Sweet Snacks Are Positively and Fruits and Vegetables Are Negatively Associated with Visceral or Liver Fat Content in Middle-Aged Men and Women.

BACKGROUND: Visceral adipose tissue (VAT) and hepatic triglyceride content (HTGC) are major risk factors for cardiometabolic diseases.
OBJECTIVE: We aimed to investigate the association of dietary intake of the main food groups with VAT and HTGC in middle-aged men and women.
METHODS: We used data from the Netherlands Epidemiology of Obesity study, a population-based study including 6671 participants aged 45-65 y at baseline. In this cross-sectional analysis, VAT and HTGC were assessed by magnetic resonance imaging and spectroscopy, respectively, as the primary outcomes. Habitual intake of main food groups (dairy, meat, fish, fruits and vegetables, sweet snacks, and fats and oils) was estimated through the use of a food-frequency questionnaire. We examined associations of intake of different food groups with VAT and HTGC by linear regression analysis stratified by sex and adjusted for age, smoking, education, ethnicity, physical activity, basal metabolic rate, energy-restricted diet, menopausal state, and total energy intake.
RESULTS: In women, a 100-g/d higher intake of dairy was associated with 2.0 cm2 less VAT (95% CI: -3.4, -0.7 cm2) and a 0.95-fold lower HTGC (95% CI: 0.90-, 0.99-fold). Moreover, a 100-g/d higher intake of fruit and vegetables was associated with 1.6 cm2 less VAT (95% CI: -2.9, -0.2 cm2) in women. Fruit and vegetables were negatively associated (0.95; 95% CI: 0.91, 1.00) with HTGC, and sweet snacks were positively associated (1.29; 95% CI: 1.03, 1.63). Patterns were weaker but similar in men. Fish intake was not associated with VAT or HTGC and plant-based fat and oil intake were only associated with VAT after adjustment for total body fat.
CONCLUSIONS: Despite some variation in the strength of the associations between men and women, dietary intake of sweet snacks was positively associated with HTGC, and fruit and vegetable intake were negatively associated with visceral and liver fat content. Prospective studies are needed to confirm these results. The Netherlands Epidemiology of Obesity study is registered at clinicaltrials.gov with identifier NCT03410316.

Introduction

Obesity, in particular abdominal obesity, is increasingly prevalent worldwide and is a major risk factor for type 2 diabetes and cardiovascular diseases (1, 2). The excess cardiometabolic risk associated with abdominal obesity is suggested to result from accumulation of fat in nonadipose tissue (2). Visceral adipose tissue (VAT) and hepatic triglyceride content (HTGC) have been associated with a cluster of metabolic risk factors, insulin resistance, coronary artery disease, and cardiovascular disease in general (3–6). Furthermore, VAT is thought to contribute to the excess cardiometabolic risk caused by high rate of release of FFA and cytokine secretion (2). In addition, high concentrations of FFA and insulin resistance are related to fat deposition in the liver (7) and are strongly related to type 2 diabetes and cardiovascular disease (8, 9).

Because of the many health-related consequences, both visceral and liver fat are potential targets in prevention or treatment of cardiometabolic disease and its consequences. In addition to physical activity, diet is a key modifiable lifestyle risk factor for obesity and chronic diseases (10–12). A recent systematic review reported that dietary patterns recognized as healthy and intake of medium-chain triacylglycerols display an inverse association with visceral and subcutaneous fat. For visceral fat only, inverse associations were also shown with dietary fiber, calcium, and phytochemicals (13). Interestingly, an overfeeding study of saturated and polyunsaturated fat showed distinct effects on visceral and liver fat (11). Most previous studies have assessed the role of nutrients in fat deposition (14–18), although higher energy intake during childhood has been suggested to be associated with greater nonalcoholic fatty liver disease risk irrespective of the macronutrients consumed (19). However, it is increasingly being recognized that studying foods and food groups rather than single nutrients may be important in relation to health outcomes, as foods are not merely the sum of their nutrients (20–23). Within a food item, there may be unknown effects of other nutrients, or interactions between the separate nutrients, and the food matrix may play a role (22–24). Countries throughout Europe as well as the United States have published dietary guidelines based on whole-food products and groups rather than single nutrients (25). Although evidence on major food groups (e.g., dairy, meat, fruit and vegetables) in relation to body weight (26) and clinical cardiometabolic outcomes including coronary heart disease (27) and diabetes (28) is increasing, knowledge of the relations between food groups and ectopic fat deposition is scarce. We therefore aimed to investigate the associations between dietary intake of the main food groups and visceral and liver fat content in a population-based cohort of middle-aged men and women.

Methods

Study design and study population

The Netherlands Epidemiology of Obesity (NEO) study is a population-based prospective cohort study in 6671 individuals aged 45–65 y, with an oversampling of persons with a BMI (in kg/m2) ≥27. Detailed information about the study design and data collection has been described elsewhere (29). Men and women aged between 45 and 65 y with a self-reported BMI ≥27 living in the greater area of Leiden (in the west Netherlands) were eligible to participate in the NEO study. In addition, all inhabitants aged between 45 and 65 y from 1 municipality (Leiderdorp) were invited irrespective of their BMI, allowing a reference distribution of BMI.

Participants visited the NEO study center of the Leiden University Medical Center after an overnight fast. Before the NEO study visit, participants completed a questionnaire about demographic and clinical information, as well as an FFQ. At the study center, the participants completed a screening form asking about anything that might create a health risk or interfere with MRI (most notably metallic devices, claustrophobia, or a body circumference of >1.70 m). Of the participants who were eligible for MRI, ∼35% were randomly selected to undergo direct assessment of abdominal fat.

