnostic models, and there are few reports of its diagnostic value in patients with CHB [10,11,13-16,44]. Our study suggests that obesity can contribute to liver fibrosis development. However, the relationship between BMI and fibrosis in CHB requires further investigation.
The main limitation of our models is that they were unable to define the precise stage of liver fibrosis. In addition, all new assessments of liver fibrosis should be combined with newly emerging noninvasive screening methods such as liver elasticity ultrasound (Fibroscan) [45] to further improve diagnostic AC. Further research, development, and validation of these models are required in a larger cohort of patients with CHB before the models can be introduced into clinical practice.
In conclusion, the Wang models provide reliable alternative to liver biopsy for assessing fibrosis. They may also help to evaluate the efficacy of antiviral and anti-liver fibrosis treatments. With further validation and development of these models in additional patient cohorts, their application may reduce the pain and risk associated with invasive liver examination, and conserve health care resources.
Acknowledgements
This study was supported by grants from the Natural Science Foundation of China (81072766, 215 program (2013-2-11)) and Beijing Science and Technology Project Grant (Z111107056811044).
Disclosure of conflict of interest
None.
References
- 1.Poynard T, Mathurin P, Lai CL, Guyader D, Poupon R, Tainturier MH, Myers RP, Muntenau M, Ratziu V, Manns M, Vogel A, Capron F, Chedid A, Bedossa P. A comparison of fibrosis progression in chronic liver diseases. J Hepatol. 2003;38:257–265. doi: 10.1016/s0168-8278(02)00413-0. [DOI] [PubMed] [Google Scholar]
- 2.Ganem D, Prince AM. Hepatitis B virus infection--natural history and clinical consequences. N Engl J Med. 2004;350:1118–1129. doi: 10.1056/NEJMra031087. [DOI] [PubMed] [Google Scholar]
- 3.World Health Organization. Hepatitis B (Revised August 2008) [Google Scholar]
- 4.Rockey DC, Bissell DM. Noninvasive measures of liver fibrosis. Hepatology. 2006;43:S113–120. doi: 10.1002/hep.21046. [DOI] [PubMed] [Google Scholar]
- 5.Intraobserver and interobserver variations in liver biopsy interpretation in patients with chronic hepatitis C. The French METAVIR Cooperative Study Group. Hepatology. 1994;20:15–20. [PubMed] [Google Scholar]
- 6.Bedossa P, Dargere D, Paradis V. Sampling variability of liver fibrosis in chronic hepatitis C. Hepatology. 2003;38:1449–1457. doi: 10.1016/j.hep.2003.09.022. [DOI] [PubMed] [Google Scholar]
- 7.Regev A, Berho M, Jeffers LJ, Milikowski C, Molina EG, Pyrsopoulos NT, Feng ZZ, Reddy KR, Schiff ER. Sampling error and intraobserver variation in liver biopsy in patients with chronic HCV infection. Am J Gastroenterol. 2002;97:2614–2618. doi: 10.1111/j.1572-0241.2002.06038.x. [DOI] [PubMed] [Google Scholar]
- 8.Maharaj B, Maharaj RJ, Leary WP, Cooppan RM, Naran AD, Pirie D, Pudifin DJ. Sampling variability and its influence on the diagnostic yield of percutaneous needle biopsy of the liver. Lancet. 1986;1:523–525. doi: 10.1016/s0140-6736(86)90883-4. [DOI] [PubMed] [Google Scholar]
