Observational and Genetic Associations of Body Mass Index and Hepatobiliary Diseases in a Relatively Lean Chinese Population

This cohort study compares observational associations with genetic associations of body mass index with hepatobiliary diseases and liver biomarkers.


Introduction
Hepatobiliary disease is a major cause of morbidity and mortality worldwide, responsible for 2.4 million deaths in 2017, with approximately 20% of these occurring in China. 1 In Western populations, alcohol consumption is the major cause of liver disease, associated with 30% to 60% of liver disease diagnoses, 2-4 whereas hepatitis C virus accounts for 20% to 30% of liver disease in the West. 2 In China, by contrast, hepatitis B virus (HBV) accounts for 50% and alcohol consumption accounts for approximately 10% of liver disease. 5,6 Gallstones affect 10% to 20% of the global adult population 7,8 and account for at least 90% of gallbladder disease in Western populations and 40% to 60% of the disease in Chinese populations. 8,9 Prospective cohort studies in Western populations 10,11 have consistently demonstrated positive associations of adiposity, mostly identified via body mass index (BMI; calculated as weight in kilograms divided by height in meters squared), with chronic liver disease and gallbladder disease. In contrast, a 2019 study in China 12 reported that associations with BMI differed by chronic liver disease subtype, with a positive association for nonalcoholic fatty liver disease, an inverse association for liver cancer, and a U-shaped association for cirrhosis. A 2015 meta-analysis 11 examined symptomatic gallbladder disease and did not differentiate between disease subtypes (ie, cholelithiasis vs cholecystitis). Previous studies in Western populations 13,14 have reported some genetic evidence for the associations of BMI with risks of chronic liver disease and symptomatic gallbladder disease.
However, previous genetic studies have been constrained by the small number of individuals the studies included, which meant they lacked power to reliably assess the associations by disease subtypes.
There is limited evidence for the association of BMI and these diseases in populations in East Asia, where mean BMI and the causes of hepatobiliary diseases differ from those in Western populations. Moreover, there is limited evidence for the genetic associations of BMI with biomarkers, including those associated with liver function, steatosis, and fibrosis. Additional evidence for these associations may provide insights into the underlying mechanisms. Understanding the genetic associations of BMI with hepatobiliary diseases may improve understanding of disease causes in diverse populations and inform prevention strategies. 15,16 We present findings from the China Kadoorie Biobank (CKB) study population of approximately 500 000 adults. The objectives of this study were to compare the observational associations between BMI and hepatobiliary diseases and liver biomarkers with the genetic associations between BMI and these factors and to assess whether the genetic associations of BMI with liver diseases differed by HBV infection status.

Study Population
Details of the CKB prospective cohort study design, survey methods, and population characteristics have been described elsewhere. 17 Briefly, 512 715 participants aged 30 to 79 years were recruited from 10 (ie, 5 urban and 5 rural) diverse areas in China from 2004 to 2008. All participants provided written informed consent. Prior international, national, and regional ethical approvals were obtained; this study was approved by the ethical committee and research council of the Chinese Center for Disease Control and Prevention and the Oxford Tropical Research Ethics Committee at the University of Oxford. At local study assessment clinics, each participant completed an intervieweradministered, laptop-based questionnaire on sociodemographic characteristics, smoking history, alcohol consumption, diet, physical activity, personal and family medical history, and current medication. Trained technicians recorded physical measurements (ie, height, weight, hip circumference, waist circumference, bio-impedance, lung function, blood pressure, and heart rate) using calibrated instruments and standard protocols. In addition, desktop analyzers were used to measure random blood glucose and hepatitis B surface antigen (HBsAg) (Acon Biotech). Trained technicians recorded all anthropometric measurements while participants wore light clothes and no shoes, with measurements taken usually to the nearest 0.1 cm or 0.1 kg. Standing height was measured using a stadiometer. Weight was measured using a body composition analyzer (TBF-300GS, Tanita), with clothing weight subtracted according to season (ranging from 0.5 kg in summer to 2.0-2.5 kg in winter). The baseline survey included quality control measures and assessment of liver biomarkers (eAppendix in the Supplement).

