A total of 58,701 volunteers (20,293 men and 38,408 women) were recruited from an urban hospital-based cohort involving multi-institutional hospitals in major cities in Korea. This cohort was part of the Korean Genome and Epidemiology Study (KoGES) conducted by the Korean National Research Institute of Health (NIH), the Korea Disease Control and Prevention Agency (KDCA), and the Ministry of Health and Welfare (MOHW) of Korea. The cohort was made aware of the public health issues associated with metabolic diseases (18). The inclusion criteria for recruiting participants were age ≥40 and the presence of mild to moderate metabolic disease states at baseline (18). The study was conducted after ethical approval from the Institutional Review Board of the National Institute of Health, Korea (KBP-2015-055) and Hoseo University, Korea (HR-034-01). The participants signed written informed consents.

On their visits to the hospital, the participants wore light clothes and removed their shoes to measure height, weight, waist, and hip circumference (19). Body mass index (BMI) was determined by dividing the body weight (kg) by height in m squared (m²). The waist circumference was measured at the midpoint between the lower border of the rib cage and the iliac crest, at the level of the umbilicus, using a flexible inch tape. Body fat and skeletal muscle mass were estimated using a machine-learning prediction model generated in the Ansan/Ansung cohort. These parameters were measured using the Inbody 3.0 measurement device (Cheonan, Korea) based on the principle of bioelectrical impedance (20). Skeletal muscle index (SMI) was calculated by dividing the appendicular skeletal muscle mass by height. A doctor measured blood pressure in the left arm in a sitting position with a sphygmomanometer.

After the subjects undertook an overnight fast, their blood was drawn using vacuum blood collection tubes with and without ethylenediaminetetraacetic acid (EDTA). Separated plasma and serum samples were used for biochemical analysis. Glucose, total cholesterol, high-density lipoprotein (HDL) cholesterol, triglycerides, creatinine concentrations, and aspartate aminotransferase (AST) and alanine aminotransferase (ALT) activities were assayed from the fasting serum or plasma samples using a Hitachi 7,600 automatic analyzer (Hitachi, Tokyo, Japan). Glycated hemoglobin or hemoglobin A1c (HbA1c) was determined using an HbA1c Analyzer from EKF Diagnostics (Manchester, United Kingdom). In the Ansan/Ansung cohort, insulin resistance was also estimated using the prediction model generated by the homeostatic model assessment for insulin resistance (HOMA-IR) equation. The HOMA-IR is calculated by multiplying fasting serum glucose concentrations (mg/dL) by insulin (mU/L) concentrations and dividing the result by 405 (21). The participants were classified into low-and high-insulin resistance with the cutoff of 2.32 HOMA-IR value (21). The serum low-density lipoprotein (LDL) cholesterol concentrations were calculated using the Friedewald equation, and subjects with serum triglyceride concentrations ≥500 mg/dl were excluded. Serum high-sensitive C-reactive protein (hs-CRP) concentrations were measured with an enzyme-linked immunoassay (ELISA) kit (R&D Systems, Minneapolis, MN, United States). The estimated glomerular filtration rate (eGFR) was calculated using the modification of diet in renal disease (MDRD) formula: 175 X serum creatinine concentration^−1.154 X age^−0.203 X [0.742 if female].

Physical exercise was determined based on the questionnaires about exercise intensity and duration, and the intensity was categorized into light, moderate, and intensive. The light exercise included walking, office work, and dishwashing; moderate exercise included brisk walking, mowing the lawn, badminton, swimming, and tennis; and intensive exercise included climbing, running, football, basketball, and volleyball. Regular physical exercise was defined as daily exercise with a weekly aggregate of over 150 min of moderate exercise or over 100 min of intense exercise. The participants who did not belong to the regular or intense exercise groups were considered part of the low-exercise group. Alcohol consumption was assessed based on the type, amount, and frequency of alcohol intake during the previous 6 months. Daily alcohol intake was calculated by multiplying the drinking frequency by the average alcohol consumed on each occasion and dividing it by the period in days. Smoking status was stratified into current, past, or never based on >100 cigarettes smoked over the lifetime and smoking during the last 6 months before joining the study.

