Hepatitis B virus infection and the risk of gastrointestinal cancers among Chinese population: A prospective cohort study

Abstract Our study aims to explore the relationship between chronic hepatitis B virus (HBV) infection and the risk of gastrointestinal (GI) cancers including liver, gastric, gallbladder or extrahepatic bile duct, pancreatic, small intestine, esophageal and colorectal cancer in the Kailuan Cohort study. We prospectively examined the relationship between HBV infection and new‐onset GI cancers among 93 402 participants. Cox proportional hazards regression models, subgroup analyses and competing risk analyses were used to evaluate the association between HBV infection and the risk of new‐onset GI cancers. During a median follow‐up of 13.02 years, 1791 incident GI cancer cases were diagnosed. Compared to HBsAg seronegative participants, a significant positive association between HBV infection and GI cancers was observed in the multivariate‐adjusted models (HR 5.59, 95% CI: 4.84‐6.45). In the site‐specific analyses, participants with HBsAg seropositive exhibited an increased risk of liver cancer (HR = 21.56, 95% CI: 17.32‐26.85), gallbladder or extrahepatic bile duct cancer (HR = 14.89, 95% CI: 10.36‐21.41), colorectal cancer (HR = 1.75, 95% CI: 1.15‐2.96) and pancreatic cancer (HR = 1.86, 95% CI: 1.10‐3.99). After taking death as the competing risk event, the associations of HBV infection with the risk of these cancers were attenuated but remained significant both in the cause‐specific hazards models, the subdistribution proportional hazards models and sensitivity analyses. Our study suggests that HBV infection is associated with the elevated risk of liver cancer and extrahepatic cancer including gallbladder or extrahepatic bile duct, pancreatic and colorectal cancer among adults in Northern China.

the risk of these cancers were attenuated but remained significant both in the cause-specific hazards models, the subdistribution proportional hazards models and sensitivity analyses. Our study suggests that HBV infection is associated with the elevated risk of liver cancer and extrahepatic cancer including gallbladder or extrahepatic bile duct, pancreatic and colorectal cancer among adults in Northern China.

| INTRODUCTION
Gastrointestinal (GI) cancer has the highest incidence and mortality among all cancer types. 1 Due to the short lifespan of its cells, the GI tract has one of the most replicative tissues in the body. Digestive system issues are constantly affected by physical, biological and chemical insults, increasing the risk of oncogenic mutations. Chronic infection with possible carcinogenic agents represents a major risk for subsequent development of cancer and was estimated to be accounted for 2 million incident cancer cases in 2008, with significant variations among regions and countries. 2 Of these, hepatitis B virus (HBV) infection is a major public health concern. With the implementation of the HBV immunization program in 1984 and the chronic viral hepatitis therapy program in 2003, 3 its prevalence rate has dropped from 8% to 2%-7% in the past few years. 4 However, the Global Hepatitis Report (2017) estimated that nearly 257 million people worldwide were still living with chronic HBV infection, 5 which would increase the risk of liver fibrosis, liver cirrhosis and hepatocellular carcinoma (HCC) in those infected. 6,7 The burden of HBV infection, as well as subsequent tissue damage, has been primarily centered in the liver. HBV deposits its life cycle for long-period persistence in the target tissues by forming a covalently closed circular DNA form (cccDNA) in the nucleus of infected cells. 8 In animal experiments, HBV DNA has been detected in nonliver tissues including the pancreas, kidney, gonads and lymphoid. 9,10 Additionally, several clinical case studies revealed that HBV-specific nucleic acid sequences and related proteins had been detected in extrahepatic tissues of patients with acute or chronic infection of HBV, providing a possible connection of HBV infection to the oncogenesis of extrahepatic cancers. 11,12 More recently, literature has offered contradictory findings concerning the effect of chronic HBV infection on the occurrence of subsequent nonliver cancers. 13,14 Despite limited population-based prospective studies focused on this issue, previous studies may also yield to several limitations including the retrospective design, minimal control of potential confounding factors, short-term follow-up, small sample size, hospital-based identification of individuals and neglect of competing risks events in epidemiologic research, all of which affect the reliability of the existing results.
HBV infection and GI cancers are highly endemic in China, providing a great opportunity to explore the association. The purpose of this investigation is to explore the relationship between chronic HBV infection and the occurrence of GI cancers including liver, gastric, gallbladder or extrahepatic bile duct, pancreatic, small intestine, esophageal and colorectal cancer by drawing data from the Kailuan Cohort study. To further test the validity of the results, stratified analysis and competing risk analysis were also conducted in the current study.

