Association between Non-Alcoholic Fatty Liver Disease and Dietary Habits, Stress, and Health-Related Quality of Life in Korean Adults

Considering the increasing prevalence of non-alcoholic fatty liver disease (NAFLD), this study aimed to evaluate the association between NAFLD and dietary habits, stress, and health-related quality of life (HRQoL) in Korean individuals by using data from the Korea National Health and Nutrition Examination Survey (KNHANES) VI 2013–2015. NAFLD was defined in individuals with a hepatic steatosis index (HSI) value ≥36. Eating habits were assessed based on the frequencies of eating and eating out; stress was assessed through the stress perception rate; and the EuroQol-5D (EQ-5D) questionnaire was used to assess the HRQoL. We performed a complex sample logistic regression analysis and estimated the odds ratios by adjusting for significant factors to evaluate associations between NAFLD and dietary habits, stress, and HRQoL. Occurrence of NAFLD was not significantly associated with meal frequencies over one week. With an increase in stress, based on the stress perception rate, the risk of NAFLD increased 1.316-fold (95% confidence interval (CI): 1.175–1.469, p < 0.05). Additionally, a decrease in the EQ-5D score by 1 increased the risk of NAFLD 3.38-fold (95% CI: 1.893–4.844, p < 0.05). Thus, NAFLD treatment should include stress management, and underlying HRQoL should be considered during treatment.


Introduction
The incidence of non-alcoholic fatty liver disease (NAFLD) has been increasing globally, including in children and adolescents. NAFLD currently accounts for 75% of the chronic liver disease cases in the United States and its prevalence has doubled in Korea, Japan, and the United Kingdom [1][2][3]. NAFLD has been linked to obesity, various metabolic diseases, and even cardiovascular diseases and mortality. NAFLD occurs independently of alcohol consumption and is associated with an abnormality in insulin sensitivity and fatty acid uptake in the liver, and metabolic alterations mediated by an inflammatory mechanism [2,4]. As with obesity, lifestyle habits, such as lack of exercise and unhealthy diet, have a great impact on the development and exacerbation of NAFLD [1,3,5]. Several studies have shown that consumption of certain nutrients, such as carbohydrates and fructose, affect NAFLD. High intake of carbohydrates, contributing to over 60% of daily energy consumption, or excessive simple sugar intake have been reported to exacerbate NAFLD [3,6,7].
Many studies have also shown that obesity and metabolic disorders can be associated with breakfast habits; particularly when the least of a day's food is eaten at breakfast, breakfast is skipped or of poor quality, or a nocturnal eating habit affects subsequent breakfast intake [8][9][10][11][12]. Notably, a small breakfast size exacerbated obesity in mouse experiments [8], and nocturnal syndrome was found to be correlated to obesity [9]. Furthermore, consumption of snacks along with skipping breakfast was related to obesity in adolescents [10]. In contrast, a regular diet reduced incidence of obesity and chronic diseases [11] and breakfast has been shown to be important for diet success [12].
The following persons were excluded: patients with excessive alcohol use (defined as >20 g/day for males and >10 g/day for females); hepatic cellular carcinoma; hepatitis virus B and/or C (HBV, HCV) with systematic history and blood tests; autoimmune hepatitis, drug-induced liver damage, or other liver diseases; liver cirrhosis; or other medical conditions that may affect HSI levels.
The subjects were divided into two groups, one group comprising of suspected NAFLD cases and the other a control group (Figure 1). The suspected NAFLD cases were defined as those with an HSI > 36, while the control group had HSI ≤ 36; where HSI was calculated as follows: HSI = 8 × (ALT/AST ratio) + BMI (+2 for women). We used HSI to define NAFLD since the fatty liver index (FLI) cannot be used as gamma glutamyl transferase data are not presented in KNHANES, and ultrasound or CT were not performed. The following persons were excluded: patients with excessive alcohol use (defined as >20 g/day for males and >10 g/day for females); hepatic cellular carcinoma; hepatitis virus B and/or C (HBV, HCV) with systematic history and blood tests; autoimmune hepatitis, drug-induced liver damage, or other liver diseases; liver cirrhosis; or other medical conditions that may affect HSI levels.
