In this study of schoolchildren living in West Sweden, we were able to observe longitudinal effects over two years and cross-sectional trends over a five-year period. Consumption of SSB increased in children with low SEP in the longitudinal sample and decreased cross-sectionally in children with high SEP; this resulted in socioeconomic differences. Sedentary behavior increased both cross-sectionally and longitudinally, while sports participation increased only longitudinally.
Age, time, and SEP aspects of lifestyle
In two recent review articles, consumption of SSB was linked to overweight and obesity in children [
19,
20]. Consumption of SSB has also been linked to difficulties in meeting dietary guidelines of several micronutrients [
21]. Therefore, consumption of SSB could be a marker for unhealthy dietary habits. In our longitudinal sample, we observed a doubling of SSB consumption from ages 7–9 years to ages 9–11 years in the low SEP group. An American study including somewhat older children than in our study, observed that mainly boys increased their consumption of SSB over five years, even after adjusting for SEP [
22]. The Swedish part of the WHO study Health Behaviour in School-aged Children (HBSC) also found that in older children than in the current study, boys more often than girls consumed SSB on a daily basis [
10]. These differing results may reflect that gender plays a greater role in lifestyle choices among older children, while younger children are more influenced by the socioeconomic position and the lifestyle of their family. Children also gain more independence as they grow older and are able to purchase their own food to a larger extent.
In our cross-sectional samples, we found that SSB consumption decreased in children with high SEP and was stable in those with low SEP at ages 7 to 9 years. In a repeated cross-sectional study where Norwegian 11- and 12-year-olds answered a questionnaire about usual beverage habits, consumption of SSB decreased in children with both low and high parental education between 2001 and 2008 [
23]. Studies from the United States have found that SSB consumption in children has decreased [
24,
25] or has been stable [
26] during the last ten years. Similar to our study, others have found that consumption of SSB was more prevalent in low SEP groups [
3,
23,
24]. An American study examining cross sectional trends in food and beverage consumption from the 1980s to 2010 found that SSB was a major contributor to energy intake in children aged 2–18 years [
27]. There was an increase in energy intake and consumption of SSB from 1989 to about 2004, at which time a decline in both was observed. Although this study did not report trends in SSB consumption by SEP groups, it was concluded that the observed decrease in energy intake only occurred in high income families and not in low income families [
27]. In that study, other top ten contributors to energy intake were foods such as pizza, snacks and deserts. This is similar to a Swedish national survey of children’s dietary habits [
28]. In 2003, 25% of children’s energy intake came from energy dense foods with low nutritional value such as SSB. Considering that we observed an increased socioeconomic gradient in consumption of SSB both in cross sectional and longitudinal observations in West Sweden, it is important to monitor changes in subgroups and on a national level.
Due to the nature of our food frequency questionnaire we could not assess the amount of fruit and vegetable children consumed [
16]. In 2008, the highest frequency available to choose from was once every day, which does not give us an estimate of habitual daily intake and is difficult to compare to the dietary recommendations for fruit and vegetable consumption, which in Sweden is 400–500 grams every day depending on age [
29]. We did, however, find that around 60% of children ate fruit daily, with a declining longitudinal trend. Girls and children with high SEP had a higher consumption than boys and children with low SEP, which is consistent with findings in one review article [
30] that also found that younger children usually have higher fruit consumption than older children. The Swedish HBSC also found that boys more seldom ate fruit on a daily basis than girls [
10].
In children followed longitudinally, the number of days participating in sports increased in all children, and there was a consistent SEP difference with higher sports participation in the high-SEP group. This is consistent with findings in a Swedish study of longitudinal changes in lifestyle over two years [
31]. That study included three cohorts in Stockholm County and observed increased sport participation in all age groups including children in grade 2 [
31]. An Australian survey using accelerometers found that moderate and vigorous after-school physical activity decreased over five years in two cohorts of 5-to-6-year-olds and 10-to-12-year-olds [
32]. The use of objective and more detailed measures of physical activity in that survey may explain the discrepancies with our survey. Furthermore, they also studied duration of physical activity while we only asked for number of days of sports participation, which does not give any information of duration or intensity [
32]. In the cross-sectional sample, the SEP gradient in sports participation disappeared due to increased sports participation in the low SEP group. A qualitative study including schools from different socioeconomic areas in a Scottish city found that lack of easy access to sport facilities and costs were factors limiting sports participation in children in low-SES areas [
33]. We could not access this kind of information about our study area, but it would be of interest to investigate whether such changes influenced the increase in sports participation in the low SEP group.
