Study population
The SUN project is a prospective, multipurpose, and dynamic cohort study conducted in Spain, which was originally designed to establish associations between diet and the occurrence of several diseases and chronic conditions including overweight and obesity [16]. Using biennial mailed questionnaires participants have been continually followed-up. Participants’ recruitment started in December 1999, and it is permanently open. All participants are university graduates, with ages ranging from 20 to 90 years.
For the present study, we included a subsample of the cohort with a minimum follow-up period of 6 years, therefore only those participants who were recruited before September 2006 could be included (n = 15,909). To avoid reverse causality bias, we excluded 4,084 participants who met at least one MetS criterion at baseline. As recommended in nutritional epidemiology, we also excluded 1,187 participants who reported values for total energy intake at baseline out of predefined limits (less than 800 kcal/d in men and 500 kcal/d in women or more than 4,000 kcal/d in men and 3,500 kcal/d in women). Further exclusions were 578 participants without any follow-up who were considered lost to follow-up (retention rate = 95%), and 1,997 participants who did not provide the relevant information about diagnostic criteria for the MetS at the 6th -year follow-up. After these exclusions a total of 8,063 participants were available for the final analysis. The retention rate of the study was above 95%. A comparison analysis between baseline characteristics of the participants who were included in the study, and those participants who did not have any information at Q_6, did not find meaningful differences between them (data not shown).
The study was approved by the Human Research Ethical Committee of the University of Navarra. Voluntary completion of the first self-administrated questionnaire was considered to imply informed consent.
Dietary assessment
A semi-quantitative food frequency questionnaire (FFQ) was included in the baseline questionnaire. The FFQ was repeatedly validated in Spain [17-19] and includes 136 food items. Nutrient composition of specified portion sizes was multiplied by the frequency of consumption in order to calculate each nutrient score. Frequencies of consumption were measured in 9 categories (ranging from never/almost never, to >6 servings/day) for each food item, in order to assess habitual dietary intakes over the previous year. We defined a Mediterranean dietary pattern “a priori” using the score proposed by Trichopoulou et al. [20]. The traditional Mediterranean dietary pattern is characterized by abundant consumption of olive oil (the major source of fat), plant foods (vegetables, fruits, cereals, and nuts), fresh fruit as daily dessert, low-to moderate intake of dairy products (cheese and yogurt), fish and poultry, low intake of red meat, and regular moderate intake of wine generally consumed during meals. For the present study the Mediterranean dietary pattern was appraised combining 8 out of 9 of items of the original score (fruits and nuts; vegetables; fish; legumes; cereals; meat and meat products; alcohol; and the ratio monounsaturated fatty acids/saturated fatty acids). Since yogurt consumption is our exposure of interest, we excluded consumption of dairy products from the score. Therefore, the score ranged from 0 to 8 points. Food composition tables valid for Spain were used to update the nutrient databank [21].
Yogurt consumption assessment
Frequency of whole-fat and low-fat yogurt consumption in the previous year was reported by each participant in the baseline FFQ. Frequency of total yogurt consumption was estimated using the sum of these two food items (whole-fat and low-fat yogurt consumption). The sample was divided into three categories according to their total, whole-fat and low-fat yogurt consumption: 0–250 g/week (0–2 servings/week), >250 to <875 g/week, (>2 to <7 servings/week) and ≥875 g/week (≥7 servings/week); (125 g was considered as one serving size).
Assessment of non-dietary variables
Other questions (46 items for men and 54 for women) were also included in the baseline questionnaire assessing the participants’ lifestyle, and socio-demographic variables (sex, age, marital status, and employment), health related habits (smoking status, physical activity during leisure time), medical history (prevalence of chronic diseases such as cancer, diabetes and cardiovascular disease), as well as anthropometric data (weight and height) previously validated in the cohort [22]. We calculated body mass index (BMI) as self-reported weight in kilograms divided by the square of the self-reported height in meters. Physical activity was ascertained through a baseline 17-item questionnaire Physical activity was previously validated in a subsample of the cohort, with weekly MET-hours shown to adequately correlate with objectively measured energy expenditure [Spearman ρ = 0.51; 95% confidence interval (CI): 0.232, 0.707] [23].
