Association of dietary patterns with the newly diagnosed diabetes mellitus and central obesity: a community based cross-sectional study

Aim The purpose of this study was to investigate the association of dietary patterns with the risk of insulin resistance (IR), diabetes mellitus (DM), and central obesity in China. Methods We performed a cross-sectional study on 1432 participants, aged 40–65 years in Hangzhou, Zhejiang province, China. Dietary intake was assessed using a semi-quantitative food frequency questionnaire. Results Factor analysis extracted four major dietary patterns: vegetable-fruits, rice–meat, seafood–eggs, and sweet–fast. The vegetable-fruits pattern was inversely associated with HOMA-IR (p < 0.001 in both genders), while sweet–fast food pattern was significantly associated with higher HOMA-IR (p = 0.002 in male, and p < 0.001 in female). The vegetables–fruits pattern was inversely correlated with visceral fat area (VFA) (p = 0.029 in males, and p = 0.017 in females), while sweet–fast food pattern presented a significant direct association (p < 0.001 in male) with VFA in males. There was no association observed between the rice–meat pattern or the seafood–eggs pattern and HOMA-IR or VFA. After adjustment for potential confounding factors, participants in the highest tertile of vegetable-fruits pattern showed a significantly lower risk of DM in both males and females (OR: 0.30, 95% CI: 0.13–0.70 in male, and OR: 0.28, 95% CI: 0.11–0.72 in female), and lower risk of central obesity was observed in males (OR: 0.50, 95% CI: 0.29–0.86 in male). Conversely, participants in the highest tertile of sweet–fast food pattern had higher risk of DM (OR: 2.58, 95% CI: 1.23–5.88 in male), and central obesity (OR: 2.85, 95% CI: 1.67–4.86 in male) only in male. While neither the rice–meat pattern nor the seafood–eggs pattern showed significant association with DM or central obesity in both genders. Conclusions Our findings indicated low risk of IR, DM, and central obesity with vegetable-fruits pattern while inverse relation with sweet–fast food pattern.


Introduction
Diabetes mellitus (DM) has become a major public health concern worldwide due to its growing epidemic prevalence. In China, the prevalence of DM dramatically increased among adults from 0.67% in 1980 to 10.9% in 2013 [1][2][3][4][5][6] . There are two major subtypes of DM: type 1 and type 2. Type 2 diabetes (T2DM) is the most common type of DM, affecting 90% of people with diabetes, and is mainly linked to insulin resistance (IR) and relative insufficiency of insulin secretion. The probability of DM was higher in overweight and obese individuals than individuals with normal body mass index (BMI), and was even higher in participants with central obesity 7 . It is well known that obesity has a multifactorial etiology which includes genetic, environmental, and dietary factors. Of these, dietary factors are thought to play a key role.
Approximately 90% of T2DM occurring is attributed to excess weight 8 . Previous studies have reported that body composition, especially body fat, is closely related to glucose metabolism in humans 9,10 . Compared with subcutaneous adipose tissue, visceral adipose tissue is known to release more proinflammatory and proatherogenic factors, leading to exacerbation of oxidative stress and IR. Studies have demonstrated that magnetic resonance imaging (MRI) and computed tomography (CT) are the reference methods to measure the accumulation of visceral fat 11 . Dietary patterns, especially western dietary pattern, have been shown to increase weight and body fat percentage 12,13 . However, few reports have demonstrated the association between dietary patterns and body fat distribution.
Traditional approach that investigates diet and disease associations mainly focuses on a single nutrient or food. However, no one consumes just one single nutrient or food, making it more difficult to separate the effect of individual dietary components. The effect of different combinatorial foods on the body can be studied by dietary pattern analyses. Cumulative studies have revealed that vegetables and fruits 14 , low-fat dairy products 15 , and regular alcohol consumption 16,17 might reduce the risk of IR. In contrast, the western dietary pattern defined by high energy and fat, and low dietary fiber intake increased the risk of glucose tolerance abnormalities [18][19][20] .
Due to the rapid social and economic development, dietary patterns, and lifestyles have changed substantially in China. Therefore, the aim of the study was to identify the major Chinese dietary patterns of individuals aged 40-65 years and evaluate their association with body fat distribution, and DM.

Study design
During the period of March to May 2010, a populationbased cross-sectional survey was carried out in the Caihe communities of Hangzhou, Zhejiang province, China. Totally 1432 Han Chinese participants aged 40-65 years were included in the present study. Subjects who had a previous history of stroke or ischemic heart disease, and who were previously diagnosed with DM were excluded. The study was conducted in accordance with the Declaration of Helsinki, and all procedures were approved by the ethics committee of Sir Run Run Shaw Hospital. All participants signed an informed consent before data collection.

