Implication of liver enzymes on incident cardiovascular diseases and mortality: A nationwide population-based cohort study

Although liver enzymes, such as γ-glutamyltransferase (GGT), alanine aminotransferase (ALT), and aspartate aminotransferase (AST), have recently been suggested as risk factors for cardiovascular diseases (CVD), impact on mortality after myocardial infarction (MI) or ischemic stroke (IS) was not previously examined. Using a population-based, nationwide cohort database, we explored the implication of GGT and aminotransferases on the development of CVD and all-cause mortality during a median 9.1 years of follow-up. Among 16,624,006 Korean adults, both GGT and aminotransferases exhibited a positive relationship with MI, IS, and mortality in a multivariate adjusted model. ALT and AST showed U-shaped associations with mortality, whereas GGT showed a positive linear relationship with mortality. The risk of 1-year mortality after MI or IS was significantly higher in the highest quartile of GGT compared to the lowest quartile (HR, 1.46; 95% CI, 1.40-1.52). The implication of GGT on MI, IS, and mortality persisted regardless of traditional cardiovascular risk parameters. This study demonstrated the unique pattern of association of ALT, AST, and GGT with the development of CVD and all-cause mortality in the Korean population. In particular, GGT showed the most robust linear relationship with mortality before and after cardiovascular events independent of risk factors.

for CVD 8 . Furthermore, to the best of author's knowledge, there is no previous report about the impact of GGT and aminotransferases on mortality after the development of MI or stroke.
The National Health Insurance Sharing Service (NHISS) manages a population-based, nationwide cohort database with information about utilization of health insurance and periodic health examination for almost all Korean adults 9 .
In the present study, the implication of GGT and aminotransferases on CVD and all-cause mortality was evaluated using the longitudinal NHISS database, which includes more than 16,000,000 Korean men and women. In addition, the influence of GGT on 30-day, 90-day, and 1-year mortality after development of myocardial infarction (MI) and ischemic stroke (IS) was explored. Furthermore, the differential impact of GGT on the development of CVD and mortality according to risk factor subgroups was examined to explore interaction of confounding risk variables.

Subjects and Methods
Study design and participants. The Korean National Health Insurance Program is a government-operated mandatory social health insurance program that covers almost the entire (about 97%) Korean population. The remaining 3% of Korean people is covered by the Medical Aid program. The National Health Insurance System (NHIS) in Korea is composed of a comprehensive set of health information including around 50 million Koreans. The NHIS contains an eligibility database, health examination database, medical treatment database, and medical care institution database 9,10 . The NHISS, which includes claims and mortality data, is open to any researchers whose protocols are approved by the NHISS review committee. The Korea University institutional review board approved this study protocol in accordance with the Declaration of Helsinki of the World Medical Association.
NHIS recommends that all eligible Korean adults undergo a standardized health checkup every two years. Anthropometric and laboratory measurements including fasting glucose, lipid profile, creatinine, liver enzymes, and urinalysis are performed. Health-related behavioral variables, such as smoking, alcohol consumption, physical activity, and detailed medical history are assessed and recorded. Blood samples are taken after an overnight fast, and quality control procedures follow the Korean Association of Laboratory Quality Control. Data from a database of 16,094,237 Korean residents aged 20 years or older who had participated at least one biennial health examination provided by NHIS was examined.
