Original Research••

Prevalence and prediction of hepatocellular carcinoma in alcohol-associated liver disease: a retrospective study of 136 571 patients with chronic liver diseases

Abstract

Background and aims To explore the incidence of alcohol-related hepatocellular carcinoma (HCC), evaluate possible synergisms between alcohol and well-known risk factors associated with HCC and establish a nomogram to predict alcohol-associated liver disease (ALD)-related HCC risk.

Methods A database of 136 571 inpatients in the Fifth Medical Center of Chinese PLA General Hospital from 2002 to 2018 with chronic liver disease was established. Data were collected by medical records review. Multivariate logistic regression was used to identify the independent high-risk factors associated with HCC, and then were incorporated into a novel nomogram. Afterward, the new established model was validated using external cohort by receiver operating characteristic curves analysis. For external cohort, 1646 patients with ALD admitted to our hospital from 2019 to 2021 were included. ALD was diagnosed on the basis of a history of sustained heavy alcohol intake greater than 40 g/day for men and 20 g/day for women for >5 years, clinical evidence of liver disease and supporting laboratory abnormalities.

Results Over the last 17 years, trends showed obviously increases in ALD. ALD-related HCC experienced a significant increase from 5.8% to 30.7%, whereas hepatitis B virus (HBV)-related HCC declined from 77.6% to 52.0%. In patients with ALD-related HCC (5119), 3816 (74.54%) cases had HBV infection, 493 (9.63%) cases had hepatitis C virus (HCV) infection, 71 (1.39%) cases were coinfected with both HBV and HCV, and 739 (14.44%) cases had neither HBV nor HCV infection. Drinking years (OR 1.009, 95% CI (1.000 to 1.017)), age (OR 1.060, 95% CI (1.051 to 1.069)), diabetes mellitus (OR 1.314, 95% CI (1.123 to 1.538)), HBV infection (OR 4.905, 95% CI (4.242 to 5.671)), liver cirrhosis (OR 4.922, 95% CI (3.887 to 6.232)) and male sex (OR 17.011, 95%CI (2.296 to 126.013)) were associated with increased risk of HCC in patients with ALD. A nomogram had a concordance index of 0.786 (95% CI 0.773 to 0.799) and had well-fitted calibration curves. These results were successfully validated both in the internal cohort and external cohort.

Conclusion The prevalence of ALD and ALD-related HCC has been increased dramatically. The nomogram model established here with its high accuracy and easy-to-use features achieved an optimal prediction of HCC development in patients with ALD, which can help clinicians to develop an individualised and precise treatment strategy.

What is already known on this topic

  • Existing research has established that the occurrence of alcohol-associated liver disease (ALD) and alcohol-associated liver cirrhosis has significantly risen. Alcoholic cirrhosis has been identified as a primary risk factor for patients with hepatocellular carcinoma (HCC). However, the prevalence and ability to predict HCC in ALD still require evaluation.

What this study adds

  • This study revealed that the prevalence of ALD-related HCC increased dramatically. Through the examination of six high-risk factors (drinking years, age, diabetes mellitus, hepatitis B virus infection, liver cirrhosis and male sex) as clarified in this study, a newly established and user-friendly nomogram model has been developed. This model demonstrates high accuracy and provides optimal predictions for the development of HCC.

How this study might affect research, practice or policy

  • This study provided the evidence that can help clinicians in formulating personalised and precise treatment strategies aimed at reducing the mortality rate associated with ALD-related HCC.

Introduction

Over the last 30 years, the incidence of liver cancer has decreased in high-incident countries1 2 and increased in low-incident countries.3 4 Viral hepatitis is considered the primary risk factor and leading cause of liver cancer. The increasing incidence in low-prevalence countries may be related to the increased hepatitis B and C virus (HBV and HCV) infections caused by migration from high-prevalence countries and injectable drug use. It has been estimated by the International Agency for Research on Cancer that East Asia and sub-Saharan Africa were the high incidence areas in 2012.5 The decreased incidence of high-prevalence areas is due to HBV vaccination in newborns.6

