Editor’s Note: Chronic hepatitis B (CHB) is the primary cause of chronic liver disease in China and other Asian countries, as well as a major contributor to liver-related morbidity and mortality. Additionally, with lifestyle changes in recent years, the prevalence of non-alcoholic fatty liver disease (NAFLD) has been steadily increasing. The prevalence of NAFLD in the Asian population is nearing 30%, becoming a significant cause of liver cirrhosis and liver cancer. Clinically, more and more patients with CHB combined with NAFLD are emerging, forming the main group of chronic liver disease in China and causing a substantial disease burden.
Patients with NAFLD often develop various metabolic disorders, especially diabetes mellitus (DM). However, it remains unclear whether DM is associated with liver inflammation, liver fibrosis, and cirrhosis in patients with CHB combined with NAFLD. At the same time, seeking a precise non-invasive diagnostic model to assess liver inflammation, fibrosis, and cirrhosis in patients with CHB combined with NAFLD has been a challenging clinical problem. It is still unknown whether machine learning (ML) can assist in clinical diagnosis and grading. At the 32nd Asia-Pacific Study of Liver Association Conference (APASL 2023) held in Taipei, China, from February 15 to 19, 2023, two studies by the team of Professor Wu Chao and Professor Li Jie from Nanjing University Medical School Affiliated Drum Tower Hospital, China, were selected for conference presentations, providing insightful answers to the aforementioned issues.
DM as an Independent Predictor of Significant Liver Inflammation or Fibrosis in Patients with CHB Combined with NAFLD (Abstract No.: FP12-65)
This study included patients with CHB combined with NAFLD who underwent liver pathology examinations in eight medical centers in China from April 2004 to October 2020. Using univariate and multivariate logistic regression analysis, the study explored the relationship between DM and significant liver inflammation (G2-G4) and significant liver fibrosis (S2-S4) (Figure 1-1).
Figure 1-1: Research Summary Graphic
The results included 869 patients with CHB combined with NAFLD (average age 40.6 ± 10.4 years, 79.9% male), of whom 71 (8.2%) had DM. The average body mass index (BMI) was 24.9 ± 3.3 kg/m^2, and the average HBV DNA level was 5.3 ± 2.0 log10 IU/mL. Approximately half of the patients (380, 46.3%) were HBeAg positive, and 42 patients (5.9%) were undergoing antiviral therapy. A significant proportion of patients (206, 24.3%) had moderate to severe NAFLD (grades 2-3). Most patients (529, 60.9%) exhibited significant liver inflammation (G2-G4), and about half (431, 49.6%) had significant liver fibrosis (S2-S4).
The analysis showed that patients with DM were more likely to have significant liver inflammation (76.1% vs 59.7%, P=0.02) or significant liver fibrosis (76.1% vs 47.3%, P<0.001, Figure 1-2) compared to those without DM.
Figure 1-2: Distribution of Significant Liver Inflammation and Fibrosis in Patients with Diabetes Mellitus vs. Non-Diabetic Patients
After adjusting for DM, hepatic steatosis, age, gender, BMI, HBV DNA levels, and HBeAg status in a multivariate logistic analysis, it was found that DM is independently associated with significant liver inflammation (OR 3.38; 95% CI 1.46-7.86; P=0.005) and significant liver fibrosis (OR 4.49; 95% CI 2.08-9.72; P<0.001) (Table 1-1).
Table 1-1: Analysis of Factors Related to Liver Inflammation or Fibrosis in Patients with CHB Combined with NAFLD
In summary, this study indicates that in the CHB combined with NAFLD population, patients with coexisting DM have a significantly higher risk of developing notable liver inflammation and fibrosis compared to those without DM. This risk is independent of age, gender, hepatic steatosis, other metabolic factors (such as BMI), and virological factors. Comprehensive management of DM should be integrated into the chronic care of CHB to reduce adverse liver outcomes.
