A new model predicts hepatocellular carcinoma in patients with HBV-related decompensated liver cirrhosis and long-term antiviral therapy: a prospective study

Patients and primary endpoint

From May 2010 to December 2021, 307 patients with LC admitted to the Third People’s hospital of Changzhou, were recruited, and were prospectively followed up. HBV-related compensated LC and DC were diagnosed according to Chinese guidelines for prevention and treatment of CHB (2019 version) (Jia et al., 2020). Patients with nodules in the hepatic parenchyma found in histological or ultrasonographic examination, or gastroesophageal varices detected by endoscopic evaluation, were diagnosed with compensated LC. DC and its stages were defined by the complications including ascites, encephalopathy, variceal bleeding, or hepatorenal syndrome (D’Amico, Garcia-Tsao & Pagliaro, 2006; D’Amico et al., 2018). HCC was diagnosed according to the imaging techniques, alpha-fetoprotein and/or histological findings (Health Commission of the People’s Republic of China, 2020). Patients co-infected with other hepatitis virus or suffered from malignant tumor, were excluded. Patients who had positive HBV DNA at the end of follow-up, developed HCC within 6 months during follow-up, or lost to follow-up were also excluded.

All the patients received NAs treatment, including Lamivudine, Adefovir, Telbivudine, Entecavir or Tenofovir at admission, and routinely underwent clinical examination, laboratory tests and ultrasonography every 3 to 6 months. Demographic and clinical data, including age, sex, alanine transaminase (ALT), aspartate transaminase (AST), total bilirubin (TBil), gamma-glutamyl transpeptidase (GGT), albumin, international standard ratio (INR), HBV serologic markers, serum HBV DNA, blood cell count, and complications were collected at the first time of admission. The endpoints of the study were HCC development or death.

The study was non-interventional and not harmful to the patients, and written consents were obtained from all the participants. The protocol was approved by the Ethics Committee of the Third People’s Hospital of Changzhou according to the Declaration of Helsinki, 2013 (No. CZSY2018-0601).

Score systems

Risk scores including CAMD (Hsu et al., 2018), PAGE-B (Papatheodoridis et al., 2016), mPAGE-B (Kim et al., 2018), aMAP (Fan et al., 2020), ALBI score (Casadei Gardini et al., 2019) were calculated as described before.

Statistical analysis

All data were analyzed using SPSS version 25.0 (Armonk, NY, USA). Continuous variables were presented as median (interquartile range, IQR), and were analyzed using the Mann-Whitney U tests. Categorical values were presented as frequencies, and were compared using the chi-square test. Correlation analysis was evaluated using the Spearman correlation test. Independent risk factors for HCC development were identified using univariate and multivariate logistic regression analysis. Accuracy of the scoring systems was compared according to the area under the receiver operating characteristic curve (AUROC), which was calculated using MedCalc version 15.2.2 software for Windows (Medcalc Software, Mariakerke, Belgium). The cutoff value was identified using MedCalc software, and then Kaplan-Meier analysis was performed using GraphPad Prism version 5.0 for Windows (GraphPad Software, San Diego, CA, USA). P value < 0.05 was considered statistically significant.

Characteristics of patients

Among the 307 patients, 78 patients were excluded, including 58 patients lost to follow-up, three patients developed HCC during 6 months of follow-up, 13 patients had positive HBV DNA at the last visit, one patient coinfected with hepatitis C virus, two patients suffered from lung cancer, and one patient suffered from kidney tumor. Data from 229 patients were analyzed, including 94 individuals with compensated LC and 135 individuals with DC (Fig. 1). Of the 135 patients with DC, 12 patients had hepatic encephalopathy, six patients had variceal bleeding, 121 patients had ascites, and one patient had renal failure. A total of 28 patients were stratified as late decompensated state based on the refractory ascites, recurrent encephalopathy, and renal failure.

The median follow-up duration was 37 (28–66) months. A total of 9 (9.57%) patients in the compensated LC group and 39 (28.89%) patients in the DC group developed HCC. The incidence of HCC was higher in the DC group ( $\mathcal{X}$² = 12.478, P < 0.01).

The characteristics of the 48 patients who developed HCC are shown in Table 1. Patients who developed HCC were older than those without HCC (P < 0.01). CAMD, PAGE-B, mPAGE-B and aMAP scores were higher, while the platelet count was lower in patients with HCC (P < 0.01).

