The ISCOLE project

The International Study of Childhood Obesity, Lifestyle and the Environment (ISCOLE) was designed to investigate the influence of behavioural settings and the physical, social, and policy environments on the observed relationships between lifestyle and weight status among school-aged children from 12 countries around the world [26]. While previous multi-country childhood obesity studies have focused on specific geographic regions (e.g. Europe), ISCOLE was designed to have global representation by including study sites from within high-, middle-, and low-income countries. Consequently, the 12 ISCOLE countries represent five major geographic regions of the world including Africa (Kenya, South Africa), the Americas (Brazil, Colombia, Canada, USA), Europe (Finland, Portugal, United Kingdom (UK)), South Asia (India), and the Western Pacific (China, Australia), with varying economic ranks [27]. The study design also incorporated comprehensive and robust indicators of lifestyle behaviours (e.g. physical activity, sedentary behaviour, food consumption, and sleep) and directly measured adiposity and physical activity. A detailed description of the ISCOLE protocol is published elsewhere [26]. Overall coordination, training in data collection methods, and data management for the ISCOLE project, including final quality control checks, was provided by the ISCOLE Coordinating Center staff at the Pennington Biomedical Research Center (Baton Rouge, USA). Data collection was conducted following ethical approval from the University of South Australia Human Research Ethics Committee (Australia), the Research Ethics Committee of the Municipal Health of Sao Caetano do Sul—Prima (Brazil), the Children’s Hospital of Eastern Ontario Research Ethics Board (Canada), the Biomedical Ethics Committee of Tianjin Women’s and Children’s Health Center (China), the Universidad de los Andes Committee on Research Ethics (Colombia), the Ethics Committee of the Hospital District of Helsinki and Uusimaa (Finland), the St. John’s Medical College and Hospital Institutional Ethical Review Board (India), the Kenyatta University Ethics Review Committee (Kenya), the Ethics Committee University of Porto (Portugal), the University of Cape Town Health Sciences Faculty Human Research Ethics Committee (South Africa), the University of Bath Research Ethics Committee for Health (UK), and the Pennington Biomedical Research Center Institutional Review Board for Research with Human Subjects (USA). Written informed consent and assent from all parents/guardians and participating children was also obtained.

ISCOLE design

Recruitment targeted a sex-balanced sample of approximately 500 children 9–11 years of age from each site. The primary sampling frame in all sites was schools in urban and suburban areas, stratified by indicators of SES status in order to maximize variability within sites. The secondary sampling frame was classrooms having children with minimum variability around 10 years of age. ISCOLE data collection was conducted during a full school year, which varied across countries; however, data collection in all sites was conducted between September 2011 and December 2013. Any variations in recruitment strategies employed in the different countries are reported elsewhere [26].

Anthropometry

Standing height was measured by trained research staff using a Seca 213 portable stadiometer (Birmingham, UK), with the participant as erect as possible and head positioned in the Frankfort horizontal plane. Weight was then measured using a portable Tanita Body Composition Analyser (Illinois, USA), after all outer clothing, heavy pocket items, shoes, and socks were removed. Body mass index (BMI) was derived from weight and height (kg/m²), and children were categorised based on the World Health Organization (WHO) BMI growth reference classification of thinness (z-score < -2); healthy weight (≥ -2 and ≤ +1); overweight (>+1 and ≤ +2); and obese (>+2) [28]. For the purpose of this study, the term “overweight” is used, and is inclusive of children classified as obese.

Accelerometry

Physical activity was objectively measured using ActiGraph GT3X+ accelerometers (Florida, USA). Accelerometers were attached firmly to belts and worn at the waist on the right mid-axillary line. Participating children were instructed to wear the devices 24 hours per day for at least 7 consecutive days (at all times except when bathing or swimming), including an initial familiarisation day, maximizing the number of children providing at least 4 days of waking-hours wear time of ≥10 hours/day, including at least one valid weekend day. Time spent sleeping at night was first identified using a published algorithm, and non-wear time within the awake portion of the day was classified as 20 consecutive minutes of ‘0’ counts [29, 30]. Data from the accelerometers were downloaded, reviewed for completeness, and data reduction conducted using cut-points validated in children and youth [31]. MVPA was defined as ≥ 574 counts per 15 seconds [31]. Children meeting the WHO physical activity guideline of ≥ 60 minutes of daily MVPA [8], based on their mean daily minutes in MVPA, were also identified.

