Background
Prevalence of childhood obesity is increasing in Europe as well as in other developed countries [
1]. This is a major concern as it has harmful physical, psychological, behavioral, social and economic consequences. Overweight and obesity are results of a disturbed energy balance in which energy intake is higher than energy expenditure. During the past decades energy expenditure among children has decreased over time [
2]. There are increased opportunities for children to be sedentary in their leisure time; especially TV viewing and electronic gaming are shown to be positively related with overweight [
3‐
5]. Time spent on sedentary behaviors is inversely associated with physical exercise among adolescents [
5,
6], particularly among girls [
4]; however, controversies exist on this issue [
7].
Before starting interventions, knowledge is needed about the prevalence of sedentary behaviors and low physical exercise, their interrelationships, possible gender differences and their association with overweight. Large variations across countries in physical activity, time spent on TV viewing and PC use by children have been reported previously [
8]. However, publications presenting gender differences regarding these prevalence rates, and gender differences in associations between overweight status and these behaviors, are scarce or not available [
9]. Moreover, only a few studies explore the patterns between these different behaviors [
9,
10]. Usually only one or a small subset of individual health behaviors per study are examined. However, there is evidence that a combination of health behaviors may introduce a health risk that is greater than would be expected from the sum of the individual factors [
11]. Instead of solely investigating associations between individual behaviors and overweight, exploring associations between health-behavior patterns and overweight can provide insights for the development of multifactorial overweight preventive strategies. Additionally, cluster analyses can identify subgroups that are at high risk for becoming overweight and might best benefit from interventions
The Pro Children Project [
12] provides recent data concerning overweight status, TV viewing, PC use and physical exercise behavior and can therefore contribute to increase our knowledge in this area. The specific aims of the present study were therefore to: (I) describe the prevalence of sedentary and physical exercise behaviors; (II) assess how these behaviors are associated and whether specific clusters of behaviors can be identified; and (III) investigate whether the potential clusters are associated with overweight status, separately for boys and girls in nine European countries.
Results
The prevalence of sedentary and low physical exercise behaviors varied by country (Table
1). TV viewing during dinner was least common in Norway and most often seen among Portuguese boys. All countries report averages higher than 2 hours/day of TV viewing for both genders, except for the Icelandic girls (1.7 ± 1.3 hours/day). Highest proportions of high TV viewing during the day were observed in the Netherlands and Belgium. High PC use was most common among Dutch boys and least common in Norwegian girls. In general, all sedentary behaviors were more common among boys than girls, with the largest differences in PC use. Girls more often reported low physical exercise behavior in all countries except for Denmark and Iceland.
Table 1
Gender-specific prevalence rates of the sedentary and low physical activity behaviors in nine European countries participating in the Pro Children study.
Austria | 1601 | 23.1 | 21.3 | 36.3 | 32.4 |
40.9
| 15.9 | 57.5 |
72.4
|
Belgium | 1295 | 28.0 | 27.8 |
50.1
| 41.9 |
35.4
| 20.0 | 57.8 |
74.7
|
Denmark | 1817 |
18.9
| 14.3 |
37.6
| 31.9 |
39.0
| 13.1 | 62.6 | 65.8 |
Iceland | 1108 | 48.5 | 44.1 |
35.2
| 23.2 |
35.7
| 11.9 | 49.4 | 51.8 |
Netherlands | 1089 |
25.8
| 21.1 | 50.2 | 45.5 |
53.2
| 26.1 | 55.0 |
70.8
|
Norway | 1091 | 14.2 | 12.3 | 38.0 | 34.6 |
23.9
| 10.1 | 66.1 |
77.5
|
Portugal | 2000 | 69.3 | 70.5 |
49.0
| 42.0 |
39.8
| 17.3 | 83.9 |
94.9
|
Spain | 1223 |
59.4
| 52.3 |
36.6
| 31.4 |
21.7
| 15.3 | 63.9 |
78.8
|
Sweden | 1314 | 21.3 | 22.9 | 31.6 | 31.0 |
34.9
| 17.5 | 57.0 |
70.5
|
Total | 12538 |
35.3
| 33.4 |
40.6
| 35.3 |
36.2
| 16.4 | 62.7 |
74.4
|
Means (SD) | | 3.5 (3.0) d/week | 3.3 (3.0) d/week | 2.5 (1.7) h/day | 2.2 (1.6) h/day | 1.5 (1.6) h/day | 0.84 (1.1) h/day | 3.2 (2.3) h/week | 2.5 (2.1) h/week |
Overall, Portuguese boys were most likely to report sedentary behaviors as well as low physical exercise. Among girls, the Portuguese were also most likely to show these behaviors, except for high PC use. Swedish boys and Danish girls most often reported low sedentary and high physical exercise behavior.