The present study is a cross-sectional analysis of the baseline measurements of the participants with a measurement of VAT. We excluded participants with self-reported diabetes before the study visit, participants with missing data on dietary intake, participants with implausibly high or low total energy intake (<600 or >5000 kcal/d), or participants with missing data on potential confounding factors. For the analyses on HTGC, we also excluded participants without assessment of HTGC and those who consumed >4 units of alcoholic beverages/d.

The study was approved by the medical ethics committee of the Leiden University Medical Center and all participants gave written informed consent. The Netherlands Epidemiology of Obesity study is registered at clinicaltrials.gov with identifier NCT03410316.

Data collection

On the questionnaire, participants reported ethnicity by self-identification in 8 categories, which we grouped into white (reference) and other ethnicity. Tobacco smoking was categorized as current, former, or never smoking (reference). The highest level of education was reported in 10 categories according to the Dutch education system and grouped into high (including higher vocational school, university, and postgraduate education) compared with low education (reference). Participants reported their medical history of diabetes and cardiovascular diseases. Body weight and percentage of body fat were assessed by the Tanita bio impedance balance (TBF-310, Tanita International Division) without shoes, and 1 kg was subtracted from the body weight. BMI was calculated by dividing the weight (in kg) by the height (in m2). The menopausal state was categorized as pre- or postmenopausal according to information on ovariectomy, hysterectomy, and the self-reported state of menopause in the questionnaire. The basal metabolic rate was calculated based on age, sex, height, and weight according to the Mifflin-St Jeor equation. Participants reported the frequency and duration of their physical activity during leisure time, which was expressed in h/wk of metabolic equivalents (MET h/wk) (30).

Dietary intake of food groups

Habitual dietary intake of all participants was estimated through use of a self-administered, semiquantitative 125-item FFQ (31, 32). In this FFQ, participants were asked about the frequency of intake of foods during the past month (times per day, times per week, times per month, or never). In addition, the serving size was estimated (spoons of potatoes, pieces of fruit). In a random subsample of 110 men and 119 women, the relative validity of the FFQ against two 24-h dietary recalls was assessed regarding total fatty acids, SFAs, MUFAs, and PUFAs. The correlation coefficients corrected for within-person variation for total fatty acids, SFAs, MUFAs, and PUFAs were ∼0.5 (33). Intake of nutrients and total energy was calculated through use of the Dutch Food Composition Table (NEVO-2011).

Based on the FFQ, products were categorized into food groups on the basis of similar source, nutrient characteristics, or hypothesized biological effects (28). Hereby, we followed the categorization of food groups of the Netherlands Nutrition Centre (34) as much as possible based on the distinctive capabilities of the FFQ used. Food groups were categorized into dairy (including milk, cheese, yogurt, and butter), meat, fruits and vegetables, sweet snacks (chocolate, cake, pie, candy bars, and candy), fish, and plant-based fats and oils (margarine, cooking oils). In addition, subdivisions were made: dairy was also subdivided into cheese, milk, butter, and yogurt; fruit and vegetables were studied separately; fats and oils were divided into margarine and oils; and sweet snacks were divided into cake and candy. The caloric intake of products within these food groups was summed and converted into percentage of total energy intake by dividing the caloric intake of the food groups by the total caloric intake per day.

VAT and HTGC assessment by imaging techniques

Imaging was performed on a 1.5 Tesla MR system (Philips Medical Systems). VAT was quantified by a turbo spin echo imaging protocol with MRI. At the level of the fifth lumbar vertebra, 3 transverse images each with a slice thickness of 10 mm were obtained during a breath-hold. Visceral fat area was converted from the number of pixels to centimeters squared with the use of in-house developed software (MASS, Medis) and the average of the 3 slices was used in the analyses (29).

HTGC was quantified by proton magnetic resonance spectroscopy (1H-MRS) of the liver (35). An 8-mL voxel was positioned in the right lobe of the liver, avoiding gross vascular structures and adipose tissue depots. Sixty-four averages were collected with water suppression. Spectra were obtained with an echo time of 26 ms and a repetition time of 3000 ms. Data points (1024) were collected with use of a 1000-Hz spectral line. Without changing any parameters, spectra without water suppression, with a repetition time of 10 s and 4 averages were obtained as an internal reference. 1H-MRS data were fitted with use of Java-based magnetic resonance user interface software (jMRUI version 2.2), as described previously (36). HTGC relative to water was calculated as the sum of the signal amplitudes of methyl and methylene divided by the signal amplitude of water ×100.

Statistical analysis

In the NEO study, there is an oversampling of persons with BMI ≥27. To correctly represent associations in the general population (37), adjustments for the oversampling of individuals with a BMI ≥27 were made by weighting individuals toward the BMI distribution of participants from the Leiderdorp municipality (38), which was similar to that of the general Dutch population (39). All results were based on the weighted analyses, and consequently, the results apply to a population-based study without oversampling of individuals with BMI ≥27. As a result of the weighted analyses, percentages and proportions are given instead of numbers of participants. Other characteristics are expressed as percentages or means ± SDs. We tested for interactions with sex and performed all analyses for the total population and for men and women separately because of major differences in body fat distribution between men and women and previously observed gender differences in the relation between food group scores and abdominal obesity (40). Linearity of the main food groups was checked by adding a quadratic term to the main multivariable adjusted model and visual inspection of scatter plots.