Definitions. MI, IS, and all-cause mortality were examined as primary outcomes. Diagnosis of MI was defined using the International Classification of Diseases, tenth revision (ICD-10) codes (I21-I22) and admission record. Diagnosis of IS was based on ICD-10 codes (I63-I64), CT or MRI claim data, and admission record. All-cause mortality was also examined. The health examination questionnaire categorized smoking status as "non-smokers, ex-smokers, or current smokers, " and alcohol drinking as "0, ≤2, 3-4, ≥5 times/week". "Strenuous physical activity that was performed for at least 20 min" was defined as regular exercise, and subjects were graded as "exercising 0, 1-2, 3-4, ≥5 times/week". Due to the change of questionnaire about physical activity after 2009, subjects were graded as "exercising 0, ≥1 times/week. " in analysis of mortality after MI or IS and forest plot analysis. Income level was dichotomized at the lower 10%. The presence of type 2 diabetes was defined based on criteria of fasting glucose level ≥7 mmol/L or the presence of at least one claim per year for a prescription of antidiabetic medication under ICD-10 codes (E11-E14). The presence of hypertension was defined based on criteria of systolic/diastolic blood pressure ≥140/90 mmHg or the presence of at least one claim per year for a prescription of antihypertensive agent under ICD-10 codes (I10-I13, I15). The presence of dyslipidemia was defined based on criteria of total cholesterol ≥6.21 mmol/L or the presence of at least one claim per year for a prescription of antihyperlipidemic agent under ICD-10 codes (E78).
Statistical analysis. Data are expressed as mean ± standard deviation, geometric mean (95% confidence interval [CI]), or percentage. Differences between groups were identified using the Student's t-test or χ 2 -test to assess differences in the distribution of categorical variables. Event incidence rates are presented as 1000 person-years. Cox proportional hazards regression analysis was used to evaluate the association between quartiles of ALT, AST, and GGT and the incidence of outcomes. Kaplan-Meier curves were plotted and compared with the log-rank test. A multiple Cox proportional hazard model was used after adjusting for age, sex, BMI, smoking status, alcohol consumption, exercise, diabetes, dyslipidemia, and hypertension. Forest plots for incident MI, IS, and mortality according to risk factor subgroups were evaluated. Interactions between variables were tested. All statistical results were analyzed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA) and P-values < 0.05 were considered statistically significant. All statistical analyses were performed by an experienced professional statistician who is one of the authors (Han K.D., Ph.D.).

Results
Baseline characteristics of study subjects. Table 1 shows baseline characteristics of study subjects according to the quartiles of GGT levels. The higher quartile of the GGT group exhibited a higher proportion of male and elderly subjects compared to the lower quartile group. Furthermore, current smokers and heavy drinkers increased in the higher quartile of the GGT group. The prevalence of chronic metabolic diseases, including hypertension, type 2 diabetes, and dyslipidemia, also increased. As expected, ALT and AST concentrations showed stepwise incremental increases according to quartiles of GGT values.
Implications of ALT, AST, and GGT in MI, IS, and all-cause mortality. In Table 2 1.04-1.06; P < 0.001) in the highest quartile of AST compared to their lowest quartile counterparts, respectively. A much stronger relationship between GGT and MI was found, with adjusted HR of 1.27 (95% CI, 1.26-1.29; P < 0.001). For IS and all-cause mortality, a similar association with ALT, AST, and GGT was observed, and GGT was the strongest risk indicator (IS: HR, 1.36; 95% CI, 1.34-1.37; P < 0.001; Mortality: HR, 1.64; 95% CI, 1.63-1.66; P < 0.001). Interestingly, both ALT and AST revealed U-shaped associations with mortality, whereas GGT demonstrated a positive linear relationship. Kaplan-Meier survival curves for MI, IS, and mortality showed clear differences and dispersion with time passage according to the quartiles of GGT (Fig. 1). The highest GGT group exhibited the worst survival as well as a higher incidence of MI and IS. Table 3 demonstrates the association of liver enzymes with 1-year mortality after the development of MI or IS after adjusting for confounding factors. Consistent with our previous results, ALT and AST showed similar U-shaped associations with mortality, whereas GGT revealed a positive linear relationship. The risk of all-cause mortality after MI or IS was 46% higher in the highest quartile of GGT compared to the lowest quartile counterparts (HR, 1.46; 95% CI, 1.40-1.52). In Kaplan-Meier survival analysis, quartile groups of GGT exhibited different 30-day, 90-day, and 1-year survival after the development of CVD, showing increased mortality in the higher GGT groups (Fig. 2). Furthermore, other confounding factors including sex, metabolic diseases, and lifestyle factors affect the relationship of GGT with CVD and all-cause mortality. However, the increased risk of MI according to GGT levels was similar between men and women (P-interaction = 0.826). Importantly, the significant relationship of GGT with MI, IS, and mortality persisted independent of other risk factors.