Clusters of sedentary and physical exercise behavior
Watching TV during dinner, TV viewing during the day and PC use were all statistically significantly positively correlated with each other (see Table
2), but not very strongly. Time spent on physical exercise was inversely correlated with TV viewing during dinner and during the day in girls, and with TV viewing during dinner in boys. Physical exercise was significantly positively associated with PC use in girls.
Table 2
Pearson's correlation coefficients between TV viewing during dinner, TV viewing, PC use and physical exercise behavior, separately for boys and girls participating in the Pro Children study.
TV dinner | - | 0.22*** | 0.10*** | -0.14*** |
TV (hours/day) | 0.25*** | - | 0.33*** | -0.04** |
PC use (hours/week) | 0.14*** | 0.33*** | - | 0.05*** |
Physical exercise (hours/week) | -0.10*** | -0.01 | 0.02 | - |
The K-means cluster analyses resulted in five clusters for boys as well as girls. Three of the five clusters had more or less identical characteristics for both sexes. These three clusters were labeled as follows: Cluster 1: healthy behavior pattern; Cluster 2: high TV viewers; Cluster 4: high PC users;. Cluster 3 was characterized by low sedentary behavior & low physical exercise behavior in girls, but had a mixed pattern in boys. In boys Cluster 5 was labeled unhealthy behavior pattern because of the high scores on the sedentary behaviors and below average score on physical exercise. In girls, this Cluster was labeled 'high sedentary and high physical exercise', because of high scores on the sedentary behaviors as well as on the physical exercise behavior.
As can be seen in Table
3, the healthy cluster (1) is characterized by low z-scores (below 0) for TV during dinner, TV viewing and PC use combined with high (above 0) z-scores for physical exercise. However, the physical exercise score in the boys' cluster was only moderate (z-score = 0.10) and lower than the score for physical exercise in the boys' high TV cluster (z-score = 0.78). The high TV viewers (Cluster 2) are characterized by high values on TV viewing during dinner and during the day (z-scores > 0.76), but boys in this cluster were also highly physically active in contrast to girls in this cluster. The mixed group (cluster 3 in boys) had a high value for TV viewing during dinner (z-score = 0.99) but not for TV viewing during the day, and also low scores for PC use and physical exercise. In girls this cluster (Cluster 3: low sedentary and low physical exercise behavior) had low scores on all four behaviors. The high PC users (Cluster 4) are characterized by high z-scores on PC use (z-score > 1.47) but low to moderate z-scores on the other three behaviors. The unhealthy behavior cluster was characterized by high scores on the sedentary behaviors and low scores on physical exercise. The girls' high sedentary and high physical exercise cluster showed high scores on the sedentary behaviors but also above average scores on physical exercise (z-score = 0.28).
Table 3
Cluster centers in z-scores for the four behaviors included in the cluster analyses, prevalence of overweight and adjusted risk for overweight, separately for boys and girls participating in the Pro Children study.