In our previous study on the nationally representative sample in 2008, we found that children’s overweight and obesity were associated with high levels of sedentary behavior, similar to other’s observations [
1,
34]. In the current study, in which the baseline data collection was part of the national survey, we observed an increase in sedentary behavior longitudinally as well as between cohorts, deriving mostly from a change in screen time (data not shown). If this increase continues into adolescence, there is a risk that it will lead to excess weight gain in the group with high screen use [
35]. Since sedentary behavior was consistently higher in the low SEP group, this could in extension lead to further social differences not only in sedentary behavior but also in obesity.
Trends in overweight and obesity
The prevalence of overweight and obesity was higher when applying the WHO 2007 reference compared to the IOTF reference. This is commonly reported and is due to differing cut off points in the two references [
15,
36]. Whether applying the WHO 2007 reference or IOTF reference, the children followed longitudinally remained in their weight class, indicating stability in the growth trajectory. Other short- and long-term surveys following children longitudinally have also reported a significant agreement between the weight classifications over time, which is in line with our findings [
37,
38]. In the repeated cross-sectional samples, prevalence of overweight and obesity was stable overall, but an increase in overweight in girls was observed when applying the IOTF reference but not the WHO 2007 reference. Both references for weight classification revealed stable SEP gradients in overweight and obesity consistent with findings in the national sample from 2008 [
11]. Other studies, including our previous regional study on a similar sample of children, indicate that the social gradient in weight status may be increasing [
15,
39]. In our previous study, an area proxy for socioeconomic status was used [
15]. Including questionnaire based information reduces the number of participants and contributes to a selected sample. However, similar socioeconomic gradients were observed whether we used the area classification or the individual SEP information as in the current study.
Strengths and limitations
The baseline of the current study includes the regional part of WHO Childhood Obesity Surveillance Initiative. Even though we were not able to implement a continuous national surveillance system, we could perform two additional data collections in the large region of West Sweden. All data collections followed the same protocol as other participating COSI countries, including standardized measurements, trained staff, and common questionnaires. We were able to detect changes in weight classification and lifestyle both overall and when stratifying by SEP and gender. In the longitudinal sample, despite the relatively short time period, we seem to have captured a pivotal age when children’s lifestyles are susceptible to change and when SEP gradient may increase. Limiting the cross-sectional part of this study is the fact that we were forced to change the method for how the FQ was returned in the last study year. This could have contributed to the lower participation rate in 2013 and to the change in distribution of maternal education. Maternal employment rates were, however, stable. Further, results in the current study differ from results when analyzing overweight in the 23 schools in West Sweden that were included all three years [
15]. As discussed earlier, this may reflect differing participation rates which in turn may bias the results. Decreasing participation rates are a growing problem in epidemiological research when using questionnaire-based information and avoiding this problem may prove difficult [
40]. However, since we have anthropometric information on children without FQs, we can attempt to account for the differing prevalence. Thus, to avoid underestimating or overestimating the trends in lifestyle variables associated with weight status, we adjusted the analyses for overweight. Furthermore, the adjustment did not affect the results. An additional limitation was the different approach in the process of parental consent among the three years. However, sensitivity analyses showed that the data collected from children measured with active consent did not affect our results. Our questionnaire data on physical activity and sedentary behavior cannot be used to assess energy expenditure or duration and intensity of the activity. However, there are several methodological issues to consider when assessing physical activity. Objective measures gain more detailed information over a short period of time, whereas questionnaire-based methods have the ability to include a larger number of participants and assess habitual activity. Finally, the study is limited by the lack of validation of the FQ developed by the WHO. We did however, find similar SEP gradients in physical activity and consumption of SSB as others have [
23], and in the previous national sample we observed association between sedentary behavior and overweight, similar to other studies [
1,
34].