ABSTRACT
Background and Aims
Liver cancer is concurrently influenced by a number of risk factors, such as chronic viral hepatitis, heavy alcohol use, and metabolic associated conditions. This study aims to examine whether there are interaction patterns among factors associated with liver cancer risk among patients with diabetes.
Methods
In this retrospective cohort study, patients who received care for type 2 diabetes in general outpatient clinics between 2010 and 2019 were identified using electronic health records of Hong Kong. Patients were followed up until a cancer diagnosis, death, or December 31st, 2019. The interaction patterns among factors associated with liver cancer risk were examined using conditional inference survival tree.
Results
A total of 1,995 and 1,969 patients with and without liver cancer were included. Chronic viral hepatitis status appeared as most dominant factor in differentiating the risk of liver cancer. In the absence of chronic viral hepatitis, low lipids and elevated ALT were identified as key risk factor in older females (> 63years) and males respectively. Differential thresholds across different age groups in males were found, with a higher threshold for the younger group (39 U/L for ≤ 53years vs 31 U/L for > 53years).
Conclusion
Chronic viral hepatitis, sex, age, ALT, and lipids appear to exhibit interaction patterns on liver cancer incidence among patients with diabetes, providing potential targets for public health prevention strategies.
Keywords: diabetes, interaction, liver cancer, public health, risk factor
1. Introduction
Liver cancer ranks third in cancer morality worldwide [1]. Major risk factors include chronic infection with hepatitis B/C virus, heavy alcohol use, metabolic conditions (obesity or diabetes), aging, and male sex [2, 3]. Prior research has shown that patients with diabetes are 2.5 times likely to develop liver cancer when compared to those without diabetes [4].
While a number of risk factors for liver cancer have been identified [2, 3], liver cancer is concurrently influenced by the presence of one or more risk factors. For example, it has been previously shown that obesity, alcohol use, and tobacco use may synergistically contribute to an increased risk of liver cancer [5]. Differential patterns of alcohol and tobacco use, together with the role of sex hormones, may partially explain why males are at higher risk of developing liver cancer [2, 3]. Moreover, metabolic syndrome is associated with an elevated risk of liver cancer [6]. However, metabolic syndrome is a cluster of factors and different components are interrelated. In addition, serum alanine aminotransferase (ALT) is a commonly used liver function indicator to detect liver disease [7]. However, beyond liver injury, other conditions such as age [7], sex [7], diabetes [8, 9], non‐hepatic medical conditions [10], and drug use [10], may also influence the level of ALT.
Furthermore, changing lipid profile may occur as a result of the presence of one or more age‐related risk factors. Adipose tissue dysfunction and lipid dysregulation occur in obesity [11, 12], diabetes [11, 12], aging [13], liver disease [14], and liver cancer [14]. The presence of these factors is characterized by a state of chronic low‐grade inflammation [11, 12, 13, 14], which may create an environment conducive to liver cancer development. Nevertheless, in the literature, the associations between lipid profile and liver cancer remain inconclusive [14, 15, 16]. While lower levels of triglycerides and low‐density lipoprotein (LDL) cholesterol are associated with favorable atherosclerotic cardiovascular outcomes, several studies have found inverse associations of triglycerides [17] or total cholesterol [17, 18] with the risk of liver cancer. This implies the complexity of resulting altered lipid profile associated with liver cancer risk due to changing lipid metabolism.
Given i) diabetes being associated with a higher risk of liver cancer; ii) the interdependent relationships among multiple factors associated with liver cancer risk; and iii) the lack of unifying strategies to identify high‐risk patients in the absence of liver diseases [19], this study aims to examine whether interaction patterns among factors associated with liver cancer risk among patients with diabetes exist using survival tree analysis approach.
2. Methods
2.1. Study Design and Setting
A retrospective cohort study was performed using electronic health records of Hong Kong's public healthcare system managed by the Hospital Authority (HA). Disease diagnoses were coded according to the International Classification of Disease 9th or 10th revision (ICD‐9 or ICD‐10), or the International Classification of Primary Care 2nd edition (ICPC‐2). Data were accessed via HA Data Collaboration Lab during 2023 to 2024.
2.2. Patients
Patient selection procedures were reported in a previous study [20] conducted among the same diabetes cohort. In brief, adult patients who were diagnosed with type 2 diabetes (ICPC‐2: T90) with a known self‐reported date of physician‐diagnosed diabetes and underwent a diabetes complication screening assessment in general outpatient clinics between 2010 and 2019 were identified. Clinical profiles of the patients were assessed during a first assessment. Those with a history of malignancy were excluded. Patients were followed up until a cancer diagnosis, death, or December 31st, 2019, whichever occurred earlier. To minimize reverse causality, those with less than 6 months of follow‐up were excluded [21]. In addition, since the diagnosis of one cancer may influence the diagnosis of another cancer [21], those who developed cancer at sites other than the liver were excluded.
