Table presents basic data on age, education, gender and setting in the full sample and the sub-sample that underwent sensory testing. Country-based descriptive statistics on television, weight status, and diet are then shown for the full sample of boys and girls separately in Table . The two countries with least overweight also had the lowest rates of eating meals while watching TV, while the pattern was less apparent for viewing time and for television in bedroom. The survey center in Italy, with the highest rates of overweight and obesity, also had the highest proportion of children who watch TV during meals and have TVs in their bedrooms, although their usual viewing time was close to the 8-country average. Regarding diet, the most notable difference across survey centers was a low propensity for consumption of high-sugar foods and beverages by Swedish children. Based on the uncertainties inherent in this type of ecological observation, all formal analyses were based on individual-level data. To take into consideration the large country-level differences across survey centers, pooled analyses of associations between individual television habits, on one hand, and weight status and diet preferences on the other, are adjusted for country in which the survey was conducted as described previously.
Selected sociodemographic characteristics of full sample and sub-sample participating in taste preference tests
Key study variables by sex, in 15,144 boys and girls aged 2–9 years included in full sample
The first analysis series examined whether television viewing patterns were associated with weight status. The television indicators were significantly related to increased odds for overweight (Table ), with only minor variation in magnitude of the association in girls and boys. Eating while watching television was significantly associated with overweight, with prevalence odds ratios of 1.20 (95 % CI 1.04–1.40) for boys and 1.35 (95 %CI 1.17–1.55) for girls in fully adjusted models. Having a television in the bedroom showed similar associations: OR 1.39 (1.19–1.61) in boys and 1.23 (1.06–1.42) in girls. Finally, watching 60 min a day or more on weekdays and/or weekends was associated with overweight to a similar degree in both sexes: OR 1.20 (1.05–1.38) in boys and OR 1.21 (1.06–1.38) in girls. These estimates were all derived from statistical models that adjusted for age, survey country, parental education, dietary propensities, parental reports of physical activity, and the other two television variables. Although not the focus of this paper, major differences across survey countries were observed for prevalence of overweight, and high parental education was a protective factor. Both of these latter results were statistically independent of the television viewing parameters under investigation here. No effect modification by age group was observed. Sensitivity analyses were also conducted using different cut-points for television viewing, i.e. rather than 60 min per day, we tested 30 min per day and 2 h per day on weekdays and/or weekends. Irrespective of the definition used, children in the higher exposure category had greater odds of overweight and there was no statistical advantage to either raising or lowering the cut-point. All of the high-risk TV indicators were significantly associated (positively) with child’s age in years and (negatively) with higher parental education; data not shown. Finally we repeated the analysis in the subset of children with accelerometry data and observed that if the positive association between television and obesity reported in Table was hardly changed after adjusting for objectively measured physical activity levels as opposed to proxy reported activity.
Table 3 Prevalence odds ratios (OR) for overweight associated with (a) eating while watching TV never or rarely vesus more frequently (b) watching TV on average at least 60 min per day on weekdays and/or weekends; and (c)having a TV/video/DVD player in (more ...)
Country-specific analyses, illustrated in Fig. , indicated that the magnitude of the associations between television habits and overweight varied among different countries. While not all associations were statistically significant, all point estimates except for Cyprus indicated positive associations.
Fig. 1 Odds ratios for overweight (including obesity) in 8 survey centers and in all countries combined. All analyses considered covariates age, sex, and parental education level; the pooled estimate (ALL) further adjusts for survey country. Each panel describes (more ...)
We next investigated whether a propensity to eat high-fat foods or high-sugar foods was associated with television habits. We found that television viewing patterns differed significantly among children with different dietary propensity scores, with strong indications of a linear gradient (Table ). Eating while watching TV was associated with a higher proportion of high-fat items and high-sugar items in the diet, in proportion to total number of foods consumed. This association was independent of education, age, sex, and country. Similar associations were observed for watching TV more than 60 min per day and having a TV in the bedroom, in relation to increasing propensity for high-fat and high-sugar foods. All three television indicators tended to be more strongly increasing across quartiles for sugar propensity ratios, compared with corresponding quartiles for dietary fat. Similar associations were seen in boys and girls. No effect modification by age group was observed.
Relation between fat/sugar propensity (quartiles) and television habits, prevalence odds ratios (95 % CI) adjusted for age, sex, survey center, and parental education
Lastly, we explored the association between the diet propensity ratios and TV habits in the sub-sample that underwent sensory testing. For simplicity we used the 4-level indicator here based on the quartile cut-points developed for the previous analysis. The main results in this subsample were similar to those seen in the full sample, although fat intake was only significantly associated with one of the TV variables (eating while watching TV). In contrast, the results for sugar propensity were robust using all three television indicators. Controlling for a preference for the sweet taste did not materially affect any of the observed associations between sweet food propensity and television habits, all of which remained similar in magnitude and significance. Similarly, the association between fat propensity and having a TV in bedroom was not explained by preference for fat in the sensory test. This indicates that taste preference is unlikely to be the mechanism linking television with propensities to consume sugar and to some extent fat. These results are shown in Table . The fully adjusted models also indicated no independent associations between high-risk TV behaviors and taste preferences per se (data not shown).
Relation between television habit and fat/sugar propensity in sub-sample participating in the taste preference tests for sweet and fat