Absence of association between dietary habits and adiposity among primary school children in Kilimanjaro Tanzania | BMC Nutrition
From August to November 2019, we conducted a school-based cross-sectional study in 20 primary schools in the Kilimanjaro region of northern Tanzania. Kilimanjaro is one of the 31 administrative regions of Tanzania and comprises seven administrative districts. Two districts, Urban Moshi and Rural Moshi, were deliberately selected to represent urban and rural settings. Moshi urban has 48 primary schools (35 public schools and 13 private schools) and 33,207 students. Rural Moshi has 272 primary schools (252 public: 20 private) and 81,297 students.
A multistage sampling technique was used to select study participants. We randomly selected eight neighborhoods from each of the two districts. The primary schools in the selected neighborhoods were stratified by type of school (public or private). Simple random sampling was used to select 20 primary schools (10 from each district). Enrollment in public or private schools was considered a proxy for lower or higher socioeconomic status. We have excluded children from boarding schools as they generally follow the same school menu. A list of all students aged 9 to 11 in grades 4 to 6 was obtained from school attendance records. Study participants were randomly selected from each school with a probability proportional to the size of the schools on the list. This age range was established based on the assumption that children aged nine or older can express themselves and complete questionnaires with guidance. Nevertheless, we excluded 24 children who could not understand the questions and express themselves.
Sample size estimation
Sample size estimates were based on the primary objective of the study: to estimate the prevalence of overweight and obesity . An accuracy of ±2.5% and a design effect of 1.3 were assumed since we recruited participants from clusters (schools). The final sample size was 1170 primary school children.
Dietary intake assessment
Pilot study: 24-hour dietary recall
We first conducted a pilot study to explore the foods usually eaten by children, using a 24-hour food recall with 51 randomly selected children (9-11 years old) from four primary schools, two of each district (urban/rural). A trained nutritionist asked the children to recall their food and drink consumption over the previous 24 hours. We followed the four-step multi-pass technique to obtain information on detailed foods eaten in the past 24 hours. . Information on snacking habits, particularly during recess at school, at lunchtime, on the way home and at friends’ homes, was used to obtain specific details on sugary drinks, snacks and low-calorie, low-nutrition snacks. Food models were used to help estimate portion sizes . We used the Tanzania Food Composition Table (TFCT)  to assign nutritional values and estimate the energy and nutrient intake of different foods. For processed foods mentioned by children who were not on the TFCT (e.g. Pringles [dehydrated processed potato crisps], and sausages), we obtained information from food labels where possible. The nutritional information for each food was calculated per 100g and then converted to the amount consumed based on self-report. For children who could not estimate their portion sizes, standardized portion sizes from the TFCT were used to estimate their intake.
Food frequency questionnaire
For the main study of food intake, the results of the 24-hour recall were used to adapt and modify the ISCOLE FFQ . This FFQ standard has been used elsewhere to explore the association between diet and obesity [25,26,27]. We have removed some foods that did not apply to Tanzanian children (e.g. skimmed milk, cheese, energy drinks: “Redbull, guru”, sports drinks “Powerade”) and added local foods/snacks commonly reported by children (eg, samosas, crisps, fried cassava, fried plantain, flavored popsicles (usually made with tamarind, sugar, and food coloring), sweetened squeezed juices, mandazi, milk, and dairy products ( whole cow’s milk and cultured milk)). The children were asked about their usual consumption of different foods during a typical week. Scores were generated from the FFQ responses for servings per day as follows: never = 0 servings/day; once a day = 1 serving a day; once or twice a week = 0.21 servings per day; 3-6 times per week = 0.64 servings per day; and 2-3 times a day / every day more than once = 2.5 servings a day . Trained research assistants guided the children step-by-step through completing the FFQ to ensure they understood each food, its contents, and responded independently.
