A hierarchical Bayesian tri-variate analysis on factors associated with anthropometric measures in a large sample of children and adolescents: the CASPIAN-IV study

Introduction

Non-communicable diseases (NCDs) such as cancer, diabetes and other cardiovascular and metabolic (cardiometabolic) conditions are the leading cause of death in the world [1]. Obese children are at increased risk of mortality and morbidity resulting from cardiovascular disease (CVD) in adulthood [2]. According to the report by the Global Burden of Disease Study, over one quarter of cases of obesity in childhood may co-exist with metabolic syndrome in developing countries [3], [4]. Furthermore, on the basis of the report by the Global Burden of Disease Study, developing countries will be affected the most by obesity and its related NCDs within the next two decades [5]. Obesity is a challenging health problem among pediatric populations, both in developed and developing countries, because of its impact on different aspects of health including physical, mental and social functioning, quality of life as well as healthcare costs [6], [7]. The health consequences of obesity in childhood highlight the importance of identification, prevention and controlling its modifiable risk factors [8], [9].

Body mass index (BMI), is an accepted tool to evaluate body fat in children [10]. However, BMI is a crude measure which cannot distinguish fat from muscle mass and present distribution of body fat. On the other hand, some factors such as height and level of sexual maturation, may distort the association between BMI and body fat in children [11]. Therefore, other anthropometric measures, such as waist circumference (WC), waist-to-height ratio (WHtR) and wrist circumference (WrC), have been proposed in the literature [12], [13]. WHtR is simple and inexpensive for measuring central adiposity, which is associated with cardiometabolic risk factors even in children [14]. It is a more sensitive universal screening tool than BMI to detect health risks [12]. WrC is a measure of skeletal frame size that is easy to measure and is less confounded by body fat variation. In addition, it is a specific measure of upper body fat in young people [15]. Adults with large WrC have excess morbidity, including diabetes and CVD risk factors [16].

The coverage of behavioral, biological, social, environmental and systems-level factors contributing to obesity requires multi-sectoral approaches across the life course [17], [18], [19]. Some factors such as breastfeeding and birth weight, genetic predisposition including family history of obesity and psychosocial disorders such as depression and socioeconomic status may be attributed to the increasing childhood obesity [20], [21], [22], [23].

Like many other developing countries, the prevalence of obesity in childhood has an escalating trend in Iran [24]. This study aimed at identifying factors associated with anthropometric measures including BMI, WHtR and WrC in a nationally representative sample of Iranian pediatric population.

Materials and methods

Results

Overall, 13,280 students were included in the study, 6543 (49.2%) of which were girls and 10,037 (75.6%) were urban residents. Tables 1 and 2 present descriptive statistics and compare means in sex groups, where girls and boys were different in all variables except in sleep hours. Overall, 12.2% (13.0% of boys and 11.2% of girls) were underweight, 9.6% (9.2% of boys and 10.0% of girls) were overweight, and 11.6% (13.2% of boys and 9.9% of girls) were obese. Abdominal obesity (WHtR ≥0.5) was observed in 20% (21.3% of boys and 18.7% of girls) of the studied population. The results of multivariate multilevel analysis are shown in Table 3. As demonstrated by 95% HPD regions, SES, family history of DM, screen hours, sleep duration, breakfast consumption status, birth weight and residence area had significant direct effect on all three measures of obesity. PA and breastfeeding had, respectively, inverse and direct effects on both BMI and WrC. A family history of CVD and being a passive smoker had significant direct effect on WHtR. History of psychological disorders and the use of junk food increased BMI, only.

Discussion

In this national study on a large sample of 6–18 year-old Iranian schoolchildren, some significant modifiable risk factors on anthropometric measures were determined in both genders. The results provide updated references on the risk factors associated with BMI, WrC and WHtR in the pediatric age group. The following are significantly associated with the three aforementioned obesity indicators, each of which aims to measure a special feature of obesity; Family history of obesity and diabetes, screen hours, SES, sleep hours, eating breakfast, weight at birth and residence area. Some of these factors are modifiable and should get higher priority in health education and programs.

The prevalence of obesity varies in different parts of Iran and age, sex and urbanization are the most important determinants [31], [32]. Childhood obesity is increasing in developing countries due to, mainly, high socioeconomic status and welfare without adopting proper lifestyle [32], [33].

It is proposed that factors such as growth, development and body shape of people living in a same habitat tend to be more similar than those selected randomly from larger populations due to, perhaps, a common impoverished or developed environment [34]. When lower level units are nested within one or more higher level strata, conventional single level regression analysis is not appropriate because observations are no longer fully independent and such dependency may lead to false inferences [34]. This may be important for health applications, for instance, if the impact of individual level risk factors varies according to geographic context. Therefore, in this study we chose individuals and provinces as the first and the second levels, respectively.

The observed associations between BMI and age, SES, family history of obesity, family history of DM, PA, screen time, duration of sleep, psychological disorders, breakfast consumption, birth weight, breastfeeding, junk food and place of residence (urban-rural) were stronger than those reported in other studies [21], [35], [36]. However, we found no association between BMI and sex, family history of CVD, family history of dyslipidemia and passive smoking.

Previous studies suggest that imbalance in energy expenditure is related to behaviors such as low PA and poor diet that results in obesity in children [8], [9], [35], [36]. Our findings showed a negative relationship between PA and BMI and WrC. In a study on determinants of BMI, it has been reported that weight was significantly related to PA, especially for girls and healthy-weight persons were more active than overweight, obese, and very obese girls [35]. In a report on 6–19 year-old subjects, a significant association was observed between high PA and lower odds of overweight and central fat in adolescents based on BMI and WHtR [37]. Similar results have been observed in other studies [36].