MetS was defined based on 2005 revised National Cholesterol Education Program-Adult Treatment Panel III (NCEP-ATP III) criteria for Asians. The cutoff of waist circumference was ≥90 cm for men and >85 cm for women, as established by the Korean Society for the Study of Obesity (KSSO), since they increase MetS risk in lower waist circumferences than Caucasians (22, 23). The persons taking medication for hyperglycemia, dyslipidemia, and hypertension were assigned to the MetS group, although they had serum glucose concentration, lipid profiles, and blood pressure within a normal range.

The KoGES committee established a semi-quantitative food frequency questionnaire (SQFFQ) for Koreans, including 106 foods commonly consumed in meals (24). The SQFFQ was validated from the 3-day food records in four seasons (25). The SQFFQ was filled by the participants based on food intake for the last 6 months and included the consumption frequency and amounts of the 106 food items with the selected serving sizes. The results of the food intake from the SQFFQ were converted into the intake of 23 nutrients, including sodium, using the computer-aided nutritional analysis program 3.0 developed by the Korean Nutrition Society (24). The nutrient contents of the processed and cooked foods in SQFFQ were calculated based on the typical recipes. Sodium intake was calculated from the food intake determined from SQFFQ (8). Lee et al. (8) calculated sodium intake using the SQFFQ, the same one used in the present study. The salt contents of the foods, such as Kimchi, Doenjang, and soup, were estimated using their regular recipes, although condiment contents, including salt, can be varied according to the individuals.

Jang, a fermented soybean paste containing salt, was included as a food group in the SQFFQ, and it included Doenjang soup, Doenjang, Chungkookjang, and Ssamjang. The daily Jang intake was calculated from the frequencies and portion sizes of these four items as the other food items. Sodium intake from the Jang intake was calculated using the computer-aided nutritional analysis program 3.0. Due to different variations of Jang and sodium intake, Jang intake was categorized into tertiles or two groups (cutoff: 1.9 g/day; 25th percentile), while sodium intake was stratified into quintiles and two groups (cutoff: 1.5 g/day; 25th percentile).

The 106 food items were classified into 29 pre-combined food groups and used as independent variables in a principal component analysis (PCA) conducted to find the optimal factors representing dietary patterns in Korea. The optimal number of the classified factors was estimated based on eigenvalues of >1.5 in the PCA. In this cohort, the number of factors that met the criteria was four. The orthogonal rotation procedure (Varimax) was applied to generate the appropriate clusters, indicating dietary patterns (26). Dietary factor-loading values of ≥0.40 were used to indicate significant contributions of food items by assigning names to the dietary patterns (26). The primary food groups in the four different dietary patterns were fish, crabs, red meat, vegetables, kimchi, pickles, seaweed, and mushrooms for the Korean balanced diet (KBD); noodles, bread, fast foods, soups, and meat for the Western-style diet (WSD); beans, potatoes, green vegetables, seaweed, milk, nuts, and eggs for the plant-based diet (PBD); and rice in the rice-main diet (RMD).

The dietary inflammatory index (DII) was calculated based on the dietary inflammatory weights of foods and nutrients having anti-or proinflammatory properties (energy, 32 nutrients, four food products, four spices, and caffeine). The inflammatory weights of the foods and nutrients in the DII equation were adopted from a previous study (27). However, the intake of the four spices, garlic, ginger, saffron, and turmeric, was removed from the DII original equation since they were not measured in the SQFFQ. The DII was calculated by multiplying the pro-(plus value) and anti-inflammatory weights (minus value) of the 38 dietary components by their daily intake and then dividing the sum of each item by 100.