Define food groups and subgroups
We used information on local foods and snacks reported by children during the 24-h dietary recall to estimate the contribution of different foods to total fat, carbohydrate, and protein in the diet. Additionally, to create food categories, we grouped the 15 foods in the FFQ into nine foods/food groups based on their nutritional composition and culinary uses (percentage contribution in children’s diets). For example, we have grouped the consumption of chocolates, cakes, sweets, cookies and donuts into sweets and sugars. Food groups/subgroups were used to identify eating habits.
The results of our study were five measures of adiposity: BMI z-scores, waist circumference, body fat percentage by bioelectrical impedance, triceps, subscapular skinfolds; and associations between dietary habits and measures of adiposity.
Anthropometric measurements were performed in duplicate according to standard procedures of the National Health and Nutrition Examination Survey (NHANES) , using calibrated equipment. Before taking measurements, children were asked to remove shoes, socks, hair ornaments, objects from pockets, jewelry, and clothing other than their usual school uniform. Two trained research assistants took independent anthropometric and adiposity measurements of each child, and the average between the two measurements was calculated. We regularly check standard operating procedures and evaluate technical error of measurement (TEM)  to estimate inter-observer measurement errors.
Height was measured to the nearest 0.1 cm using the TANITA height board with children standing straight with their heads on the Frankfurt plane. Weight and percentage of total body fat were measured using Bioelectrical Impedance Analysis (BIA) technology, model TANITA DC 430 MA. The BIA machine was set to deduct 0.5 kg (clothes weight) before measurements.
Waist circumference was measured to the nearest 0.1 cm using a non-elastic circumference tape. The children were asked to stand straight with the abdomen relaxed, the arms at the sides, the feet together and the measurement was taken from the narrowest circumference. For obese children in whom we could not identify the point of measurement, we asked them to lean to the side, and the measurement was taken from the point where the trunk bends.
Triceps and subscapular skinfolds were measured using a Harpenden precision caliper to within 0.1mm, on the right arm . The triceps skinfold was measured at the mid-back of the upper right arm. The subscapular skinfold was measured sub and laterally at the angle of the scapula with the shoulder and arm relaxed.
The analysis was performed by STATA version 15.1 (StataCorp, College Station, TX, USA). Probability plots were used to assess the distribution of variables and check for normality. Continuous variables were summarized using means and standard deviations for normally distributed data, and medians and interquartile ranges for skewed data. For categorical variables, frequencies and proportions were reported. WHO Anthroplus (STATA SE) was used to determine BMI z-scores by age and sex; slimming +1 and
Based on the scores generated by the FFQ, we performed factor analysis, a data-driven method for deriving eating habits. The factor has been loaded and rotated using varimax rotation to simplify factor interpretation. The correlation between foods/food groups was examined using a Kaiser Meyer Olkin test (KMO) = 0.83, indicating that the correlation between the variables was strong enough for factor analysis. We used a scree plot to examine the distribution of factors and select the factors to retain. We retained the factors whose eigenvalues are greater than 1; factor 1 = eigenvalue 2.53 and factor 2 = eigenvalue 1.01 [Fig. 1].
Variables (foods/groups) with an absolute loading factor ≥ 0.3 , were retained and used for the labeling of food habits. When foods have a charge greater than 0.3 in more than one factor, the factor with the highest charge was considered for factor labeling. The children were classified into terciles of low, medium or high adherence to each diet. In addition, chi-square test, analysis of variance (ANOVA), and Kruskal Wallis test were performed to explore participants’ characteristics across diet terciles.
Variables potentially associated with eating habits were selected for adjustment based on a recent overweight/obesity study conducted in Tanzanian primary schools . In Model 1, we adjusted for age and sex. Model 2 adjusted for lifestyle factors: type of school (private or public), time spent walking to school, neighborhood (rural or urban), availability of television and electronic gadgets at home and neighborhood playground.
Next, we examined the association between identified dietary pattern terciles and different measures of adiposity using multilevel linear regression. Dependent variables were logarithmically transformed before linear regression analysis due to nonnormal distributions. All analyzes were two-sided and the significance level was set at 5%.