We analyzed a cross-sectional review utilizing the 2008–2011 Korea Countrywide Wellbeing and Diet Assessment Surveys (KNHANES). KNHANES is a nationwide, representative, and population-centered study carried out every year by the Korean Centers for Illness Control and Prevention (K-CDC). Detailed info about these knowledge has been documented14,15. Topics aged < 20 years were excluded, and we conducted the following steps to select the final samples.
Step (1) We estimated the average amount of alcohol consumption per week using questionnaire-based information (covariates section). Lee et al.6 excluded men and women with alcohol consumption>140 g/week and >70 g/week, respectively. In an additional examine by Kang et al.7, topics who eaten >210 g/week (adult men) and >140 g/week (girls) ended up removed to examine NAFLD. Lee et al.16 excluded guys and girls who consume ≥30 g/working day and ≥20 g/working day of alcoholic beverages, respectively. Among the these requirements, we chosen the gender-particular cut-offs for categorizing the alcohol group decided by Lee et al.6 and excluded non-alcoholic topics who eaten much more than 140 g/7 days and 70 g/7 days in males and females, respectively.
Phase (2) Several scientific studies have conducted analyses following excluding topics with liver ailment. Lee et al.6 excluded topics with evidence of hepatitis B or C virus and experienced earlier been identified to have liver cirrhosis. In addition, various research eliminated good hepatitis B antigens or hepatitis C antibodies7,16,17. In our study, any subject having positive serologic markers for viral hepatitis and liver cirrhosis diagnosis were excluded.
Action (3) Lee et al.6 retained topics if any of the 4 hepatic steatosis indices could be calculated. In yet another review, subjects whose hepatic steatosis index could not be calculated owing to missing values of clinical and laboratory variables ended up excluded16. In the current review, we only excluded topics with out any indices for hepatic steatosis.
Step (4) Subjects who did not have a measurement of body composition were excluded.
Eventually, we chosen 12,324 subjects (gentlemen: 4201, ladies: 8123) to recognize the association among muscle mass mass and NAFLD. Take note that topics only enrolled in the KNHANES ended up used to acquire the Korean population’s generalized conclusions.
The current review was authorized by the Institutional Evaluation Board (IRB) of the K-CDC (IRB amount: 2008-04EXP-01-C, 2009-01CON-03-2C, 2010-02CON-21-C, 2011-02CON-06-C). The present research was performed in accordance with the Declaration of Helsinki.
Measurement of entire body composition and muscle mass mass
Dual-vitality X-ray absorptiometry (DEXA) was utilized to measure the body composition information gathered for the head, trunk, pelvis, arms, legs, and complete entire body. Skeletal muscle mass (SMM) was calculated by subtracting the lean system mass (g) from bone mineral content (g). The appendicular skeletal muscle mass (ASM) was measured by the sum of SMM for the two arms and two legs1,2. The skeletal muscle mass index (SMI) was calculated by dividing ASM by BMI. Based on the Foundation for the Countrywide Institutes of Wellness (FNIH) Sarcopenia Job requirements18, gender-specific LSMI was outlined as SMI < 0.789 in men and < 0.512 in women.
Calculation of indices for NAFLD
Numerous models for predicting NAFLD have been established and validated for their clinical usefulness in predicting CVD and cancers19,20. The KNHANES that we analyzed includes incomplete clinical and laboratory data for calculating certain NAFLD models. For example, because the KNHANES does not compose serum uric acid, it cannot calculate the Comprehensive NAFLD score21. Therefore, we used four indices for screening NAFLD that are computable from the KNHANES: the hepatic steatosis index (HSI)8, Framingham steatosis index (FSI)9, liver fat score (LFS)10, and fatty liver index (FLI)11.
The HSI was based on 10,724 subjects (derivation set, 5360 validation set, 5364) who visited Seoul National University Hospital Gangnam Healthcare Center8. Multivariate logistic regression was used to select risk factors for NAFLD, and the model for predicting NAFLD was constructed with training parameters using logistic regression, scaling optimized parameters, and adjusting parameters for gender. The HSI used ultrasonography (US) examinations as the actual label for NAFLD.
The FSI was established using 1181 participants in the Framingham Heart Study Third Generation Cohort. Clinical and laboratory biomarkers were determined by integrating expert knowledge (i.e., gender) and stepwise logistic regression. The model for predicting hepatic steatosis was constructed using logistic regression. A multi-detector CT scan was used to measure the level of liver attenuation9.
The LFS was based on Finnish subjects in which the ratio of type 2 diabetes is 0.23 (111/470)10. Predictors for the LFS were selected using multivariate backward stepwise logistic regression, and the prediction model for NAFLD was established using logistic regression. The liver fat content was measured using proton magnetic resonance spectroscopy.
