This study examined the clustering of activity-related behaviors and eating routines among young children, and is the first to include a literature-based range of eating routines. The fact that all these eating routines clustered with activity-related behaviors and/or other eating routines shows the value of moving beyond interpreting someone’s diet as merely what that person consumes. As such, our study shows the importance of incorporating the context of these behaviors, in order to establish a more informative typology of children with high scores for a particular pattern. At a methodological level, the inclusion of eating routines instead of dietary intake increases the compatibility of these two behavioral categories, since activity-related behavior measures also tend to include context (e.g. differentiating between sports at school and at a sports club). In addition, this study was the first to examine the longitudinal association between these patterns and weight status development.
We identified four behavioral patterns, two of which were cross-behavioral (i.e. covering eating routines as well as activity-related behaviors), while the third only covered eating routines and the fourth only activity-related behaviors. The first cross-behavioral pattern was named the ‘Television–Snacking’ pattern. Children with high scores for this pattern watch much television, often eat with the television on, have a high snacking frequency, and are more likely not to eat at the table. This pattern is similar to a pattern found in older children (9–14 years old) by Te Velde and colleagues , who also found that having the television on during dinner clustered with television viewing in general, as well as less PA. However, they did not include any eating routines other than having the television on during dinner. In addition to the study by Te Velde and colleagues and our current study, numerous clustering studies incorporating dietary intake instead of eating routines have reported similar associations between television use and snacking [1–7, 9]. The association between television viewing or having the television on while eating on the one hand and snacking on the other has previously been attributed to various mechanisms, including the idea that sedentary activities offer a context that promotes passive snacking , the stimulating influence of snack commercials , and the distracting influence of watching TV while eating, disrupting the habituation to food cues (e.g. satiety) . High scores for the Television–Snacking pattern were found to be associated with higher odds of being overweight and having a higher BMI at follow-up. This longitudinal association extends the findings of cross-sectional studies [2, 20], and is in line with previous findings in the current cohort .
A second cross-behavioral pattern we found was what we named the ‘Traditional Family’ pattern. Children with high scores for this pattern frequently use active means of transport, do not skip meals, and often eat together with the family. To our knowledge, such a pattern has not been identified previously, and we think it reflects a typical traditional Dutch family lifestyle. Dutch families traditionally eat all meals together: even during lunch, children often return home to eat together with their family (or part of the family), which is possible because primary schools are generally at walking (or at most cycling) distance from home (average distance: 700 m ). Eating together implies transport (which may include using active means) and could thus provide an explanation for the association between using active means of transport and eating as a family. The Dutch are well-known for their bicycle use, having the highest level of bicycle use in Europe . Viewing this pattern as a traditional lifestyle pattern also fits in with the finding that maternal working hours were negatively associated with the scores for this pattern.
The ‘Sports–Computer’ pattern comprises both high computer use and high levels of PA resulting from the use of active means of transport, engaging in school sports and playing sports at a sports club. This cluster might be explained by the competitive element involved in both sports and computer games, which appeals to certain children, but further research would be needed to confirm this hypothesis. A similar pattern has previously been found in a study by Jago et al. , who examined activity-related behavior patterns in 10- and 11-year-olds. In their study, the group of children having a so-called high active–high sedentary pattern accumulated the highest mean number of minutes of moderate to vigorous PA, even higher than the children in the high activity–low sedentary group. The study by Jago et al.  and our current study therefore both stress that it is important to consider PA behaviors as well as sedentary behaviors when evaluating a child’s activity-related behavior (e.g. for intervention purposes). The current study revealed an increased overweight risk at age 6–7 for children with high scores for this pattern, which could indicate that the Sports–Computer pattern may be problematic in young children. In line with this, a growing body of evidence shows that sitting time might be more predictive of weight status and health than time spent being physically active (see e.g. [42–44]). This underlines the importance of interventions focusing on reducing sedentary time, in addition to promoting physical activity. The fact that television viewing and computer use clustered within different patterns shows the importance of assessing these behaviors separately, and not as one measure of sedentary screen-based behavior. Another reason to assess screen-based behaviors separately is that previous research has reported television use in youngsters to be negatively, not positively, associated with other sedentary behaviors, including computer use .
The fourth pattern found in the current study was a pattern combining a high frequency of consuming take-out meals or eating out with a short average duration of meals. Children with high scores for this pattern were thus literally ‘fast’ food eaters. This pattern was not related to weight status at follow-up, possibly as a result of the low frequency of this behavior, as consuming take-out meals or eating out occurred, on average, only once every 2 weeks in the whole study population.
Various background characteristics proved to be related to the pattern scores. Boys had higher scores for the Television–Snacking pattern and the ‘Fast’ Food pattern, which adds to the findings of studies showing that boys are more likely to have an unhealthy intake pattern [46, 47]. In line with previous studies [1, 9, 48], parental educational level was found to be inversely associated with scores for the Television–Snacking pattern. The Television–Snacking pattern and the Sports–Computer pattern were also positively associated with parental BMI. Interestingly, these two patterns were both also associated with an increased overweight risk for the child at a later age, which could indicate a mediating role of these behavioral patterns in the relationship between parental BMI and children’s BMI, and the intergenerational transmission of overweight and obesity risk. Although previous research assessing individual energy balance-related behaviors as a mediator in the intergenerational transmission of overweight has shown little evidence for such mediation , the examination of behavioral patterns as a mediator may provide additional insights in this respect.
All data used in the current study, including those regarding eating routines, activity-related behavior and anthropometric data, were self-reported by the parents, which may have led to bias. Some have suggested the use of accelerometry instead of self-report data in studies examining clustering of activity-related behaviors in children . However, although accelerometer measurements provide objective data on the intensity and duration of activities, they do not distinguish between different activity types , and were thus unsuitable for the current study. Although the questionnaire we used to assess activity-related behavior had not yet been validated, previous research has shown that parental reports of BMI are generally quite reliable . However, since BMI does not discriminate between lean mass and fat mass, the association we found with BMI could also reflect associations with lean instead of fat mass. The use of additional measures, such as waist circumference , is therefore recommended in future studies using BMI as a measure of childhood overweight. Furthermore, the sampling approach we used meant that families with an ‘alternative’ lifestyle were overrepresented in our study population . Hence, the study population is probably not representative of the general Dutch population, warranting caution when generalizing the results. The children in the current study population had a slightly lower mean BMI than the reference population, for example. On the other hand, all regression analyses were adjusted for the recruitment channel. An additional limitation is the drop-out rate of participants between age 5 and the final follow-up at age 7–8 years, although such drop-out is inevitable in longitudinal studies. Children of mothers with a low educational level and a higher BMI were more likely to drop out, which also limits the generalizability of our findings. Another limitation is that the effect sizes found in the regression analyses were small. A final limitation lies in the analyses. PCA relies on various subjective choices which influence the outcomes, such as the choice of the cut-off point for component loadings. In line with recommendations , we used a cut-off point of 0.4, although cut-off points in previous clustering studies were found to range between 0.2 and 0.6. A different cut-off point would have led to different patterns. The findings of the current study are therefore of an indicative nature, and further examination of cross-behavioral clustering of energy balance-related behaviors in children is needed.