We performed linear regression analyses with multiple models to examine the associations between dietary intake of the food groups with visceral and liver fat content. First, we examined the crude associations of dietary intake of 100 g/d of each food group with visceral and liver fat content. Second, we adjusted the models for age, smoking, education, ethnicity, physical activity, basal metabolic rate, menopausal state, total energy intake, and adherence to an energy-restricted diet, and liver fat models were also adjusted for alcohol intake. Third, we adjusted the models for total body fat, to examine whether the associations were specific for visceral and liver fat instead of merely reflecting associations with overall adiposity. Fourth, to examine whether associations were specific for the food groups and not merely reflecting a healthy diet, we adjusted all models for the food group fruit and vegetables, and the food group fruit and vegetables model for the food group sweet snacks. As secondary analyses, we subdivided several food groups into finer categorizations: dairy into milk, cheese, yogurt, and butter; sweet snacks into cake and candy; plant-based fat and oils into margarine and oils; and fruit and vegetables into fruit and vegetables separately. We performed subgroup analyses and stratified the multivariable model (not including total body fat and markers of a healthy diet) by menopausal state (pre- and postmenopausal). This stratification was done because, for example, visceral fat may differ greatly between pre- and postmenopausal women (41). We also stratified the same multivariable models of HTGC by the rs738409 single nucleotide polymorphism in the PNPLA3 (patatin-like phospholipase domain-containing protein 3) gene to examine whether associations would be different in carriers of the risk allele that is associated with high liver fat content (42). Because of a skewed distribution of HTGC, we used the natural logarithm of this variable in the analyses. For interpretation of the results, we back-transformed the regression coefficients of HTGC toward a ratio with 95% CI, which can be interpreted as a ratio in HTGC associated with dietary intake of 100 g/d of the food groups (e.g., 1.2 can be interpreted as a 1.2-fold higher HTGC, which in a person with a HTGC of 5% would reflect an increase to 6%). The regression coefficients of VAT represent an absolute difference in VAT expressed as cm2 per 100 g/d of the food groups.

As a means of sensitivity analysis, we also performed 2 types of isocaloric substitution analyses in which dairy was specifically replaced by the other food groups: one analysis substituted 100 g dairy/d with 100 g from other food groups, and the other substituted 10% of the energy from dairy with energy derived from other food groups. In these substitution models, we included all the food groups under study (meat, fruits and vegetables, sweet snacks, fish, and plant-based fats and oils) except dairy, the food group to be substituted, in addition to all other foods consumed that were not categorized into 1 of the food groups, and all confounding factors. Finally, we performed a sensitivity analysis including a variable in our fully adjusted model that divides energy intake by basal metabolic rate, to adjust for potential under- and overreporting.

We performed all analyses with STATA statistical software (version 14, Statacorp).

Results

In total, 6671 participants were included in the NEO study between September 2008 and October 2012, of whom 2580 underwent an MRI of the abdomen. For 11 of those participants, the quality of the MRI images was insufficient for quantification of abdominal VAT. MRI was performed in a random subsample of those without contraindications. As a result, those who underwent the MRI have a slightly lower BMI (26.0 compared with 26.6) and slightly less often a medical history of cardiovascular disease (4.1% compared with 6.8%).

After exclusion of participants with a medical history of diabetes (n = 161); extreme energy intake [<600 or >5000 kcal/d (n = 19)]; an incomplete FFQ (n = 16); or missing data on smoking (n = 2), education (n = 22), ethnicity (n = 3), energy-restricted diet (n = 4), or physical activity (n = 38), 2304 participants were included in the analyses on VAT. Participants included in the analyses did not substantially differ from those excluded because of missing data regarding total body fat (30.7% for those without missing data compared with 30.6% for those with missing data), visceral fat (87.6 compared with 88.2 cm2), nor regarding dietary intake of the food groups. Liver fat was slightly lower in the participants with missing data (4.3% compared with 5.6%).

For the analyses with HTGC as an outcome, we excluded 424 participants without HTGC measurement. Because of the limited time available per subject it was not possible to check the spectra during the measurement and repeat the measurement when technical failures were present. As a consequence, 1H-MRS of the liver could not be completed in 424 participants, mainly because of technical failures. However, the failure rate of the MR spectroscopy was not related to age (56 y for participants with HTGC measurement compared with 55 y for participants without HTGC measurement), sex (47% male compared with 46%), BMI (26.0 compared with 25.8), VAT (90.1 compared with 87 cm2), total body fat (30.7% compared with 30.9%), or any of the food groups. Lastly, participants who drank >4 units of alcohol/d (n = 165) were excluded for the analyses of HTGC.

The baseline characteristics of the study population participants are shown in . Whereas women had more total body fat, men had more VAT and HTGC.

TABLE 1

Baseline characteristics of participants in the Netherlands Epidemiology of Obesity study[1]

	Total population	Men (46.4%)	Women (53.6%)
Demographic variables
Age, y	55 ± 6	56 ± 6	55 ± 6
Ethnicity, % white	96	96	95
Education level,[2] % high	47	51	43
Tobacco smoking, % current	15	16	13
Menopausal state, % post	—	—	58
Physical activity in leisure time, MET h/wk	38.3 ± 33.1	40.0 ± 38.8	36.9 ± 28.1
Dietary variables
Dairy intake, g/d	322 ± 196	341 ± 223	306 ± 172
Meat intake, g/d	83 ± 46	96 ± 50	72 ± 38
Fish intake, g/d	18 ± 17	19 ± 19	16 ± 15
Fruit and vegetable intake, g/d	326 ± 163	304 ± 169	345 ± 154
Sweet snack intake, g/d	82 ± 57	89 ± 59	75 ± 54
Fat and oil intake, g/d	35 ± 22	41 ± 26	29 ± 17
Energy restricted diet, %	10	6	14
Basal metabolic rate, MJ/d	6.3 ± 1.1	7.3 ± 0.7	5.5 ± 0.6
Body fat measures
BMI, kg/m2	25.8 ± 3.9	26.5 ± 3.5	25.2 ± 4.0
Total body fat, %	30.7 ± 8.3	24.5 ± 5.7	36.1 ± 6.1
VAT, cm2	87.6 ± 54.2	113.0 ± 58.7	65.5 ± 39.8
HTGC,[3] %	5.6 ± 7.7	6.8 ± 8.2	4.5 ± 7.2
Fatty liver (HTGC >5.56%), %	28.2	37.5	20.2
Waist circumference, cm	90.9 ± 12.6	97.3 ± 10.4	85.3 ± 11.3
CVD risk factors
CVD, %	3.8	3.9	3.7
Lipid lowering medication, %	7	11	4
Serum total cholesterol, mmol/L	5.75 ± 1.04	5.63 ± 1.04	5.86 ± 1.03
Fasting serum triglycerides, mmol/L	1.23 ± 0.82	1.43 ± 0.97	1.06 ± 0.64
Serum HDL cholesterol, mmol/L	1.58 ± 0.46	1.35 ± 0.36	1.79 ± 0.43

