There is compelling evidence about the relationship between socio-economic status (SES) and health in adulthood(Reference Richter, Vereecken and Boyce1) but whether this is also the case in adolescence is less clear owing to the homogenising effect produced by the influence of school, peers and youth culture(Reference Richter, Vereecken and Boyce1). Several studies have indicated that SES may influence certain health-related behaviours (sedentary behaviours and nutrition) at childhood and adolescence. For instance, diets with a high nutritional value (rich in fruits, vegetables and dairy) are more common in high-SES adolescents(Reference Abudayya, Stigum and Shi2), yet not all SES markers are associated with adolescents’ health-related dietary habits in the same direction. For example, in a recent study conducted in Norway, only parental education was associated with adolescents’ health-related dietary habits(Reference Nilsen, Krokstad and Holmen3). Moreover, some demographic factors, such as age and gender, seem to be important determinants of the prevalence of health-related behaviours during adolescence(Reference Alamian and Paradis4).
A growing number of studies indicate that the obesogenic effect of television (TV) viewing in young people may be attributed more to increased food intake than to decreased physical activity levels(Reference Jackson, Djafarian and Stewart5, Reference Manios, Kourlaba and Kondaki6). In fact, adolescents or families who watch TV excessively are less likely to eat fruit and vegetables(Reference Vereecken, Todd and Roberts7, Reference Campbell, Crawford and Ball8) and more likely to eat sweets(Reference Vereecken, Todd and Roberts7, Reference Campbell, Crawford and Ball8), savoury snacks(Reference Coon, Goldberg and Rogers9), soda(Reference Coon, Goldberg and Rogers9) and soft drinks(Reference Vereecken, Todd and Roberts7, Reference Campbell, Crawford and Ball8, Reference Ng, Young and Corey10). One mechanism by which excessive TV viewing may lead to the adoption of unhealthy dietary habits could be the higher exposure time to food commercials, which advertise energy-dense products with lower nutritional value(Reference Hitchings and Moynihan11, Reference Wiecha, Peterson and Ludwig12). Alternatively, it could be also possible that sociodemographic factors may explain the association between TV viewing and food energy content. It is known that families with low economic resources buy less-nutritious, energy-dense foods owing to their reduced cost(Reference Drewnowski and Darmon13). These same families may prefer watching TV during their leisure time as it is also an inexpensive behaviour. Whatever the mechanisms involved, knowledge provided by observational studies is useful for the design of future interventions.
In the present European multicentre study we investigated: (i) whether consumption of high-energy-dense food products during TV viewing was more frequent in adolescents who reported more time watching TV (>2 h/d); and (ii) the association between sociodemographic variables (age, gender and SES) and the consumption of energy-dense food products during TV viewing.
Methods
Design
The Healthy Lifestyle in Europe by Nutrition in Adolescence (HELENA) cross-sectional study (CSS) is a European Union-funded project on lifestyle and nutrition among European adolescents(Reference Moreno, González-Gross and Kersting14). Data collection took place between October 2006 and December 2007. More detailed information on the study design has been published elsewhere(Reference Moreno, De Henauw and González-Gross15). Participants were recruited at schools. To guarantee that the heterogeneity of social background of the population would be represented, schools were randomly selected after stratification on school zone or district. In the case that the selected schools refused to participate, a second list of substitute schools had already been foreseen. Up to three classes from two grades were selected per school. A class was considered eligible if the participation rate was at least 70 %. The HELENA-CSS inclusion criteria were age range 12·5–17·5 years, not participating simultaneously in another clinical trial, and free of any acute infection lasting less than 1 week before inclusion(Reference Moreno, De Henauw and González-Gross15).