The subjects were divided into two groups, one group comprising of suspected NAFLD cases and the other a control group (Figure 1). The suspected NAFLD cases were defined as those with an HSI > 36, while the control group had HSI ≤ 36; where HSI was calculated as follows: HSI = 8 × (ALT/AST ratio) + BMI (+2 for women). We used HSI to define NAFLD since the fatty liver index (FLI) cannot be used as gamma glutamyl transferase data are not presented in KNHANES, and ultrasound or CT were not performed. The usual level of stress perception was investigated with questions that required the individual to respond by selecting one of the following answers: (1) I feel very much, (2) I feel a lot, (3) I feel a little, (4). A person who responded with a (1) or (2) was given a score of 1, and those who marked answers (3) or (4) were scored 0, with 1 indicating high stress and 0 indicating low stress.
HRQoL was evaluated by the widely used EQ-5D [29]. EQ-5D consists of five multiple choice questions and one subjective health level. The health status was determined with regards to five areas: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. The individuals were asked to choose one of the following three levels: "No problem at all", "There are some problems", or "There are many problems." The responses to the five questions were subjected to the score conversion system, and expressed as a point between 1, which means excellent health, and −1, which indicates health worse than death. The EQ-5D index was then calculated using the following equation: The usual level of stress perception was investigated with questions that required the individual to respond by selecting one of the following answers: (1) I feel very much, (2) I feel a lot, (3) I feel a little, (4). A person who responded with a (1) or (2) was given a score of 1, and those who marked answers (3) or (4) were scored 0, with 1 indicating high stress and 0 indicating low stress.
HRQoL was evaluated by the widely used EQ-5D [29]. EQ-5D consists of five multiple choice questions and one subjective health level. The health status was determined with regards to five areas: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. The individuals were asked to choose one of the following three levels: "No problem at all", "There are some problems", or "There are many problems." The responses to the five questions were subjected to the score conversion system, and expressed as a point between 1, which means excellent health, and −1, which indicates health worse than death. The EQ-5D index was then calculated using the following equation: We compared the general factors, stress perception, and EQ-5D values between the normal (HSI ≤ 36) and NAFLD (HSI > 36) groups. We also examined the general factors, cardiovascular-related factors (from blood tests), dietary habits (meal patterns), and whether the individuals had companions at meals in both groups. The relationship between suspected NAFLD and control group and general characteristics, dietary habits, stress, and EQ-5D are presented as an odds ratio (OR). Furthermore, to determine the most important variables in the relationship between the NAFLD and control groups and dietary habits, stress, and EQ-5D, the other variables were corrected, and the correlations are presented as OR.

Anthropometric Measurements
Anthropometric and biochemical measurements were performed by an experienced inspector. The height (cm) and weight (kg) were measured to the nearest 0.1 value, using a Seca 225 instrument (Seca, Hamburg, Germany) and GL-6000-20 scale (G-tech, Seoul, Korea), respectively. BMI was calculated as weight (kg)/square of height (m 2 ). Waist circumference (WC) was measured between the margin under the ribs and the iliac crest and recorded to 0.1 cm.
Laboratory tests using blood tests were taken after fasting for 8 h, and samples were immediately transported to a central laboratory (Neodin Medical Institute, Seoul, Korea) for processing. The Hitachi 7600 automatic analyzer (Hitachi, Tokyo, Japan) was used to measure fasting plasma glucose, HbA 1c (hemoglobin A 1c ), total cholesterol, triglyceride (TG), high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, aspartate aminotransferase (AST), and alanine aminotransferase (ALT) levels.

Sociodemographic and Lifestyle Variables
A self-written questionnaire was used to investigate the subject's sociodemographic and lifestyle variables. Based on the recent smoking status, individuals were divided into "smokers who are currently smoking" (within two years) and "non-smokers." Alcohol consumption was expressed as grams of alcohol consumed per week. Physical exercise was measured using the International Physical Activity Questionnaire [42]. Regular exercise was defined as three or more times per week and 20 min or more per trial. The education level was divided into four categories: under elementary school graduation, middle school graduation, high school graduation, and above university graduation.