Demographic characteristics
Demographic and lifestyle covariates, such as smoking status, alcohol intake, and physical activity were collected by a questionnaire. We identified three categories for smoking: current smoker, former smoker, and nonsmoker according to the participants' cigarette use answer. A participant was defined as "drinker" in case of having drunk alcoholic beverage once per week on average during the last 4 weeks, excluding those who drank alcoholic beverage only during festivals.

Dietary assessment
A validated semi-quantitative food frequency questionnaire (FFQ) with 81 items was used to assess the food and nutritional intakes by trained dietitians during a structured interview 21,22 . Participants were asked to recall their frequency of intakes of each food item in the previous 4 weeks. The frequency of food intake was measured using seven categories which were as follows: (1) never; (2) <1 time/ week; (3) 1 time/ week; (4) 2-3 times/ week; (5) 4-6 times/ week; (6) 1 time/day; (7) ≥2 times/day. Eighty-one food items were gathered in 21 predefined food groups on the basis of similarity in nutrient profiles, culinary use among the foods and the grouping scheme used in other studies. The amount of each food item consumed on average was calculated by multiplying the frequency of intake by portion size of the food, using China Food Composition Table (2009) as the database.

Laboratory measurements
Participants were requested not to eat after 20:00 pm, and received a medical checkup from 7:00 to 12:00 am the next day. Venous blood samples were obtained at 0 and 2 hours after a 75 g glucose load for an oral glucose tolerance test (OGTT). Serum glucose concentrations, triglyceride (TG), total cholesterol (TC), high density lipoprotein-cholesterol (HDL-c), and low density lipoprotein-cholesterol (LDL-c) were determined by an autoanalyzer (Aeroset, Chicago, IL, USA). Glycosylated hemoglobin A 1 c (HbA 1 c) in the whole blood was measured by high-performance liquid chromatography (Hemglobin Testing System; Bio-Rad, Hercules, CA, USA). Insulin in the serum was measured by radioimmunology using an insulin detection kit (Beijing North Institute of Biological Technology, China). Homeostasis model of insulin resistance (HOMA-IR) sore was calculated using the following equation: [fasting serum insulin (FINS; mU/L) × fasting serum glucose (FPG; mmol/L)/22.5] 23 .

Anthropometric measurement
BMI was defined as body weight divided by height squared (kg/m 2 ). Waist circumference (WC) was measured at a level halfway between the lower rib cage and iliac crest. Hip circumference was measured at the maximal horizontal girth between the waist and thigh, and the waist-to-hip ratio (WHR) was calculated and recorded for each participant. Both measurements were performed while the participant was standing. Body fat percentage (Fat%) was determined with a bioelectrical impedance analysis system (TBF-300, Tanita Co, Tokyo, Japan). Blood pressure was calculated as the average of three readings, using a mercury sphygmomanometer.
Abdominal adipose tissue was measured using a wholebody imaging system (SMT-100, Shimadzu Co, Kyoto, Japan). Spinecho sequences, TR-500 and TE-200, with a matrix size of 256 × 256 were performed for all participants. Visceral fat area (VFA) and subcutaneous adipose tissue area (SFA) were determined at the abdominal level between L4 and L5 vertebral disc space in the supine position with MRI, which were calculated with the workstation provided by the manufacturer.
Diagnosis of diabetes and central obesity DM was diagnosed according to the World Health Organization criteria. Subjects without a history of DM were administered with a 75 g OGTT. We coded participants as DM when they had glucose level above the diagnostic threshold, which is fasting glucose level≥7 mmol/L, and or 2-h glucose level ≥11.1 mmol/L. And VFA ≥ 80 cm 2 was defined as central obesity 24 .

Statistical analysis
Sample characteristics were presented as mean ± standard deviation (SD) for normally distributed variables and frequency and percentage for categorical variables. Glucose, insulin, HOMA-IR, TG, VFA, SFA, and total energy intake were reported as median (range) due to skewness and transformed lg(x) before analysis. Sweet-fast food pattern score underwent lg(x + 10) transformation to achieve a normal distribution. To compare the general characteristics according to DM status in patients and controls, continuous variables were tested by student'stest and Pearson's χ 2 test for categorical variables. We used the factor analysis (principal component) to extract the participants' dietary patterns among the 21 predefined food groups. To improve interpretability, the factors were rotated with an orthogonal rotation (varimax rotation) to minimize the correlation between factors. Four patterns were selected after evaluating the eigenvalue (≥1.5), scree plot, and factor interpretability. Items were remained in a factor if they had an absolute correlation of >±0.25 with that pattern. Correlation analysis and linear regression analysis were used to investigate the association between each dietary pattern score and HOMA-IR, as well as VFA. Factor scores for each pattern were divided into tertiles for further analysis. Multivariate logistic regression analysis was used to examine the relationships between dietary pattern with DM status and central obesity.
All analyses were performed by use of SPSS software (version 16.0 for Windows, SPSS Inc., Chicago, IL, USA). Statistical significance was considered when two-sided p-values < 0.05.

