Dataset and study participants

This study used data from the Korea National Health and Nutrition Examination Survey (KNHANES) conducted between 2019 and 2021. Among the 22,559 respondents who participated in the KNHANES during this period, individuals aged <19 years (n = 3,868) and those who were pregnant or lactating (n = 100) were excluded. Participants with missing information on covariates, including education level, employment status, household income, marital status, alcohol consumption, smoking status, physical activity, and body mass index (BMI) (n = 2,330), were excluded. Additionally, participants with a total energy intake (<500 kcal or >5,000 kcal; n = 2,982) were excluded. Furthermore, participants who lacked data on MetS components, for whom the hepatic steatosis index (HSI) could not be calculated (n = 307), and those with a history of hepatitis (n = 197) were further excluded. The final sample comprised 12,775 participants (Fig 1). Written informed consent was obtained from all participants before study participation. The dataset was obtained from the Korea Disease Control and Prevention Agency, and the study protocol was approved by the Institutional Review Board (IRB approval no. 2018-01-03-C-A, 2018-01-03-2C-A, and 2018-01-03-5C-A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow diagram of study participant selection.

https://doi.org/10.1371/journal.pone.0346774.g001

Assessment of dairy product intake

Daily consumption of dairy products was computed from a dietary intake survey using a 1-day 24-hour dietary recall. Dairy products were identified using tertiary food codes and categorized as condensed milk, milk, goat milk, liquid yogurt, semi-solid yogurt, sorbet, ice milk, ice cream, cheese, or cream. These items were converted into serving sizes according to the dietary reference intakes of Koreans [20]. One serving size was defined as 200 g of milk (including condensed milk, milk, and goat milk), 150 g of liquid yogurt, 100 g of semi-solid yogurt, ice cream (including sorbet, ice milk, and ice cream), cream, and 20 g of cheese. Dairy product intake was classified as “0” for no intake, “0–1” for intake of less than one serving, and “≥1” for intake of one or more servings.

Assessment of nonalcoholic fatty liver disease

The hepatic steatosis index (HSI) is a simple and widely used screening tool for NAFLD. The HSI formula used in this study was as follows: HSI = 8 × (ALT/AST ratio) + BMI + 2 (women) + 2 (diabetes mellitus) [21]. Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels were measured using a LaboSpect 008AS (Hitachi, Tokyo, Japan) after fasting for at least 8 h. BMI was calculated as weight (kg) divided by height (m²). Height was measured using a Seca 274 stadiometer (Seca, Hamburg, Germany), and weight using a GL-6000–20 scale (G-tech, Seoul, Republic of Korea). Data regarding sex and diabetes status were collected using a self-administered survey. NAFLD was defined using the HSI with a cutoff value of >36.

Assessment of metabolic syndrome components

According to the National Cholesterol Education Program Adult Treatment Panel III (NCEP ATP III) guidelines, MetS was defined as present when an individual met at least three of the following five criteria: (1) fasting blood glucose ≥100 mg/dL or current use of medication or insulin treatment for hyperglycemia or type 2 diabetes; (2) waist circumference ≥90 cm for men and ≥80 cm for women; (3) systolic blood pressure (SBP) ≥130 mmHg or diastolic blood pressure (DBP) ≥85 mmHg, or current use of antihypertensive medication; (4) HDL cholesterol <40 mg/dL for men and <50 mg/dL for women; and (5) triglycerides ≥150 mg/dL [22]. Blood samples were collected from participants after a fasting period of at least 8 h. HDL cholesterol, triglyceride, and fasting blood glucose levels were measured using a Labospect 008AS (Hitachi, Tokyo, Japan). Waist circumference was assessed at the midpoint between the lowest rib and highest iliac crest using a Lufkin W606 tape measure (Lufkin, USA). SBP and DBP were measured using a non-mercury aneroid sphygmomanometer (WatchBP Office AFIB; Microlife, Switzerland), and the average of these measurements was used.