| Study population
The data for our study are drawn from the Kailuan Study, which is an ongoing, population-based, prospective cohort study based on the Kaliuan community in Tangshan City, China. 15 In our study, participants who did not meet the criteria for analysis were excluded: (a) 469 subjects with a history of malignancy; (b) 2339 subjects with missing or unclear information of hepatitis B surface antigen (HBsAg); (c) 4332 subjects without data of potential confounders including age, sex, body mass index (BMI, in kg/m 2 ), high-sensitivity C-reactive protein (hs-CRP, in mg/L), total cholesterol (TC, in mmol/L), alanine aminotransferase (ALT, in u/L), triglyceride

| Exposure assessment
After at least 8 hours fast, blood samples were collected to test for HBV infection from each participant at baseline using vacuum tubes containing EDTA. The enzyme-linked immunosorbent assay was applied to detect hepatitis B surface antigen (HBsAg) quantitatively with a standard operating procedure (Shanghai Kehua Bio-Engineering, KHB, Shanghai, China). Participants were divided into two groups according to HBV infection status: HBsAg seronegative group (n = 90 795) and HBsAg seropositive group (n = 2607).

| Outcome ascertainment
Incident GI cancers were obtained through questionnaires in the rou-

| Potential confounders
Information on age, sex, lifestyle behaviors, educational background, socioeconomic status, medical comorbidities, personal and family medical histories were collected via a standard questionnaire. Current alcohol consumer was defined as having drunk ≥100 mL/day of alcohol lasting for more than 6 months, regardless of the type of alcohol.
Current smoker was defined as having 1 cigarette/day at least for F I G U R E 1 The procedure of participants screening more than 6 months. Physical exercise was evaluated from responses regarding the frequency of physical activity (≥3 times/week, ≥30 min/ time). Dietary salt intake was self-reported and classified into three categories: low (<6 g/day), medium (6-9 g/day) or high (≥10 g/day).
Physical examinations were performed by trained workers for each participant. Height and weight were measured by trained medical staff. BMI was calculated as body weight (kg) divided by the square of height (m 2 ) and classified into normal (<24 kg/m 2 ), overweight (24.00-27.99 kg/m 2 ) or obese (≥28 kg/m 2 ). 17 Hypertension was defined as: previously diagnosed, and/or an SBP ≥140 mm Hg, and/or a DPB ≥90 mm Hg, and/or using antihypertensive medication. 18 The ultrasonic examination was used to examine the abdominal region, including liver, gallbladder, pancreas and spleen of each participant after fasting for at least 8 hours by a panel of specialists. Liver cirrhosis, fatty liver, gallstone disease and gallbladder polyp were diagnosed by abdominal ultrasonography according to previous clinically established criteria 19,20 or through medical records from the Tangshan Medical Insurance System.
All the serum samples were analyzed by an auto-analyzer (Hitachi 747; Hitachi, Tokyo, Japan) at the central laboratory of Kailuan General Hospital. Serum TC and TG were both measured by the enzymatic colorimetric method (Mind Bioengineering Co. Ltd, Shanghai, China). ALT was measured using an enzymatic rate method (Mind Bioengineering Co. Ltd, Shanghai, China). Hs-CRP was measured using a high-sensitivity nephelometry assay (Cias Latex CRP-H, Kanto Chemical Co. Inc). Serum TBil was measured using a chemical oxidation method (MedicalSystem Biotechnology, China). Diabetes mellitus was defined as fasting blood glucose level ≥ 7.0 mmol/L, and/or taking oral hypoglycemic agents or insulin, and/or a validated physician diagnosis. Hs-CRP was divided into three groups (<1, 1-3 and >3 mg/L) based on guidelines from Disease Control and Prevention and the American Heart Association. 21 Based on the tertiles of each variable, serum TG, TC, ALT and TBil were grouped into three categories.