From an initial sample of 1413 (730 girls and 683 boys) adolescents who met the HELENA inclusion criteria, 1336 adolescents (699 girls and 637 boys) had valid information from the sedentary behaviour questionnaire (with at least 75 % of answers). Non-significant differences were found in SES level between the initial sample and the group finally included in the present study. The current study collected data from Ghent (Belgium), Heraklion (Greece), Pecs (Hungary) and Zaragoza (Spain) since other cities that took part in the HELENA-CSS did not examine food consumption during sedentary behaviours. During school visits, researchers asked adolescents about their habitual TV viewing time and the concurrent food choices in a self-reported sedentary behaviour questionnaire. In addition, parental SES was registered by the adolescents. In all schools, researchers explained the aims of the questionnaires and clarified questions asked by adolescents. The study was approved by the Research Ethics Committee of each city involved. Written informed consent was obtained from the parents of the adolescents and the adolescents themselves(Reference Béghin, Castera and Manios16).
Parental socio-economic status
Family Affluence Scale
Each adolescent was asked for the number of cars and computers at home, whether the Internet was available at home, and whether or not the adolescent had his/her own bedroom. The Family Affluence Scale (FAS) was calculated based on a model developed by Currie et al.(Reference Currie, Molcho and Boyce17), but adapted slightly by replacing the item on holidays by Internet at home. FAS indicates the SES of the adolescent on a scale from 0 (very low SES) to 8 (very high SES). We recoded the scores as low SES (0–2), medium SES (3–5) and high SES (6–8)(Reference Rey-López, Vicente-Rodriguez and Ortega18). To our knowledge, the test–retest reliability of adolescents’ reports on FAS has never been measured. Despite this, the proportion of missing data on FAS items(Reference Currie, Molcho and Boyce17) was low and FAS presents good criterion validity against a national wealth indicator (Gross Domestic Product)(Reference Boyce, Torsheim and Currie19).
Parental occupation
Information on parental occupation was collected using the International Standard Classification of Occupations (http://unstats.un.org/unsd/class/family/family2.asp?Cl=224), a hierarchical scale developed to be used as a nine-class classification of occupations (armed forces are added but not really part of the classification). In social epidemiology, the scale is mainly used as a nine-class classification system, 1 being the highest category and 9 being the lowest. We recoded the occupations into three levels: (i) low (Skilled agricultural and fishery workers, Craft and manual workers, Machine operators and assemblers, Workers and elementary occupations); (ii) medium (Intermediate occupations, Administration/clerical, Business services); and (iii) high (Managerial staff, Intellectual and scientific occupations).
Parental education
The parents’ educational level was considered as the highest level of education attained by each parent. Adolescents selected one of the following levels: (i) lower education; (ii) lower secondary education; (iii) higher secondary education; or (iv) higher education/university degree.
Television viewing
Habitual TV time was estimated by the HELENA sedentary behaviour questionnaire(Reference Rey-López, Vicente-Rodriguez and Ortega18), which asked the question: ‘During weekdays, how many hours do you usually watch TV?’ Possible answers were: (i) none; (ii) less than 0·5 h; (iii) between 0·5 and 1 h; (iv) between 1 and 2 h; (v) between 2 and 3 h; (vi) between 3 and 4; (vii) more than 4 h. The reliability (1-week test–retest) of the HELENA sedentary behaviour questionnaire was studied in 183 adolescents (seventy-nine males, 104 females; age range 12·5–17·5 years). They were recruited from two schools involved in the HELENA study, Ghent (Belgium) and Zaragoza (Spain). Cohen's κ using quadratic weights showed a substantial agreement (0·71 and 0·68 during weekdays and weekend days, respectively)(Reference Landis and Koch20).
Dietary habits during television viewing
Dietary habits during TV viewing were obtained by the HELENA sedentary behaviour questionnaire with the questions: ‘How often do you drink while you watch TV?’ and ‘How often do you eat while you watch TV?’ The answers were: (i) never; (ii) less than once per week; (iii) between 1 and 2 times per week; (iv) between 3 and 4 times per week; (v) almost every day; (vi) several times per day. In addition, the adolescent were asked ‘What do you usually drink while you watch TV?’ and What do you usually eat while you watch TV?’ The answers were: nothing, water, milk, fruit juices, light drinks, soft drinks, beer, herbal infusions, coffee (for drinks); and nothing, fruits, dried fruits, savoury, pastry, sweets, sandwich, dairy (for foods). We categorized beer, soft drinks and fruit juices as energy-dense drinks; and pastry, savoury (salty products) and sweets as energy-dense foods. The categorization was based on the energy content: all energy-dense drinks contained a minimum of 159 kJ (38 kcal)/100 ml and the energy-dense foods at least 1675 kJ (400 kcal)/100 g(Reference Farran, Zamora and Cervera21). The reliability of the HELENA sedentary behaviour questionnaire was examined in a sub-sample of adolescents. Cohen's κ coefficients for most of the dietary items showed substantial agreement (0·61–0·80) between test and retest, except for beer consumption (moderate agreement, κ = 0·60)(Reference Landis and Koch20).
