Insufficient physical activity and poor nutrition are widely acknowledged as the primary mechanisms underlying the rise in excess body weight.¹² Recent studies have described geographic and socioeconomic gradients and identified aspects of children's lifestyle, including physical and sedentary activities, as risk factors for overweight and obesity.^{13,14,15,16,17,18,19,20} Most recently, Swinburn and colleagues identified risk factors related to parents, family and school as potentially significant and as requiring more evidence so that evidence-based policy and program decision-making for the prevention of excess weight can be developed.²¹The need for further research is underlined by the fact that a number of schools and jurisdictions in Canada are already implementing policies and programs without evidence of their effectiveness.

The CLASS survey included a modified version of Harvard's Youth/ Adolescent Food Frequency Questionnaire,²⁷ which gathers information on both dietary intake and habits pertaining to mealtime behaviours. It also included validated questions on the frequency of physical activities and the number of hours of sedentary activities (watching television, working on a computer, playing video games) taken from the National Longitudinal Survey of Children and Youth (NLSCY).²⁸ Information on sociodemographic factors was taken from a survey completed by parents and included the child's sex, place of birth and residency, as well as the parents' marital status, income level and educational attainment. The children's ages were not included, since the vast majority of grade 5 students were either 10 or 11 years old at the time they completed the CLASS survey.

We estimated neighbourhood income by averaging, per school, the postal-code level means of household income (available through the 2001 Canada census) of residential addresses of children attending that school. Finally, we developed and administered a short survey for school principals to collect information on school characteristics, including sales of soft drinks, presence of vending machines, type of food services, frequency of physical education classes and possible financial restraints for recreation and gymnasium equipment.

Because our observations of students are nested within those of their schools, we assessed student and school-based factors at distinct levels using multilevel statistical methods. Specifically, student and parental factors were considered as first-level covariates. All school characteristics and neighbourhood income were considered as contextual factors and treated as second-level covariates. We first applied multilevel logistic regression methods to determine to what extent risk factors were associated with overweight or obesity. Second, we grouped each risk factor into 1 of 4 groups — dietary habits, activities, and sociodemographic and school-based factors — and then adjusted for all significant risk factors within each group. These estimates are referred to as theme-adjusted odds ratios. Third, we considered all significant risk factors simultaneously to quantify their independent importance for excess body weight.

Because participation rates were slightly lower in residential areas with lower estimates of household income, we calculated response weights to overcome potential nonresponse bias. On the basis of average household incomes according to postal code data from the 2001 census for both participants and nonparticipants, we calculated response rates per decile of household incomes by postal code. These response rates were subsequently converted into response weights. Because we weighted all prevalence and risk estimates for nonresponse bias, they represent provincial population estimates for grade 5 students. We considered missing values as separate covariate categories, but we do not present their estimated values.

The prevalence of overweight among grade 5 students in Nova Scotia in 2003 was 32.9%, with 9.9% being obese. The prevalence of overweight was about the same for girls (32.9%) and boys (33.0%), whereas the prevalence of obesity was lower among girls (9.0%) than among boys (10.9%).

Among grade 5 students, 3.7% did not eat breakfast (Table 1). In a univariate analysis, these students were 1.5 times (or 50%) more likely to be overweight than students who usually ate breakfast. Missing lunch was also associated with an increased risk of excess body weight, but this risk was not statistically significant. Relative to those bringing their lunch from home, children buying lunch at school were 47% more likely to be overweight (unadjusted odds ratio [OR] 1.47, 95% confidence interval [CI] 1.23–1.76). Increasing frequencies of eating supper together at home (family supper) and decreasing frequencies of eating supper in front of the television were associated with a decreased risk of overweight. Because dietary habits are interrelated, we considered these habits simultaneously and found that lunch pattern and frequency of family supper and supper in front of the television were determinants of body weight (Table 1: theme-adjusted ORs). After adjusting for all significant risk factors across all 4 theme groups, only buying lunch at school significantly increased the risk of overweight, and family suppers both 3–4 and 5 or more times a week significantly decreased the risk (Table 1: fully adjusted ORs).