For analyzes, patients who developed liver cancer (n = 1995) and an equal number of random subset of patients who did not develop cancer during follow‐up were selected. Those who did not receive any ALT test were excluded (n = 26). Finally, a total of 1995 and 1969 patients with and without liver cancer were included.
2.3. Covariates
Candidate split variables were demographics (sex and age), duration of diabetes, disease history (chronic hepatitis B/C, liver cirrhosis, ischemic heart disease, cerebrovascular disease, heart failure, hypertension, chronic kidney disease, chronic obstructive pulmonary disease, pneumonia, and family history of diabetes), medication use (anti‐diabetic drugs, statins, aspirin, non‐steroidal anti‐inflammatory drugs, anti‐coagulants, anti‐platelets, and anti‐hypertensive drugs), behavioral factors (alcohol and tobacco use), anthropometric measurements (body mass index and waist‐to‐hip ratio), and laboratory measurements (ALT, HbA1c, fasting glucose, LDL cholesterol, high‐density lipoprotein cholesterol, triglycerides, and serum creatinine). Disease history was defined as whether patients had received a diagnosis of a disease at baseline. Anti‐diabetic drugs included metformin, sulfonylurea, insulin, and dipeptidyl peptidase‐4 inhibitors, since these drugs were commonly used among the study diabetes cohort. Medication use was defined as whether patients had been prescribed a drug at baseline. Laboratory results were taken from most recent measurements to the assessment date.
2.4. Outcome
The outcome of interest was diagnosis of liver cancer (ICD‐9: 155; ICD‐10: C22) during follow‐up.
2.5. Data Analysis
Conditional inference survival tree [22, 23] was applied to examine the interaction patterns among factors associated with liver cancer risk among patients with diabetes. The interaction patterns among risk factors for colorectal [24] and lung cancers [25] among the same diabetes cohort were previously reported. The entire set of covariates was considered as candidates for splits. The maximum depth and α level of the tree model were set at 4 and 0.01 respectively. The cumulative incidence of identified subgroups with distinct characteristics was graphically examined. Model performance was evaluated using area under the curve (AUC) as metric. In post‐hoc analyzes, the adjusted hazard ratios (aHRs) of selected split variables between comparison nodes were estimated using Cox regression. Data analyzes were conducted using R packages party and pec (software version 4.2.3; R Foundation for Statistical Computing, Vienna, Austria).
3. Results
A total of 384,121 patients were initially identified. During a median follow‐up of 6.2 years (IQR 3.3–8.0), 1,995 patients developed liver cancer. The overall liver cancer incidence was 0.92 per 1000 person‐years. The presence of chronic viral hepatitis emerged as most important risk factor for liver cancer. In the absence of preexisting chronic viral hepatitis, sex and age appeared as important factors in differentiating the risk of liver cancer. In older females and males, lipid and ALT levels were identified as key factor in differentiating the risk of liver cancer respectively (Figure 1).
Figure 1.
Survival tree diagram for liver cancer risk among patients with diabetes.
3.1. Presence Versus Absence of Chronic Hepatitis B/C
The presence of preexisting chronic viral hepatitis appeared as strongest risk factor for liver cancer. Among patients with chronic viral hepatitis, the majority were diagnosed with chronic hepatitis B alone (82.18%, 309/376), followed by chronic hepatitis C alone (17.02%, 64/376), with the rest being co‐infected with hepatitis B and C viruses (0.80%, 3/376) (Table 1). Patients with chronic viral hepatitis had a higher risk of developing liver cancer than those without chronic viral hepatitis (aHR 3.58, 95% CI: 3.18–4.02), controlling for age, sex, and duration of diabetes (Table 2).
Table 1.
Baseline characteristics of three distinct subgroups of patients by chronic viral hepatitis status and sex.