Follow-up for Mortality and Morbidity
The vital status of each participant was determined periodically through the Chinese Center for Disease Control and Prevention's Disease Surveillance Points system, 18 supplemented by regular checks against local residential records and health insurance records and by annual active confirmation via street committees or village administrators. Additional information about major diseases and any episodes of hospitalization was collected by linkage, via each participant's unique national identification number, to disease registries (for cancer, ischemic heart disease, stroke, and diabetes) and national health insurance claims databases (for hepatobiliary diseases). The databases had almost universal coverage in the study areas. Trained staff who were blinded to baseline information 18

coded all events using International Statistical Classification of Diseases and Related
Health Problems, Tenth Revision (ICD-10). 19 In the classification of hepatobiliary diseases, nonalcoholic fatty liver disease and alcoholic liver disease were classified entirely using medical records (eTable 1 in the Supplement). A total of 44 066 participants (9%) died and 4751 participants (<1%) were lost to follow-up by January 1, 2017.

Genotyping and Biochemistry Measurements
Genotyping was conducted using a custom-designed 800K-single-nucleotide variation (SNV) (formerly single-nucleotide polymorphism [SNP]) array (Axiom, Affymetrix) with imputation to 1000 Genomes Phase 3. Genotype data were available for 100 408 participants whose samples passed quality control (ie, overall call rate >99.97% across all variants). Participants with genotype data consisted of a population-based sample of 75 736 participants randomly selected from the total CKB cohort and 24 672 participants who had been selected for nested case-control studies of incident stroke, coronary heart disease, or chronic obstructive pulmonary disease (eFigure 1 in the Supplement). To avoid selection bias, only the 75 736 randomly selected participants were used for genetic analyses of hepatobiliary outcomes. Among genotyped participants who had been selected for nested case-control studies of stroke and coronary heart disease, 17 567 participants were consumption, with stratification by sex, study area (from among 10 areas), and HBsAg (for chronic liver disease).

Single-Nucleotide Variation Selection and Construction of Genetic Score
We used an unweighted genetic score for predisposition to higher BMI as an instrumental variable.
The BMI genetic score was derived by summing up the number of BMI-increasing alleles using 97 independent (r 2 Յ 0.01 in the CEU [Centre d'Etude du Polymorphism Humain -Utah residents with Northern and Western European ancestry] population) BMI SNVs reported in a genome-wide association study by the GIANT (Genetic Investigation of Anthropometric Traits) consortium. 20 Two variants (ie, rs12016871 and rs2245368) were not available in CKB, and 3 variants (ie, rs13107325, rs17024393, and rs2121279) were excluded because of the low minor allele frequency in the East Asian population (<1%), leaving 92 SNVs for the BMI genetic score. Per-allele effect sizes for BMI in CKB were estimated for the individual genetic variants in the GIANT consortium (eTable 2 in the Supplement). The BMI genetic score has been shown to perform well (F statistic, 1071; variance explained, 1.1%) and was not associated with potential confounders (eFigure 2 in the Supplement).

Mendelian Randomization
The genetic associations of BMI with hepatobiliary diseases were calculated by the 2-stage leastsquares estimator method using individual participant-level data. In the first stage, the associations between BMI genetic score and BMI were examined in 75 736 participants in the genome-wide association study population subset using linear regression, adjusting for age, age squared, sex, region, the first 12 principal components, education level, smoking history, and alcohol consumption.
In the second stage, the associations of the resulting estimated values with hepatobiliary diseases were examined using logistic regression, adjusting for the same covariates plus HBsAg (for chronic liver disease). We calculated the genetic estimates per increase in genetically predicted BMI, equivalent to 1 SD baseline BMI, on disease outcomes to compare with associations with measured BMI.