Statistical analyses
Descriptive statistics (frequency and percentage) of drinks and foods consumed during TV viewing are presented by media recommendations for children (≤2 h/d and >2 h/d)(22) but based on exclusively TV viewing. Statistical differences between groups or genders were tested by χ 2 analysis. Crude odds ratios with 95 % confidence intervals of having energy-dense dietary habits (drinks and foods) during TV viewing were calculated by logistic regression analysis. We included as independent variables age (<15 years, ≥15 years), parental education, parental occupation, FAS and time watching TV in separate univariate models. In addition, a multivariate model was performed. Adjusted odds ratios were obtained accounting for clustering of schools within cities. All analyses were carried out separately for girls and boys. Statistical analyses were performed using the STATA statistical software package version 10·0 (StataCorp., College Station, TX, USA). The significance level was set at P ≤ 0·05.
Results
The number of adolescents included in the current study is presented as Supplementary Material. Table 1 summarizes food and drink consumption during TV viewing in adolescents by time watching TV. In the group of adolescents who watched TV for >2 h/d, a higher percentage of girls reported a daily consumption of beverages (while watching TV) than boys (50 % v. 43 %, respectively; P ≤ 0·05 (χ 2 test); Table 1). The results showed that girls tended to consume more water, fruit juice and herbal infusions, whereas boys drank significantly more soft drinks and beer (all P ≤ 0·05). Regarding food items, girls had a higher frequency of eating food during TV viewing than boys (for those watching >2 h/d, 41 % of girls reported a daily consumption v. 35 % of boys; P ≤ 0·05). Girls reported a more frequent intake of sweets than boys (P ≤ 0·05). For the rest of the food items, non-significant gender differences emerged. Dietary habits also differed between groups of TV viewing time. Compared with their counterparts watching TV for ≤2 h/d, boys watching TV for >2 h/d had more frequent consumption of soft drinks, savoury, pastry, sweets and sandwiches, and less frequent intake of herbal infusions; and girls watching for >2 h/d had more frequent consumption of fruit juice, soft drinks, beer, coffee, pastry, sweets and sandwiches.
Table 1 Drink and food consumption during television viewing in adolescents aged 12·5–17·5 years from Ghent (Belgium), Heraklion (Greece), Pecs (Hungary) and Zaragoza (Spain); the Healthy Lifestyle in Europe by Nutrition in Adolescence (HELENA) study
*P ≤ 0·05 between genders.
†P for difference between ≤2 h/d and >2 h/d (χ 2 test).
Tables 2 and 3 show the odds of having energy-dense dietary habits during TV viewing by age, parental SES markers and time spent watching TV. Age was associated with the consumption of energy-dense drinks only in adolescent boys aged ≥15 years (adjusted OR = 1·65, 95 % CI 1·04, 2·62). None of the studied SES indicators was consistently associated with the consumption of energy-dense foods during TV viewing, in both genders. By contrast, girls whose mothers achieved a low educational level had an increased risk (adjusted OR = 3·22, 95 % CI 1·81, 5·72) of consumption of energy-dense drinks during TV viewing compared with those whose mothers attained the highest educational level. Similarly, girls whose mothers attained lower secondary or higher secondary education had higher odds compared with those whose mothers had the highest level (adjusted OR = 1·76, 95 % CI 1·10, 2·82 and adjusted OR = 1·85, 95 % CI 1·19, 2·88, respectively). Girls with low FAS were more likely to consume energy-dense drinks during TV viewing (adjusted OR = 2·03, 95 % CI 1·19, 3·47). However, in girls the association with energy-dense drinks was less pronounced when paternal occupation and education were tested.