| Chronic viral hepatitis | No chronic viral hepatitis, female | No chronic viral hepatitis, male | ||||
|---|---|---|---|---|---|---|
| Characteristics | (n = 376) | (n = 1433) | (n = 2155) | |||
| Number of liver cancer cases during follow‐up, n (%) | 352 | (93.62%) | 446 | (31.12%) | 1197 | (55.55%) |
| Demographics | ||||||
| Male, n (%) | 299 | (79.52%) | 0 | (0%) | 2155 | (100%) |
| Age at assessment in year, mean ± SD | 62.23 | ±8.50 | 66.01 | ±12.10 | 63.98 | ±10.78 |
| Duration of diabetes in year, median (IQR) | 6 | (2‐10.25) | 5 | (1‐11) | 4 | (1‐10) |
| Disease history | ||||||
| Chronic hepatitis B only, n (%) | 309 | (82.18%) | 0 | (0%) | 0 | (0%) |
| Chronic hepatitis C only, n (%) | 64 | (17.02%) | 0 | (0%) | 0 | (0%) |
| Coinfection of chronic hepatitis B and C, n (%) | 3 | (0.80%) | 0 | (0%) | 0 | (0%) |
| Liver cirrhosis, n (%) | 157 | (41.76%) | 58 | (4.05%) | 96 | (4.45%) |
| Ischemic heart disease, n (%) | 25 | (6.65%) | 76 | (5.30%) | 199 | (9.23%) |
| Cerebrovascular disease, n (%) | 23 | (6.12%) | 78 | (5.44%) | 166 | (7.70%) |
| Heart failure, n (%) | 6 | (1.60%) | 40 | (2.79%) | 53 | (2.46%) |
| Hypertension, n (%) | 311 | (82.71%) | 1273 | (88.83%) | 1876 | (87.05%) |
| Chronic kidney disease, n (%) | 51 | (13.56%) | 159 | (11.10%) | 292 | (13.55%) |
| Chronic obstructive pulmonary disease, n (%) | 6 | (1.60%) | 10 | (0.70%) | 25 | (1.16%) |
| Pneumonia, n (%) | 28 | (7.45%) | 44 | (3.07%) | 84 | (3.90%) |
| Family history of diabetes, n (%) | 165 | (43.88%) | 618 | (43.13%) | 910 | (42.23%) |
| Medication use | ||||||
| Anti‐diabetic drugs | ||||||
| Metformin, n (%) | 141 | (37.50%) | 663 | (46.27%) | 1041 | (48.31%) |
| Sulfonylurea, n (%) | 147 | (39.10%) | 479 | (33.43%) | 765 | (35.50%) |
| Insulin, n (%) | 81 | (21.54%) | 106 | (7.40%) | 146 | (6.77%) |
| Dipeptidyl peptidase‐4 inhibitors, n (%) | 16 | (4.26%) | 55 | (3.84%) | 69 | (3.20%) |
| Glucosidase inhibitors, n (%) | 3 | (0.80%) | 9 | (0.63%) | 12 | (0.56%) |
| Meglitinide, n (%) | 0 | (0%) | 1 | (0.07%) | 0 | (0%) |
| Glitazone, n (%) | 0 | (0%) | 5 | (0.35%) | 9 | (0.42%) |
| Sodium‐glucose cotransporter‐2 inhibitors, n (%) | 0 | (0%) | 2 | (0.14%) | 5 | (0.23%) |
| Glucagon‐like peptide‐1 receptor agonizts, n (%) | 0 | (0%) | 1 | (0.07%) | 0 | (0%) |
| Any anti‐diabetic drugs, n (%) | 252 | (67.02%) | 832 | (58.06%) | 1311 | (60.84%) |
| Statins, n (%) | 88 | (23.40%) | 659 | (45.99%) | 858 | (39.81%) |
| Aspirin, n (%) | 60 | (15.96%) | 271 | (18.91%) | 532 | (24.69%) |
| Non‐steroidal anti‐inflammatory drugs, n (%) | 175 | (46.54%) | 809 | (56.45%) | 1023 | (47.47%) |
| Anti‐coagulants, n (%) | 14 | (3.72%) | 63 | (4.40%) | 122 | (5.66%) |
| Anti‐platelets, n (%) | 15 | (3.99%) | 61 | (4.26%) | 132 | (6.13%) |
| Anti‐hypertensive drugs, n (%) | 234 | (62.23%) | 1065 | (74.32%) | 1510 | (70.07%) |
| Behaviors | ||||||
| Current drinker/ex‐drinker, n (%) | 161 | (42.82%) | 163 | (11.37%) | 1045 | (48.49%) |
| Current smoker/ex‐smoker, n (%) | 179 | (47.61%) | 96 | (6.70%) | 1263 | (58.61%) |
| Anthropometric measurements | ||||||
| Body mass index in kg/m2, mean ± SD | 25.88 | ±4.15 | 26.12 | ±4.35 | 26.02 | ±4.06 |
| Waist‐to‐hip ratio, mean ± SD | 0.95 | ±0.06 | 0.93 | ±0.07 | 0.96 | ±0.06 |
| Laboratory measurements | ||||||
| Alanine aminotransferase in U/L, mean ± SD | 47.02 | ±44.63 | 28.46 | ±30.14 | 40.72 | ±59.24 |
| HbA1c in %, mean ± SD | 7.47 | ±1.51 | 7.34 | ±1.39 | 7.48 | ±1.59 |
| Fasting glucose in mmol/L, mean ± SD | 7.89 | ±2.62 | 7.54 | ±2.12 | 7.70 | ±2.31 |
| Low‐density lipoprotein cholesterol in mmol/L, mean ± SD | 2.49 | ±0.79 | 2.68 | ±0.82 | 2.59 | ±0.75 |
| High‐density lipoprotein cholesterol in mmol/L, mean ± SD | 1.29 | ±0.41 | 1.35 | ±0.36 | 1.24 | ±0.34 |
| Triglycerides in mmol/L, mean ± SD | 1.22 | ±0.68 | 1.56 | ±1.08 | 1.49 | ±1.20 |
| Serum creatinine in µmol/L, mean ± SD | 92.36 | ±81.95 | 72.89 | ±32.95 | 90.54 | ±30.99 |
Abbreviation: HbA1c, glycated hemoglobin.