Stress status was evaluated with 18 questions about physical and psychological stress at home and work, and each question was scored from zero (lowest stress) to three (highest stress). The overall stress status scores were estimated by a summation of the scores of the 18 questions, and higher stress status scores indicated that the participant’s stress status was high.

Statistical analyses were carried out using the SAS version 9.3 software (SAS Institute, Cary, NC, United States). When the sample size was calculated using the G*Power program with effect size (0.05), significant level (α = 0.05), and power (β = 0.99), the sample size derived was 1,036. The sample size was satisfied for each gender. The descriptive statistics for the categorical variables (e.g., gender and lifestyle) were evaluated based on the frequency distributions of low-and high-Jang intake in each gender. The statistical differences in the frequency distributions were measured using the Chi-squared test. Adjusted means and standard errors of the low-Jang and high-Jang intake groups in each gender were calculated for continuous variables. The statistical differences between low-and high-Jang intake were determined using the analysis of covariance (ANCOVA) after adjusting for covariates, including age, residence area, education, income, energy intake, sodium intake, alcohol consumption, physical exercise, and smoking status. The adjusted odds ratio (ORs) and 95% confidence intervals (CI) of Jang intake with MetS risk were measured by multiple regression analysis after covariate adjustment. p values <0.05 were considered to be statistically significant.

There was no difference in age between the low-and high-Jang intake groups. The participants, both men and women, with below high school education and income below 4,000 dollars per month, had a lower Jang intake than those with over high school education and income over 4,000 dollars (Table 1). Participants who exercised regularly consumed more Jang than those with a sedentary lifestyle. An evaluation of Jang intake based on the smoking status showed that male non-smokers had a lower Jang intake than former and current smokers, but women showed the opposite trend (Table 1). Alcohol intake was higher in men than women but did not vary with the Jang intake in both men and women. The stress status was linked to Jang intake, and both men and women participants with a high Jang intake had lower stress scores.

The men and women participants with a high Jang intake consumed 4.62 ± 0.03 and 4.64 ± 0.02 g Jang per day, respectively, while those with a low Jang intake consumed 0.60 ± 0.05 and 0.65 ± 0.04 g/day, respectively (Table 2). The multigrain rice intake was higher, but white rice intake was lower in male participants with a high Jang intake compared to those with a low Jang intake. Noodle and bread intakes were also higher in the low-Jang intake group than in the high-Jang intake group for both genders (Table 2). The intake of fruits and vegetables was higher in the high-Jang intake group than in the low-Jang intake group. Furthermore, the intake of kimchi, a fermented cabbage, was much higher in the high-Jang intake group than in the low-Jang intake group for both genders (Table 2). The intakes of seaweed, fish, beans, and nuts were higher in the high-Jang intake group than in the low-Jang intake group, but meat intake showed a trend opposite to that of seaweed. There were twice as many participants with high Jang intake in the KBD and PBD compared to those with low Jang intake in both genders. However, the trend was reversed in the participants having the RMD (Table 2).

Energy intake did not meet the estimated energy requirement (EER) in men regardless of their Jang intake. In women, the EER was met only in the participants in the high-Jang intake group (Table 3). Carbohydrate intake was lower in the high-Jang intake group compared to the low-Jang intake group in both genders. However, the trends of protein and fat intakes were opposite to that of the carbohydrate intake in both genders. Fiber and calcium intakes were 1.3-fold higher in the high-Jang intake group than in the low-Jang intake group in both genders (Table 3). Along with sodium, potassium intake was higher in the high-Jang intake than in the low-Jang intake groups in both genders. However, the potassium-to-sodium intake ratio was lower in the high-Jang intake group compared to the low-Jang intake group in both genders. Dietary consumption of vitamin C did not meet the recommended intake levels (100 mg/day) in the low-Jang intake group but was met in the high-Jang intake group (Table 3). Similarly, vitamin D consumption also did not reach the recommended intake levels, although it was much higher in the high-Jang intake group than in the low-Jang intake group. DII was lower in the high-Jang intake group than in the low-Jang intake group for both genders. Moreover, the intakes of total polyphenols, flavonoids, and isoflavonoids were higher in the high-Jang intake group than in the low-Jang intake group for both genders (Table 3). These results indicated that the participants with a high Jang intake consumed a high-quality diet compared to those with a low Jang intake.