1 | 2624 (42.0%) | -0.92a
| -0.43a
| -0.44a
| 0.10a
| 12.7 (11.1–14.3) | 1 | |
2 | 1100 (17.6%) | 0.79b
| 0.76b
| -0.25b
| 0.78b
| 17.1 (14.2–19.9) | 1.25 | 0.97–1.62 |
3 | 1494 (23.9%) | 0.99c
| -0.28c
| -0.30b
| -0.73c
| 21.7 (19.0–24.5) |
1.50
| 1.18–1.90 |
4 | 601 (9.6) | -0.48d
| -0.03d
| 1.47c
| 0.01a,d
| 17.9 (13.9–21.9) |
1.43
| 1.05 – 1.96 |
5 | 436 (7.0%) | 0.83e
| 1.68e
| 2.26d
| -0.10d
| 22.2 (17.0–27.4) |
1.67
| 1.18 – 2.37 |
Girls | | | | | | | | |
1 | 1337 (21.3%) | -0.34a
| -0.30a
| -0.26a
| 1.55a
| 11.4 (9.3–13.5) | 1 | |
2 | 1339 (21.3%) | 0.95b
| 0.95b
| -0.24a
| -0.41b
| 20.0 (17.3–22.7) |
1.64
| 1.23 – 2.18 |
3 | 2794 (44.5%) | -0.38a
| -0.49c
| -0.36b
| -0.47b
| 12.6 (11.1–14.2) | 1.03 | 0.80 – 1.33 |
4 | 584 (9.3%) | 0.15c
| 0.17d
| 1.57c
| -0.18c
| 16.0 (12.4–19.7) |
1.42
| 1.00 – 2.01 |
5 | 229 (3.6%) | 0.58d
| 2.01e
| 3.23d
| 0.28d
| 15.4 (9.6–21.3) | 1.33 | 0.80 – 2.21 |
Being in the healthy behavior cluster discriminated very well for everyday TV viewing during dinner (nobody exhibits this behavior) and for low physical exercise in girls (nobody reported physical exercise < 4 hours/week) (data not shown). The unhealthy cluster discriminated very well for high TV viewing during the day (98.2% for boys; 93.9% for girls) and high PC use (100% for boys and girls)
The clusters including most children were the healthy behavior cluster in boys (42%) and the low sedentary and low physical exercise cluster in girls (44.5%). The unhealthy behavior clusters included 7.0% of the boys and about 3.6% of the girls.
Associations with overweight
The highest prevalence of overweight was found in Portugal (21.6% for boys and 19.5% for girls) and Spain (17.9% for boys and 18.4% for girls). Lowest prevalence values for overweight were found among Belgian boys (8.3%) and Dutch girls (7.9%).
High TV viewing during dinner and during the day, and low physical exercise behavior (in boys only) was associated with a risk of being overweight (see Table
4). High PC use was not associated with being overweight in boys or girls. These findings did not change after adjustments for family educational level (data not shown).
Table 4
Odds ratios (OR) and confidence intervals (CI) for being overweight (including obesity) for TV viewing during dinner, TV viewing, PC use and physical exercise behavior (adjusted for each other), separately for boys and girls participating in the Pro Children study.
TV during dinner every day | 1.31 | 1.07 | 1.59 | 1.30 | 1.06 | 1.59 |
TV viewing > 2 hours/day | 1.33 | 1.11 | 1.61 | 1.30 | 1.07 | 1.58 |
PC use > 1 hours/day | 1.15 | 0.95 | 1.39 | 1.09 | 0.86 | 1.39 |
Physical exercise ≤ 3 hours/week | 1.57 | 1.29 | 1.90 | 1.17 | 0.94 | 1.46 |
The healthy behavior cluster had the lowest proportions of overweight boys (12.7%, (11.1–14.3%)) and girls (11.4%, (9.3–13.5%)). In girls, the cluster with low sedentary and low physical exercise behavior had the second lowest proportion of overweight girls (12.6%, (11.1–14.2%)) (Table
3).
