Study participants
Children aged between 5 and 10 years were recruited from public and private schools in the Greater Beirut area, Lebanon, following a random cluster sampling design. Fourteen schools were randomly selected from the Ministry of Education list of schools in the Greater Beirut area (n = 32). In the selected schools, all students aged between 5 and 10 years received an invitation letter to participate in the study, addressed to their parents. In this letter the eligibility criteria (inclusion/exclusion) to participate in the study were outlined. These criteria were: Parents holding the Lebanese nationality or residing in Lebanon for more than 10 years, and children to be healthy (i.e., with no medical conditions, allergies nor specific dietary restrictions affecting food intake). Interested parents were contacted by trained research dietitians and were enrolled in the study between October 2011 and June 2012. Data collection was conducted between October and July, hence covering the various seasons in Lebanon. The study protocol was approved by the Institutional Review Board of the Social and Behavioral Sciences at the American University of Beirut, Lebanon. Participating parents gave a written informed consent and children above the age of seven also signed an informed assent to indicate their approval to participate in the study. All parents who signed the consent form were the mothers of the children.
Study protocol
Participants were enrolled in the study for a period of 4 weeks (Fig. 1), during which two face-to-face interviews took place at the Nutrition and Food Sciences department at the American University of Beirut. In the first face to face interview with the child and his/her mother, a socio-demographic questionnaire and the first FFQ (FFQ-1) were completed. In addition anthropometric measurements of the child were obtained. After four weeks, another face to face interview took place with the mother and the child whereby a second FFQ (FFQ-2) was completed. Four 24-HRs of the child’s diet were collected by phone. These 24-HRs were one week apart between the two interviews. For both dietary intake collection methods (FFQ and 24-HR), information was collected from the mother, in the presence of the child. For each participant, all interviews (face to face and phone) were carried by the same research dietitian. Mothers were constantly encouraged to maintain their children’s regular dietary habits. Probing and interviewing techniques were standardized to minimize interviewer bias.
Out of the 171 mothers who returned the signed invitation letter indicating their interest to participate, 7 were excluded as they were not eligible according to the inclusion/exclusion criteria of the study. Out of the remaining 164 mother-child pair, 125 showed up for their first visit. The total number of participants who completed the study was 120 (dropout rate: 4 %). The main reasons for not completing the study were lack of time and interest. This sample size is considered appropriate for studies validating dietary intake tools [10].
Data collection
Socio-demographic questionnaire
The socio-demographic questionnaire included information about the mother’s age, education, marital status, occupation, and crowding index. In addition, the child’s age, gender and school class level were recorded.
Anthropometric assessment
Anthropometric measurements of the child were obtained using standardized techniques and calibrated equipment. The InBody 230 (Biospace, Korea) and stadiometers (Seca model 213, Germany) were used to measure weight (in kg) and height (in cm), respectively. Before stepping on the InBody 230, children were asked to remove as much outerwear as possible, as well as their shoes and socks and to empty their pockets. Subjects were weighed to the nearest 0.1 kg. Height was measured to the nearest 0.1 cm with the child bare footed, using a stadiometer. Body Mass Index was calculated as weight (kg)/height (m2). All measurements were carried in duplicates and the average was used in the analysis.
Dietary intake assessment
A semi-quantitative FFQ was developed to assess dietary intake among school-aged children in Lebanon. The FFQ included three sections: the food list, the portion size and the frequency response. A multitude of approaches was followed in order to compile the food list for this questionnaire: 1) a review of previously collected 24-HRs data pertaining to a national representative sample of Lebanese children 6–10 years old was carried out (n = 200) [22]. Frequently cited food items in the 24-HRs (>5 %) were included in the food list of the developed FFQ, 2) the developed FFQ’s food list was checked by 30 mothers of 5–10 year old children for clarity, user-friendliness as well as the cultural sensitivity/appropriateness of the food items included, and 3) the food list was also compared to previously published FFQs aiming to assess dietary intake among young children. The final FFQ food list included a total of 112 food items.
For portion size, mothers were given the option to indicate their child’s food intake in function of a reference portion size or in grams. The reference portion for each food item represented one standard serving expressed in household measures (cups, spoons and plates) and/or customary packing size. In order to assist in quantifying the reference portion size, the standard two-dimensional food portion visual chart was also used. This chart has been developed by Nutrition Consulting Enterprises and validated for use amongst adult men and women aged 20 to 70+ years as part of the Framingham Heart Study, especially for the use of telephone dietary interviewing [23]. The frequency of the child’s food intake was indicated by how many times per day, week or month the child has consumed the food. Seasonal adjustments were considered for food items that were eaten at specific times of the year. The reported frequency of consumption for these seasonal foods was adjusted for the length of the specific season they were consumed within, in order to obtain frequency of their consumption over the course of a year. For all food items in the FFQ, the frequency per day was multiplied by the portion size of the food item in order to calculate the total amount of food consumed per day.