Table 2 Odds ratio and 95 % confidence interval for consuming energy-dense foods (pastry or savoury or sweets) during television (TV) viewing by sociodemographic variables and TV time in adolescents aged 12·5–17·5 years from Ghent (Belgium), Heraklion (Greece), Pecs (Hungary) and Zaragoza (Spain); the Healthy Lifestyle in Europe by Nutrition in Adolescence (HELENA) study. Crude and adjusted for clusters (schools) of each city
Ref., referent category.
Table 3 Odds ratio and 95 % confidence interval for consuming energy-dense drinks (beer or soft drinks or fruit juice) during television (TV) viewing by sociodemographic variables and TV time in adolescents aged 12·5–17·5 years from Ghent (Belgium), Heraklion (Greece), Pecs (Hungary) and Zaragoza (Spain); the Healthy Lifestyle in Europe by Nutrition in Adolescence (HELENA) study. Crude and adjusted for clusters (schools) of each city
Ref., referent category.
Discussion
The main findings of the current study are: (i) energy-dense dietary choices were more likely in adolescents who watched TV excessively (>2 h/d); and (ii) maternal educational level showed the highest impact on the adolescents’ dietary habits during TV viewing.
Gender differences in the types of drink consumed during TV viewing were particularly evident. Boys reported more frequent consumption of beer and soft drinks, whereas girls tended to consume more water, fruit juices and herbal infusions. These results concur with recent studies(Reference Nilsen, Krokstad and Holmen3, Reference Scully, Dixon and White23). For instance, in Norwegian adolescents, 15 % of girls and 29 % of boys had soft drinks daily(Reference Nilsen, Krokstad and Holmen3). In Australian adolescents, about 44 % reported consuming energy-dense drinks at least four times in the previous week. Older students and males had higher consumption of energy-dense drinks(Reference Scully, Dixon and White23). It should be noted, however, that we did not observe gender differences in food consumption during TV viewing, except for sweets, consumed more frequently by girls.
Foods and drinks consumed during TV viewing were more energy-dense in adolescents who exceeded the media recommendations (watching TV for >2 h/d). As depicted in Table 1, adolescents had more frequent consumption of energy-dense food products (sweets, bakery, soft drinks) if they watched TV for >2 h daily, which agrees with previous studies(Reference Vereecken, Todd and Roberts7–Reference Coon, Goldberg and Rogers9). Furthermore, the regression model confirmed that exceeding screen time recommendations was associated with the selection of energy-dense foods and drinks (Tables 2 and 3). Several mechanisms may explain why TV viewing associates with poorer dietary choices. Children's food preferences may be a function of copying the food preferences of their parents. Parents not only provide the food that will be eaten at home, but also the feeding style (i.e. eating in front of the TV or not). An indirect pathway could be the effect of food commercials, which produce more likely consumption of the advertised products (generally energy-dense)(Reference Hitchings and Moynihan11, Reference Wiecha, Peterson and Ludwig12). Alternatively, it is possible that TV viewing may operate as a confounding factor. There is evidence that families with reduced income have diets with a higher energy content(Reference Drewnowski and Darmon13). These same families could also spend more time watching TV. However, the effect size and the direction of our findings did not change when the odds were additionally adjusted for family affluence, which reinforces a probable causal link between TV viewing and the consumption of energy-dense drinks.
Another relevant finding was that low SES (maternal education and occupation) was associated with the consumption of energy-dense drinks (soft drinks or beer or fruit juice) while watching TV, but only in girls. Our results are in agreement with those reported by Aranceta et al.(Reference Aranceta, Pérez-Rodrigo and Ribas24), who observed that children whose mother had a low level of education were more likely to drink soft drinks. Richter et al.(Reference Richter, Vereecken and Boyce1) explored the association between socio-economic position and four different aspects of adolescents’ health behaviour in 28 countries from Europe, USA, Canada and Israel. Occupational status of the parents exerted a strong effect on the time watching TV, in agreement with a recent study published by our group(Reference Rey-López, Tomas and Vicente-Rodriguez25). Thus, adolescents with less favourable socio-economic circumstances are at risk of having unhealthy behaviours.