Statistical Analysis
The genetic estimates were compared with the corresponding observational estimates using a Cochran Q test. Odds ratios were used to approximate relative risks (RRs) of hepatobiliary diseases because of the rarity of events in CKB (<1.5%). Risk ratios and hazard ratios were used interchangeably. Therefore, we used the term RRs to refer to odds ratios and hazard ratios for consistency between the observational and genetic analyses. Statistical analyses included liver biomarker analysis, sensitivity analyses, and meta-analysis (eAppendix in the Supplement). The statistical analysis was performed January 2019 to October 2019 using R statistical software version 2.14.2 (R Project for Statistical Computing). P values are from Wald tests, and P < .05 was considered significant.

Results
Of 473 938 participants, mean (SD) BMI was 23.8 and mean (SD) age was 52 (10.9) years; 276 041 (58.2%) were women. During 10 years of follow-up, there were 5904 cases of chronic liver disease and 16 720 cases of gallbladder disease. We ascertained all nonalcoholic fatty liver disease diagnoses between 2013 and 2015 and found that 1033 of 1111 individuals hospitalized for the disease (93%) were diagnosed by ultrasonography or computed tomography. Participants with chronic liver disease were more likely to be men, whereas participants with gallbladder disease were more likely to be women ( Table 1). Participants with chronic liver disease and gallbladder disease were older and more likely to live in rural areas, have lower levels of education, and have lower household income. 60 of 2608 women [2.3%]); participants with chronic liver disease were also more likely have test results positive for HBsAg (1222 of 5904 participants [20.7%]). Participants with chronic liver disease and gallbladder disease had higher mean levels in measures of systolic blood pressure, random plasma glucose, BMI, and waist circumference and were more likely to have prevalent diabetes and hypertension. No differences in female reproductive factors were observed across disease status.

Observational and Genetic Associations of BMI With Hepatobiliary Diseases
There were positive associations of BMI with risks of chronic liver disease and gallbladder disease, and the association was log-linear in the range 15 to 50 for gallbladder disease and the range 24 to 50 for chronic liver disease (eFigure 3 in the Supplement). The adjusted RRs per 1-SD increase were 1.14 (95% CI, 1.11 to 1.17) for chronic liver disease and 1.29 (95% CI, 1.27 to 1.31) for gallbladder disease (

Subgroup and Sensitivity Analyses
The positive genetic associations of BMI with risks of chronic liver disease were consistent in participants regardless of HBV status. For noncancer chronic liver disease, RR per 1-SD increase was A. Boxes indicate the relative risks (RRs) of liver diseases associated with 1-SD greater genetically determined body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) in participants stratified by HBV status, with the size of the box inversely proportional to the variance of the log RR. B. Boxes indicate the SD differences of liver biomarkers associated with 1-SD greater genetically determined body mass index (BMI) in participants stratified by HBV status, with the size of the box inversely proportional to the variance of the SD difference. The genetic analysis was adjusted for age, age squared, sex, region, the first 12 principal components, education level, smoking history, and alcohol consumption. P values for comparison were obtained from a Cochran Q test comparing the observational and genetic estimates.
For chronic liver disease, there was a positive observational association for baseline BMI in women (RR per 1-SD, 1.23; 95% CI, 1.18 to 1.27) and men (RR per 1-SD, 1.06; 95% CI, 1.02 to 1.10), and the associations differed significantly (P for heterogeneity < .001) (eFigure 4 and eTable 3 in the Supplement). There was a genetic association of BMI with chronic liver disease in men (RR per 1-SD, 1.89; 95% CI, 1.23 to 2.91) and women (RR per 1-SD, 1.31; 95% CI, 0.88 to 1.96), and the sex differences were not significant (P for heterogeneity = .15). For gallbladder disease, observational and genetic associations for BMI did not differ by sex.
The weighted median estimates were concordant with the inverse variance weighted estimates. The Mendelian randomization-Egger estimates were less precise than the inverse variance estimates (eFigure 5 in the Supplement). The intercepts in the Mendelian randomization-Egger regressions were 0.001 (95% CI, −0.02 to 0.02) for chronic liver disease and 0.005 (95% CI, −0.05 to 0.015) for gallbladder disease. The genetic associations persisted when restricting the score to 73 SNVs that did not show different associations with BMI in East Asian populations (eFigure 5 in the Supplement).