In contrast, with improved living conditions, the incidence of alcohol-associated liver disease (ALD) and alcohol-associated liver cirrhosis has clearly increased.7 ,8 Alcohol consumption is related to many chronic diseases, such as heart disease, vascular disease, liver cirrhosis and cancer. Alcohol consumption may directly cause approximately 10% and 3% of all cancer cases in males and females, respectively. A study at the Mayo Clinic showed that the main risk factor in 29% of patients with hepatocellular carcinoma (HCC) was alcoholic cirrhosis.9 Case–control studies from different countries have also demonstrated that chronic alcohol consumption is associated with an approximately twofold increased risk of HCC.10

Facing the rising incidence of alcohol-related HCC and declining incidence of viral hepatitis-related HCC, a retrospective study was conducted to explore the incidence of alcohol-related HCC, evaluate possible synergisms between alcohol consumption and well-known risk factors associated with HCC (eg, drinking years, age, sex, diabetes mellitus, liver cirrhosis, HBV or HCV infection), and establish a nomogram to predict the risk of ALD-related HCC.

Materials and methods

Study design

For this retrospective, non-interventional cohort study, we enrolled consecutive hospitalised patients admitted to the Fifth Medical Center of Chinese PLA General Hospital, the largest hepatology tertiary referral hospital in Beijing, China, from January 2002 to December 2018 with chronic liver diseases (CLDs) according to The International Classification of Diseases (ICD)-10 coding from electronic medical records.

All of the enrolled patients with ALD were randomly assigned to the training cohort for nomogram development or the validation cohort to confirm the model’s performance at a ratio of 3:1.

ALD was diagnosed on the basis of a history of sustained heavy alcohol intake greater than 40 g/day for men and 20 g/day for women for >5 years, clinical evidence of liver disease and supporting laboratory abnormalities.11 The diagnosis of HBV infection was determined by the presence of hepatitis B surface antigen or HBV DNA. Infection with HCV was defined by the presence of anti-HCV antibodies and previously or currently detectable HCV RNA. Liver cirrhosis was determined by the characteristics and physical examinations of the patients combining several diagnostic tools, such as serum biomarkers, imaging modalities, liver biopsy and endoscopy, as non-invasive alternatives to liver biopsy, which indicated cirrhotic F4 fibrosis. If patients with liver cirrhosis had a history of sustained heavy alcohol intake greater than 40 g/day for men and 20 g/day for women for >5 years, they were also considered to have ALD cirrhosis. HCC was screened by abdominal ultrasonography, and the serum alpha-fetoprotein level was determined every 3–6 months for patients with CLD. The diagnosis of HCC was dependent on pathology and/or dynamic hepatic imaging of CT or MRI demonstrating the presence of one or more liver lesions with enhancement on arterial phase and washout on the venous/delayed phase.12

Criteria for the diagnosis of diabetes13: first, symptoms of diabetes and a casual plasma glucose ≥200 mg/dL (11.1 mmol/L). Casual was defined as any time of day without regard to the time since the last meal. The classic symptoms of diabetes include polyuria, polydipsia and unexplained weight loss. Second, fasting plasma glucose ≥126 mg/dL (7.0 mmol/L). Fasting was defined as no caloric intake for at least 8 hours. Third, 2-hour plasma glucose ≥200 mg/dL (11.1 mmol/L) during an oral glucose tolerance test. The test should be performed as described by the WHO, using a glucose load containing the equivalent of 75 g anhydrous glucose dissolved in water.

Patients

The inclusion criteria were: (1) consecutive patients admitted to the Fifth Medical Center of Chinese PLA General Hospital, the largest hepatology tertiary referral hospital in Beijing, China, from January 2002 to December 2018; (2) diagnosed with CLD according to ICD-10 coding from electronic medical records, including chronic hepatitis, chronic hepatitis C, ALD, drug-induced liver injury (DILI), autoimmune hepatitis (AIH) and primary biliary cirrhosis (PBC); (3) if the patients with HBV or/and HCV infection had a history of sustained alcohol intake greater than 40 g/day for men and 20 g/day for women for >5 years, they were classified into the ALD group to evaluate the risk factors for HCC; (4) autoimmune-induced liver diseases (AILDs) that included AIH, PBC and primary sclerosing cholangitis; (5) other liver diseases (OLDs) that included non-alcoholic fatty liver disease (NAFLD), Wilson’s disease, Buddi-Chiari syndrome and liver injury for unknown reasons.