Questionnaires

A Demographic and Family History Questionnaire was completed by one or both parent(s) or guardian(s) of the participating child, but collected information on the child’s biological parents. This questionnaire collected information on basic demographics, family health, and socioeconomic factors [26]. ISCOLE’s questionnaires were a compilation of several validated items obtained from existing questionnaires, and where no suitable alternatives were found, new questions were designed by content experts from the ISCOLE investigator team. To maximize the collection of valid and reliable data, technicians were trained on questionnaire administration in a standardised manner, and provisions were made to administer the questionnaire via an interview for participants with low literacy levels. Parental education was categorised as some college/associate degree, bachelor, or post-graduate degree (higher education level), compared to the remainder reporting that they completed high school, had some high school, or less than high school (lower education level). Parents were also categorised based on self-reported height and weight into underweight (< 18.5 kg/m²), normal weight (18.5–24.9 kg/m²), overweight (25.0–29.9 kg/m²), or obese (≥ 30.0 kg/m²) [32]. For the purpose of this study, the term “overweight” is inclusive of those classified as obese by these standards.

Missing data

A total of 7372 children participated in ISCOLE. However, children who were missing information on BMI, physical activity, or mother’s and/or father’s education level, weight, and/or height, were excluded from these analyses. This resulted in a decrease to an analytic sample of 4752 (64.5%). The largest contribution to the missing data was father’s BMI; however, this was not country specific. The Fisher’s exact test was used to investigate whether there was a statistical difference between excluded children and those in the analytic sample. We found no differences in key variables including sex (p = 0.77), overweight status (p = 0.16), or meeting MVPA guidelines (p = 0.17).

Statistical analysis

The statistical analyses presented in this paper were conducted using SAS version 9.3 (SAS Institute Incorporated, Cary, North Carolina, USA). Descriptive statistics (e.g. frequencies, means, and standard deviations) were calculated. The correlates included were maternal education, paternal education, maternal overweight and paternal overweight. Sex was included as a covariate in the final models. Multilevel modeling analysis was used to obtain parameter estimates for the associations between childhood overweight or meeting the WHO MVPA guidelines and parental education level or parental overweight status. This analytical method allows for the simultaneous examination of the effects at the individual (child), school, and country (site) levels. That is, it accounts for the hierarchical nature of the data. Multilevel general linear models (PROC GLIMMIX), including site as a fixed effect, were used to determine correlates of childhood overweight or meeting MVPA guidelines.

Parental weight status and child overweight

Maternal and child overweight were positively related in all sites except Kenya. A similar positive association between paternal and child overweight was found for seven of the twelve sites (Brazil, China, Colombia, Finland, India, Kenya, and Portugal). These results are aligned with a number of past studies showing that parental overweight, and perhaps more predominantly maternal overweight status, is associated with a higher likelihood of childhood overweight [10–18]. Similarly, our findings showed that maternal overweight was more consistently (in more of the sites) associated with child overweight compared to paternal overweight. Further, and as previously determined, the relationship between parental and child overweight status was fairly consistent across the different sites given our older age group of children 9–11 years [10–12, 15].