Dietary habits were investigated for the number of breakfasts, lunches, and dinners during a week over the past year and the presence or absence of companions at the time of the meals.

Statistical Analysis
Statistical analyses were performed using PASW (Predictive Analytics SoftWare) statistics software (SPSS version 24.0, IBM SPSS Inc., Chicago, IL, USA). Overall, frequency analysis was performed using the complex sample plan for frequencies analysis. A complex sample Rao-Scott adjusted chi-square test and a complex sample generalized linear model were used to compare general characteristics and dietary habits, stress, HRQoL, and NAFLD. Values are presented as the number (%) or the mean ± standard deviation. A complex sample logistic regression test was used to correlate NAFLD with general characteristics and dietary habits, stress, and HRQoL. We performed a complex sample Rao-Scott adjusted Chi-square test and complex sample generalized linear model t-test (Table 1) to investigate the differences between general characteristics and dietary habits, stress, HRQoL, and NAFLD.
We performed a complex sample logistic regression analysis and estimated the OR and 95% confidence intervals (CI) by adjusting for statistically significant factors ( Table 2) to investigate the associations between dietary habits, stress, and HRQoL and potential NAFLD. All data are presented as means ± standard errors (SEs) or percentages (%) for categorical variables. p values < 0.05 were considered statistically significant.

Results
In total, 17,726 subjects (50.6% male; 49.4% female) were recruited in this study, with an average age of 43.93 ± 0.25 years. Table 1 summarizes the differences between the general characteristics, dietary habits, stress perception, and HRQoL among individuals with and without NAFLD. The rate of practice of aerobic exercise was lower among individuals with than in those without NAFLD (41.7% vs. 42.1%; p < 0.0001). In a weekly dinner frequency survey over the past year, the rate of almost no daily consumption of dinner was slightly but not significantly lower in the NAFLD group than in the non-NAFLD group. The correlation analysis between EQ-5D and stress perception rate showed a strong correlation with the parameter estimation value of −0.994. In other words, it was found that EQ-5D decreased whenever the stress perception rate increased.  Table 2 delineates the association between NAFLD risk and dietary habits, stress, and HRQoL. The risk of developing NAFLD was higher when skipping breakfast and in the absence of a companion for breakfast (OR: 1.165 and CI: 1.016-1.337, OR: 1.008 and CI: 0.903-1.126, respectively, p < 0.05). The NAFLD risk was higher when eating out than when not eating out (OR: 1.250 and CI: 1.006-1.552, p < 0.05). Every 1-in increase in WC and an increase in BMI increased the risk of NAFLD (OR: 1.237 and CI: 1.227-1.247, OR: 2.389 and CI: 2.309-2.471, respectively, p < 0.05). Every 1-unit increase in FBS (fasting blood sugar), HbA 1c , total cholesterol, HDL cholesterol, triglycerides, and LDL cholesterol also increased the NAFLD risk (OR: 1.021 and CI: 1.018-1.023, OR: 1.829 and CI: 1.707-1.96, OR: 1.008 and CI: 1.006-1.009, OR: 0.948 and CI: 0.944-0.952, OR: 1.004 and CI: 1.003-1.005, and OR: 1.008 and CI: 1.005-1.011, respectively, p < 0.05). Furthermore, the NAFLD risk was higher in the stress-perceiving group (OR: 1.316 and CI: 1.203-1.44, p < 0.05). When the EQ-5D score decreased by 1, the risk of developing NAFLD increased (OR: 2.255 and CI: 1.533-3.317, p < 0.05). Table 3 delineates the association between NAFLD and eating habits, HRQoL, and stress after correcting for significant variables. The frequency of breakfast, lunch, and dinner for one week were not significantly associated with NAFLD occurrence. In the stress-perceiving group, the risk of developing NAFLD was higher (OR: 1.314 and CI: 1.175-1.469, p < 0.05), and when EQ-5D decreased by 1, the risk of developing NAFLD increased (OR: 3.028 and CI: 1.893-4.844, p < 0.05).