Assessment of covariates

Demographic characteristics included sex, age, BMI, education level, household income, employment status, marital status, household type, and region. Sex was classified as “men” or “women,” and education level was categorized as “≤elementary school,” “middle school,” “high school,” or “≥college.” Household income was categorized into quartiles as “low” (quartile 1), “low-medium” (quartile 2), “medium-high” (quartile 3), and “high” (quartile 4). Employment status was classified as “employed” or “unemployed,” and marital status was classified as “single” or “married.” Household type was categorized as “single-person” or “multi-person,” and region was classified as “urban” or “rural.” Health behaviors included alcohol consumption, smoking status, and physical activity. Alcohol consumption was classified as “none” (never or <1 time/month), “moderate” (1–4 times/month), or “high” (more than 2 times/week). Smoking status was classified as “never,” “past,” or “current.” Physical activity was classified as “yes” or “no” based on engagement in at least 2.5 h of moderate-intensity activity, 1.25 h of vigorous-intensity activity, or a combination of both per week.

Statistical analyses

The complex sampling design of the survey was determined by applying sampling weights, strata, and clusters. For analyses combining multiple survey years, the integrated weight was calculated by multiplying the sampling weight for each year by the integration ratio. To compare general characteristics according to NAFLD status, continuous variables are presented as means and standard errors (SEs) using t-tests and general linear models, and categorical variables are presented as frequencies (n) and percentages (%) using chi-square tests. Multivariable logistic regression was performed to estimate the adjusted odds ratios (AORs) and 95% confidence intervals (CIs) for the association of dairy product intake with NAFLD, dairy product intake with MetS components, and MetS components with NAFLD. To analyze the association between MetS components, each MetS component was categorized into two groups using the same criteria applied in the definition of MetS. All models were adjusted for age, education level, household income, employment status, marital status, household type, region, alcohol consumption, smoking status, total energy intake, physical activity, and BMI. To evaluate the mediating effects of MetS components on the relationship between dairy product consumption and NAFLD, multiple regression analyses were conducted using a bootstrapping method in an SAS macro (n = 1,000). All variables included in the mediation analysis were treated as continuous. MetS components were analyzed as continuous clinical measurements, and NAFLD was represented by a continuous HSI score. Using a continuous HSI score allowed us to capture the full spectrum of hepatic steatosis severity, preserve variability, and satisfy the linearity assumptions of the bootstrapping approach implemented in the SAS macro. All statistical analyses were conducted using SAS software (version 9.4; SAS Institute, Cary, NC, USA). Two-sided p-values < 0.05 were considered statistically significant.

Summary of study findings

This study examined the mediating effects of MetS components on the association between dairy product consumption and NAFLD in Korean adults. Among women, consumption of at least one serving of dairy products per day was associated with a lower prevalence of NAFLD as well as a reduced prevalence of several MetS components, including hyperglycemia, abdominal obesity, low HDL cholesterol, and elevated triglycerides, compared with women who did not consume dairy products. In contrast, no significant associations were observed among men. Additionally, all MetS components were independently associated with NAFLD in both men and women. Mediation analyses further indicated that the protective association between dairy intake and NAFLD in women was significantly mediated by waist circumference, SBP, and HDL cholesterol levels, whereas no significant mediation effects were observed among men.

Association between dairy product consumption and NAFLD

The inverse association between dairy product consumption and NAFLD observed in this study is consistent with the findings of previous studies conducted in Asian populations. For example, the consumption of ≥ 50 g of full-fat dairy products was inversely associated with the prevalence of NAFLD in Thailand [18]. Similarly, among Chinese individuals, those who consumed seven servings of dairy products per week exhibited a lower incidence of NAFLD than those who did not consume dairy products [23].

Several biological mechanisms can explain this association. Dairy products with high calcium content can reduce the circulating levels of 1,25-dihydroxy vitamin D, resulting in limited uptake of calcium into cells. Lower intracellular calcium levels stimulate lipolysis and inhibit the expression of fatty acid synthase, which in turn inhibits lipid synthesis and prevents fat accumulation in the liver [13]. Additionally, dairy components such as inulin-type fructans and whey protein may modulate the gut microbiota, potentially contributing to the inhibition of NAFLD progression by regulating the gut-liver axis [24].

