Controlled attenuation parameter value-based diagnostic algorithm improves the accuracy of liver stiffness measurement in chronic hepatitis B patients

Patient characteristics

A total of 441 CHB patients who had a liver biopsy and LSM for primary liver disease were recruited for this study, and 330 CHB patients were ultimately enrolled in the study (Figure 1).

Figure 1. Flowchart of the study population selection.

The mean age was 36 years, with a predominance of males (61.21%). Overall, 41.52% of the patients had significant liver necroinflammation (G2-4), and 31.21% of the patients had significant fibrosis (S2-4). The baseline characteristics of the patients are shown in Table 1.

Factors associated with CAP

CAP was significantly correlated with sex (r=-0.216, P <0.001), ALT (r=0.232, P<0.001), aspartate aminotransferase (AST) (r=0.151, P=0.006), glutamyl transpeptidase (GGT) (r=0.384, P<0.001), alkaline phosphatase (ALP) (r=0.207,P<0.001), total cholesterol(TC) (r=0.125, P=0.024), liver fibrosis (r=0.119, P=0.03), body mass index (BMI) (r=0.663, P <0.001) and LSM (r=0.279, P <0.001), while there was no correlation with age, total bilirubin(TB) or liver necroinflammation (P>0.05 for all). Based on the univariate linear regression analysis, sex (male), ALT, GGT, ALP, BMI and LSM were all correlated with CAP (P<0.05). However, only BMI was independently significantly correlated with CAP (P<0.01) (Table 2) by multivariate linear regression analysis.

Factors associated with LSM

The median LSM values were 6.63 (5.25-8.22) for S0, 6.67 (5.28-8.93) for S1, 9.32 (7.08-12.38) for S2, 10.66 (8.14-17.8) for S3 and 16.45 (7.88-21.24) for S4. LSM was significantly correlated with ALT (r=0.329, P<0.001), AST (r=0.292, P<0.001), GGT (r=0.358, P<0.001), ALP (r=0.161, P=0.003), TB (r=0.101, P=0.046), liver necroinflammation (r=0.337, P<0.001), liver fibrosis (r=0.435, P<0.001), CAP (r=0.279, P<0.001) and BMI (r=0.264, P<0.001); there was no correlation with age, sex or total cholesterol (P>0.05 for all).

Univariate and multivariate linear regression analyses were performed to identify factors correlated with LSM in the entire CHB cohort and in the fibrosis stage subgroups of S0-1 and S2-4 (Table 3). LSM values were significantly correlated with ALT (P<0.001) and GGT (P=0.007) in the entire cohort. ALT (P=0.004) and CAP (P=0.041) were significantly correlated with LSM values in the S0-1 subgroup but not in the S2-4 subgroup.

Effect of CAP on LSM values

Because CAP was associated with LSM, the effect of CAP on LSM values in the same stages of liver fibrosis was evaluated. Based on the CAP value, the patients were further classified into the normal CAP group (< 248 dB/m) and the elevated CAP group (≥ 248 dB/m). The median LSM values were 6.42 and 8.41 kPa (P<0.001) in CHB patients at stage S0; 6.50 kPa and 7.96 kPa (P=0.02) in CHB patients at stage S1; 9.32 kPa and 9.28 kPa (P=0.29) in CHB patients at stage S2; 10.66 kPa and 13.38 kPa (P=0.51) in CHB patients at stage S3; and 10.26 kPa and 19.11 kPa (P=0.20) in CHB patients at stage S4 (Figure 2). Higher LSM values were correlated with higher CAP values, especially in lower-stage patients (S0-1). This finding suggested that the LSM values could be overestimated in CHB patients with higher CAP values.

Figure 2. LSM values in normal and elevated CAP groups according to fibrosis stage.

Effect of CAP on LSM diagnostic performance

The study further evaluated the effect of CAP on LSM diagnostic performance. In the entire CHB cohort, the area under the receiver operating characteristic (AUROC) curve of LSM for the diagnosis of S2-4 was 0.77 (0.72-0.83). Using the Youden Index of transient elastography at 8.9 kPa, the sensitivity and specificity were 64.08% and 79.30%, respectively. For S3-4, the AUROC of LSM was 0.77 (0.69-0.85). Using the Youden Index of transient elastography at 10.1 kPa, the sensitivity and specificity were 61.36% and 82.87%, respectively (Table 4, upper).