Independent risk factors for HCC development

As shown in Table 2, the univariate logistic analysis showed that age, DC status and platelet were associated with HCC development. Multivariable analysis showed that age and DC status were independent risk factors for HCC development (both P < 0.01). Then a prediction models were developed: Model (Age_DC) (y = 1) = $\frac{\mathrm{e}\mathrm{x}\mathrm{p}\left(-3.658+0.049\ast \mathrm{A}\mathrm{g}\mathrm{e}-1.107\ast \mathrm{D}\mathrm{C}\left(1\right)\right)}{1+\mathrm{e}\mathrm{x}\mathrm{p}\left(-3.658+0.049\ast \mathrm{A}\mathrm{g}\mathrm{e}-1.107\ast \mathrm{D}\mathrm{C}\left(1\right)\right)}$. Considering that TBil and PLT were important indexes of several non-invasive models, another model incorporating TBil and PLT into Model (Age_DC), Model (Age_DC_PLT_TBil) (y = 1) = $\frac{\mathrm{e}\mathrm{x}\mathrm{p}\left(-2.406+0.051\ast \mathrm{A}\mathrm{g}\mathrm{e}-1.090\ast \mathrm{D}\mathrm{C}\left(1\right)-0.022\ast \mathrm{T}\mathrm{B}\mathrm{i}\mathrm{l}-0.008\ast \mathrm{P}\mathrm{L}\mathrm{T}\right)}{1+\mathrm{e}\mathrm{x}\mathrm{p}\left(-2.406+0.051\ast \mathrm{A}\mathrm{g}\mathrm{e}-1.090\mathrm{D}\mathrm{C}\left(1\right)-0.022\mathrm{T}\mathrm{B}\mathrm{i}\mathrm{l}-0.008\ast \mathrm{P}\mathrm{L}\mathrm{T}\right)}$, was developed. The AUROC of the two models were compared, and Model (Age_DC_PLT_TBil) was more potent than Model (Age_DC) (AUROC: 0.760 and 0.718). For patients with compensated LC, AUROC of Model (Age_DC_PLT_TBil) was larger than Model (Age_DC) (AUROC: 0.778 and 0.684). For patients with DC, AUROC of Model (Age_DC_PLT_TBil) was also larger than Model (Age_DC) (AUROC: 0.691 and 0.629). Patients who developed HCC had significantly higher Model (Age_DC_PLT_TBil) scores than those patients who did not develop HCC (Z = 5.538, P < 0.01).

Comparison of ALBI, aMAP, CAMD, PAGE-B, mPAGE-B and Model (Age_DC_PLT_TBil)

As shown in Fig. 2, the AUROC of ALBI, aMAP, CAMD, PAGE-B, mPAGE-B and Model (Age_DC_PLT_TBil) scores were 0.512, 0.667, 0.638, 0.663, 0.679 and 0.760, respectively. There was no significant difference in AUROC between CAMD, aMAP, PAGE-B and mPAGE-B (all P > 0.05). Moreover, AUROC of Model (Age_DC_PLT_TBil) was larger than the other five models (all P < 0.05). Moreover, the AUROC of Child-Pugh score in predicting HCC development was just 0.529, which was comparable with ALBI.

For patients with compensated LC, AUROC of ALBI, aMAP, CAMD, PAGE-B, mPAGE-B and Model (Age_DC_PLT_TBil) were 0.548, 0.641, 0.662, 0.665, 0.637 and 0.778, respectively. Model (Age_DC_PLT_TBil) had larger AUROC than mPAGE-B (P < 0.01) and ALBI (P = 0.02).

For patients with DC, AUROC of ALBI, aMAP, CAMD, PAGE-B, mPAGE-B and Model (Age_DC_PLT_TBil) were 0.574, 0.623, 0.607, 0.642, 0.638 and 0.691, respectively. AUROC of Model (Age_DC_PLT_TBil) seemed to be larger than those of ALBI (P = 0.05) and CAMD (P = 0.07).

Correlation between ALBI, aMAP, CAMD, PAGE-B, mPAGE-B and Model (Age_DC_PLT_TBil)

Model (Age_DC_PLT_TBil) positively correlated with CAMD, PAGE_B, mPAGE_B and aMAP (all P < 0.01), but not with ALBI (P = 0.99 and 0.10) in patients with compensated LC or DC. Moreover, for patients with compensated LC or DC, ALBI correlated with mPAGE_B and aMAP (all P < 0.05), but not with other three models (Fig. 3).

Risk stratification for cumulative incidence of HCC

With an optimal cut-off value of 0.236, Model (Age_DC_PLT_TBil) achieved 70.83% sensitivity, 76.24% specificity. Then the patients were divided into two groups: the low-risk group (Model (Age_DC_PLT_TBil) <0.236) and high-risk group (Model (Age_DC_PLT_TBil) ≥0.236). Patients with DC in high-risk group had a poor prognosis (P < 0.01) (Fig. 4).