From May 2013 to May 2015, 337 consecutive CHB patients who underwent liver biopsy to evaluate the severity of fibrosis and necroinflammation at Huashan Hospital, Fudan University were enrolled in the study. CHB (defined as chronic necroinflammatory disease of the liver caused by persistent infection with HBV) is diagnosed according to the Asian-Pacific clinical practice guidelines on the management of hepatitis B.2 Demographic and laboratory data of all subjects were collected from the medical records within 3 days of liver biopsy. The exclusion criteria in this study include: co-infection of HCV, hepatitis D virus or HIV; coexistence of alcoholic liver disease, autoimmune hepatitis, drug-induced liver injury, metabolic liver disease, hepatocellular carcinoma (HCC); liver transplantation. Patients who received anti-HBV treatment were also excluded. The study was approved by the Ethical Committee of Huashan Hospital, Fudan University.

Percutaneous liver biopsy was performed using 16G biopsy needle and guided by ultrasonography after written informed consent. Scheuer's scoring system was adopted to assess the degree of liver necroinflammation activity and fibrosis stage. Significant liver fibrosis was defined as fibrosis stage >S2 while severe necroinflammation activity was defined as necroinflammation grade >G2.

Fib-4=(Age (years)×AST (IU/L))/(platelets count (109/L)×(ALT (IU/L))1/2)

APRI=(AST (IU/L)/ULN)×100/platelets count (109/L)

GPR=(GGT (IU/L)/ULN)×100/platelets count (109/L)

RPR=RDW (%)/platelets count (109/L)

Data analysis was performed by using SPSS version 19.0 software. Continuous variables were presented as mean±standard deviation. Categorical variables were expressed as frequency and percentage. Comparisons between groups were performed using Student's t-test for continuous variables and the Chi-square test for categorical variables. The correlation coefficients (r) were assessed by Spearman correlation. The performance of non-invasive models was evaluated and compared by receiver operating characteristic (ROC) curves and the area under the ROC curves (AUROCs), Youden index was engaged to determine the optimal cut-off. P values <0.05 were considered statistically significant.

A total of 337 chronic hepatitis patients who had undergone liver biopsy were enrolled in the study. After exclusion of patients with prior or current antiviral treatment (n=101), hepatitis C virus co-infection (n=54), insufficient data (n=12), HCC (n=5), autoimmune hepatitis (n=3) and alcoholic liver disease (n=2), 160 patients were finally subjected to further analysis, thirty-eight of the patients had a fibrosis stage of S0, 42 had S1, 51 had S2, 14 had S3, and 13 had S4; for necroinflammation grade, 15 had G0, 46 had G1, 65 had G2, 31 had G3, and 3 had G4. The baseline characteristics for all subjects are shown in Table 1. Next, a comparison of baseline characteristics between patients who meet the criteria with either S>2 or G>2 and those who do not meet the criteria were performed (Table S1). The results revealed that in patients who meet the criteria, the levels of AST (114.7±120.1 vs 60.1±80.1IU/L, p=0.013), ALP (89.8±41.6 vs 70.7±24.2IU/L, p=0.011), GGT (85.0±68.4 vs 35.7±42.8IU/L, p=0.000) and total bilirubin (15.9±8.7 vs 12.8±6.1μmol/L, p=0.045) were significantly higher than that in patients who do not meet the criteria, while the levels of albumin (40.9±7.7 vs 44.7±6.6g/L, p=0.003), RBC (4.6±0.9 vs 4.9±0.7 1012/L, p=0.030) and platelet (167.7±65.8 vs 196.5±54.0 109/L, p=0.007) were significantly higher in patients who do not meet the criteria than that in patients who meet the criteria. However, HBV DNA levels between the two groups had no significant difference.

Table 2 shows a comparison of factors between patients (n=131, 82%) with absence or mild fibrosis (Stage 0–2) and patients (n=29, 18%) with significant fibrosis (Stage 3–4). In patients with absence or mild fibrosis, the levels of HBV DNA (6.0±1.8 vs 5.1±1.6 log10 IU/mL, p=0.012), albumin (44.4±6.5 vs 41.5±8.8g/L, p=0.049) and platelet count (197.9±55.8 vs 153.1±54.1 109/L, p<0.001) were significantly higher than that in patients with significant fibrosis, while the GGT level in patients with significant fibrosis was higher than that in patients with absence or mild fibrosis (73.0±61.8 vs 41.4±50.4IU/L, p=0.014). For established non-invasive fibrosis models, a higher score of Fib-4 (1.08±0.81 vs 2.15±1.82, p=0.004), GPR (0.39±0.49 vs 0.90±0.81, p=0.003) and RPR (0.07±0.03 vs 0.09±0.04) was observed in patients with significant fibrosis, while the APRI score showed no significant difference between the two groups.

