Dietary patterns, insulin resistance, and prevalence of the metabolic syndrome in women

Abstract

Background: Although individual foods and nutrients have been associated with the metabolic syndrome, whether dietary patterns identified by factor analysis are also associated with this syndrome is not known.

Objective: We aimed to evaluate the association of major dietary patterns characterized by factor analysis with insulin resistance and the metabolic syndrome among women.

Design: Usual dietary intakes were assessed in a cross-sectional study of 486 Tehrani female teachers aged 40–60 y. Anthropometric and blood pressure measurements were performed, and fasting blood samples were taken for biomarker assessment. The metabolic syndrome was defined according to Adult Treatment Panel III guidelines, and insulin resistance was defined as the highest quartile of the homeostasis model assessment scores.

Results: We identified 3 major dietary patterns by factor analysis: the healthy dietary pattern, the Western dietary pattern, and the traditional dietary pattern. After control for potential confounders, subjects in the highest quintile of healthy dietary pattern scores had a lower odds ratio for the metabolic syndrome (odds ratio: 0.61; 95% CI: 0.30, 0.79; P for trend < 0.01) and insulin resistance (0.51; 0.24, 0.88; P for trend < 0.01) than did those in the lowest quintile. Compared with those in the lowest quintile, women in the highest quintile of Western dietary pattern scores had greater odds for the metabolic syndrome (1.68; 1.10, 1.95; P for trend < 0.01) and insulin resistance (1.26; 1.00, 1.78; P for trend < 0.01). Higher consumption of traditional dietary pattern was significantly associated only with abnormal glucose homeostasis (1.19; 1.04, 1.59; P < 0.05).

Conclusion: Significant associations exist between dietary patterns identified by factor analysis, the metabolic syndrome, and insulin resistance.

INTRODUCTION

The term metabolic syndrome describes a clustering of risk factors for cardiovascular disease (CVD). Its pathophysiology is believed to include insulin resistance (1); but its definition is controversial (2). It is now established that this syndrome predicts the development of type 2 diabetes and CVD (3). Persons with the metabolic syndrome are also at greater risk of premature death due to CVD or all-cause mortality (4, 5). Cross-sectional and longitudinal epidemiologic studies have provided prevalence and incidence data on the syndrome, but estimates vary according to the criteria used (6, 7). The syndrome is common in the United States, particularly among Mexican Americans (8). However, it is not just that an epidemic of the metabolic syndrome is occurring in developed countries; rates of the metabolic syndrome in developing countries are also high. In Tehran, Iran, one-third of adults (9, 10) and one-tenth of the adolescent population (11) are affected.

The metabolic syndrome is a multifactorial disorder, and diet plays an important role in its development (12). Diet can be considered in terms of dietary patterns, an approach that has been used to investigate diet-disease relations (13–15). Dietary patterns address the effect of the diet as a whole and thus may provide insight beyond the effects described for single nutrients or foods (13).

Although several dietary factors have been associated with the metabolic syndrome (16–18), few studies have examined the association between dietary patterns and the metabolic syndrome. We are aware of only 2 reports that evaluated dietary patterns directly in relation to the metabolic syndrome (19, 20). Both used cluster analysis to identify dietary patterns, and it remains unknown whether dietary patterns identified by factor analysis are also associated with the metabolic syndrome. Factor analysis and cluster analysis are statistically different procedures, and each identifies dietary patterns with different food compositions (21). Although dietary patterns derived from both factor and cluster analysis have been associated with risk of chronic diseases (19–23), some evidence supports the possibility that a person's dietary patterns would be best represented by using factor analysis (21, 24). The current study was conducted to assess the relation of major dietary patterns identified by factor analysis to insulin resistance and the metabolic syndrome in a group of Tehrani female teachers aged 40–60 y.