| Statistical analyses
Continuous variables and categorical variables were presented as the mean ± SD and absolute value with percentage. The comparisons of continuous or categorical characteristics were examined using the ttest or χ 2 test. Person-years were calculated as the time from baseline examination to the data of cancer diagnosis, death or 31 December 2019, whichever event came first. The Cox proportional hazards regression was used to calculate the hazard ratios (HRs) and their 95% confidence intervals (CIs) for determining the association between HBV infection and cancer development. Three models were fitted as follows: model 1 was an unadjusted analysis; model 2 was adjusted for sex and age (every 10 years); model 3 was further adjusted for BMI, levels of TC, TG, hs-CRP, TBil and ALT, diabetes, family income, educational background, marital status, salt consumption, smoking status, drinking status, physical exercise and family history of cancer based on model 2. In the pooled GI cancer analyses, we only included the first reported cancer type. However, site-specific analyses were conducted for all patients with multiple relevant GI cancers. In the site-specific analyses, model 3 was additionally adjusted for liver cirrhosis and fatty liver in the model of liver cancer, while gallstone disease and gallbladder polyp were further adjusted in the analyses of the gallbladder and biliary cancer. Subgroup analyses were performed for each specific cancer site stratified by sex, age, BMI, smoking status, drinking status, cirrhosis, fatty liver and gallstone disease. The interactions between HBV infection status and these variables were further tested using multiplicative models.
During follow-up, death may occur before the occurrence of GI cancers. Due to the existence of competing risk events (death), the observation of new-onset GI cancer cases and further interventions can be hindered. Conventional methods for survival analysis such as standard Cox regression may neglect the competing events and overestimate the risk of the disease. Thus, competing risk analysis should be applied to epidemiologic research. The selection of approach should be determined by the scientific purpose. In general, epidemiological studies focus on two types of issues: (a) Aetiological research is designed to explore the causal relationship between risk factors and an outcome, and applying the cause-specific hazards (CS) models would be more applicable. (b) Prognostic research is used to predict the probability of the outcome, and applying the subdistribution proportional hazards models (SD) model would be more appropriate. In the current study, CS models and SD models were used to calculate HR CS and HR SD of the specific site of GI cancers with the existence of competing risk, but only if a significant association was found previously in the Cox regressions.
Statistical computations were performed using a commercially available software program (SAS software, version 9.4). Reported P-values are two-sided, and the significance level was set at P < .05.

| Sensitivity analyses
Previous studies found patients with liver cirrhosis were associated with an elevated risk of digestive system diseases, 22,23 therefore, we excluded participants with cirrhosis at baseline and reanalyzed the association to further test the robustness of our findings. Although there was a clear temporal sequencing between HBV exposure and the occurrence of GI cancers, participants who were diagnosed with cancer during the first year of follow-up were also excluded, because a positive HBV test may lead to additional tests, identifying prevalent cancer. We also estimated an HR for cancers diagnosed within 1 to 3 years, an HR for cancers diagnosed between 3 and 5 years and an HR for cancers detected >5 years after baseline to see how the magnitude of the effect of HBV infection changes over time.

| Characteristics of the study population
Of the 93 402 participants, the mean ± SD age was 51.52 ± 12.43 years with 74 637 (79.91%) males and 18 765 (20.09%) females. The overall age-and sex-standardized HBV infection rate was 3.01% and significantly higher in men (3.10%) than in women   F I G U R E 2 Stratified analysis of the association of HBV infection with the risk of GI cancers. All models were adjusted for age BMI (every 10 years), sex (normal, overweight and obesity), TG, TC, hs-CRP (<1, 1-3 and >3 mg/L), TBil, diabetes, family income, educational background, marital status, salt consumption, current smoker, drinking status, physical activity and family history of cancer. Liver cancer models were further adjusted for liver cirrhosis and fatty liver disease. Gallbladder or extrahepatic bile duct cancer models were further adjusted for gallstone disease and gallbladder polyp. (A) Age (every 10 years), BMI (normal, overweight, obesity), current smoker, ALT and drinking status were further adjusted when participants were stratified by gender. (B) Sex, BMI (normal, overweight, obesity), current smoker, ALT and drinking status were further adjusted when participants were stratified by age. Age was also adjusted within each age stratum to prevent residual confounding. (C) Age (every 10 years), sex, ALT, current smoker and drinking status were further adjusted when participants were stratified by BMI. (D) Age (every 10 years), BMI (normal, overweight, obesity), sex, ALT and drinking status were further adjusted when participants were stratified by smoking status. (E) Age (every 10 years), BMI (normal, overweight, obesity), sex and smoking status were further adjusted when participants were stratified by drinking status with an elevated risk of liver cancer within all stratified analyses.