Table 2.
Comparisons of selected split variables between comparison nodes.
| Comparison i, Metformin use in younger female | Comparison ii, LDL cholesterol in older female | Comparison iii, ALT in younger male | Comparison iv, ALT in older male | Comparison v, Chronic viral hepatitis status | ||
|---|---|---|---|---|---|---|
| Node | Characteristics | aHR (95% CI) | aHR (95% CI) | aHR (95% CI) | aHR (95% CI) | aHR (95% CI) |
| Node 5 | Absence of chronic viral hepatitis, female aged ≤ 63 years with metformin use | 1.10 (0.70–1.75) | ||||
| Node 6 | Absence of chronic viral hepatitis, female aged ≤ 63 years without metformin use | 1 | ||||
| Node 8 | Absence of chronic viral hepatitis, female aged > 63 years with LDL cholesterol ≤ 2.531 mmol/L | 2.01 (1.62–2.48) | ||||
| Node 9 | Absence of chronic viral hepatitis, female aged > 63 years with LDL cholesterol > 2.531 mmol/L | 1 | ||||
| Node 12 | Absence of chronic viral hepatitis, male aged ≤ 53 years with ALT ≤ 39 U/L | 1 | ||||
| Node 13 | Absence of chronic viral hepatitis, male aged ≤ 53 years with ALT > 39 U/L | 4.80 (3.14–7.32) | ||||
| Node 15 | Absence of chronic viral hepatitis, male aged > 53 years with ALT ≤ 31 U/L | 1 | ||||
| Node 16 | Absence of chronic viral hepatitis, male aged > 53 years with ALT > 31 U/L | 2.37 (2.10–2.68) | ||||
| Node {5,6,8,9,12,13,15,16} | Absence of chronic viral hepatitis | 1 | ||||
| Node 17 | Presence of chronic viral hepatitis | 3.58 (3.18–4.02) |
Note: All comparisons were adjusted for age and duration of diabetes. For comparison ii and v, statin use and sex was additionally controlled respectively. aHR in bold indicates statistical significance at an alpha level of 0.05.
Abbreviations: aHR, adjusted hazard ratio; ALT, alanine aminotransferase; LDL, low‐density lipoprotein.
3.2. Absence of Chronic Hepatitis B/C, Male Sex, Age, and Alanine Aminotransferase
In the absence of chronic viral hepatitis, male sex was identified as most important risk factor for liver cancer. Across both sexes, age symmetrically appeared as strongest risk factor for liver cancer. Males were subsequently separated into younger (≤ 53 years) and older (> 53 years) age groups to differentiate their risk of liver cancer. Across two age groups, ALT symmetrically emerged as most dominant factor in differentiating the risk of liver cancer. Nevertheless, differential thresholds of elevated ALT for increased liver cancer risk were identified (> 39 U/L for the younger; > 31 U/L for the older) (Figures 1 and 2). Among younger and older males, those with elevated ALT had a higher risk of developing liver cancer (aHRs for younger males: 4.80, 95% CI = 3.14–7.32; older males: 2.37, 95% CI = 2.10–2.68), adjusting for age and duration of diabetes (Table 2).
Figure 2.
Cumulative liver cancer incidence across four distinct subgroups of patients with distinctive characteristics in comparison with patients with preexisting chronic viral hepatitis. (A) Older male; (B) Younger male; (C) Older female; (D) Younger female. CVH, chronic viral hepatitis; LDL‐C, low‐density lipoprotein cholesterol. Shading indicates 95% confidence interval.
3.3. Absence of Chronic Hepatitis B/C, Female Sex, Age, and Low‐Density Lipoprotein Cholesterol
On the other hand, in the absence of chronic viral hepatitis, females were partitioned into younger (≤ 63 years) and older (> 63 years) age groups to differentiate their risk of liver cancer. In older females, LDL cholesterol appeared as key factor in differentiating the risk of liver cancer (Figures 1 and 2). Older females with a lower level of LDL cholesterol (≤ 2.531 mmol/L) had a higher risk of developing liver cancer than those with a higher level of LDL cholesterol (> 2.531 mmol/L) (aHR 2.01, 95% CI: 1.62–2.48), controlling for age, duration of diabetes, and statin use. Nevertheless, no split variable was identified as key factor for liver cancer among younger females after adjusting for age and duration of diabetes (Table 2).
3.4. Model Performance
The AUCs of the tree model at 2, 5, and 7 years were 0.705 (95%CI: 0.695–0.715), 0.697 (95%CI: 0.687–0.707), and 0.696 (95%CI: 0.686–0.706) respectively.
