Participants
From January 2008 to April 2009 a cross-sectional study was conducted in a cluster random sample of adolescents in an urban area (n = 606), whilst all students from 8th, 9th and 10th grade willing to participate in the rural area were included (n = 173). A mean energy intake of 1700 kcal/d among school-going Ecuadorian children (E. Segarra, unpublished data, 2006) was used for sample size calculations. A precision of 10% and a cluster effect of 2 were taken into account for this. Anticipating possible not response, the original sample size was increased by14%. In total, 779 adolescents, aged 10–16 years old, from an urban (Cuenca) and rural (Nabón) area in Ecuador participated. Blood samples were collected in a subsample of 334 volunteers. A more detailed methodology of this study has been previously published [14].
The study protocol was approved by both the Ecuadorian and Belgian Ethical Committees (Nr: CBM/cobi-001 - 2008/462). An additional protocol for biochemical assessment was approved by the Ghent University Hospital Ethical Committee (Nr 2008100–97). Written consents were obtained from participants and their parents or guardians. Adolescents that were pregnant, had a concomitant disease, or were following a special diet were excluded from the study.
Dietary assessment
Food intake was measured using an interview-administered 2-day 24-hour dietary recall. Days were randomly allocated to include one recall on a weekday and a second one on a weekend day per each participant [15]. Locally used utensils were calibrated and used to estimate portion sizes in order to quantify the amounts of food consumed. If the participants did not supply detailed information on the ingredients used and/or cooking methods of a recipe, recipes were prepared in triple by local volunteering housewives. For each of these recipes the ingredients and their weights were measured. The average was calculated which served as the final estimate for the ingredients and their weight of the recipe. In the case of uncommon recipes, such as some desserts or traditional dishes, an experienced cook was asked to prepare the recipe. The ingredients and their weights were recorded and calculated from this.
Since an up-to-date Ecuadorian food composition database does not exist, a compiled food composition database was developed using a pre-determined procedure. During the first stage all the 24-hour dietary recall forms were revised to make a list of all consumed recipes or ingredients. The second stage involved constructing a database using the following procedure: we searched the U.S. (USDA, 2012) database, when food items were not available the Mexican database (INNSZ, 1999) was used. If data could not be obtained, the Central American (INCAP/OPS, 2012) and Peruvian (CENAN/INS, 2008) databases were searched to compose the final food composition database. For locally processed and pre-packed food items, food labels were used to obtain the composition. A total of 13 food items were not available in any of the searched databases and data were obtained from analysis in our lab [16].
Nutrients and recommendations
Added sugar was estimated by including all sugars used in any type of processed or prepared food (recipes), sugar naturally present in food was excluded from the analysis [8]. From a total of 556 identified food items, sugar content was available for 408 of them; either from the USDA or the Mexican food composition databases. For the remaining 148 food items, sugar content was estimated as follows: 58 food items were coded as containing natural sugar, since they were fruit, vegetables, grains, tubers or maize. For the remaining food items, either the sugar content of food items with similar nutritional characteristics was extrapolated (n = 35 food items), food labels were used (n = 42 food items), or the information of standardized recipes was used to quantify the sugar content (n = 12 food items). The sugar content of one food item (“Chicha de Jora”) could not be determined and was excluded from the analysis as it was consumed by only one participant. The following sources of added sugar were identified as described by Wesh et al.: Sweets (candies and gums, soda, added sugar and syrups, fruit flavoured drinks, pre-sweetened coffees and tea, sport drinks and energy drinks), grains (cake and cookies, ready to eat cereals, bread and muffins and other grains), fruits and vegetables, dairy (dairy desserts, milk, yogurt, and other dairy), protein sources (meat, egg and beans) and fat and oils [8].
Macronutrients and added sugar energy percentages per day (E%/day) were calculated by dividing the energy of each variable by the total energy intake per day. Macronutrient E% and sodium intake were compared with Dietary Reference Intakes (DRIs) for sex and age [17]. Since a clear definition of recommended added sugar intake is not available, we used the US recommendation of consuming <15% E% intake of solid fats and added sugar [18]. Therefore the variable solid fat was generated as described by the U.S. Department of Agriculture: all excess fat from the milk and meat and beans and solid fats added to foods in preparation or at the table, including cream, butter, stick margarine, regular or low-fat cream cheese, lard, meat drippings, cocoa, and chocolate [19]. The energy of added sugar and solid fats was summed up, expressed as E%/day and compared with the threshold of 15%. Total fruit and vegetable consumption were compared with the 400 g/d recommendation of the World Health Organization [20].
Food groups
The food groups used in this study were based on the classification as proposed by the Health Department of Mexico [21] since this classification is in concordance with the objectives of this study and both the ingredients and recipes were comparable to those in Ecuador. After adapting this classification we identified a total of 20 main food groups: (i) white rice, (ii) refined wheat (bread, pasta, wheat powder), (iii) other refined cereals (tapioca, maize powder, banana powder and any other kind of powder different from wheat), (iv) whole grain cereals (quinoa, oat, barley), (v) maize, (vi) tubers, (vii) plantain, (viii) legumes, (ix) total fruit (including raw fruit and fruit used in juices or any other preparation), (x) vegetables, (xi) poultry, (xii) red meat (including processed meat), (xiii) fish and seafood, (xvi) dairy products (milk, yogurt, flavored milk, cheese), (xv) oilseeds (nuts, peanuts, almonds), (xvi) vegetable oils, (xvii) animal fat (butter, mayonnaise, crackling), (xviii) coffee, (xix) spices and (xx) processed food rich in salt, fat or added sugar included the following subgroups based on the Mexican classification which comprised (a) table sugar and sweets (honey, candies, chocolates, ice creams, sweet cookies, traditional sweet desserts and sugar added to juices, coffee, etc.), (b) salty snacks and fast food (all packaged salty snacks, salty cookies, French fries, pizza, hamburgers) (c) soft drinks (soda, artificial sweetened juices, energy drinks) and (d) any other packaged food (ketchup, packaged soups, gelatine). Food groups are presented as E%/day.