Strengths of our study lie in the use of a large and cultural diverse sample of European adolescents (Belgium, Hungary, Greece and Spain). It should be noted that the sedentary behaviour questionnaire showed a good reliability. Another strength is that we studied the specific types of foods and drinks consumed while viewing TV. By contrast, previous studies have used FFQ(Reference Vereecken, Todd and Roberts7, Reference Campbell, Crawford and Ball8) or 24 h dietary recalls which cover intake over the whole day to compare ‘high TV viewers’ with ‘low’ ones(Reference Coon, Goldberg and Rogers9, Reference Ng, Young and Corey10). Although food intake during TV viewing represents between approximately 17 and 26 % of daily energy intake(Reference Matheson, Killen and Wang26), the direction of our findings does not differ from that of a recent study which included an FFQ to assess dietary health-related behaviours in Norwegian adolescents(Reference Nilsen, Krokstad and Holmen3), which strengths our results. Thus, perhaps the nutritional quality of snack foods during TV viewing may be a proxy of habitual dietary patterns. As main limitations, conclusions could be drawn only for TV viewing and not for other modern types of sedentary behaviour. Moreover, we collected information only about the consumption of food when watching TV and not during the whole day. Nevertheless, there is evidence that reductions in screen time are related with reduced total energy intake(Reference Epstein, Roemmich and Robinson27). Another limitation is the assessment of SES using adolescents’ self-reports. Parental education and family affluence were easily reported by adolescents. In contrast, researchers observed that some adolescents had more problems in indicating their parents’ occupation. If missing or unclassifiable responses were unequally distributed among social groups, it might have influenced the results. Finally, we did not assess the amounts of food or drinks consumed during TV viewing which would have provided more detailed knowledge on this topic.
As previously commented, intervention studies have shown that reductions in TV viewing are related to changes in energy intake(Reference Epstein, Roemmich and Robinson27). To what extent a reduction in the time spent watching TV also modifies diet quality should be studied. A complementary way to reduce adolescents’ energy intake would be to limit TV food advertising. According to a mathematical model, TV food advertising may be responsible for 15–40 % of the obesity prevalence among 6- to 12-year-old US children(Reference Veerman, Van Beeck and Barendregt28). Although the aetiology of obesity may differ among countries, if TV fast-food advertising for children were banned, lower overweight/obesity rates could be observed(Reference Chou, Rashad and Grossman29). Our study has an important public health message. Young people should minimize the time spent watching TV, especially socio-economically disadvantaged adolescents. Theoretically this would allow a decrease in the consumption of unhealthy beverages; but due to the cross-sectional design of our study, we cannot conclude on this issue. Intervention studies should be conducted to replicate our findings.
In conclusion, excessive TV viewing during adolescence was associated with more likely consumption of energy-dense food and drink products. Energy-dense foods were not associated with parental SES. Maternal education (in girls) was associated with the consumption of energy-dense drinks. In developed countries, efforts to promote healthy eating behaviours should prioritize children whose parents have low SES.
Acknowledgements
The HELENA Study was carried out with financial support of the European Community Sixth RTD Framework Programme (Contract FOOD-CT-2005-007034). The content of this article reflects only the authors’ views, and the European Community is not liable for any use that may be made of the information contained herein. The present analysis was also supported (J.P.R.-L., L.A.M.) by the Spanish Ministry of Health's Maternal, Child Health and Development Network (grant no. RD08/0072) and the Spanish Ministry of Education (grant nos. EX-2007-1124, DEP2007-29933-E and AP2006-02464). J.P.R.-L. was also supported by the Fundación Cuenca Villoro (Spain). The authors declare no conflict of interest. J.P.R.-L. was involved in the acquisition, analysis and interpretation of the data. L.A.M., J.R.R., F.B.O., I.d.B. and M.S. made substantial contributions to the conception and design of the manuscript. L.A.M., I.d.B. and M.S. gave final approval of the version to be published. All authors read and approved the final manuscript.