Sedentary activity of more than 1 hour per day was associated with a significantly increased risk of overweight, as was being driven to school longer than 30 minutes (Table 2: theme-adjusted ORs). Participating in physical activity more than 7 times a week was associated with a decreased risk of overweight (Table 2: theme-adjusted ORs). When considered in conjunction with other determinants, frequency of physical activity appears to be the only activity-related factor independently associated with overweight. However, sedentary activity was independently associated with overweight when physical activity was not considered (data not shown), which indicates a correlation between these 2 covariates.

With respect to sociodemographic factors, children whose parents had attained higher levels of education and had an income over $60 000 were at a decreased risk of overweight, as were children who resided in urban areas and who resided in neighbourhoods where the income was in the middle or highest one-third (Table 3).

Of all children, 29.3% attended schools where lunches were provided by a foodservice company. This may include on-site catering or fast food delivery. Children attending such schools were 12% more likely to be overweight than children who attended schools where no foodservice companies were used, but this difference was not statistically significant (Table 4). Also, no substantial or statistically significant differences were observed among children attending schools where lunches were distributed by staff and volunteers or as part of a prepaid lunch program.

Of all children, 15.3% attended schools where soft drinks were sold. These children drank an average of 4.0 cans of soda pop per week compared with 3.6 cans drunk by children attending schools without such sales (p < 0.01). Children from schools with and without sales of soft drinks consumed an average of 33.5 and 32.5 g of sucrose per day respectively (p = 0.11). Availability of soft drinks at schools was not associated with significantly increased risks of overweight (Table 4).

Children attending schools with more frequent physical education classes were increasingly more likely to have normal body weight. Financial restraints on the purchase of recreation and gymnasium equipment were not statistically associated with increased weight (Table 4).

Swinburn and associates called for the investigation of risk factors relating to parents, family and school environment.²¹We found that children from families that eat together regularly are less likely to be overweight or obese. One reason for this is that these children generally eat a more healthy diet.^25,30,31 A second reason is that family meals prevent children from eating in front of the television, which may lead to “mindless eating” and higher energy intake.^32,33,34 There may also be broader benefits of better communication and relationships between children and parents resulting from the daily interactions during meals.³⁵

In terms of school environment, the association between obesity levels and frequency of physical education classes was striking. This finding supports policies that increase the frequency of physical education classes, such as those recently announced for all elementary schools in Alberta. School lunches represent another opportunity to reduce overweight and obesity. Halting the sales of soft drinks in schools has been suggested to reduce the intake of refined sugars^25,35 and is an advocated strategy to prevent overweight.³⁶ We observed that children attending schools that sell soft drinks consumed somewhat more soft drinks and sugar, but the amounts were likely insufficient to bring about differences in body weight. Previously we showed that integrated school programs that include more physical education, healthy lunches, health and nutrition education, training of staff, parental involvement, and a halt to the sales of soft drinks were successful in improving children's diets and reducing overweight by 59% and obesity by 72%.²⁴

In our study, as in other studies of the effect of risk factors related to children's lifestyle, normal-weight children were more physically active and engaged less in sedentary activities.^15,16,17 Similarly in keeping with other studies, we observed a gradient whereby children of socioeconomically disadvantaged families were more likely to be overweight or obese.^13,14,23 Studies in the United States and Europe have shown neighbourhood differentials with respect to overweight.^37,38,39 Our study confirms this in a Canadian setting, with childhood obesity rates that were twice as high in low-income neighbourhoods as in high-income neighbourhoods. Because of these dramatic differences, children and schools in low-income neighbourhoods should receive priority in public health initiatives to reduce future socioeconomic inequalities in health.

Strengths of our study include the population-based and multidisciplinary approach. Other strengths include measurements of participants' height and weight, the adjustment for nonresponse bias and the consideration of various potential confounders. Although most of the questions and questionnaires we administered were validated, responses remain subjective and prone to error. An additional limitation is the cross-sectional design, which introduces uncertainties regarding the sequence of cause and effect of the observed associations. Longitudinal studies overcome this limitation and will strengthen the evidence regarding risk factors for childhood overweight and obesity.