4. Discussion
Using survival tree analysis approach, the current study identified the interaction patterns among chronic viral hepatitis, sex, age, liver enzyme, and lipid profile on liver cancer incidence within a diabetes cohort in primary care. While chronic viral hepatitis plays a dominant role in determining the risk of liver cancer, in the absence of preexisting chronic viral hepatitis, differential profiles of ALT and lipids across age‐sex subgroups may become important factors in differentiating the risk of liver cancer. In addition, the optimal cutoffs for age, ALT, and LDL cholesterol to differentiate the risk of liver cancer in the absence of chronic viral hepatitis among subgroups of the study diabetes cohort were identified.
In the absence of chronic viral hepatitis, differential thresholds of elevated ALT were identified as key factor in differentiating the risk of liver cancer across younger and older males, where a higher threshold was suggested for the younger. While elevated ALT is an indicator of liver damage [26], findings support that the threshold may depend on age, sex, and preexisting medical conditions. In a recent study conducted among the German population, strong age‐dependency was observed for males, but much weaker for females [27]. Two studies (cross‐sectional [28] and longitudinal) [29] from the same cohort in the United States also found a decreasing trend of ALT with age after controlling for multiple confounders. In an earlier study [30] performed in Isreal, ALT exhibited an inverted U‐shaped association with age, peaking at around 50 years old. Other previous research [7] suggested that ALT may increase up to the fourth decade of life and decrease afterwards. The decrease in ALT threshold from younger (≤ 53 years) to older (> 53 years) males in the present study was consistent with the declining trends observed in previous studies. In addition, the thresholds identified in males (> 39 U/L for ≤ 53 years; > 31 U/L for > 53 years) were largely consistent with the literature [7, 31, 32]. Findings of the present study may provide evidence to support differential thresholds to differentiate the risk of liver cancer by age, sex, and preexisting liver disease status.
Moreover, in the absence of chronic viral hepatitis, decreased LDL cholesterol appeared as important factor associated with an increased risk of liver cancer among older females. While higher LDL cholesterol is known to be associated with adverse atherosclerotic cardiovascular outcomes, its association with liver cancer risk remains controversial. Under liver disease conditions such as fatty liver disease or chronic viral hepatitis, lipid metabolism of the liver may change, resulting in altered circulating lipid profile [14, 15]. Prior research demonstrated mixed findings on the associations between cholesterol and liver cancer risk [14, 15, 16]. Some studies suggested that lower levels of triglycerides [17] or total cholesterol [17, 18] could be associated with an elevated risk of liver cancer. Nevertheless, the observed inverse association between LDL cholesterol and liver cancer risk among older females in the absence of chronic viral hepatitis was consistent with the general opposite associations between lipids and several site‐specific cancers (including the liver) found in a previous study [33] among the same diabetes cohort. It was found that on average, every 1 mmol/l‐increment of LDL cholesterol was associated with a 25% decreased risk of liver cancer, accounting for age, sex, duration of diabetes, smoking, alcohol use, metabolic factors, disease history and medication use [33]. Possible mechanisms include impaired liver function resulting in inhibition of cholesterol synthesis, and weakened immune function associated with low lipid levels [16]. Adipose tissue dysfunction, hepatic fat accumulation, and changing lipid profile due to lipid dysregulation in diabetes [11, 12, 13], aging [34, 35], and liver conditions [14, 15] may simultaneously influence the risk of liver cancer. Findings of the current study suggest that changing lipid profile in the presence of diabetes, aging, and potential underlying tumor growth could be associated with the risk of liver cancer.
Findings of the present study are consistent with those of another study [36] on the comparisons of interaction patterns between patients with incident diabetes and those without preexisting diabetes from the same population. Specifically, in another study [36], it was shown that in the absence of chronic viral hepatitis, sex became the dominant factor in differentiating the risk of liver cancer, where ALT and statin use were identified as key factor in determining the risk of liver cancer among females and males respectively. Furthermore, differential thresholds of ALT for elevated liver cancer risk were found in younger and older females using 54 years as cutoff. Several similarities in the interaction patterns among risk factors for liver cancer across two studies were shown: i) chronic viral hepatitis remains the most dominant risk factor; ii) sex appears as the most important factor in the absence of chronic viral hepatitis; iii) differential thresholds of ALT for elevated risk across younger and older individuals within the same sex were identified, where higher thresholds were observed for younger individuals, and separately, for females; iv) the cutoff age for differential thresholds was 53 and 54 years across two sexes in two separate studies; and v) statin use or lipid level were identified as key factor. While two separate studies demonstrated consistent interaction patterns among risk factors, several key differences in the study design include i) the present study included a prevalent diabetes cohort receiving routine diabetes management in primary care, whereas another study included an incident diabetes cohort from the broader general population; ii) the homogeneity of patients included in the present study; iii) the availability of alcohol use and adiposity indicators (body mass index and waist‐to‐hip ratio) as candidate split variables in the present study but not in another study. Overall, the consistent interaction patterns across two separate studies with different target study populations and study design support findings of the present study. The exchange of ALT and lipid (or lipid‐lowering drugs) as most dominant factor in differentiating the risk of liver cancer across both sexes in two separate studies may imply the generalizability of the importance of differential ALT thresholds and lipid profile across both sexes.