Meta-analysis of CKB and UKB
The UK Biobank (UKB) study found positive genetic associations of BMI with risk of chronic liver disease (RR per 1-SD increase, 1.55; 95% CI, 1.27 to 1.89; P for heterogeneity = .95) and gallbladder disease (RR per 1-SD increase, 1.61; 95% CI, 1.50 to 1.72; P for heterogeneity = .16), with no heterogeneity between disease subtypes (eTable 4 in the Supplement). In the meta-analysis of CKB and UKB estimates, there were positive genetic associations of BMI with hepatobiliary diseases across subtypes, with little heterogeneity between the 2 studies (Figure 2). Each 1-SD higher genetically determined BMI was associated with a 55% greater risk of chronic liver disease (RR, 1.55; 95% CI, 1.30 to 1.84) and a 42% higher risk of gallbladder disease (RR, 1.42; 95% CI, 1.22 to 1.64).    Boxes indicate the relative risks (RRs) of hepatobiliary diseases associated with 1-SD higher genetically determined body mass index in the China Kadoorie Biobank (CKB) and UK Biobank (UKB) studies, with the size of the box inversely proportional to the variance of the log RR; diamonds, the summary RRs for CKB and UKB; and No., the number of individuals with the disease in the biobank sample.

Discussion
In this cohort study of a relatively lean Chinese population, there were concordant positive associations of measured BMI and genetically instrumented BMI with hepatobiliary diseases. There were genetic associations of BMI with liver biomarkers, including liver enzymes (ie, ALT and γ-glutamyl transferase), as well as steatosis and fibrosis scores, consistent with the observational associations. The genetic associations of BMI with liver diseases and biomarkers were consistent regardless of HBsAg status. The genetic associations of BMI with hepatobiliary diseases in the CKB study were also consistent with the findings in the UKB study and a previous Mendelian randomization study in the European population, 13,14 suggesting that general adiposity may be a risk factor for hepatobiliary diseases in relatively lean Chinese adults regardless of the underlying disease causes.
To date, 1 study has assessed the genetic associations of BMI with hepatobiliary diseases, as In our study's Chinese population, the genetic associations of BMI with chronic liver disease and gallbladder disease were not significantly different from the observational associations. The associations for the genetically instrumented BMI may reflect the longer duration of adiposity. In addition, the observational estimates could have been affected by residual confounding and reverse causality, particularly for chronic liver disease. Smoking is a major risk factor for cirrhosis and liver cancer and is associated with lower BMI 21,22 and therefore is a negative confounder between BMI and chronic liver disease.
Approximately 70% of men and less than 3% of women smoked in CKB, so the association of BMI with chronic liver disease may have been confounded by smoking. There was no significant difference between the genetic estimate in men and that in women. In addition, our previous reports in CKB showed that the inverse associations of BMI with liver cancer and cirrhosis attenuated toward the null and lost significance when excluding the first 5 years of follow-up. 12,23 This finding suggests that participants may have subclinical or undetected diseases at baseline that may affect adiposity at baseline, resulting in reverse causality.
For the observational estimates, the associations for BMI differed by chronic liver disease subtypes. This is likely because reverse causality disproportionally affects liver cancer and cirrhosis (accounting for 65% of the total chronic liver disease population in CKB) and these subtypes have latency periods of more than 10 years. 24,25 When restricting the analysis to individuals who were never regular smokers or when excluding the first 5 years of follow-up, the inverse associations of BMI with risk of cirrhosis and liver cancer in the BMI range 15 to 24 were not significant and there were positive associations in the BMI range 24 to 50 (eFigure 6 in the Supplement).
Obesity leads to increased fat deposition within the hepatocytes and to the development of hepatic steatosis. 26 In addition, obesity is associated with higher levels of oxidative stress and inflammation and subsequent liver fibrosis, cirrhosis, and liver cancer. 26 In populations in the West, where nonalcoholic fatty liver disease is the major cause of liver cancer, obesity plays an important role in causing liver cancer. 24 However, HBV is the major cause of liver cancer in East Asia, where up to 40% of individuals with HBV-related liver cancer do not have underlying cirrhosis; among these individuals, cancer is caused by integration of the HBV genome into the host's intracellular DNA. 27 Although fatty liver index has been used as a proxy for liver fat, this approach has not been validated among patients with HBV. Nonetheless, a 2007 in vivo study 28 showed that chronic HBV infection may have a synergistic effect with obesity on hepatic lipid accumulation and the development of steatosis, and these effects may confer higher risks of cirrhosis and liver cancer.
Future studies are warranted to develop markers of liver fat among individuals with HBV infection and to assess the genetic association of BMI with liver fat.
Approximately 90% of gallstones are cholesterol stones in Western populations, whereas the proportion is approximately 80% in Chinese populations. 29,30 In individuals with obesity, the activity of 3-hydroxy-3-methyl-glutaryl-coenzyme A reductase is upregulated and cholesterol is hypersecreted from the liver into the bile. 31,32 The high rate of cholesterol secretion is associated with supersaturation of cholesterol in the bile and disruption of the normal function of the gallbladder. 33 In addition, individuals with obesity have higher gallbladder volumes and lower gallbladder motility, conditions associated with higher risks of gallstones. 34,35 The strengths of the CKB study included the prospective design, large and diverse study population, detailed adjustment for risk factors for hepatobiliary diseases, validity of the genetic scores developed for BMI, and ascertainment of hepatobiliary diseases through linkage to hospital records in addition to death and cancer registries.