Discussion
In this cohort study including a huge Korean population participating in a regular public health examination program, GGT showed a positive linear relationship with risk of CVD and all-cause mortality, whereas both ALT and AST had U-shaped associations with mortality during a median 9.1 years of follow-up. Overall, GGT had a more resolute influence on CVD and mortality compared to aminotransferases. Moreover, this study first demonstrated that higher GGT levels reflect increased risk of 1-year mortality after the development of MI or IS after adjusting for conventional risk factors. Previous studies have suggested a potential link between NAFLD and CVD. Lee et al. reported the relationship between severity of NAFLD and estimated 10-year CVD risk using the pooled cohort equation 11 . Accumulation of liver fat and elevated liver enzymes are associated with type 2 diabetes and hypertension 12,13 , both of which are representative risk factors for CVD. We observed the synergistic impact of NAFLD and metabolic syndrome on subclinical atherosclerosis 14 . Furthermore, using 18 F-fluorodeoxyglucose positron emission tomography (FDG-PET), we found that NAFLD is associated with vascular inflammation, which may reflect rupture-prone vulnerable atherosclerotic plaques 15 . Both visceral fat accumulation and risk of metabolic syndrome are significantly correlated with GGT and ALT levels 16,17 . Liver enzyme assays are a low-cost, simple, sensitive, and standardized method, which may support to assess cardiovascular risk 4 .
ALT is mainly produced by the liver due to increased hepatic inflammation or injury, whereas AST originates from liver and muscle cells and rises with myocardial cell injury or hepatic dysfunction 2 . There is a significant amount of data linking liver enzymes and incident diabetes. In a recent meta-analysis, a HR of 1.85 is related to 1 logged IU/L increment of ALT, whereas HR of 1.92 is associated with 1 logged IU/L increment of GGT after adjustment for important diabetes risk factors 12 . Moreover, mild NAFLD diagnosed using ultrasonography was found to be a determinant of incident diabetes 12 . However, the association between liver transaminase enzymes and CVD has been debated. The association between baseline ALT levels and CVD events was attenuated in multivariate models in the Framingham Offspring Study 18 . In a Korean study of 37,085 patients who underwent health examination at a hospital, elevated ALT levels were associated with increased CVD-or diabetes-related mortality 19 . However, there was no subgroup analysis using the non-diabetic group for CVD-related mortality. In a study of 455 community dwelling elderly men, Elinav et al. showed that low ALT levels are associated with increased long-term mortality 20 . Recently, Oh et al. reported a U-shaped relationship between serum ALT concentrations and all-cause death in individuals aged ≥ 60 years 21 . However, a meta-analysis failed to demonstrate a reliable relationship between ALT and CHD or stroke 3 . Decisively, a review based on prospective data summarized that current evidence does not support the linear relationship between ALT and CVD events 8 . The present study including an enormous Korean population provides evidence for the relationship of ALT and AST with CVD and mortality, although GGT was more closely associated with CVD or mortality than aminotransferases. Interestingly, the present study demonstrated a U-shaped relationship between aminotransferases and all-cause mortality. The difference in the individual mechanism of GGT and aminotransferases has not been clearly elucidated, and further research is needed. Several prospective studies have reported an association between GGT and development of cardiovascular events independent of alcohol intake. In a study of 7,613 middle-aged British men, Wannamethee et al. reported that GGT levels were positively associated with all-cause mortality, mortality from ischemic heart disease, and cardiovascular risk factors such as BMI, total cholesterol, and diabetes 22 . In another longitudinal study, GGT was an independent risk factor for cardiovascular mortality in 163,944 Austrian adults 23 . Interestingly, higher GGT was associated with