SUBJECTS AND METHODS

Participants

This cross-sectional study was conducted in a representative sample of Tehrani female teachers aged 40–60 y selected by a multistage cluster random-sampling method. The sample of 583 female teachers was invited to participate in the current study; 521 of the women agreed to do so. Participants with a history of CVD, diabetes, cancer, or stroke were excluded because of possible disease-related changes in diet. We also excluded women who had left >70 items blank on the food-frequency questionnaire (FFQ), who reported a total daily energy intake (EI) outside the range of 800–4200 kcal, and who were taking medications that would affect serum lipoprotein concentrations, blood pressure, and carbohydrate metabolism. These exclusions left 486 women for the current analysis.

Written informed consent was obtained from each participant. The study was approved by the research council of the National Nutrition and Food Technology Research Institute, Shaheed Beheshti University of Medical Sciences.

Assessment of dietary intake

Usual dietary intake was assessed by using a 168-item semiquantitative FFQ. All of the questionnaires were administered by a trained dietitian. The FFQ consisted of a list of foods with standard serving sizes commonly consumed by Iranians. Participants were asked to report their frequency of consumption of a given serving of each food item during the previous year on a daily (eg, bread), weekly (eg, rice or meat), or monthly (eg, fish) basis. The reported frequency for each food item was then converted to a daily intake. Portion sizes of consumed foods were converted to grams by using household measures (25). Total EI was calculated by summing up EIs from all foods. Because of the large number of the food items relative to the number of participants, we assigned each food item into 1 of 41 defined food groups (Table 1). The basis for placing a food item in a certain food group was the similarity of nutrients. Some food items were considered individually as a food group because their nutrient profiles were unique (eg, eggs, margarine, coffee, and tea) or their consumption was considered to reflect a distinct dietary pattern [eg, garlic, broth, or doogh (an Iranian yogurt preparation with a consistency similar to that of whole milk)]. A previous validation study of this FFQ revealed good correlations between dietary intakes assessed by a similar FFQ and those from multiple days of 24-h dietary recalls completed during an earlier year-long study (26).

Assessment of anthropometric measures

Weight was measured while the subjects were minimally clothed and not wearing shoes; weight was measured with digital scales and recorded to the nearest 100 g. Height was measured by using a tape measure while the subjects were standing, were not wearing shoes, and had the shoulders in a normal position. Body mass index (BMI) was calculated as weight (in kg) divided by height (in m²). Waist circumference (WC) was measured at the narrowest level, and hip circumference was measured at the maximum level over light clothing, by using an unstretched tape measure, without any pressure to body surface; measurements were recorded to the nearest 0.1 cm. Because the measurements were taken over light clothing, participants were asked to remove belts and tight or loose garments intended to alter the shape of the body, and the person measuring was asked to inspect the tension of the tape on the subject's body to ensure that the tape had the proper tension—ie, neither too loose nor too tight. Although the narrowest waist is easy to identify in most persons, in some, no single narrowest waist can be identified because of a large amount of abdominal fat or extreme thinness (27). In the current study, when the narrowest point of the waist was difficult to identify (particularly in obese participants), we measured WC immediately below the end of the lowest rib, because in most persons the narrowest waist is at the lowest rib (27). To reduce error, all measurements were taken by the same technician.

Assessment of biomarkers

A blood sample was drawn between 0700 and 0900 into evacuated tubes after an overnight (12 h) fast. Blood samples were taken while the subject was sitting and according to a standard protocol; the samples were centrifuged for 10 min at 500 × g and at 4 °C within 30–45 min of collection. Samples were analyzed by using an autoanalyzer (Selectra 2; Vital Scientific, Spankeren, Netherlands). Fasting plasma glucose (FPG) was measured on the day of blood collection by the enzymatic colorimetric method and using glucose oxidase. Serum triacylglycerol concentrations were assayed with triacylglycerol kits (Pars Azmoon Inc, Tehran, Iran) by using enzymatic colorimetric tests with glycerol phosphate oxidase. HDL cholesterol was measured after precipitation of the apolipoprotein B–containing lipoproteins with phosphotungistic acid. Serum insulin concentrations were measured by using enzyme-linked immunosorbent assay kits and an enzyme-linked immunosorbent assay reader (Tecan Sunrise, Salzburg, Austria). The interassay and intraassay CVs of this method were <10%.