Establishment and Validation of a Machine Learning-Based Diagnostic Model for Liver Fibrosis and Cirrhosis in Patients with CHB Combined with NAFLD (Abstract No.: FP08-41)
This study included patients with CHB combined with NAFLD who underwent liver pathology and laboratory examinations in eight medical centers in China from April 2004 to October 2020. Pearson correlation coefficients were used to explore the relationship between patient clinical characteristics and liver fibrosis staging. Eventually, 20 clinical features of patients were selected and included in a machine learning (ML) model [including random forests (RF), logistic regression (LR), and gaussian naive bayes (GNB)] to predict significant liver fibrosis (≥S2), advanced liver fibrosis (≥S3), and cirrhosis (S=4). The study also compared the diagnostic accuracy of the ML models with the fibrosis-4 score (FIB-4), aspartate aminotransferase to platelet ratio index (APRI), and NAFLD fibrosis score (NFS) for assessing the degree of liver fibrosis and cirrhosis.
A total of 790 patients with CHB combined with NAFLD who underwent liver pathology were included in the analysis. The average age of the patients was 40 years (range 33-48), with 633 males and 157 females. There were 392 patients (49.6%) with significant liver fibrosis (S≥2), 221 patients (28.0%) with advanced liver fibrosis (S≥3), and 101 patients (12.8%) with cirrhosis (S=4). All 20 clinical features of the patients were included in the ML models (Figure 2-1).
Figure 2-1: Correlation between Patient Clinical Characteristics and Liver Fibrosis Staging
In this analysis, the Random Forest (RF) model demonstrated higher diagnostic accuracy for significant liver fibrosis, advanced liver fibrosis, and cirrhosis, with Area Under the Curve (AUC) values of 0.723 (0.647-0.800), 0.768 (0.696-0.841), and 0.826 (0.761-0.891), respectively. These values were significantly higher than those for the fibrosis-4 score (FIB-4), aspartate aminotransferase to platelet ratio index (APRI), and NAFLD fibrosis score (NFS), with all P-values being less than 0.05 (Figure 2-2).
Figure 2-2: Comparison of AUC between Machine Learning Models and Non-invasive Diagnostic Scores
(A) Significant Fibrosis; (B) Advanced Fibrosis; (C) Cirrhosis
Furthermore, among the three machine learning (ML) models, the Random Forest (RF) model demonstrated superior diagnostic accuracy for the staging of liver fibrosis and cirrhosis in patients with CHB combined with NAFLD, compared to the fibrosis-4 score (FIB-4), aspartate aminotransferase to platelet ratio index (APRI), and NAFLD fibrosis score (NFS).
The findings of this research suggest that ML could serve as an effective tool in the clinical assessment of liver fibrosis and cirrhosis in patients with CHB combined with NAFLD.
Acknowledgments: Special thanks to Professor Junping Shi from Hangzhou Normal University Affiliated Hospital, Professor Qi Zheng from Fujian Medical University First Affiliated Hospital, Professor Qinglei Zeng from Zhengzhou University First Affiliated Hospital, and Professor Ze Bao He’s team from Taizhou Enze Medical Center for their support in this study!
Reference:
1. Jie Li, Fajuan Rui, Brian Nguyen, Qi Zheng, Qinglei Zeng, Zebao He, Junping Shi, Chao Wu, Mindie H. Nguyen. Diabetes mellitus (DM) is an independent predictor of significant inflammation or fibrosis in chronic hepatitis B (CHB) patients concurrent with nonalcoholic fatty liver disease (NAFLD). APASL 2023. Abrasts FP12-65.2. Jie Li, Yayun Xu, Fajuan Rui, Qi Xue, Qi Zheng, Qinglei Zeng, Zebao He, Yunliang Chen, Junping Shi, Chao Wu. Establishment and Validation of a diagnostic model for liver fibrosis and cirrhosis in chronic hepatitis B (CHB) concurrent with nonalcoholic fatty liver disease (NAFLD) based on machine learning (ML). APASL 2023. Abrasts FP08-41.
TAG: APASL 2023, Voice of China, HBV, NAFLD, Liver fibrosis