- 9.Poynard T, Morra R, Ingiliz P, Imbert-Bismut F, Thabut D, Messous D, Munteanu M, Massard J, Benhamou Y, Ratziu V. Assessment of liver fibrosis: noninvasive means. Saudi J Gastroenterol. 2008;14:163–173. doi: 10.4103/1319-3767.43273. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Forns X, Ampurdanes S, Llovet JM, Aponte J, Quinto L, Martinez-Bauer E, Bruguera M, Sanchez-Tapias JM, Rodes J. Identification of chronic hepatitis C patients without hepatic fibrosis by a simple predictive model. Hepatology. 2002;36:986–992. doi: 10.1053/jhep.2002.36128. [DOI] [PubMed] [Google Scholar]
- 11.Kim BK, Kim SA, Park YN, Cheong JY, Kim HS, Park JY, Cho SW, Han KH, Chon CY, Moon YM, Ahn SH. Noninvasive models to predict liver cirrhosis in patients with chronic hepatitis B. Liver Int. 2007;27:969–976. doi: 10.1111/j.1478-3231.2007.01519.x. [DOI] [PubMed] [Google Scholar]
- 12.Chung IK, Lee JH, Lee SH, Kim SJ, Cho JY, Cho WY, Hwangbo Y, Keum BR, Park JJ, Chun HJ, Kim HJ, Kim JJ, Ji SR, Seol SY. Therapeutic outcomes in 1000 cases of endoscopic submucosal dissection for early gastric neoplasms: Korean ESD Study Group multicenter study. Gastrointest Endosc. 2009;69:1228–1235. doi: 10.1016/j.gie.2008.09.027. [DOI] [PubMed] [Google Scholar]
- 13.Shaheen AA, Myers RP. Diagnostic accuracy of the aspartate aminotransferase-to-platelet ratio index for the prediction of hepatitis C-related fibrosis: a systematic review. Hepatology. 2007;46:912–921. doi: 10.1002/hep.21835. [DOI] [PubMed] [Google Scholar]
- 14.Imbert-Bismut F, Ratziu V, Pieroni L, Charlotte F, Benhamou Y, Poynard T. Biochemical markers of liver fibrosis in patients with hepatitis C virus infection: a prospective study. Lancet. 2001;357:1069–1075. doi: 10.1016/S0140-6736(00)04258-6. [DOI] [PubMed] [Google Scholar]
- 15.Cales P, Oberti F, Michalak S, Hubert-Fouchard I, Rousselet MC, Konate A, Gallois Y, Ternisien C, Chevailler A, Lunel F. A novel panel of blood markers to assess the degree of liver fibrosis. Hepatology. 2005;42:1373–1381. doi: 10.1002/hep.20935. [DOI] [PubMed] [Google Scholar]
- 16.Adams LA, Bulsara M, Rossi E, DeBoer B, Speers D, George J, Kench J, Farrell G, McCaughan GW, Jeffrey GP. Hepascore: an accurate validated predictor of liver fibrosis in chronic hepatitis C infection. Clin Chem. 2005;51:1867–1873. doi: 10.1373/clinchem.2005.048389. [DOI] [PubMed] [Google Scholar]
- 17.Zhou K, Gao CF, Zhao YP, Liu HL, Zheng RD, Xian JC, Xu HT, Mao YM, Zeng MD, Lu LG. Simpler score of routine laboratory tests predicts liver fibrosis in patients with chronic hepatitis B. J Gastroenterol Hepatol. 2010;25:1569–1577. doi: 10.1111/j.1440-1746.2010.06383.x. [DOI] [PubMed] [Google Scholar]
- 18.Zeng MD, Lu LG, Mao YM, Qiu DK, Li JQ, Wan MB, Chen CW, Wang JY, Cai X, Gao CF, Zhou XQ. Prediction of significant fibrosis in HBeAg-positive patients with chronic hepatitis B by a noninvasive model. Hepatology. 2005;42:1437–1445. doi: 10.1002/hep.20960. [DOI] [PubMed] [Google Scholar]
- 19.Hui AY, Chan HL, Wong VW, Liew CT, Chim AM, Chan FK, Sung JJ. Identification of chronic hepatitis B patients without significant liver fibrosis by a simple noninvasive predictive model. Am J Gastroenterol. 2005;100:616–623. doi: 10.1111/j.1572-0241.2005.41289.x. [DOI] [PubMed] [Google Scholar]