The exclusion criteria were: (1) acute liver diseases and (2) liver injury caused by other diseases (ie, heart, vascular disease and malnutrition). A database of 136 571 patients (including 33 371 patients with HCC) with CLD was established. Among them, 20 643 patients were diagnosed with ALD, and 5119 HCC cases were ALD-related HCC (figure 1). For external validation, 1646 eligible patients admitted to our hospital from 2019 to 2021 were included. The average age and drinking years of the external cohort were 55.1±9 years and 27.2±9.7 years, respectively, with 1606 (97.6%) males, 1 (0.1%) patient with diabetes mellitus, 690 (41.9%) patients with HBV infection and 1450 (88.1%) patients with liver cirrhosis.

Figure 1
Request permissionrequest permission icon
Figure 1

Schema of the study. AILD, autoimmune-induced liver disease (including autoimmune hepatitis, primary biliary cirrhosis and primary sclerosing cholangitis); ALD, alcohol-associated liver disease; DILI, drug-induced liver injury; HB, hepatitis B; HC, hepatitis C; HCC, hepatocellular carcinoma; OLD, other liver diseases (including non-alcoholic fatty liver disease, Wilson’s disease, Buddi-Chiari syndrome, liver injury with unknown reasons, etc).

In China, if patients with HBV or HCV infection have a history of heavy drinking, they would be diagnosed with HBV or HCV infection combined with ALD. In our study, to evaluate the risk factors for HCC in patients with ALD, we classified the patients with HBV and/or HCV infection with a history of sustained alcohol intake greater than 40 g/day for men and 20 g/day women for >5 years into the ALD group, and use the following terms: ALD with HBV infection, ALD with HCV infection, or ALD with HBV and HCV coinfection.

Data collection

All patient information, such as epidemiological, demographic and outcome data, was collected from medical records based on the ICD-10 coding system. The data in source documents were confirmed independently by at least two researchers.

Statistical analysis

Continuous variables were expressed as the means±SD and compared using the unpaired, two-tailed Student’s t-test or medians (IQR) and compared with the Mann-Whitney test. Categorical variables are presented as the numbers (percentage) and were compared by the Χ2 test. A p value of <0.05 was considered significant for all statistical tests. The statistical analyses were performed using R software, V.3.6.1 (http://www.r-project.org/). The performance of the novel nomogram model was assessed using receiver operating characteristic (ROC) curves, which are a plot of sensitivity versus 1−specificity for all possible cut-off values. The area under the ROC curve (AUROC) and optimal cut-off values were determined and assessed by the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) using MedCalc (V.19.0, MedCalc Software, Acacialaan 22, Ostend, Belgium).

The significance of each variable in the training cohort was assessed by univariate and multivariate logistic regression analyses to investigate the independent high-risk factors for HCC with its OR and 95% CI. All variables at a significant level were candidates to formulate a nomogram, which is based on proportionally converting each multivariate regression coefficient to a 0-point to 100-point scale. The predictive performance of the nomogram was measured by the concordance index (C-index) and calibration with 1000 bootstrap samples to decrease the overfit bias using the rms package in R. Then, the newly established model was validated using an external cohort by ROC analysis.

Results

Prevalence of ALD

Overall, the most common aetiology in the enrolled patients was hepatitis B, which declined from 68.8% in 2002 to 44.2% in 2018. This was followed by OLDs (such as Wilson’s disease, Buddi-Chiari syndrome, NAFLD and liver injury for unknown reasons) (14.0–23.6%), ALD (8.2–17.6%), hepatitis C (3.6–10.7%), DILI (1.5–6.7%), AILD (1.3–4.7%) and coinfection of HBV and HCV (0.2–1.3%) (figure 2A). Over the last 17 years, trends have shown obvious decreases in HBV cases and increases in ALD and OLD cases. HCV cases increased mildly before 2012 and then showed an observable decreasing tendency thereafter. There were no marked changes in the coinfection of HBV and HCV cases.

Figure 2
Request permissionrequest permission icon
Figure 2

Prevalence of chronic liver diseases (LDs) and HCC by year. (A) Change of aetiology in the entire enrolled patients with chronic liver diseases. (B) Change of aetiology in the patients with alcohol-associated liver disease (ALD). (C) Change of aetiology in the patients with HCC. (D) Change of aetiology in the patients with ALD-related HCC. AILD, autoimmune-induced liver disease; DILI, drug-induced liver injury; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus.