Parental education and child overweight

Children from the Colombian and Kenyan sites were more overweight if their mothers had attained a higher education level (positive association). Kenyan children were also more overweight if their fathers had a higher education level. In contrast, children from the Brazil and USA sites were less overweight if their fathers had attained some college education or higher (negative association). While studies conducted in other lower economic status countries like Colombia and Kenya have reported a similar positive association between parental education level and child BMI [9, 19, 22], results from wealthier countries such as Brazil and the USA have reported a negative relationship between these factors [20, 21]. We speculate that these findings may be potentially explained by a higher level of awareness and knowledge among the more educated parents in higher economic status countries in regard to the positive effects of maintaining a healthy body weight. Having been exposed to an obesogenic environment earlier on [1], strategies to prevent overweight may be better established in these countries. Our findings further seem to indicate that the positive influence of parental higher education on child overweight in higher economic status countries is stronger with fathers than with mothers. In contrast, in lower economies, higher parental education is associated with a higher likelihood of their children being overweight [22]. This may be a consequence of social norms, whereby, in some developing countries, a more overweight child may be perceived as a "healthy child", with adequate food and food security aspired to, particularly for families with more disposable income. In such families, children may also have access to more motorised transport, and engage in less active travel [33]. This, coupled with minimal knowledge about the associated health risks of being overweight, may explain the positive association between parental education level and child BMI in lower economic status countries. These contrasting findings among some lower and higher income countries provides further evidence of the varying status of a physical activity and nutritional transition occurring in these countries [34, 35].

Parental weight status and child MVPA

Paternal overweight was negatively associated with Kenyan children meeting the MVPA guidelines. This finding is supported by systematic review evidence from sub-Saharan African countries showing a negative relationship between parental BMI and child physical activity measures [23]. However, it is important to note the limitations inherent in self-reported height and weight among adults and the need for directly measured data to confirm these findings.

Parental education and child MVPA

The results also showed that children from Colombia, Kenya, and Portugal had lower odds of meeting MVPA guidelines if their mothers had attained a higher education level. Kenyan and South African children were also at a lower odds of meeting MVPA guidelines if their fathers had some college education or higher. Fewer studies have investigated the relationship between parental education level and child physical activity; however, evidence points to a negative association between parental education and child physical activity in lower economic status countries [9, 23], and a positive relationship between these factors in higher economic status countries [24], a pattern which is supported by the present study. It is noteworthy that mothers’ education level is more often examined and relationships consistently found in these papers [9, 24, 25]. Our results seem to indicate that the negative influence of higher maternal education on child physical activity is stronger than higher paternal education level in lower economic status countries. We speculate that this may be the result of mothers with higher education in lower economic status countries, having greater access to motorized transport as a consequence of greater SES, and being primarily responsible for the transport means used by their children.

Household coexistence of parent-child overweight

Interestingly, we found that Kenya had the lowest proportion of overweight children, the highest proportion of overweight mothers, and the highest proportion of child-parent (and child-mother) overweight dyads in the same household compared to other sites. China had the highest proportion of overweight children, the lowest proportion of overweight mothers, and the lowest proportion of child-parent (including both child-mother and child-father) overweight dyads. The highest proportion of overweight among fathers was in the USA, which also had the highest proportion of child-father overweight compared to other sites. These findings reinforce the significant relationships between both mother and father overweight and child overweight. The seemingly higher proportions of child-parent overweight coexistence in middle and high economic status countries in the total sample of children may be a result of longer lasting (over time) pathways of familial susceptibility among children and their parents in higher economic status countries. Familial aggregation has also been shown with physical activity [36]. Since many higher economic status countries have experienced an obesogenic environment for longer periods than low- and middle-income countries [1], it is reasonable to expect higher levels of child-parent overweight coexistence in the same household; however, the inverse parental education-child overweight relationship in higher income countries suggests a conflicting paradigm.

There are a few notable limitations to these analyses including the cross-sectional design with all of its intrinsic weaknesses. Mothers or fathers proxy-reporting for the other parent, or guardians proxy-reporting for biological parents of the child, may also reduce accuracy of results. Excluding children with missing data on any of our variables of interest in the analyses reduced the sample size as previously discussed; however, Fisher’s exact tests were performed to ascertain that excluded children were not statistically different from included children. In addition, the influence of differing social or cultural norms between countries was not captured in this study, for instance, overweight could be a desired trait in some countries [37]. Nonetheless, since parental overweight and education (among other measures of SES) are related to child overweight and MVPA in different ways in developing verses developed countries (countries in different developmental stages), ISCOLE presented an ideal opportunity to examine these relationships.