Association between dairy product consumption and metabolic syndrome components

The relationship between dairy product consumption and MetS components has also been examined by previous studies [14,18,25,26]. A 200 g/day increase in milk consumption was associated with a 12% lower risk of abdominal obesity, and a 100 g/day increase in yogurt consumption was associated with a 16% lower risk of hyperglycemia [26]. In addition, when dairy product intake was divided into quartiles, individuals in the highest quartile had significantly lower odds of developing abdominal obesity and MetS than those in the lowest quartile [27].

This association can be attributed to several biological pathways. Whey proteins in dairy products increase plasma amino acid levels, including isoleucine, leucine, valine, and lysine [28]. This increase stimulates the production of glucagon-like peptide-1 and gastric inhibitory polypeptide [15], which subsequently enhances insulin secretion and leads to reduced blood glucose levels [16]. Furthermore, dietary calcium intake may increase HDL cholesterol by reducing lipid absorption via binding to bile acids and fatty acids [17,29]. Moreover, lactotripeptides in dairy proteins inhibit angiotensin-converting enzymes, thereby reducing blood pressure [30].

However, this study did not find a significant association between dairy product consumption and blood pressure. Previous studies have reported inconsistent results regarding this relationship. Nevertheless, several studies have identified an inverse association between dairy intake and blood pressure across diverse age groups, including adults, older adults, children, and adolescents [31–33]. Particularly, low-fat dairy consumption has been reported to be more effective in reducing blood pressure, whereas full-fat dairy products have shown little or no effect [34]. In contrast, a Dutch study reported that overall dairy consumption was not significantly associated with changes in blood pressure [35]. Moreover, a Mendelian randomization analysis incorporating data from 32 observational and cohort studies found no evidence of a causal effect of dairy consumption on SBP or the risk of hypertension [36]. Similarly, no significant association was observed between the consumption of different dairy products and the prevalence or incidence of hypertension [37]. Overall, this existing evidence does not provide a clear consensus on the relationship between dairy consumption and blood pressure. Therefore, further longitudinal and clinical studies are warranted to clarify the potential role of dairy consumption in the regulation of BP.

Association between metabolic syndrome components and NAFLD

NAFLD and MetS are highly interrelated, with strong associations between NAFLD and individual components of MetS [38]. Insulin resistance disturbs blood glucose control, increases liver fat accumulation, and inhibits lipolysis, leading to increased very-low-density lipoprotein (VLDL) levels and decreased HDL cholesterol levels, thereby contributing to the progression of both MetS and NAFLD [39–41]. Elevated blood pressure is also associated with an increased risk of NAFLD, with individuals with prehypertension and hypertension showing approximately 1.3-fold and 1.8-fold higher odds of NAFLD, respectively, than those with normal blood pressure [42]. Moreover, abdominal obesity has been strongly associated with a higher risk of NAFLD [43], and individuals with NAFLD exhibit significantly higher triglyceride and lower HDL cholesterol levels than those without NAFLD [44].

Sex-specific associations of dairy consumption with metabolic syndrome and NAFLD

The present study demonstrated that the associations between dairy consumption, NAFLD, and MetS components varied by sex, with significant associations observed only among women. Similar sex-specific patterns have been reported previously. A cohort study found that higher dairy consumption was strongly associated with a lower prevalence of NAFLD in women [45], a finding supported by a meta-analysis of observational studies that reported consistent results [46]. Regarding MetS components, a 10-year cohort study from the Korean Genome and Epidemiology Study (KoGES) showed that dairy consumption was associated with a reduced risk of low HDL-cholesterol among women [47], and a cross-sectional study using the KNHANES observed a lower MetS prevalence among Korean women consuming ≥1 serving/day of dairy products [25]. Moreover, MetS and its individual components are more strongly associated with NAFLD in women than in men [48].