When considering the effect of CAP, the AUROC of LSM for the diagnosis of S2-4 was 0.78 (0.71-0.85) in the normal CAP group and 0.72 (0.61-0.84) in the elevated CAP group. When a cutoff value of 8.9 kPa was used, the sensitivity and specificity of LSM were 61.64% and 84.57% in the normal CAP group and 70% and 59.62% in the elevated CAP group, respectively. The diagnostic accuracy was 77.82% in the normal CAP group and 63.41% in the elevated CAP group. For the diagnosis of S3-4, the AUROC of LSM was 0.77 (0.68-0.87) in the normal CAP group and 0.80 (0.66-0.94) in the elevated CAP group. When a cutoff value of 10.1 kPa was used, the sensitivity and specificity of LSM were 58.82% and 87.38% in the normal CAP group and 70% and 69.44% in the elevated CAP group, respectively. The diagnostic accuracy was 83.47% in the normal CAP group and 69.51% in the elevated CAP group. In summary, patients with normal CAP values had a higher diagnostic accuracy than patients with elevated CAP values. CAP influenced the LSM diagnostic performance.

Optimal cutoff values of LSM based on different CAP values

The Youden Index and at least 90% sensitivity and 90% specificity were used as the optimal cutoff values for LSM to evaluate the liver fibrosis stage [15]. The values for the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and diagnostic accuracy (DA) are listed in Table 4. In the normal CAP group, for S≥S1, the cutoff values ranged from 4.5-9.0 kPa; for S≥S2, the cutoff values ranged from 5.5-9.5 kPa; for S≥S3, the cutoff values ranged from 5.9-10.9 kPa; and for S=S4, the cutoff values ranged from 6.0-12.1 kPa. In the elevated CAP group, for S≥S1, the cutoff values ranged from 5.7-14.8 kPa; for S≥S2, the cutoff values ranged from 8.0-12.6 kPa; for S≥S3, the cutoff values ranged from 8.3-19.0 kPa; and for S=S4, the cutoff values ranged from 16.4-20 kPa.

Diagnostic flowchart using LSM accounting for CAP

Based on the results of the above analysis, we developed a liver fibrosis diagnostic flowchart using the LSM and CAP. For patients with normal CAP values, an LSM of ≤5.5 kPa was defined as no significant fibrosis (no or mild fibrosis), an LSM of 5.5-9.5 kPa was defined as a gray area, and an LSM of > 9.5 kPa was defined as severe fibrosis or cirrhosis. For patients with elevated CAP values, an LSM of ≤8 kPa was defined as no or mild fibrosis, an LSM of 8.0-14.8 kPa was defined as a gray area, and an LSM of >14.8 kPa was defined as severe fibrosis or cirrhosis (Figure 3).

Figure 3. Diagnostic flowchart for diagnosis and exclusion liver fibrosis in chronic hepatitis B virus infection based on LSM and CAP values.

Comparison of diagnostic performance using LSM, CAP or ALT level

The ALT-based LSM algorithm was recommended to diagnose or exclude severe liver fibrosis using LSM [5]. In the study, 179 patients had normal ALT levels, and 151 patients had elevated ALT levels. Based on the ALT-based LSM algorithm, in the normal ALT group, 64 patients had no or mild fibrosis (LSM<6 kPa), 74 patients had an uncertain diagnosis (6

Based on the CAP-based LSM algorithm, 248 patients had a normal CAP value, and 82 patients had an elevated CAP value in the study. Based on the CAP algorithm, the normal CAP group included 66 patients with no or mild fibrosis (LSM≤5.5 kPa), 126 patients with an uncertain diagnosis (5.59.5 kPa). The elevated CAP group included 29 patients with no or mild fibrosis (LSM<8 kPa), 37 patients with an uncertain diagnosis (8.014.8 kPa). We found that 79 (31.85%) patients had no significant fibrosis or a correct diagnosis of severe liver fibrosis/cirrhosis, 126 (50.81%) patients had an ambiguous diagnosis, and 43 (17.34%) patients had a misdiagnosis in the normal CAP group. In the elevated CAP group, 32 (39.02%) patients had no significant fibrosis or a correct diagnosis of severe liver fibrosis/cirrhosis, 37 (45.12%) patients had an ambiguous diagnosis, and 13 (15.85%) patients had a misdiagnosis in the elevated CAP group (Figure 4).

Figure 4. The comparison of diagnostic performance between the ALT-based LSM algorithm and CAP-based LSM algorithm. The comparison of diagnostic accuracy (A) and diagnosis error rate (B) in different subgroups. The comparison of diagnostic accuracy (C) and diagnostic error rate (D) of the ALT-based LSM algorithm and CAP-based LSM algorithm.