There are some potential implications of the present study. First, the predominance of chronic viral hepatitis over sex and age in differentiating the risk of liver cancer may imply the more important role of physiological age than chronological age in determining the risk of liver cancer. Second, the differential thresholds of ALT for elevated liver cancer risk across age groups in males may imply the potential usefulness of establishing age‐sex thresholds for identifying susceptible liver cancer cases early among patients with diabetes who are not known to have chronic viral hepatitis or preexisting liver disease. Third, the emergence of lipids as key factor in differentiating the risk of liver cancer may imply changing lipid profile in liver cancer risk among patients with preexisting diabetes. The inverse association between LDL cholesterol and liver cancer risk may warrant further investigation. Overall, from clinical perspectives, the identification of ALT and lipid levels as key factors in determining the risk of liver cancer may imply the potential use of ALT and lipid levels as stratification factor, alone or with other factors, for further economic evaluation and public health policy making. Moreover, in the absence of chronic viral hepatitis, the emergence of ALT and lipid levels as dominant factor in differentiating the risk of liver cancer in males and older females respectively may imply the differential thresholds of ALT and lipid levels for elevated liver cancer risk by age, sex, and liver disease status, and the thresholds of lipid levels for increased liver cancer risk and cardiovascular risk may differ. Establishing differential thresholds by age, sex, and liver disease status may provide more precise estimates of liver cancer risk levels for risk stratification or targets for cancer prevention. Furthermore, chronic viral hepatitis, age, sex, ALT and lipids may collectively characterize subgroups of patients with distinct clinical profiles across the risk spectrum, providing potential target groups for public health policy.
Several limitations may potentially exist in the present study. First, information on serological testing for hepatitis B or C virus was not available and there could be some undiagnosed asymptomatic cases. Nevertheless, the diagnosis of liver diseases (cirrhosis and chronic viral hepatitis) and liver function indicator (ALT) were included as candidate split variables. In addition, chronic viral hepatitis status and ALT were identified as key variables for liver cancer risk prediction. Patient profile of liver disease status and liver function indicator partially reflects their liver health status and severity of chronic viral hepatitis, if present. Second, information on use of antiviral treatments for hepatitis B/C was not available. Third, information on aflatoxin exposure and fat content of the liver [3] was not available. Fourth, baseline profile of laboratory parameters such as lipids and ALT may change over time, and changing profile during follow‐up was not captured. Fifth, the performance of the tree model was moderate. Future research may explore the availability of a more comprehensive list of input variables in order to enhance its performance. Sixth, the study diabetes population is largely homogeneous Chinese. Future research is warranted to examine generalizability of the findings to other ethnic or geographical populations due to genetic and environmental influences.
5. Conclusions
The current study suggests the interaction patterns among chronic viral hepatitis, sex, age, ALT, and lipids on liver cancer incidence among patients with diabetes. While chronic viral hepatitis is a known dominant risk factor for liver cancer, in the absence of chronic viral hepatitis, differential profiles of ALT and lipids across age‐sex subgroups may potentially become key factors in differentiating the risk of liver cancer. Findings of the study may provide potential targets for public health prevention strategies.
Author Contributions
Sarah Tsz Yui Yau: conceptualization, methodology, writing – original draft, data curation, formal analysis. Eman Yee Man Leung: writing – review and editing. Chi Tim Hung: writing – review and editing. Martin Chi Sang WONG: writing – review and editing. Ka Chun CHONG: writing – review and editing. Albert Lee: writing – review and editing. Eng Kiong Yeoh: writing – review and editing, supervision.
Ethics Statement
Ethics approval for secondary data analysis was provided by the Joint Chinese University of Hong Kong – Survey and Bahavioral Research Ethics Committee (reference number: SBRE‐22‐0386; year of approval: 2022). Patient consent was waived since individuals were not identifiable in this study.
Conflicts of Interest
The authors declare no conflicts of interest.
Transparency Statement
The lead author Sarah Tsz Yui Yau, Eng Kiong Yeoh affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.
Acknowledgments
The authors have nothing to report.
Data Availability Statement
Data is not available for sharing due to access restriction.