Limitations
Our study had several limitations. First, it is plausible that a subset of SNVs in the BMI genetic score affect hepatobiliary diseases independently of BMI, potentially violating the assumptions of Mendelian randomization. However, we showed that Mendelian randomization-Egger estimates and weighted median estimates were broadly consistent with the inverse variance weighted estimates in CKB. In addition, our findings are generally concordant with previous Mendelian randomization studies conducted in Denmark using different genetic variants to construct the BMI genetic score.
Second, gallbladder disease and nonalcoholic fatty liver disease were primarily ascertained through linkage to medical records, which might lead to underdiagnosis. However, our results are probably valid because: (1) of all nonalcoholic fatty liver disease diagnoses between 2013 and 2015, we found that over 90% of individuals hospitalized for the disease were diagnosed by ultrasound or computed tomography; (2) our risk estimates agreed with previous studies of ultrasound-detected nonalcoholic fatty liver disease and symptomatic gallstone disease; and (3) hospitalization for nonalcoholic fatty liver disease has been shown to be a valid diagnosis in previous CKB reports 12,36 on adiposity and diabetes. We defined a suspected nonalcoholic fatty liver disease case as 1 in which the individual had elevated ALT levels (Ն33 U/L for men and Ն25 for women; to convert to microkatal per liter, multiply by 0.0167) in approximately 18 000 participants with clinical biochemistry data and negative HBsAg test results, finding a similar genetic estimate for elevated ALT and hospitalized nonalcoholic fatty liver disease (eMethods in the Supplement). Third, BMI is a marker of general adiposity, but central adiposity is also a risk factor associated with hepatobiliary diseases. Future studies are warranted in Chinese populations to assess the genetic associations of central adiposity with hepatobiliary diseases. Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: Project staff based at Beijing, Oxford, the 10 regional centers, the Chinese Center for Disease Control and Prevention, and its regional offices assisted with fieldwork. BGI assisted in conducting DNA extraction and genotyping.