Discussion
This study investigated the association between NAFLD occurrence and dietary habits, stress, and HRQoL among Korean adults. In particular, stress and HRQoL were associated with NAFLD. Herein, the HSI was considered for diagnosing NAFLD, since it is a simple and effective diagnostic index often used among Korean individuals. In a cohort study, a logistic regression model was performed using high serum ALT to serum AST ratio, high BMI, and diabetes mellitus as variables, which are independent risk factors for NAFLD. HSI was shown to have an AUROC of 0.812 [39]. Another study reported that the optimal cut-off value of HSI was 35 (sensitivity, 81%; specificity, 60%) [43]. Several studies have also been conducted considering HSI > 36 to predict NAFLD; we used the same definition in the current study Numerous studies have investigated the association between NAFLD and nutrient intake, lack of exercise, and carbohydrate intake [44][45][46][47][48][49]; however, few studies have focused on the pattern of dietary habits and NAFLD, or have reported conflicting findings [50][51][52]. To determine whether breakfast consumption influences NAFLD risk, similar to weight loss, it is necessary to assess its association with meal patterns. Herein, the NAFLD risk was lower upon daily breakfast consumption; however, it appeared irrelevant after correcting for the relevant variables, suggesting that total calories, daily carbohydrate intake, and the lack of essential nutrients are better associated with NAFLD than with dietary habits including the frequency of breakfast consumption, dinner consumption, and the frequency of eating out.
The relevance of breakfast in metabolic diseases and obesity has been adequately clarified [10][11][12]53,54]. Furthermore, the frequency of breakfast consumption is low among obese children with a high-energy intake at dinner [55]. The reduction in breakfast consumption in obesity has emerged as a global trend [11]. People do not have breakfast and cannot control their appetite, and so they binge. However, in the current study, the frequency of breakfast consumption and NAFLD risk in Korean subjects were not associated after correcting for related variables. A previous study has examined the composition of meals and NAFLD by investigating the connection between the Mediterranean diet and NAFLD [56]. The study involved 584 subjects with one or more cardiovascular disease risk factors. In the study, NAFLD diagnosis was performed by ultrasound, and adherence to the Mediterranean diet was classified as low, intermediate, or high. The study results indicated that the intermediate and high adherence groups had a lower risk of developing NAFLD than the low adherence group. The study examined the composition of the meal and diagnosed NAFLD using ultrasound, which is different from our study.
Although further studies are required, the following mechanisms could potentially promote the pathogenesis of NAFLD. Consumption of high-energy foods stimulates insulin secretion to regulate blood glucose levels. Insulin resistance results in fat breakdown in peripheral fat tissue to increase the generation of free fatty acids, which eventually increases the intrahepatic influx of free fatty acids. Furthermore, insulin increases fat biosynthesis in the liver, thus influencing the pathogenesis of fatty liver [1][2][3]. There is also a contribution of diet to NAFLD, as opposed to increasing insulin. The results of a mouse study show that intermittent fasting is effective for NAFLD [57]. Intermittent fasting in obese diet mice reduced hepatic lipogenesis and reduced β-oxidation markers to alleviate hepatic steatosis and inflammation.
Herein, the pathophysiological basis is defined as the interaction between psychological status, insulin resistance-related metabolic disorders, and NAFLD [16][17][18][19][20]. Psychological issues are associated with metabolic diseases via the accumulation of visceral fat through hormonal dysregulation, including that of insulin. Psychological factors including stress and depression promote corticotrophin-releasing hormone and cortisol secretion through the hypothalamic-pituitary-adrenal (HPA) axis [16][17][18][19][20]. Dysregulation of the HPA axis leads to the accumulation of visceral fat through hormone secretion and the secretion of inflammatory cytokines like tumor necrosis factor (TNF) [16][17][18][19][20]58]. Inflammatory substances undergo a vicious cycle that causes the relapse of metabolic abnormalities, again through insulin resistance and leptin disruptions [58]. These metabolic abnormalities are a major factor accelerating NAFLD progression [1][2][3]. This hypothesis suggests a direct and indirect link between psychological factors and NAFLD, depending on personal susceptibility.