References
- 1. Bray F., Laversanne M., Sung H., et al., “Global Cancer Statistics 2022: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries,” CA: A Cancer Journal for Clinicians 74, no. 3 (2024): 229–263. [DOI] [PubMed] [Google Scholar]
- 2. El–Serag H. B. and Rudolph K. L., “Hepatocellular Carcinoma: Epidemiology and Molecular Carcinogenesis,” Gastroenterology 132, no. 7 (2007): 2557–2576. [DOI] [PubMed] [Google Scholar]
- 3. Llovet J. M., Kelley R. K., Villanueva A., et al., “Hepatocellular Carcinoma,” Nature Reviews Disease Primers 7, no. 1 (2021): 6. [DOI] [PubMed] [Google Scholar]
- 4. El–Serag H. B., Hampel H., and Javadi F., “The Association Between Diabetes and Hepatocellular Carcinoma: A Systematic Review of Epidemiologic Evidence,” Clinical Gastroenterology and Hepatology 4, no. 3 (2006): 369–380. [DOI] [PubMed] [Google Scholar]
- 5. Marrero J. A., Fontana R. J., Fu S., Conjeevaram H. S., Su G. L., and Lok A. S., “Alcohol, Tobacco and Obesity are Synergistic Risk Factors for Hepatocellular Carcinoma,” Journal of Hepatology 42, no. 2 (2005): 218–224. [DOI] [PubMed] [Google Scholar]
- 6. Jinjuvadia R., Patel S., and Liangpunsakul S., “The Association Between Metabolic Syndrome and Hepatocellular Carcinoma: Systemic Review and Meta‐Analysis,” Journal of Clinical Gastroenterology 48, no. 2 (2014): 172–177. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Kim B. K., Han K. H., and Ahn S. H., “Normal” Range of Alanine Aminotransferase Levels for Asian Population,” Journal of Gastroenterology and Hepatology 26, no. 2 (2011): 219–220. [DOI] [PubMed] [Google Scholar]
- 8. Kunutsor S. K., Apekey T. A., and Walley J., “Liver Aminotransferases and Risk of Incident Type 2 Diabetes: A Systematic Review and Meta‐Analysis,” American Journal of Epidemiology 178, no. 2 (2013): 159–171. [DOI] [PubMed] [Google Scholar]
- 9. Wang Y. L., Koh W. P., Yuan J. M., and Pan A., “Association Between Liver Enzymes and Incident Type 2 Diabetes in Singapore Chinese Men and Women,” BMJ Open Diabetes Research and Care 4, no. 1 (2016): e000296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Pratt D. S. and Kaplan M. M., “Evaluation of Abnormal Liver‐Enzyme Results in Asymptomatic Patients,” New England Journal of Medicine 342, no. 17 (2000): 1266–1271. [DOI] [PubMed] [Google Scholar]
- 11. Guilherme A., Virbasius J. V., Puri V., and Czech M. P., “Adipocyte Dysfunctions Linking Obesity to Insulin Resistance and Type 2 Diabetes,” Nature Reviews Molecular Cell Biology 9, no. 5 (2008): 367–377. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Hajer G. R., van Haeften T. W., and Visseren F. L. J., “Adipose Tissue Dysfunction in Obesity, Diabetes, and Vascular Diseases,” European Heart Journal 29, no. 24 (2008): 2959–2971. [DOI] [PubMed] [Google Scholar]
- 13. Ou M. Y., Zhang H., Tan P. C., Zhou S. B., and Li Q. F., “Adipose Tissue Aging: Mechanisms and Therapeutic Implications,” Cell Death and Disease 13, no. 4 (2022): 300. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Paul B., Lewinska M., and Andersen J. B., “Lipid Alterations in Chronic Liver Disease and Liver Cancer,” JHEP Reports 4, no. 6 (2022): 100479. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Jiang J., Nilsson‐Ehle P., and Xu N., “Influence of Liver Cancer on Lipid and Lipoprotein Metabolism,” Lipids in Health and Disease 5 (2006): 4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Zhou F. and Sun X., “Cholesterol Metabolism: A Double‐Edged Sword in Hepatocellular Carcinoma,” Frontiers in Cell and Developmental Biology 9 (2021): 762828. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Chiang C. H., Lee L. T., Hung S. H., et al., “Opposite Association Between Diabetes, Dyslipidemia, and Hepatocellular Carcinoma Mortality in the Middle‐Aged and Elderly,” Hepatology 59, no. 6 (2014): 2207–2215. [DOI] [PubMed] [Google Scholar]
- 18. Borena W., Strohmaier S., Lukanova A., et al., “Metabolic Risk Factors and Primary Liver Cancer in a Prospective Study of 578,700 Adults,” International Journal of Cancer 131, no. 1 (2012): 193–200. [DOI] [PubMed] [Google Scholar]
- 19. Nguyen A. L. T., Nguyen H. T. T., Yee K. C., Palmer A. J., Blizzard C. L., and de Graaff B., “A Systematic Review and Narrative Synthesis of Health Economic Evaluations of Hepatocellular Carcinoma Screening Strategies,” Value in Health 24, no. 5 (2021): 733–743. [DOI] [PubMed] [Google Scholar]