In boys three clusters had a significantly increased risk of being overweight (Table
3). Only the high TV viewers, who also had high scores on physical exercise, did not significantly differ in risk of being overweight from the boys in the healthy behavior cluster. In girls, the high TV viewers and the high PC users had the highest proportions of overweight, and significantly increased risks of being overweight. The odds ratios (ORs) for the low sedentary and low physical exercise cluster and the high sedentary and high physical exercise cluster were also >1, but not significant. The estimated ORs were not influenced by adjustments for family educational level (data not shown)
Discussion
This international study showed that the prevalence of sedentary and physical exercise behavior varies across countries and by gender, with boys exhibiting more typical recreational sedentary behaviors but also being more physically active than girls. Moreover, these behaviors, except PC use, were associated with overweight status. Furthermore, among boys the unhealthy behavior cluster had the highest risks of being overweight.
The pronounced differences between boys and girls in sedentary and physical exercise behaviors were observed across all countries and have been reported by others [
9,
22‐
24]. However, most studies to date have data only on either sedentary or physical activity behavior, of which some report gender differences and others not [
3,
24,
25]. Moreover, as with the current study, often only typical sedentary behaviors were assessed, such as TV and video viewing and PC use, while Jago et al. [
26] showed that girls engage in activities that are often not assessed, for instance personal care and social interactions. The gender difference in physical activity seems more consistent in the literature [
24,
27‐
29], while contradictory results have been reported regarding TV viewing [
24,
25,
29,
30].
Notable in the present results are the high proportions of low physical exercise behavior among girls. This may be an overestimation caused by the way the question was formulated and the typical physical activities in which girls normally engage. Children can be active in many different ways. Boys may be more often enrolled in organized sports and competitive, team, and high intensity sports while girls spend time on unorganized, non-competitive and medium to low intensity sports and exercises [
9,
28,
29,
31], which were not assessed asked for in the questionnaire. Nevertheless, the HBSC study used a more extensive questionnaire and a cut-off of ≥ 5 days/week of 60 minutes of vigorous physical activity, and their results do not differ greatly from ours [
8]. Adjusting the cut-off in the present study to < 2 hours/week, still results in 76.1% of the Portuguese girls being characterized as low physically active. A study conducted in Porto, Portugal [
32] also reported a high proportion of inactive girls (71.7%) and the HBSC study reported that Portugal was consistently in the bottom quartile regarding physical activity in any age group [
23]. Prevalences of the other behaviors and proportions of overweight in the current study is in line with observations from the HBSC study [
8,
23].
Associations between sedentary behavior and physical exercise are subject to debate [
7,
9,
33]. The 'displacement hypothesis' proposes that sedentary behavior displaces physical activity; however several studies, including ours, only find weak or no associations [
9,
24,
34]. In line with this, we found that boys in the high TV viewers cluster had above average physical exercise levels, and the low sedentary and low physical exercise cluster in girls indicates that a large proportion of girls spends little time on TV viewing and PC use as well as little time on physical exercise. Likewise a small part of the girls sample combines high sedentary behavior with above average involvement in physical exercise (i.e. the high sedentary and high physical exercise cluster). This latter finding supports the idea that some sedentary behaviors and physical activity can co-exist. Physical activity may be executed typically in the afternoon while sedentary behaviours are more prevalent later at night [
7]. The multifaceted character of sedentary behavior and the fact that we measured only a selection of possible sedentary behaviors may have caused the apparently conflicting.
We found that among girls the high TV viewers and the high PC users had an increased risk of being overweight. That TV viewing is positively associated with BMI or overweight status in girls is in line with other studies [
3,
24]. Both clusters also had high proportions of girls reporting low physical exercise (≥ 85%) and almost half of the high PC users spent more than 2 hours/day on TV viewing (data not shown). The main difference between these two clusters and the low sedentary and low physical exercise cluster is the time spent on TV viewing and PC use. Since the low sedentary & low physical exercise cluster did not show an increased risk of being overweight, it appears that for girls sedentary behavior is a more important factor with regard to overweight status than is physical exercise. This hypothesis is also supported by the observation that the high sedentary and high physical exercise cluster showed increased odds (OR>1), although not statistically significant, for being overweight and the proportion of overweight girls in this cluster was the second highest. Absence of statistical significance may be due to a low number of girls in this cluster.