The proportion of patients with ALD with concurrent HBV infection decreased from 63.3% in 2002 to 41.7% in 2018, whereas ALD without HBV or HCV infection increased from 29.6% to 51.4%. ALD with HCV infection increased slightly before 2013 and then decreased gradually. The cases with coinfection of HBV and HCV in patients with ALD remained relatively stable (figure 2B).

Prevalence of HCC in ALD

In all patients with HCC from 2002 to 2018, ALD-related HCC experienced an overall increase from 5.8% to 30.7%, whereas HBV-related HCC continuously declined from 77.6% to 52.0%. There were no marked changes in patients with HCC caused by HCV, AILD (eg, autoimmune hepatitis, primary biliary cholangitis), DILI, OLDs (for example, Wilson’s disease, NAFLD and liver injury with unknown reasons) or coinfection of HBV and HCV (figure 2C).

Among patients with ALD-related HCC (5119), 3816 (74.54%) had HBV infection, 493 (9.63%) had HCV infection and 71 (1.93%) were coinfected with both HBV and HCV. ALD with concurrent HBV infection was still the most common reason for HCC during the study period, whereas the proportion decreased from 86.8% to 62.0%. ALD-related HCC with neither HBV nor HCV infection remained the second most common reason, increasing from 9.1% to 25.1% during the study period. ALD-related HCC with HCV infection seemed to increase slightly. Coinfection with both HBV and HCV showed no changes from 2002 to 2018 (figure 2D).

Independent high-risk factors for HCC in patients with ALD

In total, 20 643 patients with ALD were enrolled in this subanalysis; 75.0% (15 483 of 20 643) were assigned to the training cohort, and 25.0% (5160 of 20 643) were assigned to the validation cohort. There were no differences between these two cohorts (table 1). In the training cohort, 24.8% (3833 of 15 483) of patients developed HCC. To identify the high-risk factors, patients in the training or validation cohort were divided into a non-HCC group and an HCC group. Clinical characteristics were compared, and the results showed that age, male sex, diabetes mellitus, HBV infection, HBV and HCV coinfection, drinking years and liver cirrhosis were significantly different between the non-HCC and HCC groups in both the training and validation cohorts (table 2). HCV infection and body mass index (BMI) showed no differences. Further univariate and multivariate logistic analyses of the training cohort showed that drinking years (OR 1.009, 95% CI (1.000 to 1.017)), age (OR 1.060, 95% CI (1.051 to 1.069)), diabetes mellitus (OR 1.314, 95% CI (1.123 to 1.538)), HBV infection (OR 4.905, 95% CI (4.242 to 5.671)), liver cirrhosis (OR 4.922, 95% CI (3.887 to 6.232)) and male sex (OR 17.011, 95% CI (2.296 to 126.013)) were associated with an increased risk of HCC in patients with ALD (table 3).

Table 1
|
Characteristics of patients with ALD
Table 2
|
Characteristics of patients in training and validation cohorts
Table 3
|
High-risk factors associated with ALD-HCC

Predictive nomogram for the probability of HCC

A predictive nomogram was formulated based on the above independent high-risk factors associated with HCC development and validated using bootstrapping internally in the training cohort and externally in the validation cohort. The nomogram demonstrated good accuracy in estimating the risk of HCC development, with a C-index of 0.786 (95% CI 0.773 to 0.799) and a total score of 0–240 points. In addition, calibration plots graphically showed good agreement between the estimated and actual HCC development analysed by bootstrap sampling (figure 3).

Figure 3
Request permissionrequest permission icon
Figure 3

Nomogram to estimate the risk of HCC incidence in patients with ALD. (A) The nomogram-based prediction scoring model established by using logistic regression. (B) Validity of the predictive performance of the nomogram internal in the training cohort. (C) Validity of the predictive performance of the nomogram external in the validation cohort. ALD, alcohol-associated liver disease; C-index, concordance index; DM, diabetes mellitus; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; LC, liver cirrhosis.