Assessment of blood pressure

For blood pressure measurements, participants were first asked to rest for 15 min. Then, a trained physician measured the blood pressure 3 times in seated participants by using a standard mercury sphygmomanometer, and thereafter the mean of 3 measurements was considered as the participant's blood pressure. Systolic blood pressure was defined as the appearance of the first sound (Korotkoff phase 1), and diastolic blood pressure was defined as the disappearance of the sound (Korotkoff phase 5) during deflation of the cuff at a 2–3-mm/s rate of decrement of the mercury column.

Assessment of other variables

Data on physical activity were obtained by using an interview-based questionnaire and expressed as metabolic equivalent hours per week (MET-h/wk) (28). Additional covariate information regarding age, smoking habits, menopausal status, medical history, and current use of medications was obtained with questionnaires.

Definition of terms

The metabolic syndrome was defined as the presence of ≥3 of the following components as recommended by Adult Treatment Panel III (ATP III; 29): abdominal adiposity (WC >88 cm); low serum HDL cholesterol (<50 mg/dL); high serum triacylglycerol concentrations (≥150 mg/dL); elevated blood pressure (≥130/85 mm Hg); and abnormal glucose homeostasis (fasting plasma glucose ≥ 110 mg/dL). Insulin resistance was estimated on the basis of fasting glucose and insulin concentrations by using the homeostasis model assessment for insulin resistance (HOMA-IR) method (30) and was defined as the highest quartile of the HOMA-IR scores.

Statistical analysis

To identify major dietary patterns based on the 41 food groups, we used principal component analysis, and the factors were rotated by orthogonal transformation. The natural interpretation of the factors in conjunction with eigenvalues >1 and the Scree test (31) determined whether a factor should be retained. The Scree plot is a plot of the eigenvalues of derived factors. The eigenvalues of the factors dropped substantially after the third factor and remained more similar to each other after the fourth factor. The derived factors (dietary patterns) were labeled on the basis of our interpretation of the data and of the earlier literature. The factor score for each pattern was calculated by summing intakes of food groups weighted by their factor loadings (31), and each participant received a factor score for each identified pattern.

We categorized participants by quintile of dietary pattern scores. One-way analysis of variance with Tukey's post hoc comparisons was performed to evaluate significant differences in general characteristics (eg, age, anthropometry, and physical activity) across quintile categories of dietary pattern scores; the distribution of qualitative variables across quintiles was evaluated by using chi-square tests. Age- and energy-adjusted means for dietary variables across quintiles of dietary pattern scores were calculated. We also calculated multivariate-adjusted means (ie, age, physical activity, smoking, menopausal status, total EI, and current estrogen use) for insulin and features of the metabolic syndrome. Analysis of covariance with Bonferroni correction was used to compare these means.

To determine the associations of dietary patterns with insulin resistance and the metabolic syndrome, we used multivariable logistic regression. First we obtained age-adjusted ORs, and then we adjusted for cigarette smoking (yes or no), physical activity (MET-h/wk), current estrogen use (yes or no), menopausal status (yes or no), and family history of diabetes and stroke (yes or no). We also adjusted for EI (kcal/d) in the third model, and finally we added BMI (kg/m²) to the logistic regression model to examine whether the relation was mediated by obesity. In all multivariate models, the first quintile of dietary patterns score was considered as a reference. To derive an estimate of association that better represents the relative risk, all ORs derived from logistic regression models were corrected by using the formula suggested by Zhang and Yu (32). The Mantel-Haenszel extension chi-square test was used to assess the overall trend of ORs across increasing quintiles of dietary pattern scores.

Because using cutoffs for defining the metabolic abnormalities involves some loss of information, we also studied relations between dietary pattern scores and metabolic risks as continuous variables by using partial correlation coefficients. All analyses were adjusted for age, EI, cigarette smoking, physical activity, current estrogen use, menopausal status, and family history of diabetes and stroke. In addition, we adjusted all models for BMI. We used SPSS software (version 9.05; SPSS Inc, Chicago IL) for all statistical analyses.