There have been mouse studies reporting that chronic stress has a negative effect on NAFLD [59]. It was reported that hepatic TNFα, MCP-1, and HMOX mRNA expression levels were high in chronic stress-exposed mice, which was attributed to accelerated liver oxidative stress and inflammatory mechanisms [59]. Studies have revealed that NAFLD is a stress-sensitive disease, as an alteration of the HPA axis exacerbates NAFLD [60,61]. Chronic psychiatric problems are closely related to NAFLD and obesity, and some studies report that psychiatric treatment must be considered when treating NAFLD [61]. This is consistent with the results of the current study, which suggest that stress management is necessary for NAFLD treatment.
NAFLD itself results in other metabolic diseases and is associated with stress and depression. Moreover, its treatment and clinical management are costly [62]. This is speculated to result in the deterioration of the HRQoL in NAFLD, and is supported by our results. Although not presented in a table, additional analysis in the current study showed that EQ-5D had a negative correlation with stress index, which is expected because EQ-5D evaluates various quality of life parameters, including stress.
A previous study has shown that NAFLD leads to economic losses, as shown by health economics, a tool to measure the economic and social consequences of the disease [63]. In addition, various metabolic conditions that can accompany NAFLD can lead to poor HRQoL [64][65][66]. Conversely, people with poor HRQoL are more likely to develop chronic diseases, especially metabolic diseases [64,65,67,68]. These studies, along with our current study, suggest that NAFLD is closely related to HRQoL.
The EQ-5D used in this study as an HRQoL assessment tool has already been demonstrated in Korean studies [69]. EQ-5D was developed by the EuroQoL group for the purpose of clinical and economic evaluation. It is a simple and widely used tool for assessing overall HRQoL. QoL scores have been previously reported in NAFLD including NASH (non-alcoholic steatohepatitis) and cirrhosis in a previous study, which investigated the association between NAFLD and QoL using the Non-alcoholic Steatohepatitis Clinical Research Network and the Short-Form 36 (SF-36) survey [26]. Previous studies investigated the association between NAFLD severity and QoL and reported a poor QoL among individuals with NAFLD [26,62], consistent with the present study. A previous study compared the relative effects of NAFLD on the HRQoL with those of other chronic liver diseases. NAFLD patients had a significantly lower HRQoL than hepatitis B or hepatitis C patients, as revealed through multiple domains of scores of the 29-item Chronic Liver Disease Questionnaire, suggesting that the HRQoL was severely impaired in NAFLD patients [24]. Herein, the EQ-5D indicator was used to evaluate the HRQoL. The closer this score is to 1 point, the higher the QoL [70,71]. Nam et al. (2007) used the EQ-5D to estimate the weightage of the QoL questionnaire, and this indicator was extensively used to assess the QoL among Korean individuals [71]. Furthermore, herein, the probability of NAFLD occurrence was 2.255-fold among the HRQoL indicators after correcting for related variables. Multiple approaches should be considered for the clinical management of NAFLD for improving stress and the HRQoL should be accompanied by classical treatments including pharmacotherapy, lifestyle improvement, and diet.
This study has some limitations. First, since this study included cross-sectional data, it is difficult to explain the causal relationship between stress dietary habits, HRQoL, and NAFLD. Second, the gold-standard diagnosis of NAFLD is liver biopsy and abdominal ultrasound; however, it was difficult to conduct these procedures when surveying the entire population. Furthermore, although the FLI is a commonly used NAFLD indicator, it was difficult to use here because γ-glutamyl transferase levels were not determined. Third, all other variables potentially affecting NAFLD were not considered herein. Despite these limitations, this study has the following strengths. First, this study delineated the importance of psychological aspects by elucidating the association between stress and NAFLD. Second, the number of breakfast, lunch, and dinner meals per week as a dietary habit was not associated with NAFLD. In the future, more specific dietary habits should be investigated to confirm the association.
Most importantly, this study shows that the HRQoL is associated with NAFLD among Korean adults. These results may provide essential strategies to facilitate the treatment of NAFLD.
Based on the results of this study, future studies should investigate these associations using more accurate tools for NAFLD diagnosis, while evaluating and comparing HRQoL using various tools. These additional studies may provide evidence for confirming the causal relationship between NAFLD and quality of life.