ROC analysis was used to assess the performance of this nomogram model both in the internal and external cohorts, and the AUROC showed no difference between these two cohorts (0.782 (95% CI 0.773 to 0.792) vs 0.756 (95% CI 0.735 to 0.777), p=0.056) (figure 4A).

Figure 4
Request permissionrequest permission icon
Figure 4

Validation and prediction performance of the nomogram scores in patients with ALD. (A) Validation of the nomogram scores. The nomogram score of each patient in both internal and external cohort was calculated and plotted by ROC. (B) HCC risk stratification by the nomogram. Patients with ALD could be categorised into low-risk group (0–90 points) with less than 5% probability of HCC, intermediate-risk group (91–160 points) with about 20.0–50% probability of HCC and high-risk group (161–240 points) with over 50% probability of HCC. ALD, alcohol-associated liver disease; AUC, area under the curve; HCC, hepatocellular carcinoma; NPV, negative predictive value; PPV, positive predictive value; ROC, receiver operating characteristic curve.

In order to facilitate the clinical use, we defined the low cut-off value to roll out HCC with relatively high NPV (sensitivity), and the high cut-off value to indicate HCC with relatively high PPV (specificity). Using a cut-off value of 90 points to roll out HCC, the NPV (95% CI) was 98.2% (95.3% to 99.3%) in the internal cohort and 97.7% (85.6% to 99.7%) in the external cohort. Using a cut-off value of 140 points to indicate HCC, the PPVs (95% CI) were 54.6% (52.7% to 56.4%) in the internal cohort and 64.0% (60.3% to 67.4%) in the external cohort. Other detailed parameters are also shown (figure 4A).

Using this nomogram, patients with ALD could be categorised into a low-risk group (0–90 points) with less than 5% probability of HCC, a medium-risk group (91–160 points) with approximately 20.0% probability of HCC and a high-risk group (161–240 points) with over 50% probability of HCC (figure 4B).

Discussion

HCC, which is the sixth most common malignancy, has become the third most frequent cause of cancer-related death globally.14 The distribution of HCC is uneven around the world. Approximately 80% of HCC cases occur in either sub-Saharan Africa or Eastern Asia. In 2020, approximately 50% of all new cases occurred in China.5 The present study was a retrospective study of 136 571 patients with CLDs. The prevalence of HCC in patients with ALD was analysed, and a novel nomogram for the prediction of HCC was established. To the best of our knowledge, this is the largest cohort study of its kind.

First, chronic HBV infection is an important risk factor for the development of HCC.15 HBV may cause both fibrosis and cirrhosis and directly lead to the development of HCC.16 In recent years, the decreasing incidence of HBV infection has been attributable to the wide use of newborn HBV vaccination and antiviral drugs.17 This is consistent with the results of our study. In recent years, increasing attention has been given to AILD, DILI and OLDs. Their proportions to all CLDs are increasing gradually (figure 2A). In our study, the percentage of AILD-related HCC changed from 0.2% to 1% (figure 2C). It has been recognised that obesity and its associated metabolic syndrome are risk factors for HCC.18 In line with this, the ratio of OLD-related HCC fluctuated between 4% and 10% in our study. In all patients with HCC, no obvious changes were observed in the proportion of DILI-related HCC.

Chronic alcohol consumption may lead to liver fibrosis, cirrhosis and HCC. There are no nationwide epidemiological investigations of ALD in China. However, the drinking population and incidence of ALD showed increasing trends in a regional epidemiological study.8 19 In our study, the patient preselection might have resulted in high number of HCC cases among the patients with ALD, because it was the constituent ratio but not morbidity of ALD-related HCC in inpatients. But the trend showed that from 2002 to 2018, the proportion of ALD-related HCC to all patients with HCC increased considerably.

Heavy alcohol consumption and concurrent HCV or HBV infection showed a synergistic effect by some evidence. These factors may work together to promote liver cirrhosis and increase the risk of HCC. In our study, the proportion of patients with ALD-related HCC with concurrent HBV infection decreased dramatically, while ALD-related HCC with neither HBV nor HCV infection increased. A possible explanation for this is that the prevalence of HBV infection decreased, and the heavy drinking population increased, as discussed above. Coinfection with both HBV and HCV and its related HCC probably remained relatively stable due to the minority infection population.