RESULTS

We identified 3 major dietary patterns by using factor analysis: the healthy dietary pattern (high in fruits, tomatoes, poultry, legumes, cruciferous and green leafy vegetables, other vegetables, tea, fruit juices, and whole grains), the Western dietary pattern (high in refined grains, red meat, butter, processed meat, high-fat dairy products, sweets and desserts, pizza, potatoes, eggs, hydrogenated fats, and soft drinks and low in other vegetables and low-fat dairy products), and the traditional dietary pattern (high in refined grains, potatoes, tea, whole-grains, hydrogenated fats, legumes, and broth). The factor-loading matrixes for these dietary patterns are shown in Table 2. Other minor dietary patterns have also been identified by the factor analysis, but because of the small variances they explained, we did not consider them in the subsequent analyses.

Characteristics of the study participants across quintile categories of the dietary pattern scores are shown in Table 3. Compared with participants in the lowest quintile, those in the highest quintile of the healthy dietary pattern had significantly lower BMI and significantly lower prevalence of the metabolic syndrome, were significantly more physically active, and were significantly less likely to be obese. Conversely, in comparison with participants in the lowest quintile, those in the highest quintile of the Western dietary pattern had significantly higher BMI, were significantly less likely to exercise, and had significantly higher prevalence of obesity and the metabolic syndrome. Participants in the highest quintile of the traditional dietary pattern were significantly older, slightly more physically active, and significantly less likely to be obese than were those in the lowest quintile. No significant difference was found in the distribution of current smokers and estrogen users across quintile categories of dietary patterns. Those in the highest quintile of the healthy dietary pattern had significantly lower intakes of energy and cholesterol and significantly higher intakes of vitamin B-6, magnesium, and fiber, whereas those in the highest quintile of the Western dietary pattern had significantly higher intakes of energy and cholesterol and significantly lower intakes of vitamin B-6, magnesium, and fiber. Participants in the highest quintile of the traditional dietary pattern had slightly lower EI than those in the lowest category, but their nutrient intakes were not significantly different in most cases.

ORs for the metabolic syndrome and insulin resistance across quintile categories of dietary pattern scores are presented in Table 4. After control for age, participants in the highest quintile of the healthy dietary pattern score had lower odds of the metabolic syndrome (OR: 0.55; 95% CI: 0.27, 0.74) and insulin resistance (0.47; 0.18, 0.89) than did those in the lowest quintile, whereas those in the highest quintile of the Western dietary pattern score had greater odds of the metabolic syndrome (OR: 1.73; 1.11, 2.06) and insulin resistance (1.33; 1.05, 1.84) than did those in the lowest quintile. Further adjustment for other potentially confounding variables attenuated these associations, but they remained significant. Even after additional control for BMI, the inverse association of the healthy dietary pattern score and the positive association of that Western dietary pattern score with the metabolic syndrome remained significant. However, after adjustment for BMI, the positive association of the Western dietary pattern with insulin resistance disappeared. No significant overall associations were seen between the traditional dietary pattern score and the metabolic syndrome or insulin resistance.

In the multivariate models, participants in the highest quintile of the healthy dietary pattern score had lower odds for components of the metabolic syndrome (in the range of 0.50 for elevated blood pressure to 0.83 for abnormal glucose homeostasis) (Table 5). In contrast, those in the highest quintile of the Western dietary pattern score had significantly higher odds for components of the metabolic syndrome (in the range of 1.28 for low HDL cholesterol to 2.17 for elevated blood pressure). However, the association was not significant for abnormal glucose homeostasis (OR: 1.11; 0.95, 1.46). Although the trends of ORs across quintiles of the traditional dietary pattern score were significant in most cases, the highest category of this dietary pattern was associated significantly only with abnormal glucose homeostasis (OR: 1.19; 1.04, 1.59).