- 20.Mohamadnejad M, Montazeri G, Fazlollahi A, Zamani F, Nasiri J, Nobakht H, Forouzanfar MH, Abedian S, Tavangar SM, Mohamadkhani A, Ghoujeghi F, Estakhri A, Nouri N, Farzadi Z, Najjari A, Malekzadeh R. Noninvasive markers of liver fibrosis and inflammation in chronic hepatitis B-virus related liver disease. Am J Gastroenterol. 2006;101:2537–2545. doi: 10.1111/j.1572-0241.2006.00788.x. [DOI] [PubMed] [Google Scholar]
- 21.Fung J, Lai CL, Fong DY, Yuen JC, Wong DK, Yuen MF. Correlation of liver biochemistry with liver stiffness in chronic hepatitis B and development of a predictive model for liver fibrosis. Liver Int. 2008;28:1408–1416. doi: 10.1111/j.1478-3231.2008.01784.x. [DOI] [PubMed] [Google Scholar]
- 22.Croagh CM, Bell SJ, Slavin J, Kong YX, Chen RY, Locarnini S, Desmond PV. Increasing hepatitis B viral load is associated with risk of significant liver fibrosis in HBeAg-negative but not HBeAg-positive chronic hepatitis B. Liver Int. 2010;30:1115–1122. doi: 10.1111/j.1478-3231.2010.02267.x. [DOI] [PubMed] [Google Scholar]
- 23.Tran TT. Immune tolerant hepatitis B: a clinical dilemma. Gastroenterol Hepatol (N Y) 2011;7:511–516. [PMC free article] [PubMed] [Google Scholar]
- 24.Lok AS, McMahon BJ. Chronic hepatitis B: update 2009. Hepatology. 2009;50:661–662. doi: 10.1002/hep.23190. [DOI] [PubMed] [Google Scholar]
- 25.Fattovich G. Natural history and prognosis of hepatitis B. Semin Liver Dis. 2003;23:47–58. doi: 10.1055/s-2003-37590. [DOI] [PubMed] [Google Scholar]
- 26.McMahon BJ. The natural history of chronic hepatitis B virus infection. Semin Liver Dis. 2004;24(Suppl 1):17–21. doi: 10.1055/s-2004-828674. [DOI] [PubMed] [Google Scholar]
- 27.Hoofnagle JH, Doo E, Liang TJ, Fleischer R, Lok AS. Management of hepatitis B: summary of a clinical research workshop. Hepatology. 2007;45:1056–1075. doi: 10.1002/hep.21627. [DOI] [PubMed] [Google Scholar]
- 28.Lok AS, McMahon BJ. Chronic hepatitis B. Hepatology. 2007;45:507–539. doi: 10.1002/hep.21513. [DOI] [PubMed] [Google Scholar]
- 29.Hadziyannis SJ, Papatheodoridis GV. Hepatitis B e antigen-negative chronic hepatitis B: natural history and treatment. Semin Liver Dis. 2006;26:130–141. doi: 10.1055/s-2006-939751. [DOI] [PubMed] [Google Scholar]
- 30.Scheuer PJ. Classification of chronic viral hepatitis: a need for reassessment. J Hepatol. 1991;13:372–374. doi: 10.1016/0168-8278(91)90084-o. [DOI] [PubMed] [Google Scholar]
- 31.European Association for the Study of the Liver. EASL Clinical Practice Guidelines: Management of hepatitis C virus infection. J Hepatol. 2012;57:167–185. doi: 10.1016/j.jhep.2012.02.010. [DOI] [PubMed] [Google Scholar]
- 32.Wong GL, Wong VW, Choi PC, Chan AW, Chan HL. Development of a non-invasive algorithm with transient elastography (Fibroscan) and serum test formula for advanced liver fibrosis in chronic hepatitis B. Aliment Pharmacol Ther. 2010;31:1095–1103. doi: 10.1111/j.1365-2036.2010.04276.x. [DOI] [PubMed] [Google Scholar]
- 33.Swets JA. Measuring the accuracy of diagnostic systems. Science. 1988;240:1285–1293. doi: 10.1126/science.3287615. [DOI] [PubMed] [Google Scholar]