Participants were men and women aged between 45 and 65 y who were not using glucose-lowering therapy. Values are means ± SDs unless otherwise indicated. Results are based on analyses weighted toward the BMI distribution of the general population (n = 2304, 1191 men and 1113 women). CVD, cardiovascular disease; HTGC, hepatic triglyceride content; MET, metabolic equivalent of task; VAT, visceral adipose tissue.

Low education: none, primary school, or lower vocational education as highest level of education.

Mean HTGC only calculated in persons with HTGC measurement (n = 1880).

Dietary intake of the main food groups in relation to VAT

We assessed the reproducibility of the dietary intake of the food groups in 100 participants who completed the FFQ twice with ∼3 mo in between. The individual measurement intraclass correlation coefficients of the food group dairy were 0.80, of fruit and vegetables 0.56, of meat 0.83, of sweet snacks 0.59, of fish 0.64, and of fats and oils 0.65. The individual intraclass correlation coefficient for total energy intake was 0.68.

In the total population, after adjustment for confounding factors, total body fat, and a marker for an (un)healthy lifestyle, dietary intake of fruit and vegetables was associated with 1.2 cm2 (95% CI: −2.4, 0.0 cm2) less VAT (). Intake of plant-based fats and oils was also associated with 13.9 cm2 less VAT (95% CI: −23.7, −4.1 cm2). Dietary intake of dairy, fish, meat, and sweet snacks was not associated with VAT (Table 2).

TABLE 2

Difference in VAT per 100 g/d consumption of the food groups in participants in the Netherlands Epidemiology of Obesity study[1]

	Crude	Multivariable	Multivariable + TBF	Multivariable + TBF + healthy diet
Food groups	Difference in VAT, cm2 (95% CI)	Difference in VAT, cm2 (95% CI)	Difference in VAT, cm2 (95% CI)	Difference in VAT, cm2 (95% CI)
Dairy
Total	0.1 (−1.3, 1.6)	−1.4 (−2.5, −0.3)*	−0.6 (−1.6, 0.4)	−0.6 (−1.6, 0.4)
Men	−0.6 (−2.6, 1.3)	−0.8 (−2.5, 0.9)	0.0 (−1.4, 1.5)	0.0 (−1.4, 1.5)
Women	−1.3 (−2.7, 0.2)	−2.0 (−3.4, −0.7)*	−1.3 (−2.5, 0.0)	−1.2 (−2.5, 0.0)
Meat
Total	26.1 (20.2, 31.9)*	5.4 (0.1, 10.6)*	1.5 (−3.1, 6.0)	1.0 (−3.5, 5.6)
Men	10.9 (2.3, 19.4)*	3.9 (−4.3, 12.1)	−1.1 (−7.8, 5.6)	−1.4 (−8.0, 5.3)
Women	15.7 (8.7, 22.7)*	6.6 (0.7, 12.6)*	3.4 (−2.0, 8.7)	2.8 (−2.6, 8.2)
Fish
Total	30.5 (13.4, 47.5)*	6.1 (−5.8, 18.1)	3.7 (−6.8, 14.2)	6.2 (−4.7, 17.1)
Men	14.6 (−8.3, 37.5)	5.2 (−14.3, 24.8)	1.9 (−15.2, 18.9)	3.0 (−14.6, 20.6)
Women	25.4 (6.8, 44.0)*	6.7 (−6.1, 19.5)	5.4 (−5.7, 16.6)	9.1 (−2.6, 20.9)
Fruit and vegetables
Total	−3.1 (−4.8, −1.4)	−1.7 (−3.0, −0.4)*	−1.2 (−2.4, −0.0)*	−1.2 (−2.4, 0.0)
Men	−2.3 (−4.8, 0.2)	−1.8 (−4.2, 0.5)	−1.0 (−3.2, 1.2)	−0.8 (−3.1, 1.4)
Women	−0.4 (−2.1, 1.4)	−1.6 (−2.9, −0.2)*	−1.1 (−2.3, 0.1)	−1.2 (−2.4, 0.1)
Sweet snacks
Total	4.3 (−0.7, 9.2)	−0.3 (−4.7, 4.2)	0.9 (−3.1, 5.0)	0.2 (−3.9, 4.4)
Men	−1.6 (−9.1, 5.8)	−0.6 (−8.1, 7.0)	4.0 (−2.3, 10.3)	3.7 (−2.6, 10.0)
Women	−0.0 (−5.0, 5.0)	0.4 (−4.8, 5.6)	0.6 (−4.5, 5.6)	−0.4 (−5.7, 5.0)
Fat and oils
Total	33.5 (21.2, 45.9)	−8.2 (−20.2, 3.8)	−12.5 (−22.2, −2.7)*	−13.9 (−23.7, −4.1)*
Men	3.7 (−12.5, 19.8)	−13.1 (−30.9, 4.6)	−18.3 (−31.3, −5.3)*	−19.8 (−33.1, −6.4)*
Women	6.8 (−9.0, 22.6)	−1.2 (−14.4, 11.9)	−7.3 (−19.9, 5.3)	−8.3 (−20.9, 4.3)

Participants were men and women aged between 45 and 65 y who were not using glucose-lowering therapy. Multivariable: adjusted for age, total energy intake, smoking, education, ethnicity, physical activity in leisure time, basal metabolic rate, menopause, and energy-restricted diet. Results are based on analysis weighted toward the BMI distribution of the general population (n = 2304; 1191 men and 1113 women). *P < 0.05. TBF, total body fat; VAT, visceral adipose tissue.