Dietary patterns have changed in recent decades and increased consumption of junk food is claimed to be associated with increasing childhood and adolescent obesity [8], [9]. Unhealthy dietary habits, such as increasing consumption of junk food (soft drinks, snacking, sweets and chocolates) increase overweight and obesity in adolescents [9]. In this regard, we found a positive association between BMI and junk food consumption which confirms previous findings. Junk food consumption has been reported to be associated with an increased risk of being overweight amongst 6–11 year-old children [37]. However, there are some contradictory reports on the association between sweets intake, and junk food and the risk of being overweight where the sweets intake frequency was lower in overweight youth in comparison with normal weight youth [8].

We observed reverse relation between breakfast consumption and all three anthropometric measures. A study on Australian children (5–16 years) showed that children who did not consume breakfast daily were more likely to be overweight or obese. There was also a significant positive association between not having breakfast daily and WHtR [38]. Consumption of breakfast has an important role in maintaining normal weight in childhood and adolescence as the risk of being overweight has been reported to be lower among who have breakfast regularly than those who skip it [18]. People who miss breakfast are more likely to consume greater amounts of food at lunch time and this may lead to more weight gain [18].

Smoking is another factor associated with metabolic disorders such as obesity; however the relationship is complicated as it connects with both lower and higher BMI [17], [39]. Smokers usually have lower BMI and larger WrC in comparison to non-smokers [39], [40]. However, these associations are controversial among adolescents [39]. Saarni et al. [39] argued that smoking status in adolescents could lead to abdominal obesity among young adults. In the current study, we only evaluated passive smoking which had a significant effect on WHtR.

There has been growing interest in determining the role of sleeping status in weight changes. The National Sleep Foundation recommends night sleep duration of 9–11 h for children aged 6–13 years and 8–10 h for teenagers [41]. Sleep duration of ≤6 h/night has been reported to be associated with higher BMI [19]. As reported by several studies, short sleep duration may be a risk factor for increased BMI in both children and adults [21]. According to a results of longitudinal study of NHANES, 7 h of night sleep is even insufficient [19]. In accordance with previous studies, our results showed that sleep hours were negatively associated with all three anthropometric measures. Longitudinal study on Mexican American children (8–10 year-olds) showed that children with lesser sleep hours were more likely to have a higher BMI and WHtR [42].

In the current study, we observed that screen time had a positive relationship with BMI and WrC. Some studies have revealed a dose–response relationship between the hours of watching TV or using computer and prevalence of obesity in children [22]. Children who watch TV are exposed to more consumption of less nutritious junk food such as salty snacks (chips, cheese curls, popcorn and pretzels), sweets (biscuits, cookies, cakes, chocolate and candies), sweetened beverages (soda, soft drinks) and fast food (hot dogs, hamburgers, cheeseburgers, fried chicken and pizza) which increase obesity risk [22]. Having a TV in the bedroom and not having limits set on screen-time were factors associated with overweight and obesity among children and adolescents [38].

The results of our study showed a positive relationship between SES and anthropometric measures. There are some reports on the reverse association between SES and rates of inactivity and obesity. Freedman et al. [23] reported a positive association between SES and childhood overweight, but this relationship was different between races or ethnicities. The results of Nasreddine et al.’s study document significant associations between certain parental socioeconomic characteristics and adiposity amongst 6–11 year-old children but not amongst adolescents. These findings highlight the importance of the home environment in adjusting the lifestyle and dietary habits and hence obesity risk in children [37].

Psychological disorders, such as depression, could contribute to the development of obesity through unfavorable health behaviors, such as physical inactivity [20]. Depression may increase the risk of obesity by triggering biochemical changes in the body [20]. In contrast to non-significant findings from previous studies in young adults, we observed a direct association between psychological disorders (depression, anxiety and insomnia) and BMI in this population [20].

Birth weight is known as a risk factor of both childhood and adulthood obesity [43], [44]. In this study, we observed a positive relationship between birth weight and all three anthropometric measures. Some studies have shown that breastfeeding is protective against childhood overweight and obesity and reduces the risk by 20%–60% depending on the duration of breastfeeding [45], [46]. Nevertheless, results of our study showed that breastfeeding has a positive effect on both BMI and wrist measures. A recent review strongly supports the relationship between exclusivity of breastfeeding and lower risk of obesity [47].

In the current study, we applied Bayesian multivariate multilevel model to assess factors associated with three anthropometric measures (BMI, WHtR and WrC) in a large sample of school-aged children. The current study had some limitations, too. First, due to its cross-sectional nature, a causal relationship between lifestyle and obesity could not be determined and the observed associations indicate that possibly more reliable evidence from longitudinal and interventional studies is warranted. Second, we used self-reported questionnaires and bias may be present due to inaccurate self-reporting or misunderstanding of the questionnaire items. Third, we gathered information on frequency of dietary patterns and estimated the total calorie intake.

In the present study, obesity in children was significantly associated with breakfast consumption, PA, screen time, duration of sleep, junk food, SES and breastfeeding. Early interventions seem to be essential for prevention of childhood obesity and associated complications. Intervention strategies could include the increase of PA and consumption of fruits, vegetables and breakfast and restricting junk food and TV or screen time. In addition, family-based interventions are needed to instill healthy lifestyles and dietary habits in children.