The FLI had been curated using 496 Italian subjects (216 subjects with and 280 without suspected liver disease) enrolled in the Dionysos Nutrition & Liver Study11. By integrating bootstrap and stepwise logistic regression, potential predictors were selected, and the model for predicting fatty liver was constructed using logistic regression. This study used US examinations to diagnose NAFLD.
We reviewed about 20 covariates previously used for multivariate logistic or Cox proportional hazard regression models in four studies6,7,16,17 (Table 1). Among them, we did not include C-reactive protein, homeostatic model assessment for insulin resistance (HOMA-IR), or serum levels of vitamin D, hemoglobin A1c (HbA1c), and uric acid as covariates in our study because these values were missing in about half of the subjects in KNHANES 2008–2011. Race and ethnicity were not included in this study because only Koreans were analyzed in the dataset.
The presence of physical activity (PA) or regular exercise was defined by Lee et al.6 as vigorous exercise ≥ 20 min at a time and ≥ 3 times per week. Peng et al.17 determined PA as 3 to 6 metabolic equivalents (METs) ≥ 5 times per week or more than six METs ≥ 3 times per week. Lee et al.16 defined vigorous exercise as ≥ 20 min at least three days per week and moderate exercise as ≥ 30 min (or walking) at least 5 days per week. We determined the presence of regular exercise when a subject satisfied one or more of the above criteria6,16,17.
Smoking status is typically categorized into three groups: non-, ex-, and current smokers6. Peng et al.17 defined a person who smokes at least 100 cigarettes during his or her lifetime as a smoker. In the eminent atherosclerotic cardiovascular disease (ASCVD) models22,23,24, smoking status is considered one of the crucial predictors for ASCVD and is implicated as the binary form of current smoking or not. The present study used the binary form of smoking (current vs. ex- and non-smoker) as a covariate.
For the definition of drinking status, we measured the total amounts of alcohol consumption in each subject, similar to the study by Lee et al.6. However, several studies initially excluded subjects with high alcohol consumption and did not include alcohol consumption as a covariate6,7,17. Therefore, we did not consider it. For daily protein intake (g/day), the 24-h recall method was conducted by well-trained dietitians. Nutritional contents, such as daily intakes of carbohydrates and protein, were measured using the Korean Food Composition Table provided by the Rural Development Administration of Korea25.
We defined subjects with hypertension based on the Seventh Report of the Joint National Committee as follows: systolic blood pressure (BP) ≥ 140 or diastolic BP ≥ 90 mmHg previous diagnosis of hypertension by a medical doctor and antihypertensive medications26. Diabetes mellitus was defined as serum fasting glucose level ≥ 126 mg/dL previous diagnosis by a medical doctor or glucose-lowering drugs. Kang et al.7 used both questionnaire- and laboratory-based information for the status of dyslipidemia. However, too many subjects missing data on serum level of high-density lipoprotein (HDL) cholesterol. Therefore, we only depended on the questionnaire-based information, including previous diagnoses by a doctor and anti-lipidemic drug administration to determine the status of dyslipidemia.
Several studies used obesity status as a covariate7,16. One study used obesity defined by BMI7, and another determined obesity by waist circumference (WC)16. However, because three of the four models for predicting fatty liver included BMI as a predictor (Table 2), the BMI-based definition of LSMI was strongly predicted to be significantly related to the BMI-based fatty liver indices. Therefore, we did not use obesity as a covariate. Serum creatinine level is known to be significantly associated with muscle mass27. Thereby, it was included as a confounder.
All data in the KNHANES are presented as mean ± standard error for continuous variables and as frequency and percentage (%) for categorical variables. For continuous variables, we used a one-way analysis of variance to test for linear trends of the covariates after determining the mean values of each quartile group of the NAFLD index (e.g., HSI) as continuous variables. A chi-square test was used for categorical variables to compare differences among quartile groups. The comparison between two continuous variables (e.g., HSI vs. FLI) was performed using Pearson’s correlation coefficient (PCC).
Sample weights assigned to subjects were used to represent the whole Korean population. We used a linear regression model to identify the association between two continuous variables. In detail, we set muscle mass and each continuous form of an index for NAFLD as dependent and independent variables, respectively. For the relationship between the binary form of muscle status, including normal and LSMI (dependent variable) and each index for NAFLD (independent variable), we used a logistic regression model. For the multivariate model (i.e., linear and logistic regressions), we determined age, PA, current smoking, daily protein intake, hypertension, type 2 diabetes, dyslipidemia, cardiovascular disease (CVD), systolic BP, serum fasting glucose, total cholesterol, and serum creatinine as the confounding variables. All statistical analyses were conducted using R language (version 4.1.2). We set a p-value less than 0.05 as the significance level.
The present study was approved by the Institutional Review Board (IRB) of the K-CDC (IRB number: 2008-04EXP-01-C, 2009-01CON-03-2C, 2010-02CON-21-C, 2011-02CON-06-C).
Consent to participate (ethics)
Informed consent had been obtained from all participants in the KNHANES.