Tests for an interaction between the food groups and sex were nonsignificant, but we decided a priori to perform the analyses separately for men and women because of the large differences in body fat distribution. All associations were attenuated in the stratified analyses, although in women intake of dairy remained associated with VAT (−1.2 cm2, 95% CI: −2.5, 0.0 cm2) (Table 2).

After a finer categorization of the food groups, yogurt seemed to drive the negative association between dairy and VAT in women () .

Dietary intake of dairy, meat, and fruit and vegetables was more strongly associated with VAT in postmenopausal women than in premenopausal women (P values for interactions: 0.56, 0.09, and 0.21) ().

The results remained similar after substituting dairy with other food groups () and when including participants with diabetes (). The results did not differ when adjusting for potential under- or overreporting (data not shown).

Dietary intake of main food groups in relation to HTGC

In the total population of 1715 participants with HTGC measurements, after adjustment for confounding factors, total body fat and a marker for an (un)healthy lifestyle, dietary intake of sweet snacks were associated with a 1.19-fold (95% CI: 1.04-, 1.37-fold) higher HTGC (). The intake of dairy, fruit and vegetables, fish, meat, and fats and oils was not associated with HTGC (Table 4). In men and women separately, the associations were attenuated (Table 4).

TABLE 4

Relative difference in HTGC per 100 g/d consumption of the food groups in participants in the Netherlands Epidemiology of Obesity study[1]

	Crude	Multivariable	Multivariable + TBF	Multivariable + TBF + healthy diet
Food groups	Relative difference in HTGC (95% CI)	Relative difference in HTGC (95% CI)	Relative difference in HTGC (95% CI)	Relative difference in HTGC (95% CI)
Dairy
Total	0.98 (0.95, 1.02)	0.97 (0.93, 1.00)*	0.97 (0.94, 1.01)	0.97 (0.94, 1.01)
Men	0.98 (0.94, 1.03)	0.98 (0.94, 1.03)	0.99 (0.94, 1.03)	0.99 (0.94, 1.03)
Women	0.96 (0.91, 1.02)	0.95 (0.90, 0.99)*	0.96 (0.92, 1.01)	0.96 (0.92, 1.01)
Meat
Total	1.48 (1.28, 1.71)*	1.14 (0.98, 1.33)	1.06 (0.92, 1.22)	1.05 (0.91, 1.20)
Men	1.16 (0.95, 1.41)	1.06 (0.86, 1.29)	0.97 (0.83, 1.15)	0.97 (0.82, 1.14)
Women	1.45 (1.16, 1.80)*	1.21 (0.97, 1.52)	1.10 (0.89, 1.36)	1.07 (0.87, 1.32)
Fish
Total	1.18 (0.78, 1.80)	0.82 (0.57, 1.19)	0.79 (0.58, 1.08)	0.85 (0.61, 1.16)
Men	1.03 (0.63, 1.70)	0.96 (0.57, 1.63)	0.93 (0.59, 1.47)	0.96 (0.61, 1.50)
Women	1.08 (0.61, 1.90)	0.71 (0.45, 1.14)	0.69 (0.46, 1.04)	0.76 (0.48, 1.19)
Fruit and vegetables
Total	0.95 (0.91, 0.99)	0.96 (0.92, 0.99)*	0.97 (0.93, 1.00)*	0.97 (0.94, 1.01)
Men	0.95 (0.90, 1.01)	0.97 (0.92, 1.02)	0.98 (0.94, 1.02)	0.99 (0.94, 1.03)
Women	0.98 (0.93, 1.03)	0.95 (0.91, 1.00)	0.96 (0.92, 1.00)	0.97 (0.93, 1.02)
Sweet snacks
Total	1.17 (1.03, 1.33)*	1.22 (1.05, 1.42)*	1.21 (1.06, 1.39)*	1.19 (1.04, 1.37)*
Men	1.06 (0.91, 1.22)	1.13 (0.94, 1.35)	1.17 (1.01, 1.35)*	1.16 (0.99, 1.35)
Women	1.17 (0.95, 1.44)	1.29 (1.03, 1.63)*	1.26 (1.01, 1.57)*	1.23 (0.97, 1.54)
Fats and oils
Total	1.84 (0.39, 2.44)	1.20 (0.88, 1.64)	1.16 (0.88, 1.53)	1.12 (0.85, 1.48)
Men	1.30 (0.91, 1.86)	1.26 (0.84, 1.88)	1.22 (0.86, 1.74)	1.20 (0.84, 1.71)
Women	1.42 (0.87, 2.30)	1.21 (0.75, 1.94)	1.08 (0.68, 1.70)	1.04 (0.66, 1.65)

Participants were men and women aged between 45 and 65 y who were not using glucose-lowering therapy. Multivariable: adjusted for age, total energy intake, smoking, education, ethnicity, physical activity in leisure time, basal metabolic rate, menopause, alcohol consumption, and energy-restricted diet. Results are based on analysis weighted toward the BMI distribution of the general population (n = 1715; 831 men and 884 women). *P < 0.05. HTGC, hepatic triglyceride content; TBF, total body fat.

After the finer categorization, vegetables were more strongly associated with HTGC than fruit, and yogurt was most strongly associated with liver fat of all the dairy components (). The association between sweet snacks and HTGC was stronger in premenopausal women than in postmenopausal women (P value for interaction 0.59) ().