- 20. Yau S. T. Y., Leung E. Y. M., Hung C. T., et al., “Scoring System for Predicting the Risk of Liver Cancer Among Diabetes Patients: A Random Survival Forest‐Guided Approach,” Cancers 16, no. 13 (2024): 2310. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Johnson J. A., Carstensen B., Witte D., Bowker S. L., Lipscombe L., and Renehan A. G., “Diabetes and Cancer (1): Evaluating the Temporal Relationship Between Type 2 Diabetes and Cancer Incidence,” Diabetologia 55, no. 6 (2012): 1607–1618. [DOI] [PubMed] [Google Scholar]
- 22. Hothorn T., Hornik K., and Zeileis A., “Unbiased Recursive Partitioning: A Conditional Inference Framework,” Journal of Computational and Graphical Statistics 15, no. 3 (2006): 651–674. [Google Scholar]
- 23. Ramezankhani A., Tohidi M., Azizi F., and Hadaegh F., “Application of Survival Tree Analysis for Exploration of Potential Interactions Between Predictors of Incident Chronic Kidney Disease: A 15‐year Follow‐Up Study,” Journal of Translational Medicine 15, no. 1 (2017): 240. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Yau S. T. Y., Hung C. T., Leung E. Y. M., Lee A., and Yeoh E. K., “Survival Tree Analysis of Interactions Among Factors Associated With Colorectal Cancer Risk in Patients With Type 2 Diabetes: Retrospective Cohort Study,” JMIR Public Health and Surveillance 11 (2025): e62756. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Yau S. T. Y., Hung C. T., Leung E. Y. M., Lee A., and Yeoh E. K., “The Interactions Among Factors Associated With the Risk of Lung Cancer Among Diabetes Patients: A Survival Tree Analysis,” NPJ Primary Care Respiratory Medicine 35, no. 1 (2025): 20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Green R. M. and Flamm S., “AGA Technical Review on the Evaluation of Liver Chemistry Tests,” Gastroenterology 123, no. 4 (2002): 1367–1384. [DOI] [PubMed] [Google Scholar]
- 27. Petroff D., Bätz O., Jedrysiak K., Kramer J., Berg T., and Wiegand J., “Age Dependence of Liver Enzymes: An Analysis of over 1,300,000 Consecutive Blood Samples,” Clinical Gastroenterology and Hepatology 20, no. 3 (2022): 641–650. [DOI] [PubMed] [Google Scholar]
- 28. Dong M. H., Bettencourt R., Barrett‐Connor E., and Loomba R., “Alanine Aminotransferase Decreases With Age: The Rancho Bernardo Study,” PLoS One 5, no. 12 (2010): e14254. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Dong M. H., Bettencourt R., Brenner D. A., Barrett‐Connor E., and Loomba R., “Serum Levels of Alanine Aminotransferase Decrease With Age in Longitudinal Analysis,” Clinical Gastroenterology and Hepatology 10, no. 3 (2012): 285–290.e1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Elinav E., Ben‐Dov I. Z., Ackerman E., et al., “Correlation Between Serum Alanine Aminotransferase Activity and Age: An Inverted U Curve Pattern,” The American Journal of Gastroenterology 100, no. 10 (2005): 2201–2204. [DOI] [PubMed] [Google Scholar]
- 31. Hyeon C. K., Chung M. N., Sun H. J., Kwang H. H., Kyu Oh D., and Suh I., “Normal Serum Aminotransferase Concentration and Risk of Mortality From Liver Diseases: Prospective Cohort Study,” BMJ 328, no. 7446 (2004): 983. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Prati D., Taioli E., Zanella A., et al., “Updated Definitions of Healthy Ranges for Serum Alanine Aminotransferase Levels,” Annals of Internal Medicine 137, no. 1 (2002): 1–10. [DOI] [PubMed] [Google Scholar]
- 33. Yau S. T. Y., Leung E., Wong M. C. S., et al., “Metabolic Dysfunction‐Associated Profiles and Subsequent Site‐Specific Risk of Obesity‐Related Cancers Among Chinese Patients With Diabetes: A Retrospective Cohort Study,” BMJ Open 14, no. 4 (2024): e082414. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Johnson A. A. and Stolzing A., “The Role of Lipid Metabolism in Aging, Lifespan Regulation, and Age‐Related Disease,” Aging cell 18, no. 6 (2019): e13048. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Mutlu A. S., Duffy J., and Wang M. C., “Lipid Metabolism and Lipid Signals in Aging and Longevity,” Developmental Cell 56, no. 10 (2021): 1394–1407. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Yau S. T. Y., Hung C. T., Leung E. Y. M., Lee A., and Yeoh E. K., “Comparison of Risk Factors, Their Interaction Patterns, and Scoring Systems for Liver Cancer Between Patients With and Those Without Diabetes: Retrospective Cohort Study Using Electronic Health Records and Tree‐Structured Algorithms,” JMIR Medical Informatics 13 (2025): e72239. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Data is not available for sharing due to access restriction.