These sex differences may partly arise because the metabolic effects of diets and dietary constituents are expressed in a sex-specific manner and highly dependent on the disease context [49]. Clinical evidence supports the existence of sex-specific metabolic responses to dairy consumption. In a 6-week randomized controlled trial (RCT), low-fat dairy intake differentially improved MetS parameters according to sex, with a modest but significant reduction in fasting blood glucose in men and reductions in waist circumference, body weight, and BMI in women [50]. Another RCT demonstrated distinct postprandial lipoprotein responses to dairy between sexes, with women exhibiting smaller increases in VLDL particles and greater increases in HDL particle concentrations than men [51]. Biological sex also strongly influences adipose tissue distribution and metabolic functions. Women generally have greater amounts of subcutaneous adipose tissue than men [52]. Although abdominal subcutaneous fat is associated with adverse glucose and lipid profiles [53], subcutaneous fat may provide partial protection against metabolic dysfunction, potentially reducing the risk of MetS in women [54]. In addition, insulin sensitivity differs by sex, with women typically exhibiting greater adipose tissue insulin sensitivity than men [55]. Dairy intake has consistently been shown to inhibit intestinal fat absorption, improve lipid profiles, reduce blood pressure, and enhance glucose metabolism [56–58], thereby contributing to improvements in NAFLD. Sex hormones further modulate these processes, with estrogens promoting metabolically protective pathways and androgens supporting male-specific metabolic environments [59]. Collectively, these biological and clinical differences suggest that dairy exposure may involve distinct metabolic pathways in men and women, contributing to the observed sex-specific associations.

Sex-specific mediating effects of metabolic syndrome components in the dairy-NAFLD association

The present study found that the association between dairy consumption and NAFLD operates primarily through MetS components, rather than through a direct pathway. Additionally, in women, waist circumference, SBP, and HDL cholesterol appeared to serve as significant mediators of the dairy-NAFLD association, whereas no significant mediating effects were observed in men. This sex-specific mediation pattern supports the idea that the metabolic benefits of dairy intake are expressed differently according to sex. Women in this study consumed more dairy products than men, and consumption of more than one dairy product per day was significantly associated with lower odds of MetS in Korean women [25]. In addition, although dairy consumption did not differ between men in the NAFLD and non-NAFLD groups, women in the non-NAFLD group reported significantly higher dairy consumption than those with NAFLD. This pattern may explain why the beneficial metabolic effects of dairy intake, such as enhanced lipid excretion and lower blood glucose levels, are more pronounced in women. Importantly, the mediation analysis did not identify any significant direct effects of dairy consumption on NAFLD, indicating that dairy intake predominantly influences NAFLD through its effects on several metabolic dysfunctions, including elevated waist circumference, SBP, and low HDL cholesterol levels, rather than through a direct hepatic pathway in women.

Study strengths and limitations

This study has certain limitations. First, it was a cross-sectional study using the KNHANES dataset, which does not allow for the establishment of a clear causal relationship. Additionally, although the KNHANES dataset includes information on whether participants were taking medication for dyslipidemia, it lacks detailed information regarding medication type, dosage, therapeutic targets (e.g., LDL cholesterol, HDL cholesterol, and triglycerides), and combination therapy. Moreover, the dataset does not provide information on liver fibrosis stage or liver enzyme levels, such as gamma-glutamyl transferase (GGT) or alkaline phosphatase (ALP), which may also influence NAFLD risk. Thus, additional longitudinal studies are needed to clarify the causal relationships among dairy products, NAFLD, and MetS. Furthermore, the use of noninvasive methods for diagnosing NAFLD may result in misclassification and diagnostic errors. Despite these limitations, this study has several strengths. To the best of our knowledge, this is the first study to identify the mediating effects of MetS components on the association between dairy product consumption and NAFLD. The serving sizes used in this study (200 g for milk, 100–150 g for yogurt, and 20 g for cheese) were consistent with the standard portions of international dietary guidelines [26,60,61] and represent realistic targets for clinical dietary counseling. Although dairy product preferences differ across countries, the focus on total dairy intake in the primary mediation analysis supports the broader applicability of our findings. Sensitivity analyses examining individual dairy types allow for interpretation across diverse dietary contexts, although the generalizability of cheese-specific findings may be limited, given the relatively low cheese consumption in our Korean cohort compared to Western populations.