| Stratified analysis
Smoking and drinking status also showed an effect on the association between HBV infection and the occurrence of liver cancer. Positive associations between HBV infection and gallbladder or extrahepatic bile duct cancer were also observed in all stratified analyses, but none of the tests for interactions were statistically significant. In the analysis of colorectal cancer, significant associations were only observed in those who were male, younger, middle-aged, with normal BMI, smoking or nondrinker, but not in those who were female, elder, overweight, obese, nonsmoker or current drinker. No interactive effects were revealed within each analysis. In the analysis of pancreatic cancer, significant associations were found only in the participants who were male, young, middle-aged, obese, nonsmoker and their interactions did not appear to be significant.

| Sensitivity analyses
In the sensitivity analysis, after excluding individuals diagnosed with GI cancers within the first year of follow-up or liver cirrhosis at baseline, the association between HBV infection and the risk of liver, gallbladder or extrahepatic bile duct, colorectal and pancreatic cancer remained significant in the multivariate analysis, the association between HBV infection and the risk of liver, gallbladder or extrahepatic bile duct, colorectal and pancreatic cancer remained significant in the multivariate analysis (Table S1). Table S2 shows the association of HBV infection with the risk of subsequent GI cancers stratified by the time window of diagnosis. We observed a significant short-term association (<3 and 3-5 years) of HBV infection with liver cancer risk, but not for extrahepatic cancer including gallbladder or extrahepatic bile duct, pancreatic or colorectal cancer. In addition, positive association were found for live, gallbladder or extrahepatic bile duct, pancreatic and colorectal cancer diagnosed >5 years after baseline.

| DISCUSSION
In this large population-based prospective cohort study, we found that participants with chronic HBV infection suffered from a higher risk of GI cancers. Results from the specific-site analysis showed Our study has several strengths. First, it is a large populationbased prospective cohort study, which makes it more efficient for testing a risk factor in the aspect of a direct causal relationship, and is less prone to recall bias than retrospective studies. Second, due to the large sample size and long follow-up, a good number of incident GI cancer cases were identified which increased the accuracy and creditability of the results. Third, the broad assessments of potential confounders which had been well addressed in our study including cirrhosis, fatty liver, gallstones were not considered in most cohort studies. Fourth, because the Tangshan medical insurance system and the Kailuan social security system contain all the health information of participants, the follow-up rate was almost 100% in the current study.
Last, in the presence of competing risk settings, the implementation of competing risks models makes our results more robust.
Several limitations should also be noted. First, we only use a single measurement of HBsAg. It is likely that there is some degree of variability in HBsAg status over the length of the follow-up period in a prospective study and may cause a possibility of misclassification.
Second, we do not have information on any other HBV marker such as HBeAg and HBV DNA, which hindered us from classifying the population more accurately. Third, specific cancer-related causal factors including hepatitis C virus (HCV) infection for HCC, and the consumption of cereal, vegetables and high-fiber food as well as Hp infection for stomach cancer were not collected in our study, which hindered us from assessing the absolute risk of HBV infection more precisely.
Though the prevalence of hepatitis C core antibody is only 0.43% in China, 43 the prevalence of HCV or human immunodeficiency virus (HIV) co-infection among chronic HBV infected individuals is several times higher compared to people who are not infected with HBV. 44 The nonexclusion of HCV (and HIV) co-infected individuals may weaken the results presented in our study. In addition, eating habits are closely related to BMI, TC and TG levels. 45 Since BMI, TC and TG were adjusted in the multivariate analysis, missing data on dietary habits may have little influence on the results. Results from a systematic review and meta-analysis showed the pooled Hp prevalence estimate for the general population was 55.8% (95% CI: 51.8-59.9) in China. 46 A cross-sectional study conducted in Tangshan

| CONCLUSIONS
Our study suggests that HBV infection is associated with the risk of liver cancer, extrahepatic cancers including gallbladder or extrahepatic bile duct, pancreatic and colorectal cancer among adults in Northern China. These findings highlight the importance of early screening for GI cancers in individuals infected with HBV. Future researches need to better assess the existence of HBV DNA and antigens in GI cancers tissues, and elucidate the potential mechanisms of HBV for carcinogenesis.

ETHICS STATEMENT
Our study was approved by the ethics committee of Kailuan General Hospital and followed the Declaration of Helsinki. Informed consent forms were signed by the participants or their legal representatives.

DATA AVAILABILITY STATEMENT
Data will be made available upon reasonable request.