Subject characteristics
The subject characteristics are presented in Table 1. A total of 575 males and 857 females were surveyed. The prevalence of newly diagnosed DM was observed to be 10.6% in males and 6.1% in females, respectively. There were no significant differences in education, family income, smoking, drinking, physical activity, BMI, systolic blood pressure (SBP), diastolic blood pressure (DBP), 2h insulin (2h INS), TC, LDL-c, SFA, total energy intake between participants with and without DM in both males and females. WC, WHR, Fat%, FPG, 2h postprandia glucose (2h PG), FINS, HOMA-IR, HbA1 C , HDL-c, VFA were significantly higher in the DM than control group. Compared with participants who did not have DM, those who had DM tended to be older and with higher TG in females, but not in males.

Dietary patterns
We derived four major dietary patterns among 21 predefined food groups, and the associated factor loading scores with ≤−0.25 or ≥0.25 are shown in Table 2. The "vegetables-fruits" dietary pattern was represented by a high consumption of vegetables, beans, mushrooms, tubers, fruits, coarse cereals, seaweeds, wheat, nuts, and dairy products. The "rice-meat" pattern was characterized by high intake of red meat, white rice, poultry, animal organ meat, condiments, eggs, and beans. The "seafoodeggs" pattern included eggs, seafood, dairy products, nuts, fruits, and tea, and low intake of beverages. The "sweetfast food" pattern was loaded heavily on fast foods, beverages, desserts, alcoholic beverages, and was inversely related on white rice and dairy products. Each dietary pattern showed a 12.33%, 12.29%, 8.41%, and 6.07% of the variation in food intake, respectively.
Correlation of dietary pattern score with insulin resistance and fat accumulation Table 3 showed the association between each dietary pattern scores with HOMA-IR. The vegetables-fruits pattern showed an inverse association (R = −0.250; p < 0.001 in male; and R = −0.152, p < 0.001 in female) with HOMA-IR. While sweet-fast food pattern presented a significant and positive association (R = 0.162, p = 0.002 in male, and R = 0.225, p < 0.001 in female) with HOMA-IR, when adjusted for age, smoking, drinking, education, economic income, total energy intake, physical activity, and BMI. In contrast, we found no significant association in the adjusted analysis for either the rice-meat pattern or the seafood-eggs pattern. Analyses of association between each dietary pattern score with VFA were also presented in Table 3. Vegetable-fruits pattern was negatively correlated with VFA in both males and females (R = −0.111, p = 0.029 in male, and R = −0.099 p = 0.017 in female). While sweet-fast food pattern was closely correlated with VFA in males but not in females (R = 0.200, p < 0.001 in male, and R = 0.012, p = 0.775 in female) after adjustment for age, smoking, drinking, education, economic income, total energy intake, and physical activity. Similarly, there was no association observed between the rice-meat pattern or the seafood-eggs pattern and VFA.  association with DM or central obesity in both males and females.