chronic forms of coronary heart disease and stroke, but not AMI in men 23 . In 2007, Lee et al. showed the longitudinal relationship between GGT and metabolic syndrome, CVD, and mortality in the Framingham Offspring Study 24 . They found that a 1-SD elevation of log-GGT is associated with a 13% higher risk of CVD after adjusting for established CVD risk factors 24 . In a prospective study of 6,997 men, risk of coronary heart disease and CVD mortality was elevated in the highest quartile of GGT, whereas risk of stroke showed only a tendency of increased risk 25 . In a meta-analysis of 10 prospective studies, 1 IU/L higher GGT was associated with a HR = 1.20 (95% CI, 1.02-1.40) for CHD and HR = 1.54 (95% CI, 1.20-2.00) for stroke 3 . However, previous studies showed considerable heterogeneity. The present study includes the largest number of subjects and demonstrated that GGT showed a positive linear relationship with CVD and all-cause mortality even after adjusting for confounding factors. Furthermore, the Kaplan-Meir survival graph exhibits clear separation and dispersion of the survival curves of MI, IS, and mortality after the passage of time according to GGT quartiles. Interestingly, the present study demonstrated an increased risk of 1-year mortality according to higher GGT levels after MI or IS.

Myocardial infarction
GGT has been established as a sensitive biomarker of liver function that reflects NAFLD, hepatitis, and alcohol ingestion 2 . Previous studies demonstrated a close relationship between GGT and hepatic triglyceride   Table 3. Hazard ratios and 95% confidence intervals for mortality according to quartiles of liver enzymes after myocardial infarction or ischemic stroke. MR, mortality rate (per 1000 person-years); HR, hazard ratio; CI, confidence interval; ALT, alanine aminotransferase; AST, aspartate aminotransferase; GGT, γ-glutamyltransferase. Model 1: HRs were analyzed after adjusting for age and sex. Model 2: HRs were analyzed after adjusting for age, sex, body mass index, smoking, alcohol, exercise, diabetes, hypertension, and dyslipidemia.
SCIEnTIFIC REPORTS | (2018) 8:3764 | DOI:10.1038/s41598-018-19700-8 accumulation in obesity, insulin resistance, metabolic syndrome, and type 2 diabetes 26,27 . Recent studies have suggested that hepatokines, predominantly liver-derived proteins, may link obesity, metabolic syndrome, and CVD 28 . As mediating mechanisms, GGT has been known as a proinflammatory indicator and a marker of oxidative stress 2 . Kawamoto et al. reported the synergistic effects of higher high-sensitivity C-reactive protein (hsCRP) and GGT levels on metabolic syndrome and insulin resistance in the general population 29 . GGT is the enzyme responsible for catabolism of glutathione, the major antioxidant in humans 30 . Moreover, GGT impacts LDL cholesterol oxidation to influence plaque progression and rupture 31 . Paolicchi et al. found active GGT staining in human coronary atherosclerotic plaques 32 . They suggested the pathogenic role of GGT as an independent and synergistic factor in addition to conventional cardiovascular risk parameters. Although exact mechanism is not clear, the association between GGT and lipoproteins suggests that LDL lipoprotein can convey GTT activity inside the atherosclerotic plaque 31,33 . The GTT-mediated oxidative stress may have a role in the development and vulnerability of plaque, such as plaque erosion and rupture, enhance plate aggregation, and thrombosis 31,34 .
Previous studies have shown the close relationship between GGT and surrogate markers of CVD. Celik et al. reported significantly higher GGT levels in patients with coronary plaque than in controls using coronary computed tomography angiography (CCTA) 35 . In the Multi-ethnic Study of Atherosclerosis (MESA) study, continuous strong, positive associations between GGT and CRP, interleukin-6 (IL-6), and soluble intercellular adhesion molecule-1 (sICAM-1) were observed in multivariable models 36 . Cho et al. demonstrated the independent association between elevated serum GGT levels and coronary artery calcification (CAC) progression 37 . Recently, low bilirubin and high GGT were reported as potential biomarkers for coronary atherosclerosis in Korean men 38 .