TABLE 5

Relative difference in HTGC per 100 g/d consumption of the food groups in participants in the Netherlands Epidemiology of Obesity study[1]

	Crude	Multivariable	Multivariable + TBF	Multivariable + TBF + healthy diet
Food groups	Relative difference in HTGC (95% CI)	Relative difference in HTGC (95% CI)	Relative difference in HTGC (95% CI)	Relative difference in HTGC (95% CI)
Cheese
Total	1.11 (0.85, 1.46)	1.01 (0.78, 1.32)	1.01 (0.80, 1.29)	1.00 (0.79, 1.28)
Men	1.34 (0.94, 1.89)	1.34 (0.96, 1.86)	1.10 (0.82, 1.48)	1.09 (0.81, 1.47)
Women	0.94 (0.64, 1.38)	0.82 (0.56, 1.20)	0.98 (0.67, 1.44)	0.97 (0.67, 1.43)
Milk
Total	1.00 (0.95, 1.05)	0.97 (0.94, 1.01)	0.98 (0.95, 1.01)	0.98 (0.95, 1.01)
Men	0.99 (0.93, 1.05)	0.98 (0.93, 1.04)	0.99 (0.94, 1.04)	0.99 (0.94, 1.04)
Women	0.98 (0.92, 1.04)	0.97 (0.92, 1.02)	0.98 (0.93, 1.02)	0.97 (0.93, 1.02)
Yogurt
Total	0.89 (0.81, 0.97)	0.89 (0.82, 0.98)*	0.91 (0.85, 0.99)	0.92 (0.85, 1.00)*
Men	0.91 (0.84, 0.99)*	0.94 (0.85, 1.04)	0.95 (0.87, 1.03)	0.95 (0.88, 1.03)
Women	0.88 (0.77, 1.02)	0.82 (0.72, 0.94)*	0.87 (0.76, 0.99)*	0.88 (0.78, 1.01)
Cream butter
Total	0.58 (0.25, 1.34)	0.75 (0.38, 1.46)	0.79 (0.43, 1.47)	0.79 (0.43, 1.46)
Men	0.57 (0.13, 2.41)	0.90 (0.20, 4.11)	1.07 (0.26, 4.41)	1.09 (0.23, 5.08)
Women	0.50 (0.20, 1.22)	0.74 (0.34, 1.59)	0.72 (0.37, 1.39)	0.69 (0.35, 1.35)
Fruit
Total	0.94 (0.89, 0.99)*	0.96 (0.92, 1.01)	0.98 (0.94, 1.03)	0.98 (0.94, 1.03)
Men	0.94 (0.89, 1.00)	0.96 (0.90, 1.02)	0.98 (0.93, 1.04)	0.99 (0.93, 1.04)
Women	0.98 (0.91, 1.06)	0.96 (0.89, 1.03)	0.98 (0.90, 1.05)	0.98 (0.91, 1.06)
Vegetables
Total	0.94 (0.88, 1.01)	0.94 (0.88, 1.00)*	0.93 (0.88, 0.98)	0.94 (0.89, 1.00)*
Men	0.96 (0.85, 1.07)	0.98 (0.89, 1.07)	0.97 (0.89, 1.05)	0.98 (0.90, 1.07)
Women	0.96 (0.89, 1.05)	0.93 (0.85, 1.00)	0.92 (0.86, 0.99)*	0.93 (0.86, 1.00)
Cake
Total	1.28 (1.00, 1.62)*	1.21 (0.96, 1.51)	1.11 (0.91, 1.37)	1.10 (0.89, 1.35)
Men	1.05 (0.78, 1.42)	1.09 (0.78, 1.53)	1.00 (0.76, 1.32)	1.00 (0.76, 1.33)
Women	1.38 (0.97, 1.96)	1.24 (0.91, 1.69)	1.19 (0.88, 1.60)	1.15 (0.85, 1.55)
Candy
Total	1.18 (0.97, 1.43)	1.20 (0.98, 1.47)	1.23 (1.03, 1.46)*	1.20 (1.00, 1.44)
Men	1.12 (0.92, 1.37)	1.11 (0.88, 1.39)	1.20 (1.00, 1.45)*	1.19 (0.99, 1.44)
Women	1.21 (0.88, 1.67)	1.26 (0.93, 1.71)	1.21 (0.91, 1.61)	1.18 (0.87, 1.58)
Margarine
Total	2.02 (1.47, 2.79)*	1.19 (0.85, 1.68)	1.15 (0.86, 1.53)	1.09 (0.81, 1.47)
Men	1.28 (0.84, 1.96)	1.15 (0.72, 1.82)	1.18 (0.80, 1.72)	1.14 (0.77, 1.69)
Women	1.64 (0.99, 2.71)	1.22 (0.75, 1.98)	1.03 (0.66, 1.61)	0.97 (0.61, 1.52)
Oils
Total	1.18 (0.54, 2.53)	1.03 (0.49, 2.17)	0.88 (0.47, 1.65)	0.96 (0.51, 1.82)
Men	1.20 (0.46, 3.08)	1.26 (0.47, 3.41)	0.88 (0.39, 2.02)	0.92 (0.40, 2.11)
Women	0.98 (0.32, 3.01)	1.06 (0.37, 3.07)	1.10 (0.40, 2.99)	1.29 (0.47, 3.56)

Participants were men and women aged between 45 and 65 y who were not using glucose-lowering therapy. Multivariable: adjusted for age, total energy intake, smoking, education, ethnicity, physical activity in leisure time, basal metabolic rate, menopause, alcohol consumption, and energy-restricted diet. Results are based on analysis weighted toward the BMI distribution of the general population (n = 1715; 831 men and 884 women). *P < 0.05. HTGC, hepatic triglyceride content; TBF, total body fat.

Substituting dairy with other food groups showed similar results to the multivariable analyses (), as did the analyses including participants with diabetes (). The results did not differ when adjusting for potential under- or overreporting (data not shown).

Correlation coefficients between the food groups are shown in .