- 34.Poynard T, Bedossa P. Age and platelet count: a simple index for predicting the presence of histological lesions in patients with antibodies to hepatitis C virus. METAVIR and CLINIVIR Cooperative Study Groups. J Viral Hepat. 1997;4:199–208. doi: 10.1046/j.1365-2893.1997.00141.x. [DOI] [PubMed] [Google Scholar]
- 35.Aster RH. Pooling of platelets in the spleen: role in the pathogenesis of “hypersplenic” thrombocytopenia. J Clin Invest. 1966;45:645–657. doi: 10.1172/JCI105380. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Jiao YF, Okumiya T, Saibara T, Kudo Y, Sugiura T. Erythrocyte creatine as a marker of excessive erythrocyte destruction due to hypersplenism in patients with liver cirrhosis. Clin Biochem. 2001;34:395–398. doi: 10.1016/s0009-9120(01)00242-9. [DOI] [PubMed] [Google Scholar]
- 37.Li XM, Ma L, Yang YB, Shi ZJ, Zhou SS. Analyses of prognostic indices of chronic liver failure caused by hepatitis virus. World J Gastroenterol. 2005;11:2841–2843. doi: 10.3748/wjg.v11.i18.2841. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Myers RP, Tainturier MH, Ratziu V, Piton A, Thibault V, Imbert-Bismut F, Messous D, Charlotte F, Di Martino V, Benhamou Y, Poynard T. Prediction of liver histological lesions with biochemical markers in patients with chronic hepatitis B. J Hepatol. 2003;39:222–230. doi: 10.1016/s0168-8278(03)00171-5. [DOI] [PubMed] [Google Scholar]
- 39.Ruhl CE, Everhart JE. Elevated serum alanine aminotransferase and γ-glutamyltransferase and mortality in the United States population. Gastroenterology. 2009;136:477. doi: 10.1053/j.gastro.2008.10.052. [DOI] [PubMed] [Google Scholar]
- 40.Giannini E, Ceppa P, Botta F, Fasoli A, Romagnoli P, Cresta E, Venturino V, Risso D, Celle G, Testa R. Steatosis and bile duct damage in chronic hepatitis C: distribution and relationships in a group of Northern Italian patients. Liver. 1999;19:432–437. doi: 10.1111/j.1478-3231.1999.tb00074.x. [DOI] [PubMed] [Google Scholar]
- 41.Bressler BL, Guindi M, Tomlinson G, Heathcote J. High body mass index is an independent risk factor for nonresponse to antiviral treatment in chronic hepatitis C. Hepatology. 2003;38:639–644. doi: 10.1053/jhep.2003.50350. [DOI] [PubMed] [Google Scholar]
- 42.Lonardo A, Adinolfi LE, Loria P, Carulli N, Ruggiero G, Day CP. Steatosis and hepatitis C virus: mechanisms and significance for hepatic and extrahepatic disease. Gastroenterology. 2004;126:586–597. doi: 10.1053/j.gastro.2003.11.020. [DOI] [PubMed] [Google Scholar]
- 43.James OF, Day CP. Non-alcoholic steatohepatitis (NASH): a disease of emerging identity and importance. J Hepatol. 1998;29:495–501. doi: 10.1016/s0168-8278(98)80073-1. [DOI] [PubMed] [Google Scholar]
- 44.Wai CT, Greenson JK, Fontana RJ, Kalbfleisch JD, Marrero JA, Conjeevaram HS, Lok AS. A simple noninvasive index can predict both significant fibrosis and cirrhosis in patients with chronic hepatitis C. Hepatology. 2003;38:518–526. doi: 10.1053/jhep.2003.50346. [DOI] [PubMed] [Google Scholar]
- 45.Sandrin L, Fourquet B, Hasquenoph JM, Yon S, Fournier C, Mal F, Christidis C, Ziol M, Poulet B, Kazemi F, Beaugrand M, Palau R. Transient elastography: a new noninvasive method for assessment of hepatic fibrosis. Ultrasound Med Biol. 2003;29:1705–1713. doi: 10.1016/j.ultrasmedbio.2003.07.001. [DOI] [PubMed] [Google Scholar]