Outcome assessment
We defined MetS according to the American Heart Association (AHA) and the International Diabetes Federation (IDF) criteria as outlined in the harmonized definition for MetS [4]. According to this harmonized definition, the MetS diagnosis requires the presence of at least 3 of the 5 criteria: central adiposity (in our study waist circumference: ≥94 cm in males and ≥80 cm in females), hypertriglyceridemia (≥150 mg/dL or condition-specific medication), low levels of high-density lipoprotein cholesterol (<40 mg/dL for men and < 50 mg/dL for women), elevated blood pressure (systolic ≥ 130 and/or diastolic ≥ 85 mmHg or antihypertensive drug treatment in a patient with a history of hypertension), and impaired glucose metabolism (fasting glucose ≥ 100 mg/dL or drug treatment of elevated glucose). For a sensitivity analysis we used a higher cut-off point for central adiposity (waist circumference: ≥102 cm in males and ≥88 cm in females) as proposed by the WHO for Caucasian populations [4].
Self-reported information about each specific MetS criterion was collected in the Q_6 (6th year follow-up questionnaire). With the Q_6 questionnaire, a measuring tape and an explanation of how to measure their own waist were sent to each participant. Diagnosis of MetS itself, the same as each MetS criterion was previously validated in a subsample of the cohort [24,25]. The validation study for MetS diagnosis found agreement between self-reported MetS diagnosis and MetS diagnosis from participant medical records to be 91.2% for confirmed MetS (95% CI: 80.7-97.1%) and 92.2% for non-confirmed MetS (95% CI: 85.1-96.4%) [24].
Incident cases of MetS were defined as all participants who did not have any MetS criterion at baseline, and reported 3 or more criteria of MetS in the 6th year follow-up questionnaire.
Statistical analyses
Three categories according to their baseline yogurt consumption were used to classify the participants included in the study: 0–250 g/week (0–2 servings/week), >250 to <875 g/week (>2 to <7 servings/week), and ≥875 g/week (≥7 servings/week). The group with the lowest frequency of consumption (0–250 g/week) was considered as the reference category.
Logistic regression models were fitted to assess the relationship between total, whole-fat or low-fat yogurt consumption at baseline and the risk of developing incident MetS during the first 6 years of follow-up. The specific relationship between each of the 5 components of the MetS and total, whole-fat and low-fat yogurt consumption were also assessed. Odds Ratios (OR) and their 95% CI were calculated. Tests for linear trend across increasing categories of yogurt consumption were conducted by assigning the median consumption within each category and treating this variable as a continuous variable.
The interaction between adherence to Mediterranean diet [low (0–4 pts)/high (5–8 pts)], and fruit consumption [low (under the median: <264.5 g/d)/high (above the median: ≥264.5 g/d)] and total yogurt consumption were tested using likelihood ratio tests comparing the fully adjusted model and the same model with the interaction product-term (2 degrees of freedom).
Additional analyses were conducted to test the association of the joint exposure to total yogurt consumption and fruit consumption with the risk of MetS, therefore we built six categories of the joint combined exposure to total yogurt consumption (low, medium and high consumption) and fruit consumption (under the median and at the median or higher) considering as the reference category those who had low yogurt consumption (<250 g/week) and low fruit consumption (under the median).
We fitted a first model without any adjustment (crude), a second model adjusted for age and sex, and a third multivariable-adjusted model after additional adjustment for the following potential confounders: baseline weight (kg), total energy intake (kcal/d), alcohol (g/d), soft drink consumption (ml/d), red meat consumption (g/d), consumption of French fries (g/d), consumption of fast food (g/d), Mediterranean dietary pattern adherence (3 categories: low, medium, and high adherence), physical activity (METs-hours/week), sitting hours (hr/d), sedentary behavior (hr/d), smoking status (non-smoker, smoker, former smoker), snacking between meals (yes/no), and following a special diet (yes/no). Snacking between meals was defined as those participants who responded affirmatively to the question “Do you eat between main meals (snacking)?” We did not consider if it was healthy or unhealthy snacking. The percentage of participants who reported snacking between meals was 33.3%. We considered that a participant followed a special diet, if she/he answered yes to the question “Do you follow any special diet”. If they had answered yes, they were asked which kind of special diet they followed (such as hypocaloric diet, vegetarian diet, low-fat diet, low sodium diet, high protein diet, among others). We used this variable dichotomously. The percentage of participants following a special diet was 6.1%.
For the selection of the potential confounders in the multivariate model, and as its currently recommended [26], we took into account the previously published scientific literature including our own results based on the cohort about risk factors for hypertension, overweight/obesity, type 2 diabetes, avoiding exclusively the statistical approach, the stepwise procedures, or the changes in the point estimates after adjusting for potential confounders.
All p values presented are two-tailed; p < 0.05 was considered statistically significant. Analyses were performed using STATA/SE version 12.0 (StataCorp, College Station, TX, USA).