Discussion

In this population-based study of participants aged 40–65 y without contraindications for an MRI, we examined for the first time to what extent dietary intake of the main food groups was specifically associated with visceral and liver fat content, assessed with MRI and 1H-MRS. In the total population, dietary intake of fruit and vegetables and plant-based fats and oils was associated with less visceral fat, and intake of sweet snacks was associated with more liver fat. Although CIs were wide, both in the total population and in men and women separately, similar patterns of positive associations could be observed with intake of sweet snacks, and patterns of inverse associations with intake of fats and oils, dairy, and fruit and vegetables.

The observed associations were largely explained by total body fat. On the one hand, the remaining observed associations may suffer from residual confounding because of imperfectly measured total body fat. On the other hand, the results of the associations of fruit and vegetables and plant-based fats and oils with visceral fat and that of sweet snacks with liver fat that remained in the total population after multivariate adjustment including total body fat and a marker for an (un)healthy diet, support the presence of specific associations of certain food groups with visceral and liver fat, but need to be confirmed in larger studies.

Although few studies have investigated the association between food groups and VAT and HTGC, our findings are in accordance with the current literature on food groups in relation to cardiometabolic diseases and the current food group-based dietary guidelines in the European region (25), and they support the dietary patterns of the Dietary Approaches to Stop Hypertension (DASH) diet and Alternative Healthy Eating Index (43). Dietary intake of meat and sugar-sweetened beverages has been associated with an increased risk of type 2 diabetes (28, 44) and intake of dairy and fruits with a lower risk of type 2 diabetes (28, 45). In a recent meta-analysis, dietary intake of fish and fruit and vegetables has also been associated with a decreased risk of all-cause mortality, whereas red meat and processed meat were associated with an increased risk (46).

Dairy was negatively associated with visceral fat in women, and this association was mostly driven by yogurt, which supports previous results from the Women's Health Initiative Observational Study showing that high yogurt consumption significantly decreased diabetes risk (47). When butter was excluded from the dairy food group, the associations remained similar, indicating that butter intake did not contribute to the inverse association.

In our study, we did not observe an association between fish intake and liver or visceral fat. Although the point estimate for fish intake and VAT was positive, CIs were very wide. It must be noted that we could not distinguish between fresh fish and fried fish on the basis of our FFQ, and thus, this food group was relatively heterogeneous. However, a recent meta-analysis showed that fish intake was negatively associated with diabetes in Asian populations but positively associated with diabetes in Western populations, in which no distinction was made between fresh and fried fish (48).

In this study, meat was not associated with visceral or liver fat. However, our FFQ did not make a distinction between poultry, red meat, and processed meat. Associations with red meat and processed meat might be stronger than those observed with total meat. Even though the exact mechanism remains unidentified, the dietary cholesterol, protein, heme-iron, advanced glycation products, or preservatives such as sodium and nitrites/nitrates in meat have been hypothesized to be responsible for the positive association with VAT and diabetes. Regarding dairy, calcium, vitamin D, magnesium, fatty acids, protein, and the effect on satiety are hypothesized to underlie the beneficial effect (45). However, dairy products are often differentially categorized across different studies (23), making comparisons difficult. Different dairy products, such as fermented dairy or low- and high-fat dairy, might be associated differently with cardiometabolic outcomes, but all are categorized as dairy. Other nutrients and dietary aspects have already been shown to be associated with measures of adiposity, such as dietary fiber with less VAT (49), high-glycemic index diets with higher waist circumference (50) and high protein (either animal or plant) and n–6 PUFAs with less HTGC (51, 52). However, food groups instead of single nutrients in relation to VAT and HTGC have not yet been studied.

Strengths of this study include the direct assessment of visceral fat and HTGC by MRI and 1H-MRS, respectively, in a relatively large sample size. In addition, the extensive phenotypic measurements allowed for adjustment for many potential confounding factors, and the large study population enabled investigation of possible sex differences. A limitation of this study is that the FFQ was self-administered and therefore prone to measurement error. When assessing reproducibility in a random subsample, the ICCs of fruit and vegetables and sweet snacks were moderate to low, which could be a result of seasonal variation, but might also indicate potential over- or underreporting. Furthermore, a limitation of studying food groups may be that they cover a broad range of food products and might comprise both relatively healthy and unhealthy products. As we could not distinguish between white meat, red meat, and processed meat, this might have attenuated our associations because of regression dilution. Moreover, the observational cross-sectional design of this study precludes any causal inference, and residual confounding, for example because of unmeasured or insufficiently measured lifestyle factors, may still be present despite our efforts to minimize confounding as much as possible. In addition, potential selection bias might have occurred because of missing data. However, the number of participants excluded because of missing data was limited (n = 96), and the failure rate of liver fat measurement was not dependent on sex, age, or body fat measurements, so we do not expect this factor to substantially alter our results. Lastly, our study population consisted primarily of white, middle-aged participants, and there might be differences in dietary habits (53) and VAT (54) and HTGC accumulation (55) between different ethnic populations. Therefore, our findings need to be confirmed in prospective studies and in other ethnic groups.

In conclusion, in this population-based study in middle-aged men and women without contraindications for an MRI, dietary intake of plant-based fats and oils and of fruits and vegetables was associated with less VAT. Intake of sweet snacks was associated with more liver fat. Larger prospective studies on the relation between a food group and ectopic fat accumulation are needed to confirm whether associations between dietary intake of certain food groups are specifically associated with visceral or liver fat. In addition, intervention studies are needed to establish to what extent dietary changes can specifically reduce ectopic fat accumulation and the risk of cardiometabolic disease.

Click here for additional data file.