Second, the present study clarified that longer years of drinking, older age, diabetes mellitus, HBV infection, liver cirrhosis and male sex were the high-risk factors for ALD-related HCC by multivariate analysis, which is in agreement with recent findings.20–22 Although BMI and HCV infection were considered as important risk factors in HCC, there were no differences between the non-HCC and HCC groups in both the training and validation cohorts in our study. A well-established risk factor for HCC is increasing age, which may cause the accumulation of more concomitant lifestyle, behavioural, or environmental risk factors and increase the possibility of developing cirrhosis.23 Therefore, it is clear that the risk of HCC increases as the drinking population ages and the alcohol consumption and drinking time increase. The incidence of HCC in males is higher than that in females, with an average ratio between 2:1 and 4:1. It is hypothesised that the development of HCC is influenced by androgens rather than exposure to risk factors for different sexes. Endogenous non-environmental factors, such as higher androgenic hormone levels and higher BMI, may play a negative role in males. Diabetes is also a moderately strong risk factor for HCC according to available data.24 A large cohort study of patients with and without diabetes on the incidence of HCC indicated that the incidence of HCC doubled in patients with diabetes.25 Some evidence has shown the synergistic effect of heavy alcohol intake with HCV or HBV, which may operate together to promote the development of HCC. Because of a positive synergistic effect, it is not difficult to imagine that HBV is a risk factor for ALD-related HCC. Donato et al reported that the risk of HCC increased in a linear pattern with daily intake of more than 60 g in alcohol drinkers. However, HCC risk doubled in patients with concomitant HCV infection compared with those with alcohol drinking alone, which meant a positive synergistic effect.26 However, HCV infection was excluded as a risk factor in both univariate and multivariate analyses in our study. In the development of liver cirrhosis among patients with HCV, host and environmental factors may be more important than viral factors. Host and environmental factors include heavy alcohol ingestion (50 g/day), male sex, older age at the time of acquisition of infection, obesity, diabetes and coinfection with HIV.27 There is no strong evidence to support the notion that HCV viral factors, including load, genotype or quasi-species, are important in the development of cirrhosis or HCC. Whether coinfection with HBV and HCV may increase the HCC risk remains to be determined. Some meta-analyses have suggested that coinfection with HBV and HCV had an additive effect on the risk of HCC.28 29 However, a meta-analysis also showed that the risk of HCC did not increase in HBV–HCV coinfection,30 which was consistent with our study. Another reason might be the small sample size of patients with HBV–HCV coinfection in the present study (0.9%).

Third, in the present study, a novel nomogram model scoring from 0 to 240 points was established based on logistic regression. The nomogram demonstrated good accuracy in estimating the risk of HCC development, with a C-index of 0.786 (95% CI 0.773 to 0.799). Patients with ALD could be categorised into low-risk, medium-risk and high-risk groups for HCC development using this nomogram and could be managed more conveniently and efficiently to improve surveillance and treatment.

Our study had several limitations that should be considered. First, it was a preliminary cohort study and data collection was limited by the ICD coding system. We did not analyse other potential risk factors such as smoking, genetics and dietary factors, which have been reported in earlier papers. Second, the nomogram model we developed requires validation in a prospective study to confirm its reliability. Third, adding other specific biochemical parameters could improve the sensitivity and specificity of the model. Finally, we did not focus on NAFLD-related HCC in this study, as it is difficult to identify patients with NAFLD in a retrospective analysis. However, we are currently conducting relevant examinations to collect appropriate data for future research on NAFLD and NAFLD-related HCC. We will report the results in another paper.

In summary, this large-scale retrospective study demonstrated that, while chronic HBV infection remained the most common aetiology among patients with CLD and HCC, there was noticeable decrease in the number of HBV cases and increase in ALD cases. The prevalence of ALD-related HCC increased dramatically during the study period. Based on six high-risk factors (drinking years, age, diabetes mellitus, HBV infection, liver cirrhosis and male sex) clarified in the present study, a novel established and easy-to-use nomogram model with high accuracy achieved an optimal prediction of HCC development, which can help clinicians develop an individualised and precise treatment strategies to reduce the mortality of ALD-related HCC.

Overview

Article metrics
Altmetricopen-url
Dimensionsopen-url