After adjustment for potential confounding variables in partial correlation analysis, the healthy dietary pattern score was associated positively with serum HDL cholesterol and inversely with other metabolic variables (Table 6). Even when the models were further adjusted for BMI, all associations remained significant except the positive association of this score with HDL cholesterol and its inverse association with plasma glucose. The Western dietary pattern score also was independently related to various metabolic risk factors except plasma glucose concentration. These associations were inverse for HDL cholesterol and positive for other metabolic variables. When we further adjusted for BMI, the associations of this pattern score with WC and HDL cholesterol disappeared. The traditional dietary pattern score was not significantly associated with any metabolic variables (Table 6).

DISCUSSION

As stated in Results, we observed 3 major dietary patterns in this population: the healthy dietary pattern, the Western dietary pattern, and the traditional dietary pattern. Further analysis suggested that these major dietary patterns are related to insulin resistance and features of the metabolic syndrome. The healthy dietary pattern was associated with lower risks of insulin resistance and the metabolic syndrome, whereas the Western dietary pattern was associated with higher risks of insulin resistance and the metabolic syndrome. We found no significant association between the traditional dietary pattern and these conditions. All associations were independent of other lifestyle factors. To our knowledge, this is the first investigation in which major dietary patterns identified by factor analysis have been associated directly with the metabolic syndrome.

Although insulin resistance and the metabolic syndrome are underlying causes of major chronic diseases (33), few studies have assessed dietary patterns in relation to these conditions. In the Framingham Offspring Study (20), higher and lower prevalences of the metabolic syndrome have been reported in women with the “empty calorie” and the “wine and moderate eating” dietary patterns, respectively. In the Malmö Diet and Cancer Cohort (19), features of the metabolic syndrome were more prevalent in women with the “white-bread” dietary pattern and less prevalent in women with the “milk-fat” pattern. In a cross-sectional study in a British population (34), a dietary pattern characterized by high consumption of fruit and vegetables and low consumption of processed meat and fried foods was inversely associated with features of the metabolic syndrome. However, that study was limited by lack of control for physical activity, which tends to be associated with dietary patterns (35). Data on the association of dietary patterns with insulin resistance or insulin sensitivity are sparse, as they are for an association with the metabolic syndrome. Whereas some studies have considered the association with insulin sensitivity as their main objective (36), others have reported it as an accessory finding (37, 38). In a cross-sectional study of a multiethnic cohort of 980 subjects aged 40–69 y, Liese et al (36) found that subjects with the “white bread” pattern (identified by cluster analysis and high in white breads, tomatoes, cheese, dried beans, eggs, meats, fats and oils, and beer) had lower levels of insulin sensitivity, whereas those with the “dark bread” pattern (with high intakes of dark bread and high-fiber cereal, rice and pasta, cruciferous vegetables, other vegetables, potatoes, low-fat milk, fish, nuts and seeds, and tofu) and with the “wine” pattern (with high intakes of wine and mixed drinks) had greater insulin sensitivity than did subjects with the other patterns identified. Higher scores on the Western dietary pattern (identified by factor analysis) were associated with higher insulin concentrations in the Health Professionals Follow-up Study (37). The same results have also been observed in the third National Health and Nutrition Examination Survey (38) and in an Irish population (39).