Discussion
In the present study, we investigated the association of dietary patterns with the risk of DM and central obesity in the middle-aged Chinese population. We extracted four major dietary patterns using factor analysis: vegetables-fruits food pattern, rice-meat pattern, seafood-eggs pattern, and sweet-fast food pattern. Vegetables-fruits food pattern and sweet-fast food pattern were associated with DM and central obesity; however, neither rice-meat food pattern nor seafood-eggs pattern were associated with DM or central obesity.
The vegetables-fruits pattern is generally considered as a healthy pattern. Previous studies have reported that the fruits and vegetables, particularly intake of high green leafy vegetables, contribute to a reduced risk of T2DM and obesity [25][26][27][28] . The protective effects of this pattern may be attributable to the healthy constituents, which contain abundant dietary fiber but a low glycemic index 29,30 . In addition, vegetables and fruits were considered as potent antioxidants 31 , and some studies have supported this hypothesis 32,33 . Overall, these findings suggested that vegetables and fruits provided benefits for decreasing the prevalence of DM and central obesity among Chinese population.
Sweet-fast food pattern characterized by high consumption of desserts, fast foods, and beverages was significantly associated with a higher prevalence of DM and central obesity. The results are consistent with the existing studies. Fast food tends to be energy-dense, high in glycemic load, poor in fiber and low in micronutrients. Positive association between the fast food and obesity and T2DM have been observed in several previous studies [34][35][36] . In addition, it was assessed that people who consumed more sweetened beverages and desserts showed a higher risk of developing central obesity, IR and DM 28,37-39 , because of high glycemic index and glycemic load. However, the study participants were restricted to the middleaged Chinese population, who rarely consume fast foods, dessert, and drinks, and this may partly affect the outcomes.
In our study, we found no association of rice-meat pattern or seafood-eggs pattern with DM or central obesity. Red meat and white rice were the major components of rice-meat pattern. Red meat, which contains huge amount of saturated fat and cholesterol, is considered to be an energy-dense food and excess consumption may contribute to a surplus intake of energy, which may in turn increase the risk of T2DM and obesity 40,41 . However, few studies showed no relationship between red meat and T2DM or obesity 22 . This may be due to high content of zinc in red meat. Zinc content is highest in animal-source foods, and low in refined cereals and vegetables 42 . Zinc deficiency may aggravate IR in non-insulin dependent DM over a 6 year follow-up study 43 . According to a previous systematic review conducted among the Asian populations, rice which had a high glycemic index was positively associated with the risk of developing T2DM and weight gain 30,44,45 . However, studies aiming to investigate the association between rice intake and propensity to develop obesity are still inconclusive [45][46][47] . Furthermore, study participants with obesity and DM risk might have been advised to limit their consumption of carbohydrate and fat, which would influence the results. Egg and seafood had a high factor loading in the seafood-eggs pattern. Egg is a nutrient-dense food containing high-quality protein and amount of beneficial nutrients. However, dietary cholesterol from eggs can also lead to a modest increase in blood liquids. The results of previous studies focused on egg consumption and IR and T2DM were controversial. A meta-analysis by Djousse et al. showed no significant effect of egg consumption on T2DM 48 . While the China National Nutrition Survey reported that egg consumption more than one time per day increased the T2DM occurrence in females and that TG and TC were significantly higher in females taking two or more eggs per week 49 . On the other hand, a study conducted in Korean reported that consumption of more than one egg per day could decrease the risk of metabolic syndrome (MetS) 50 . Different results on the association were likely due to differences in dietary patterns across countries. Data from previous studies in relation to the intake of fish with DM risk also are inconclusive. Ecological studies in several countries have shown protective effects of fish intake against the body weight gain and the development of DM 51,52 . However, the beneficial effects were not evident in other studies 53,54 , which was consistent with our study. A recent systematic review and meta-analysis of fish consumption and risk of T2DM suggested heterogeneity between geographical regions in the observed association 55 . The different associations of fish consumption with DM and obesity in different populations could be partly explained by diverse fish preparation methods, different type and amount of cooking fat used, and various kind of condiments.
In the present study, we found that the effect of food patterns might be greater in males than in females for DM and central obesity. Difference between the genders in the physiological response to different food patterns is uncertain, one plausible explanation is the different anatomy and physiology between males and females. For example, female may convert more α-linoleic acid into docosahexanoic acid than male do. In addition, the study participants are middle-aged Chinese population, and at times few female participants are involved in peri-and postmenopausal period. Previous studies proved that menopause-related loss of estrogen and increased androgens appear to be key hormonal changes that is associated with the development of MetS, which were independent of aging and other standard cardiovascular disease risk factors 56 . The study power of stratified analysis was limited due to the number of participants. Additional large population are needed in the future.
There were several potential limitations of this study that need to be highlighted. Firstly, because our study is a cross-sectional design study, no cause-and-effect relationships between dietary patterns and the risk of DM and central obesity may be established. Secondly, several arbitrary and subjective decisions in the use of factor analysis should be taken into consideration. Thirdly, even though we have adjusted for confounding factors that are related to the risk of DM and obesity, we cannot exclude the influence of uncontrolled confounders on our observed results. Finally, the study sample was composed of the middle-aged Chinese population in Hangzhou, Zhejiang Province, eastern China, and these results cannot be generalized to the entire Chinese population. In conclusion, this study has identified that vegetables-fruits  food pattern was significantly associated with a lower risk of DM and central obesity, while the reverse was apparent for sweet-fast food pattern among Chinese population aged 40-65 years. Nevertheless, further prospective studies are needed to understand the causal relationships between dietary patterns and the risk of central obesity and DM. For DM, multivatiate model adjusted for age, smoking, drinking, education, economic income, total energy intake, physical activity, and BMI. For central obesity, multivatiate model adjusted for age, smoking, drinking, education, economic income, total energy intake, and physical activity.