Food groups considered as protective (fruit, vegetables, oilseeds and fish) [7, 9] against CVD were further analysed in subgroups. As the type of fish could be important in determining this protective effect [22], we divided the reported fish consumption by processing methods, i.e. (i) fresh fish including steamed and roasted fish, (ii) fried fish and (iii) canned fish. As preparation methods may play a role in nutritional value [23], the most common preparation methods of fruit, vegetable, and oilseeds were also identified.
Mealtimes
Mealtimes were defined as: breakfast, morning refreshment, lunch, afternoon refreshment, dinner and evening refreshment. The schools in this study have morning and afternoon schedules. In general, morning schools have classes from 7:00 until 13:00 and afternoon schools from 12:00 to 18:00. Mealtimes during the week were defined in accordance to these school hours. For morning schools the times were set as follows: breakfast between 5:00–7:00, morning snack between 7:00–13:00, lunch from 13:00–16:00 and afternoon snack from 16:00–18:00. For afternoon schools the timings were breakfast from 5:00–8:00, morning snack from 8:00–11:00, lunch from 11:00–12:00, afternoon snack from 12:00–18:00. Dinner and night snack were set equally for the whole sample from 18:00–21:00 and any hour later than 21:00 respectively. The weekend’s timing was set equally for all the participants. They were as follows: breakfast from 5:00–9:00, morning refreshment from 9:00–12:00, lunch from 12:00–15:00, and afternoon refreshment from 15:00–18:00.
Cardiovascular risk factors
Body weight, height and waist circumference were measured in duplicate by trained researchers using calibrated equipment. Body mass index (BMI) was calculated as weight/height2 (kg/m2). Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were measured on-site by trained staff in triplicate after a 10-min seated rest using a portable sphygmomanometer. A fourth measurement was taken if initial values were above 120/80 mmHg.
Blood samples were collected after overnight fasting. Blood serum was used to determine the total lipid profile. Total cholesterol (COLT), triglycerides (TG), and high-density lipoprotein cholesterol (HDL) were quantified using standardized methods [14]. For Low-density lipoprotein cholesterol (LDL) calculation the Friedewald formula [24] was used. Precision and accuracy were estimated: the global analytical coefficient of variation for COLT and TG was 3.1% and 5.7%, respectively. The % bias ranged from 0% to 8%.
Socioeconomic status
Sociodemographic characteristics were assessed using the tool developed by the Integrated Social Indicator System for Ecuador [25]. This system defines poverty based on Unsatisfied Basic Needs (UBN). A household is considered as poor if one or more deficiencies in access to education, health, housing, urban services and employment are reported. Therefore adolescents were allocated to one of two groups; “Poor group” if at least one deprivation was present or “Better-off group” if no deprivation was reported.
Data analysis
Food intake data was entered using an online software designed to analyse 24-hour recall data (Lucille software 0.1, 2010, Gent University; http://www.foodscience.ugent.be/nutriFOODchem/foodintake). Anthropometric, socioeconomic and blood lipid data were entered in duplicate into Epidata by two independent researchers. Any discrepancy was corrected using the original forms. Data are reported as mean with standard deviation (SD) or as median with 25–75 percentiles. When appropriate, continuous variables were transformed into a normal distribution. A significance level of 5% was determined for all statistical tests. Energy intake of macronutrients and food groups was adjusted for total energy intake using the nutrient residual model [26].
Factor analysis was carried out to identify dietary patterns including twenty food groups expressed as E%/day. The number of factors retained was based on a scree-plot. Food groups with a loading factor below 0.10 were removed from the analysis. Foods with a factor loading above 0.3 were identified as the main contributors to each pattern [27]. A score of each dietary pattern was calculated for each participant and split into tertiles. For each tertile of the dietary pattern score, the median of the macronutrients E%, added sugar and sodium contribution were reported as well as the mean of the cardiovascular risk factors.
Linear regression models were used with two purposes: (i) to identify differences in energy intake, macronutrient E%, added sugar, sodium and food group energy sources, by UBN and place of residence (ii) to compare, macronutrients E%, added sugar, sodium and cardiovascular risk factors across the tertiles of the dietary patterns score. The comparisons of dietary components and cardiovascular risk factors among the dietary patterns score tertiles were split into subgroups (UBN or place of residence) only if interactions were significant. Outcomes not following a normal distribution were log transformed before inclusion in the models, and beta coefficients were back transformed and expressed as percentage differences (estimate-1*100). All the models were adjusted for sex, UBN and place of residence when necessary.
Logistic regression models were used to determine UBN and place of residence differences in: (i) the amount of participants exceeding the macronutrient, sodium and added sugar recommended intake per day, (ii) the number of consumers of protective groups (subgroups) such as fish (roasted, fresh, canned) and fruit (preparation methods) as well as in (iii) the number of consumers of the main sources of added sugar. All the models were adjusted for sex, UBN and place of residence. Results are reported as odds ratio (OR) and 95% confidence interval (CI).