In addition to the 112 food items, the FFQ included an open-ended section in which participants may provide information on additional foods or beverages consumed on a regular basis and which do not appear on the questionnaire’s food list. When completing the FFQ, mothers were asked to refer to their child’s diet during the 12 months prior to the interview. The completion of the FFQ lasted for approximately 30 min.
The 24-HRs dietary recalls were carried using the Multiple Pass Food Recall (MPR) 5-step approach, developed by the United States Department of Agriculture (USDA) [24]. This approach has consistently showed attenuation in the 24-HRs’ limitations [24, 25]. The five steps followed included 1) quick food list recall, 2) forgotten food list probe 3) time and occasion at which foods were consumed, 4) detailed overall cycle and 5) final probe review of the foods consumed. The four 24-HRs collected represented three weekdays, in which the child attended a regular day at school, and one weekend day (either Saturday or Sunday). For each 24-HR, the research dietitian obtained information related to the time of each meal’s intake, the food consumed by the child, its portion size, preparation methods, and the brand of the food and beverages consumed, if applicable. In order to maintain regular eating habits for their children, mothers were unaware of the day the 24-HR would be conducted. Upon calling, the interviewer made sure that the child was present with the mother during the completion of the 24-HRs as the child was probed and asked details of his/her intake in case of foods he/she has eaten away from home such as in the school. The mean of dietary intakes estimated by the four 24-HRs were used as the reference method against which the FFQ was validated.
The Nutritionist Pro software (version 5.1.0, 2014, First Data Bank, Nutritionist Pro, Axxya Systems, San Bruno, CA) was used for the analysis of the dietary intake data and to estimate energy, macro- and micronutrients’ intakes. For composite and mixed dishes, standardized recipes were added to the Nutritionist Pro Software using single food items. Within the Nutritionist Pro, the USDA database was selected for analysis (SR 24, published September 2011). Food composition of specific Lebanese foods (not included in the Nutritionist Pro software database) was obtained from local food composition tables [26].
Statistical analysis
Frequencies and percentages as well as means and standard deviations (SD) were used to describe categorical and continuous variables, respectively. For the validity assessment of the FFQ, dietary intakes derived from the FFQ-1 were compared to the mean of the four 24-HRs using energy-adjusted Spearman correlation coefficient (r). Energy adjustment was carried out using the residual method as per Willet et al. [27]. Distribution of the study participants according to quartiles of intake was calculated and the degree of agreement between FFQ-1 and mean 24-HRs was evaluated using contingency tables of quartiles. Furthermore, the analysis proposed by Bland & Altman [28] was used to graphically examine the agreement between the two methods. For this analysis, the difference in intake between the two methods (FFQ-1 and mean 24-HRs) was plotted against the mean intake of the two measures ((FFQ-1+ mean 24-HRs)/2). The plots include lines for the mean difference and the Limits of Agreement (LOA), defined as mean difference ± 1.96 x SD. The calibration of the FFQ was conducted as per the method described by Araujo et al. [29] and which involved the derivation of calibration coefficients relating dietary intakes estimated from the reference method (mean 24-HRs) to the test method (FFQ-1). These coefficients are obtained from linear regression equations using dietary intake values from mean 24-HRs as dependent variables and those of the FFQ-1 as independent variables. The regression constant (α) and the slope of regression (β) were estimated. The calibrated values for each nutrient were estimated based on α and β coefficients using the following formula:
$$ \mathrm{Calibrated}\ \mathrm{dietary}\ \mathrm{intakes} = \upalpha +\upbeta \kern0.5em \mathrm{F}\mathrm{F}\mathrm{Q}\hbox{-} 1 $$
where, α: regression constant; β: slope of regression; FFQ-1: Dietary intakes as estimated by FFQ-1.
For the reproducibility of the FFQ, Spearman correlation (adjusted for energy) between FFQ-1 and FFQ-2 was used. In addition, the Intraclass Correlation Coefficient (ICC) was calculated to examine the agreement between FFQ-1 and FFQ-2 in ranking individuals according to their energy, macro- and micronutrients intake. The degree of agreement between FFQ-1 and FFQ-2 was further evaluated using contingency tables of quartiles and weighted Kappa (κw) test with values of κw between 0.40 and 0.59 considered moderate, 0.60 to 0.79 substantial and 0.8 outstanding as per Landis & Koch [30].
Data entry was carried out using Statistical Package for Social Sciences 22.0 (SPSS for Windows, 2013, Chicago: SPSS Inc.). A p-value less than 0.05 was considered statistically significant. In this manuscript, the results for energy and selected nutrients are presented (energy, proteins, carbohydrates, fats, monounsaturated fatty acids (MUFA), polyunsaturated fatty acids (PUFA), trans-fatty acids, calcium, iron, fiber, and sugar). Results pertinent to the remaining micronutrients are found in Additional file 1.