Risk factors associated with overweight and obesity among urban school children and adolescents in Bangladesh: a case–control study

We conducted this study in seven schools located in Dhanmondi, Mohammadpur and Siddeswari area in Dhaka, Bangladesh during October through November 2007. The highest numbers of schools in Dhaka are situated in these places. We selected these schools based on three criteria: large number of students, having reputation for better education facility, and availability of play-ground within school premises. Each of seven schools was private and English was the medium of teaching in three of seven schools.

The study was designed as case–control study with cases being overweight children^a and controls being healthy/normal weight children. The study participants were students of age 10-15 years corresponding to class 5 to class 10 in the selected schools.

We adopted purposeful sampling method to select the cases and controls. The teachers at each school made initial selection of 12–15 cases by visually assessing whether the students were overweight and also selected equal number of controls who visually seemed normal weight. Then, research assistants used standard procedures of anthropometry to measure standing height and weight (including shoes) of each primary selected participant and calculated body mass index (BMI) as weight in kg/height in meter². Investigators used age- and sex-specific growth chart by the US Centers for Disease Control and Prevention [22] to categorize initially selected participant as cases or controls. Any participant falling into the underweight category according to growth chart was excluded from the study. We selected equal numbers of cases and controls from each school.

We used a structured questionnaire to collect demographic information and information on exposures of interest: regular habit of physical activities, duration of sedentary activities which included watching television, use of computer and video games, and duration of sleep/day of each study participant. In order to know regular habit of physical activity, we asked participants to report how long did they spend every day in a typical week doing activities which require physical activity. These activities included playing games such as football, cricket and other outdoor games, running in field and cycling. WHO recommends to spend at least 60 minutes of physical activity per day for children aged 5–17 years [23]. Since we conducted this study among school children who might be involved in physical activities during their school stay, we split this recommended duration of physical activity per day into two bouts, in school and at home. We assumed that splitting the physical activity habit into school and home would give us the opportunity to have better understanding of actual physical activity pattern of students and also to help determining if any particular habit serves as risk factor for overweight problem in children. We asked the respondents to report the actual physical activity habit both in school and at home and we added to find the duration of total physical activity. We recoded the response into following categories: <60 minutes and ≥60 minutes per day for total physical activity and no time, <30 minutes and ≥30 minutes per day for physical activity at home and school. Based on the median duration of sedentary activities per day of all participants, we categorized length of sedentary activities into two groups. We communicated over the phone with parents of each participant to obtain respondent’s (case or control child) birth weight, parental weight and height to estimate BMI, and highest level of education completed by mother.

We derived descriptive statistics to determine the distribution of demographic information including age, sex, education level, sleeping time, mother’s highest level of education and BMI among case and control groups. For continuous data, we estimated mean (standard deviation, SD) if the data were normally distributed or estimated median (interquartile range, IQR) otherwise. In order to compare mean or median of variables between cases and controls, we used t-test or Wilcoxon-rank sum test, respectively. We used Z-test to compare proportions between cases and controls. Initially, we determined association between outcome (overweight/obese) and exposure variables using simple logistic model (unadjusted model). Any exposure variable in simple logistic model with a beta coefficient at significant level 0.3 was selected for the multiple logistic regression model (adjusted model) [24]. We considered a p-value of <0.05 as significant level in multiple logistic regression model.

We obtained verbal informed consent to enroll students from the principals of respective schools followed by verbal consent from the parents of each participant before interviewing the participant. We obtained ethical clearance from BRAC University for conducting this study.

From October through November 2007, we enrolled 198 students in this study; 99 cases and 99 controls. The participants were selected from class 5 to class 10 in order to tally with specific age groups. The mean age of study respondents was 13.1 years (SD: 1.2). Of 198 respondents, 107 (54%) were male. The mean birth weight of respondents was 3.1 kilogram (SD: 0.8). The average sleeping time was 7.5 hours (SD: 1.0) and was similar across two groups. The median sedentary activities per day of all participants was 4 hours (IQR: 4–6). The median monthly household expenditure of respondents’ families was US$522 (IQR: 373–709). Eighty-five (44%) of 198 respondents had at least one overweight parent. One hundred twenty-four (62%) mothers reported to complete graduation degree or more (Table 1).

Unadjusted regression model revealed that three factors were independently associated with overweight or obesity in children (Table 2). Having at least one overweight parent and spending >4 hours each day on sedentary activities such as watching television and/or computer games were risk factors for overweight or obesity. Physical activity at home for ≥30 minutes each day was protective against childhood overweight or obesity. We did not find association between childhood overweight or obesity and duration of total physical activity for ≥60 minutes per day, duration of physical activity at school, sex, or maternal education level and these variables were not included in multiple logistic model. The multiple logistic regression analysis showed that children who had at least one overweight parent were nearly three times more likely to be overweight or obese (OR = 2.8; 95% CI: 1.5–5.2) compared to children whose parents were not overweight, and children who spent >4 hours on sedentary activities each day were two times more likely to be overweight or obese than children who spent less time on sedentary activities (OR = 2.0, 95% CI: 1.1–3.7). Children who spent ≥ 30 minutes each day with outdoor games at home that involved physical exercise had decreased odds (OR = 0.38, 95% CI: 0.1–0.8) of being overweight or obese compared to children who did not exercise at home (Table 2).

Several limitations of this study should be noted. One important limitation was the relatively short duration, 2 months, for data collection which resulted in lower than estimated respondent enrollment which could reduce study power. However, the majority of our findings were consistent with previously published reports, suggesting that lower sample size likely had little impact on our assessment. We did not use any previously validated physical activity questionnaire to measure physical activity of study participants. Use of valid questionnaire could provide robust assessment of physical activity habit and it’s relationship with risk of overweight or obesity. However, the list of activities that we used to measure physical activity of children included those that Bangladeshi children used to do in this context. Yet, the importance of using validated physical activity questionnaire cannot be ignored. We did not collect data on individual games they played at home or in school rather we asked them to report us their engagement in the activities that we included in the list (e.g. football, cricket, and other outdoor games, running and cycling). Therefore, we could not inform if the activities done at home by the healthy children (controls) were likely to be of a higher intensity compare to overweight or obese children (cases). The cases were selected by teacher’s visual assessment. Therefore, it could happen that the most overweight children in the class were selected. However, as we found that the BMI of cases were normally distributed (data not shown), we suspect visual selection of cases might had modest impact on study outcome. Although it would be ideal to assess height and weight of parents, since the interviews were time-consuming and conducted after school hours, we communicated with each parent over the telephone and relied on self-reported anthropometric measurement of parents. Therefore, the possibility of error during measurement of height and weight remained valid. We conducted this study in seven schools in Dhaka city. So, the risk factors that we identified may not be representative of every urban city in Bangladesh.