Waist circumferences and fat mass were higher in the high-sodium group than in the low-sodium group in both genders. Interestingly, SMI was also higher in the high-sodium group than the low-sodium group (Table 4). There were no significant differences in serum glucose, LDL, HDL, AST, and ALT concentrations, HbA1c levels, insulin resistance by HOMA-IR, and systolic blood pressure (SBP) between the two groups. Serum triglycerides and diastolic blood pressure (DBP) were slightly higher in the high-sodium group than in the low-sodium group (Table 4). eGFR was also higher in the high-sodium group than in the low-sodium group, indicating that excess sodium from the diet can be removed through urine excretion.

The highest sodium intake (≥3.31 g/day) was positively associated with MetS risk by 1.3 times based on the lowest sodium intake (<1.33 g/day) in the quintile categories of all participants (Figure 1A). The association was not shown in gender difference (Figure 1A). However, sodium intake was not associated with MetS risk in all the participants when daily sodium intake was divided into two groups by 1.5 g/day. The cutoff was assigned conservatively, less than WHO recommended sodium intake (2 g sodium/day; 5 g NaCL/day) since the sodium intake measured from the SQFFQ could be underestimated. Sodium intake was associated with some components of MetS in all participants: A positive association with waist circumference and fat mass and an inverse association with serum HDL concentrations were observed with sodium intake. However, there was no association of sodium intake with serum glucose and triglyceride concentrations and blood pressure in all participants (Figure 1B). There was no association between sodium intake and the risk of MetS and its components in men. However, sodium intake was positively associated only with waist circumferences and fat mass only in women (Figure 1B).

BMI, waist circumference, SMI, fat mass, serum glucose concentrations, HbA1c levels, and insulin resistance were not significantly different between the low-and high-Jang intake groups in both genders (Table 5). In the lipid profiles, the serum triglyceride concentrations showed an interaction with gender types, but there was no significant difference between the low-and high-Jang groups in both genders (Table 5). There were no significant differences in the serum total cholesterol, HDL, and LDL concentrations in the low-and high-Jang intake groups. Serum hs-CRP concentration, an inflammation index, was lower in the high-Jang intake group than in the low-Jang intake group. SBP, DBP, eGFR, and serum AST and ALT concentrations were not significantly different between the low-and high-Jang intake groups, although they showed a gender difference (Table 5).

Jang, a fermented soybean containing salts, provides a salty and umami taste to food to enhance the flavors in Korean-style cooking. Overall, the Jang intake in tertiles was inversely associated with MetS risk and was inversely associated with specific components of MetS, such as waist circumference, fat mass, and serum glucose and hypo-HDL concentrations, after adjusting for age, energy intake, residence area, education, income, alcohol intake, smoking status, physical activity, and sodium intake. Figure 2A. The association between MetS and Jang intake was seen in gender differences. In the tertile classification of Jang intake, Jang intake was inversely associated with MetS only in men but not women (Figures 2B,C). In MetS components, Jang intake showed an inverse association with waist circumference, fat mass, and blood pressure in men (Figure 2B). However, in women, it was inversely associated with fat mass and serum glucose and hypo-HDL concentrations (Figure 2C).

When Jang intake was categorized into two groups, the high Jang intake was inversely associated with waist circumference, fat mass, and serum glucose and hypo-HDL concentrations, but not blood pressure, in all participants (Figure 2D). In men, Jang intake showed an inverse association with waist circumferences, fat mass, and blood pressure whereas, in women, Jang intake was inversely associated only with serum hypo-HDL concentrations (Figure 2D).