Data sources
The National Health Insurance Service-National Sample Cohort 2.0 (NHIS-NSC 2.0) data set, which comprises more than 97% of Korean citizens affiliated with the obligatory public health insurance program of the National Insurance Health Service (NIHS) was analyzed for this study. The total population (n=48,222,537) was classified into 2,142 groups according to their age, sex, residing cities, eligible status, and income levels. From each stratum, randomly selected 2.1% individuals (n=1,021,208) in the year of 2006 were included in NHIS-NSC 2.0 data. In addition, retrospective and prospective follow-up data collected from 2002 to 2015 were included in the data set [13]. The individuals included in NHIS-NSC 2.0 are representative of the entire Korean population.
The cohort data was composed of four data sets: (1) sociodemographics; (2) medical claims including information on diagnosis based on the 10th revision of the International Classification of Disease (ICD-10) codes, admission, and treatment; (3) the results of the National Health Screening; (4) information on the medical institutions. The Korean national health check-ups comprise of the questionnaires on medical history and health-related behaviors such as smoking and alcohol consumption, chest X-rays, physical examinations, and blood tests. About 72.1% of the eligible population had completed the National Health screening programs, according to the 2013 NHIS statistics [13]. The cohort data also included mortality statistics from the death registration database of Statistics Korea.
The study protocol was approved by the NHIS review committee for accessing NHIS-NSC 2.0 data and also approved by the Institutional Review Board of the Chungnam National University Hospital in Daejeon, Korea (IRB No. 2019-10-053). The board waivered the requirement for informed consent, for NHIS-NSC 2.0 data was public.
Study population
Study subjects were any Korean citizens who had completed the National Health check-up at least once from 2009 to 2014. Data from the first check-up was defined as the index data, and the year when the index check-up was done was the index year. Subjects with age ³ 20 years old were included, whereas subjects meeting the pre-specified exclusion criteria were excluded. All subjects previously diagnosed with HF were excluded for the purpose of the study. Other exclusion criteria were comorbid conditions that could affect the onset of HF, including hypertension, diabetes, atrial fibrillation, cerebrovascular disease, ischemic heart disease, peripheral vascular disease, valvular heart disease, chronic kidney disease, and chronic pulmonary disease. Subjects on prescribed medications such as oral hypoglycemic agents, antihypertensives, or lipid-lowering drugs within one year before the index check-up, subjects with hypertension (systolic blood pressure ³ 140mmHg or diastolic blood pressure ³ 90mmHg), and subjects with elevated fasting blood glucose ³ 126mg/dL at the index check-up were excluded. Subjects with risk factors that could affect the FLI, such as liver disease and autoimmune disease diagnosed before the index year, were excluded. Finally, those with any missing data in the index check-up were also excluded.
Definition of HF
The primary outcome of this study was HF incidence in relation to the FLI. The incidence of HF was defined as the first HF occurrence documented on the medical record at least two separate days of outpatient visits and admission, or one-time record of death due to heart failure. The diagnosis of HF was screened in data from questionnaires and the 1-year medical claim data before the index year. The presence of HF was defined using the following ICD-10 codes in the medical claim dataset: “Hypertensive heart disease with HF” (I10.0), “Hypertensive heart disease with hypertensive kidney disease with HF” (I13.0), “Hypertensive heart disease with hypertensive kidney disease with HF and kidney failure” (I13.2), “Ischemic cardiomyopathy” (I25.5), “Dilated cardiomyopathy” (I42.0), “Cardiomyopathy, unspecified” (I42.9), “Cardiomyopathy in diseases classified elsewhere” (I43), “HF” (I50) including “Congestive HF” (I50.0), “Left ventricular failure” (I50.1), and “HF, unspecified” (I50.9).
Definition/Ascertainment of covariates
Body mass index (BMI) was weight (kg) over height (m)-squared. The population with a BMI of ≥ 25 kg/m2 was regarded as obese according to the World Health Organization guideline for the Asian population [14]. Smoking status was classified into three categories: non-smoker, ex-smoker, and current smoker. Alcohol consumption and physical activity were reported using standard self-reporting questionnaires. The total amount of alcohol consumption was calculated by multiplying the number of days per week of drinking and the amount of alcohol consumed per day. Levels of physical activity were estimated by summing up the intensity levels of workout in METs multiplied by the number of days per week of each intensity level : (1) 30-min of light exercise (2 METs); (2) 30-min of moderate exercise (3 METs); (3) 20-min of vigorous exercise (6 METs).
Blood samples were analyzed in a number of qualified institutions, which were annually audited by the Korean Association of External Quality Assessment Service. Data were reviewed at the time of HF occurrence, disqualification of the NHIS (death or immigration), or the end of the study (December 31th, 2015).
Calculation of the fatty liver index
The FLI is a well-validated surrogate marker for identifying patients with NAFLD [13]. The FLI was calculated with triglycerides (TG), BMI, gamma-glutamyl transferase (GGT), and waist circumference (WC) with the following equation:
FLI = (e 0.953´loge (TG) + 0.139´BMI + 0.718´loge (GGT) + 0.053´WC - 15.745) / (1 + e 0.953´loge (TG) + 0.139´BMI + 0.718´loge (GGT) + 0.053´WC - 15.745) ´ 100
The original study on the FLI used 60 as the cutoff value, and the FLIs higher than 60 indicated fatty liver with a positive likelihood ratio of 4.3 in the general population [15]. Although the FLI is simple to calculate and easy to screen fatty liver disease, there has been insufficient evidence to use the same value in Asians because of lower BMI and WC than other ethnic populations [16]. In this study, instead of using the absolute value of the calculated FLI, subjects were grouped into quartiles according to their FLI range, and each quartile group was used in the statistical analysis to determine the relationship between FLI and new-onset HF.
Statistical analysis
Continuous variables were used as mean ± standard deviation and categorical parameters as the numbers with a percentage. All statistical analyses were performed using R software version 3.3.3 (R Foundation for Statistical Computing, Vienna, Austria; www.r-project.org). Statistical differences between the FLI quartiles were estimated using the Chi-square test and one-way analysis of variance tests. Cumulative event rates of the FLI quartiles were calculated with the Kaplan-Meier method and compared with each other using a log-rank test. Adjusted hazard ratios (HR) and 95% confidence interval (CI) for HF incidence were estimated using Cox proportional hazard regression analysis. In the multivariate analysis, age and sex were adjusted in Model 1. In Model 2, clinical characteristics associated with new-onset HF of borderline statistical significance (P<0.100) were adjusted in addition to age and sex. Confounding factors, such as hypertension and diabetes, were excluded because of their strong association with NAFLD. Using NAFLD with other risk factors in a multivariate model raises the possibility of introducing multicollinearity into the model. The variance inflation factor (VIF) in all models (Supplementary Table 1) was examined, and VIF levels were less than 10, which resolved the multicollinearity. We adjusted covariates adjusted for the subgroup analysis. The effect size represent HRs for 1 unit increment in LogeFLI. Boxes and spikes indicate HRs and corresponding 95% CIs. P-values of < 0.05 were considered statistically significant.