In boys, all clusters had increased risks of being overweight compared with the healthiest behavior cluster, except the high TV viewers. This latter cluster also had the lowest proportion of boys reporting low physical exercise and since physical exercise was associated with overweight status, this may explain the absence of elevated risk. So, in contrast to girls, in boys both sedentary behavior and physical exercise seem to be important with regard to overweight.
It can be argued that the variable TV viewing during dinner is not a real indicator of sedentary behavior and should thus not be included in the present analyses. However, we included this variable, because it can provide new information. Since this variable has not been included in this line of exploration of specific energy-balance behaviors before, since it has been argued that eating while watching television may be one of the reasons for the positive association between TV viewing and overweight, and because in weight management programs it is advised to not engage in distracting activities such as TV viewing while eating. It can be argued that the variable 'TV during dinner' may be associated with weight management and thus overweight. Analyses of the data without this variable, resulted in the same conclusions for boys, but showed a less clear pattern in girls (data not shown). Although, again, in girls TV viewing behavior appeared to be more important than physical exercise behavior with regard to overweight status, which is in line with the present results.
Adjustment for family educational level did not change the estimated ORs for being overweight, indicating that family educational level did not confound these associations. This can probably be explained by the fact that in the multilevel logistic regression analyses results were already adjusted for country, and that the different educational levels were not equally distributed over the nine countries. Spain and Portugal showed higher proportions of lower educational levels, and also showed higher levels of overweight.
Some limitations of the present study should be addressed. Since we solely rely on cross-sectional data, we cannot draw conclusions about causal relations. Children who are highly sedentary might become overweight, or overweight children might become more sedentary. Secondly, we used self-reported data for the behaviors, which is less accurate than observed measures. If children tend to underestimate their time spent on sedentary behaviors, the prevalence of these behaviors would be higher than presented here. If underestimation is more common among overweight children, the observed effect estimates would be higher if more accurate measures were used. However, Schmitz et al. [
35] showed that a brief self-administered questionnaire is as good for group comparisons as a detailed seven-day logbook regarding TV viewing and PC use.
Thirdly, we used parent-reported data to calculate children's BMI. In general people tend to overestimate their height and underestimate their weight, resulting in an underestimation of BMI [
36], which is also more common when parental reporting is used [
37,
38]. A Norwegian study validated a methodology very similar to the Pro Children survey but using self-report by children, completed at home [
39]. In that study, there was a relatively good correspondence between reported and measured height and weight in children of a comparable age to that in our study.
The non-response on BMI was different between the clusters for boys, with the highest rates in the unhealthy behavior cluster (43.1%) and the mixed cluster (42.7%) and lowest in the healthy behavior cluster (36.2%). This suggests that an even clearer pattern would probably have emerged without this selective non-response. The non-response rates on BMI were lower in girls, and also varied less between the clusters (32.7% – 36.6%, not significant).
Besides these limitations, a strength of this study is that it provides representative information from a large sample from nine countries across Europe. A special feature of this study was the cluster analyses, investigating associations with overweight between patterns of behaviors instead of individual behaviors. In our analyses we also took into account that children were nested within countries and observed no random effects by country (data not shown). So, despite the geographical, socio-economic, cultural and climate differences, the associations shown are consistent across all countries.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SJtV carried out the analyses and drafted the manuscript; IdB participated in the design of the study, supported in the statistical analyses and helped to draft the manuscript; IT participated in the design of the study and helped to draft the manuscript; MR helped to draft the manuscript; MH helped to draft the manuscript; KIK developed and directed the overall study and helped to draft the manuscript; JB developed and directed the overall study and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.