TABLE 3

Difference in VAT per 100 g/d consumption of the food groups in participants in the Netherlands Epidemiology of Obesity study[1]

	Crude	Multivariable	Multivariable + TBF	Multivariable + TBF + healthy diet
Food groups	Difference in VAT, cm2 (95% CI)	Difference in VAT, cm2 (95% CI)	Difference in VAT, cm2 (95% CI)	Difference in VAT, cm2 (95% CI)
Cheese
Total	6.9 (−4.2, 17.9)	−0.2 (−8.7, 8.3)	−0.1 (−7.7, 7.6)	−0.2 (−7.9, 7.5)
Men	16.8 (2.2, 31.4)*	7.8 (−5.8, 21.4)	0.4 (−10.7,11.6)	0.2 (−10.9, 11.4)
Women	−6.6 (−19.5, 6.3)	−8.6 (−19.0, 1.9)	−4.2 (−13.8, 5.4)	−4.3 (−13.9, 5.3)
Milk
Total	1.5 (−0.2, 3.2)	−0.3 (−1.5, 0.8)	0.2 (−0.9, 1.2)	0.1 (−0.9, 1.2)
Men	0.19 (−2.1, 2.5)	−0.1 (−2.0, 1.8)	0.7 (−0.9, 2.4)	0.7 (−1.0, 2.3)
Women	−0.4 (−1.9, 1.1)	−0.6 (−1.7, 0.5)	−0.2 (−1.4,0.9)	−0.3 (−1.4, 0.8)
Yogurt
Total	−8.8 (−13.1, −4.5)*	−7.4 (−11.0, −3.9)*	−5.2 (−7.8, −2.6)*	−5.0 (−7.6, −2.4)*
Men	−7.4 (−12.6, −2.2)*	−6.0 (−11.3, −0.6)*	−3.9 (−7.8, 0.0)	−3.7 (−7.6, 0.2)
Women	−5.9 (−10.0, −1.7)*	−9.2 (−12.3, −6.1)*	−6.6 (−9.4, −3.7)*	−6.3 (−9.1, −3.4)*
Cream butter
Total	−3.5 (−48.0, 40.9)	−9.6 (−37.8, 18.5)	−6.8 (−33.8, 20.2)	−7.3 (−34.6, 20.0)
Men	−29.8 (−85.7, 26.1)	−43.8 (−97.2, 9.6)	−30.3 (−69.7, 9.0)	−27.8 (−67.0, 11.4)
Women	−7.4 (−52.4, 37.5)	13.0 (−23.5, 49.5)	10.4 (−27.3, 48.1)	10.8 (−27.0, 48.7)
Fruit
Total	−5.2 (−7.7, −2.7)	−2.3 (−4.2, −0.4)	−1.2 (−2.9, 0.5)	−1.2 (−2.9, 0.5)
Men	−3.9 (−7.3, −0.4)*	−2.6 (−5.9, 0.8)	−0.6 (−3.5, 2.3)	−0.5 (−3.4, 2.3)
Women	−1.7 (−4.1, 0.8)	−1.9 (−3.9, 0.0)	−1.3 (−3.1, 0.5)	−1.3 (−3.1, 0.5)
Vegetables
Total	−0.8 (−3.6, 2.1)	−1.2 (−3.3, 1.0)	−1.5 (−3.4, 0.4)	−1.5 (−3.4, 0.4)
Men	−0.4 (−5.0, 4.2)	−0.8 (−5.0, 3.5)	−2.1 (−5.9, 1.6)	−1.8 (−5.6, 1.9)
Women	1.5 (−1.3, 4.3)	−1.5 (−3.7, 0.7)	−1.3 (−3.2, 0.7)	−1.3 (−3.3, 0.8)
Cake
Total	4.7 (−4.9, 14.3)	−1.8 (−9.9, 6.3)	−3.3 (−10.1, 3.5)	−3.8 (−10.6, 3.1)
Men	−2.1 (−16.5, 12.2)	0.1 (−13.7, 13.9)	−0.1 (−10.6, 10.5)	0.1 (−10.5, 10.7)
Women	3.6 (−6.8, 14.1)	−1.6 (−10.4, 7.2)	−2.2 (−10.3, 5.9)	−3.3 (−11.5, 4.8)
Candy
Total	1.5 (−6.2, 9.2)	0.1 (−6.2, 6.5)	1.7 (−4.0, 7.4)	1.0 (−4.9, 6.9)
Men	−1.4 (−12.6, 9.9)	−2.4 (−14.1, 9.2)	1.9 (−7.6, 11.3)	1.6 (−7.9, 11.1)
Women	−0.4 (−7.2, 6.3)	2.4 (−4.2, 9.0)	2.7 (−3.8, 9.3)	1.8 (−4.9, 8.6)
Margarine
Total	34.1 (20.0, 48.2)	−10.5 (−23.0, 2.0)	−14.2 (−24.3, −4.1)*	−16.6 (−26.8, −6.4)*
Men	−2.7 (−21.0, 15.6)	−18.3 (−36.3,−0.4)*	−19.2 (−32.6, −5.8)*	−21.1 (−34.9, −7.3)*
Women	11.1 (−7.1, 29.3)	2.1 (−11.7, 15.8)	−7.0 (−20.1, 6.0)	−9.4 (−22.5, 3.8)
Oils
Total	27.1 (−4.3, 58.5)	10.0 (−15.4, 35.4)	2.1 (−20.1, 24.3)	5.6 (−17.2, 28.3)
Men	24.0 (−14.9, 63.0)	13.2 (−23.9,50.4)	−10.4 (−41.6, 20.9)	−8.2 (−40.0, 23.7)
Women	4.1 (−35.1, 43.4)	5.4 (−27.6, 38.4)	9.8 (−20.3, 39.9)	13.8 (−16.5, 44.2)

Participants were men and women aged between 45 and 65 y who were not using glucose-lowering therapy. Multivariable: adjusted for age, total energy intake, smoking, education, ethnicity, physical activity in leisure time, basal metabolic rate, menopause, and energy-restricted diet. Results are based on analysis weighted toward the BMI distribution of the general population (n = 2304; 1191 men and 1113 women). *P < 0.05. TBF, total body fat; VAT, visceral adipose tissue.