Several studies have suggested that age influences the effect of GGT on CVD and mortality. In a nested case-control study, the association between GGT and incident CVD mortality was not found in subjects aged  reported that a stronger association between GGT change and CVD mortality was observed in younger participants 40 . Wannamethee et al. also showed a stronger association between GGT and CVD mortality in younger men (<55 years) when stratified by age group 25 . However, Mahady et al. demonstrated that neither ALT nor GGT was associated with increased risk of all-cause and cardiovascular mortality in a younger age group (≤59 years), but both were associated with excess risk in older individuals 41 . The present study confirmed that age is involved in the relationship between liver enzymes, CVD, and mortality. Furthermore, our study revealed that individuals with age <65 demonstrated a remarkably higher risk of MI, IS, and mortality associated with GGT compared to those with age ≥65. On the other hand, a previous Austrian and British study did not find an association between GGT and stroke in women 3,23 , whereas a Japanese study reported a positive relationship only in women 42 . In this study, men exhibited a higher risk of IS and mortality associated with GGT than women, whereas the risk of MI was similar between men and women. Other cardiovascular risk factors, such as BMI, diabetes, hypertension, and dyslipidemia, also showed significant interactions in the association between GGT, CVD and mortality. However, their relationship was maintained regardless of the existence of risk factors. Consistent with our results, in a study including 17,852 adults from three British cohorts, higher GGT levels are associated with CVD mortality in both individuals with and without diabetes 43 . GGT has been used as a sensitive biomarker of alcohol consumption as well as hepatic inflammation and NAFLD. Subgroup analysis exhibited a persistent association of GGT with CVD and mortality independent of alcohol consumption, consistent with the results of previous studies 3, 44 .
Dyslipidemia represents a major risk factor in CVD and small, dense LDL in nonalcoholic steatohepatitis (NASH) may increase risk for atherosclerosis and CVD 45 . Statins are most effective and commonly described drugs for dyslipidemia, although statin intolerance limits effective treatment 46 . Furthermore, statin treatment is usually tolerable in patients with chronic liver disease such as NAFLD and stable viral hepatitis 47 . Recently, approaches to minimize cardiovascular risk among patients with statin intolerance are recommended 48 . The potential link between statins, liver enzymes, and CVD might be needed to consider. MI, IS, and mortality according to statin usage at baseline are shown in Supplementary Table 1. Although statin therapy decreases the risk of outcome variables, interaction between statin therapy and liver enzymes was not found.
This study has some limitations. First, although we tried to adjust for multiple covariates to influence CVD and mortality, it was impossible to eliminate residual or unmeasured confounding factors. Second, only serum liver enzyme concentrations and other variables measured at baseline were analyzed. Third, despite exclusion of patients with medical histories of liver disease or diagnosis and treatment based on Korean insurance claim data, a few patients with rare causes of liver disease may have been included. However, our study has its own unique strengths. The present study included the largest amount of subjects ever, with more than 16,000,000 subjects and a follow-up period of 9.1 years. In addition, this study was performed based on a credible database covering almost the entire Korean population and including socio-demographic variables, laboratory information, medical diagnosis, medication, and mortality data. Therefore, comprehensive analyses regarding the effects of GGT, ALT, and AST on incident MI, IS, and mortality as well as subgroup analyses are possible after adjustment for important confounding factors.

Conclusions
The present study demonstrated that liver enzymes, in particular GGT, are associated with long-term risk of CVD and mortality independent of other established CVD risk factors. GGT might be a simple and practical tool for the assessment of individual risk of CVD and mortality, and may provide a basis for the appropriate individual intervention to prevent CVD and mortality.