As was found in other studies, we found in the current study that the healthy dietary pattern is associated with lower risk of metabolic abnormalities, whereas the Western dietary pattern is related to higher risk of adverse metabolic risk factors. The inverse association between the healthy dietary pattern and the metabolic syndrome could be attributed to that pattern's healthy constituents, including whole grains (17), fiber (40), fruit and vegetables (40, 41), and magnesium (42). The mechanisms by which greater intakes of these foods and nutrients may contribute to the inverse association between the healthy dietary pattern and the metabolic syndrome are not fully understood, but they are likely to be many (17, 40–42). Constituents of fruit, vegetables, and whole grains, including dietary fiber, vitamin E, folate, and magnesium, have been independently associated with reduced metabolic risks related to metabolic syndrome. An additional protective effect of other constituents of these foods or their interactions may also explain their beneficial effects. Reduced insulin demand may be another protective mechanism associated with higher intakes of these foods. In general, because of their physical form and their viscous fiber content, these foods tend to be slowly digested and absorbed, and thus they have relatively low glycemic indexes. Furthermore, most foods in the healthy dietary pattern have a low glycemic load, which has been documented to be associated with lower risk of insulin resistance (43). The healthy dietary pattern we identified in the current study is somewhat similar to the patterns that have been labeled “prudent dietary pattern” in other studies (44, 45). Our healthy dietary pattern was also similar to the Dietary Approaches to Stop Hypertension eating plan, which has been recommended for decreasing blood pressure (46) and improving features of the metabolic syndrome (47). The positive association between the Western dietary pattern and the metabolic syndrome could be attributed to the lower amounts of beneficial foods and nutrients that this pattern contains. Higher intakes of refined grains (17) and saturated fat (48) in this dietary pattern also could explain part of this association. We observed no association between the traditional dietary pattern and the risk of the metabolic syndrome or insulin resistance. The complex nature of this pattern may explain this finding to some extent. This dietary pattern was loaded with both healthy (whole grains, tea, and legumes) and unhealthy (refined grains, potatoes, and hydrogenated fats) foods. Whereas healthy foods of this pattern have been reported to be protectively associated with the metabolic syndrome (17, 49, 50), the pattern's unhealthy constituents have adverse effects on metabolic markers (17, 48, 51).

Some of the relations remained even after control for BMI. This shows that general obesity cannot explain all associations between diet and chronic diseases and that other factors, such as abdominal adiposity, may be responsible. We have not controlled for WC, as a measure of abdominal adiposity, in our analysis, because abdominal adiposity is one feature of the metabolic syndrome. However, our previous investigation in Tehrani women showed that WC is a better index than BMI to use in explaining metabolic abnormalities (52).

The dietary pattern approach is complementary to analyses using individual foods or nutrients, which are limited by biologic interactions and colinearity among nutrients. The logic behind the dietary pattern approach is that foods and nutrients are not eaten separately but are eaten in the form of specified dietary patterns. However, all statistical methods that have been used for data reduction have limitations. For example, using factor analysis for dietary data reduction has been criticized for its subjectivity in nature and for the difficulty of replicating the results in other populations (53). However, similar dietary patterns derived by factor analysis have been observed in different populations. It appears that the dietary patterns observed in this Iranian population are similar to those in Western populations. This is not surprising, because, during the past few years, Iran has experienced a socioeconomic transition coupled with westernization in diet and lifestyle (54, 55).

Several limitations need to be considered in the interpretation of our findings. We assessed dietary patterns by using food intake data only, whereas the inclusion of eating behaviors such as meal and snack patterns in dietary pattern analysis has been recommended (56). Limitations of the FFQ also apply to dietary pattern analyses that are based on dietary information collected by this method. The other limitation of our study is its cross-sectional nature. Thus, the association between these dietary patterns and the metabolic syndrome remains to be confirmed in prospective analyses. We cannot generalize our findings to all Iranian populations, because, in Iran, teachers have a socioeconomic status higher than that of the general population. However, participants in the current study were selected from 4 large, socioeconomically diverse districts of Tehran, so that a broad range of dietary habits were represented.

In conclusion, the current findings indicate that a dietary pattern characterized by high consumption of fruit, vegetables, poultry, and legumes is associated with reduced risk of insulin resistance and the metabolic syndrome in Tehrani female teachers. In contrast, a dietary pattern with high amounts of refined grains, red meat, butter, processed meat, and high-fat dairy products and low amounts of vegetables and low-fat dairy products is associated with a greater risk of the metabolic syndrome.

We thank the participants of the study for their enthusiastic support. None of the authors had any personal or financial conflicts of interest.

AE and LA designed the study, collected and analyzed the data, and wrote the manuscript; MK served as a supervisor and YM as advisor for this research; and FBH and WCW reviewed the study and contributed to manuscript preparation.

REFERENCES

FOOTNOTES