A similar comparison of factors between subjects with low necroinflammation activity (Grade 0–2, n=126, 79%) and severe necroinflammation activity (Grade 3–4, n=34, 21%) is shown in Table 3. Patients with severe necroinflammation activity presented a higher level of ALT (282.4±339.1 vs 108.4±160.0IU/L, p=0.006), AST (141.2±130.8 vs 54.3±70.5IU/L, p=0.001), GGT (99.1±67.1 vs 33.1±39.6IU/L, p<0.001) and direct bilirubin (7.3±6.6 vs 4.7±3.1μmol/L, p=0.035), while a lower level of albumin (40.7±8.0 vs 44.7±6.5g/L, p=0.003), red cell count (4.6±1.0 vs 4.9±0.7 1012/L, p=0.036) and platelet count (171.6±67.1 vs 194.7±54.5 109/L, p=0.039). In addition, all the scores of APRI (2.26±2.06 vs 0.79±1.24, p<0.001), Fib-4 (2.24±1.76 vs 1.02±0.71, p<0.001), GPR (1.10±0.79 vs 0.31±0.39, p<0.001) and RPR (0.08±0.04 vs 0.07±0.03, p=0.044) were higher in patients with severe necroinflammation activity.

The associations between established models and liver fibrosis stages or necroinflammation grades were analyzed using Spearman test. Box plots of serum markers in relation to fibrosis stage and necroinflammation grade are presented in Figs. 1 and 2. Statistically significant positive correlations were observed between fibrosis stage and GGT (r=0.401, p<0.001), AST (r=0.339, p<0.001), RDW (r=0.286, p<0.001), GPR (r=0.444, p<0.001), APRI (r=0.424, p<0.001), RPR (r=0.286, p<0.001), Fib-4 (r=0.444, p<0.001), while a negative correlation with fibrosis stage was found in platelet count (r=−0.288, p<0.001). Similarly, necroinflammation grades were positively correlated with GGT (r=0.483, p<0.001), AST (r=0.542, p<0.001), RDW (r=0.402, p<0.001), GPR (r=0.499, p<0.001), APRI (r=0.585, p<0.001), RPR (r=0.314, p<0.001), Fib-4 (r=0.453, p<0.001) but negatively related to platelet count (r=−0.295, p<0.001).

Figure 1.

Box plots of gamma-glutamyl transpeptidase (GGT), aspartate transaminase (AST), red cell distribution width (RDW), platelet (Plt), GGT to platelet ratio (GPR), aspartate transaminase-to-platelet ratio index (APRI), RDW to platelet ratio (RPR), and fibrosis-4 index (Fib-4) according to fibrosis stage. r is the correlation coefficient.

(0.2MB).

Figure 2.

Box plots of gamma-glutamyl transpeptidase (GGT), aspartate transaminase (AST), red cell distribution width (RDW), platelet (Plt), GGT to platelet ratio (GPR), aspartate transaminase-to-platelet ratio index (APRI), RDW to platelet ratio (RPR), and fibrosis-4 index (Fib-4) according to necroinflammation grade. r is the correlation coefficient.

(0.19MB).

For predicting significant fibrosis, AUROC of GPR (0.761, 95% CI 0.671–0.850) was significantly higher than that of APRI (0.680, 95% CI 0.585–0.774, p=0.034), but comparable with RPR (0.766, 95% CI 0.667–0.865, p=0.906) and Fib-4 (0.746, 95% CI 0.656–0.836, p=0.703) (Table 4). The optimal cut-off value of GPR for the prediction of significant fibrosis was 0.274 (Table 4). Further analyses revealed that AUROC of GPR (0.726, 95% CI 0.607–0.846) was comparable with Fib-4 (0.728, 95% CI 0.606–0.850) and RPR (0.736, 95% CI 0.583–0.889) for predicting significant fibrosis in HBeAg positive patients, but in HBeAg negative patients, GPR (0.827, 95% CI 0.705–0.948) had the best performance (Table S2). For predicting severe necroinflammation activity, the performance of GPR (AUROC=0.869, 95% CI 0.800–0.937) was better than APRI (AUROC=0.816, 95% CI 0.740–0.892, p=0.085), RPR (AUROC=0.694, 95% CI 0.589–0.800, p<0.001) and Fib-4 (0.792, 95% CI 0.711–0.873, p=0.023) with the optimal cut-off value of 0.564 (Table 4). Further analyses confirmed that both in HBeAg positive (0.888, 95% CI 0.796–0.979) and negative (0.828, 95% CI 0.716–0.939) patients, GPR had the best performance for predicting severe necroinflammation activity (Table S2).

To assess the influence of liver inflammatory activity on diagnostic performance, a comparison between patients with ALT below or above 5× ULN was performed. As shown in Table S3, all of the non-invasive tests performed better in patients with ALT below 5× ULN than in those with ALT above 5× ULN for predicting either significant fibrosis or severe necroinflammation activity. In addition, GPR had the largest AUROC for predicting either significant fibrosis (0.809, 95% CI 0.715–0.903) or severe necroinflammation activity (0.858, 95% CI 0.762–0.954) in patients with ALT below 5× ULN.