By comparing the diagnostic performance between the CAP-based LSM algorithm and the ALT-based LSM algorithm, we found that 112 (33.94%) patients had no significant fibrosis or a correct diagnosis of severe liver fibrosis/cirrhosis, 151 (45.76%) patients had an ambiguous diagnosis, and 67 (20.30%) patients had a misdiagnosis based on the ALT algorithm. However, based on the CAP algorithm, we found that 111 (33.64%) patients had no significant fibrosis or a correct diagnosis of severe liver fibrosis/cirrhosis, 163 (49.39%) patients had an ambiguous diagnosis, and 56 (16.97%) patients had a misdiagnosis. In this study, we found that the CAP-based LSM algorithm had a similar correct diagnosis rate and a lower misdiagnosis rate than the ALT-based LSM algorithm (Figure 4).

Patients

CHB patients who underwent liver biopsy and transient elastography from January 2015 to August 2018 at Shengjing Hospital affiliated with China Medical University were enrolled in the study.

The enrolled CHB patients were positive for hepatitis B surface antigen (HBsAg) and HBV DNA for ≥6 months with persistently or intermittently abnormal ALT levels (> 40 IU/L). The exclusion criteria were as follows: (1) age less than 18 years; (2) other causes of liver disease; (3) other comorbid diseases (tumors, hematological diseases or human immunodeficiency virus infection); (4) significant alcohol consumption (> 20 g/d for women and > 30 g/d for men) for more than 5 years; (5) insufficient clinical data; and (6) serum ALT levels more than 5 times the upper limit of normal (UNL) (≥200 U/L).

Ethics approval and consent

The study was approved by the Medical Ethics Committee of Shengjing Hospital of China Medical University (2018PS544K). All procedures in studies involving human participants were performed in accordance with the ethical standards of the institutional and/or national research committee and the 1975 Declaration of Helsinki and its later amendments or comparable ethical standards. Written informed consent was waived as this was a retrospective study using de-identified and aggregate data for analysis.

Transient elastography

LSM and CAP were measured using transient elastography following the manufacturer’s instructions and were performed by trained and experienced operators. LSM value was defined as a valid measurement of at least 10 shots. LSM value was defined as unreliable with less than 10 valid shots, a success rate <60% and/or an interquartile range/median value (IQR/M) ≥30% of the measurement [24]. All enrolled patients underwent LSM within 6 months before liver biopsy. Hepatic steatosis was defined as more than 248 dB/m [25].

Liver biopsy

Liver tissues were obtained by an experienced physician with strict adherence to the liver biopsy protocol. A liver sample was considered adequate if it was longer than 15 mm and contained six portal tracts or more. The specimens were fixed in formalin, processed for routine histopathological analysis, and reviewed by two experienced pathologists who were unaware of the patient’s clinical information.

The histopathological evaluation was performed by two trained and experienced pathologists who were blinded to the patient’s clinical data and transient elastography results. In case of disagreement, they reviewed the slides together to reach consensus. Liver fibrosis stage was classified from 0 to 4 according to Scheuer’s classification (for staging fibrosis: S0 = fibrosis was absent, S1 = portal fibrosis without septa, S2 = portal fibrosis with few septa, S3 = severe fibrosis, and S4 = cirrhosis; significant fibrosis was defined as stage S2-4, advanced fibrosis as S3-4, and cirrhosis as S4). Liver necroinflammation grade was classified from 0 to 4 according to Scheuer’s classification (for necroinflammation: G0 = no or minimal portal inflammation, G1 = portal inflammation, lobular inflammation without necrosis, G2 = mild piecemeal portal necrosis, focal lobular necrosis or acidophil bodies, G3 = moderate piece meal portal necrosis, severe local lobular cell damage, and G4 = severe piecemeal portal necrosis, damage includes bridging necrosis) [26].

Clinical and laboratory data

Clinical data including age, sex, weight, height, dyslipidemia and alcohol consumption were recorded for all included patients. The day before liver biopsy, a 12-hour overnight fasting blood sample was collected to detect the levels of ALT, AST, ALP, GGT, triglycerides, TB, and total lipoprotein cholesterol. Serum HBsAg and hepatitis B e-antigen (HBeAg) were detected by microparticle enzyme immunoassay, and HBV DNA was measured using a TaqMan polymerase chain reaction assay.

Statistical analysis

The results analysis was performed using SPSS, version 22.0 (SPSS Inc., Chicago, IL, United States). Data are reported as the medians (interquartile range, IQR) for continuous and nonnormally distributed variables and as frequencies or percentages for categorical variables. Spearman’s or Pearson’s correlation analysis was performed to analyze the relationship between CAP or LSM and clinical data. Data were compared between two groups of continuous variables using the Mann-Whitney U test. The diagnostic performance of transient elastography for the diagnosis of different stages of fibrosis was described by the AUROC, 95% confidence intervals (95%CIs), sensitivity (Sn), specificity (Sp), PPV, NPV and DA. A two-sided P < 0.05 was considered to indicate statistical significance.