Introduction

Obesity is a global problem and is no longer a disease that affects adults only (Washington, 2008). In children and adolescents who are obese, not only metabolic and cardiovascular health is impaired (TNO, 2011), but psychological well-being is often decreased as well (Gately, Cooke, Butterly, Mackreth, & Carroll, 2000; Van der Baan, Nollet, Weller, Benninga, & Van Aalderen, 2010).

Three major factors contribute to childhood obesity: genetic predisposition, unfavorable environmental conditions, and a disturbed energy balance/imbalance of energy homeostasis (Ebbeling, Pawlak, & Ludwig, 2002). This is the result of a combination of a high energy intake and relatively low energy expenditure (Kremers, Visscher, Seidell, Van Mechelen, & Brug, 2005). In line with these findings, physical activity programs have been introduced to combat overweight and obesity. However, school-based meta-analyses show that these interventions have little effect on body weight or body mass index (BMI; Harris, Kuramoto, Schulzer, & Retallack, 2009), blood pressure (Guerra, Nobre, Silveira, & Taddei, 2013), or physical activity level (Metcalf, Henley, & Wilkin, 2012).

Materials and Methods

In the current study, we investigated associations between weight status and physical activity, and psychosocial determinants. We included measures of strength and aerobic exercises (Ten Hoor et al., 2014, 2016), and impulsive and reflective determinants (Houben & Wiers, 2008). Each child’s BMI was recoded into a BMI z-score (this is the BMI score but then age and gender specific, compared to the national reference population; Cole, Bellizzi, Flegal & Dietz, 2000; Cole, Flegal, Nicholls, & Jackson, 2007). Although use of the term ‘obesity’ implies dichotomization of youngsters into ‘obese’ and ‘non-obese’ youngsters, the underlying variable is continuous; in our study we will refer to this continuous underlying variable as weight status. For descriptive purposes, we used the cut off points as defined by the WHO Growth reference 5–18 years, to determine weight status (BMI z-score <1 = normal weight, > 1= overweight/obese. All research materials, data, syntax files, and output files are available in Additional File 1.

Measures

The implicit association test

In this study, two Implicit Association Tests (IATs; [26]) were completed. The first measured whether participants implicitly preferred physical inactivity or physical activity. In the second IAT, we measured preferences for either aerobic, or strength exercises. The IAT is a computerized response latency measure that rests on the assumption that the more strongly two concepts are associated with each other, the faster and more accurately they can be paired together. In Figure 1 (schematic display of seven computer screens) the IAT procedure is explained. The IAT consisted of five practice blocks, and two trial blocks. In the first practice block (first picture of Figure 1), participants learned to respond to attribute pictures (appearing in the center of the computer screen) by classifying these pictures into two categories. In this study, the attribute categories were a picture of a happy and sad emoticon, which were positioned on the left and right side at the top of a computer screen. The attribute pictures that appeared one by one in the center of the computer screen had to be classified into the attribute categories with either the left (E) or right (I) response key. In this way, pictures were linked to relevant concepts. The instruction for participants was to assign the pictures as quickly and as accurately as possible to the correct category by pressing the appropriate response key. The assignment of the left key (E) and right key (I) was counterbalanced across participants. With the practice blocks, all images were linked to the relevant concepts. In practice block 2 (picture 2), participants continued categorizing pictures, but now they had to sort pictures of physical activity and physical inactivity into the correct category (in the first IAT; in the second IAT, pictures of aerobic exercises and strength exercises were used). In practice block 3 (picture 3), and test block 4 (picture 4), youngsters continued to assign pictures to the correct category, but now the categories of block 1 and 2 were combined. Reaction times in block 4 were used to calculate one’s implicit association. In practice block 5 (picture 5), youngsters responded again to happy and sad emoticon, but now, the emoticons were mirrored (contrary to block 1,2,3,4, the happy emoticon was on the right, and the sad emoticon on the left). In practice block 6 and test block 7 (picture 6 and 7), the pictures of physical activity and physical inactivity (or aerobic exercises and strength exercises in IAT 2) were added, but not mirrored. Where in block 3 and 4 the picture of aerobic exercises was combined with a sad emoticon, it was now combined with a happy emoticon. Where in block 3 and 4 the picture of physical activity (or strength exercises) was combined with a happy emoticon, this was now combined with a sad emoticon. Based on the difference in mean sorting reaction time between block 4 (happy/physically active vs. sad/physically inactive) and block 7 (sad/physically active vs. happy/physically inactive), corrected for incorrect responses, a difference score was calculated based on the d-600 algorithm (Greenwald, Nosek & Banaji, 2003); see also SPSS syntax in additional file 1). For the first IAT, a calculated difference score < 0 reflected a more positive implicit reaction to physical inactivity compared to physical activity. A score > 0 reflected a more positive implicit reaction to physical activity compared with physical inactivity. A score of 0 reflects no, or ambivalent preferences for both physical activity and physical inactivity. For the second IAT, a score < 0 means a more positive association with aerobic exercises, and a score > 0 a more positive association with strength exercises.

Figure 1 

Phases in the Implicit Association Task.

The Questionnaire

Demographics (gender, age, height, and weight), and reflective determinants of physical activity, strength exercises and aerobic exercises were measured using a self-report questionnaire. All items were measured on a 100-point visual analogue scale, except the demographics, BMI, how often they exercised, and self-image. When needed, scores were recoded such that a higher score reflected more agreement with the variable. Scores on items that measured the same construct were averaged into one scale when the internal consistency was sufficient (alpha > .60). The questionnaire about the reflective determinants contained the variables listed below, derived from the Reasoned Action Approach, and the Integrated Behavioral Model (Montaño & Kasprzyk, 2015). For all questions, pictures were added to make sure that the children understood what was meant by the question (see Additional File 1: Complete Study Design).

Self-image

Questions were asked about self-image: 1) “do you see yourself more as a physically active person” (score = 1) or “a non-physically active person” (score = 2), and 2) “do you see yourself more as someone more suited to aerobic exercises” (score = 1) or “strength exercises” (score = 2).

Attitude

Participants’ evaluations of 1) physical activity, 2) physical inactivity, 3) aerobic exercises, and 4) strength exercises, were each measured with two unipolar sematic-differentials with the labels “unpleasant–pleasant” and “bad-good”.

Subjective norms

Two items measured the extent to which participants believed important others think they should exercise: “My parents want me to exercise” and “My friends want me to exercise”. Response categories ranged from “absolutely disagree” to “absolutely agree”.

Perceived behavioral control and intention

For 1) physical activity, 2) aerobic exercises, and 3) strength exercises, a total of three questions were asked about perceived behavioral control: “If you want to, could you be: 1) more physically active 2) carry out more aerobic exercises, and 3) carry out more strength exercises”, and three questions about the intention: “Are you going to be 1) more physically active, 2) carrying out more aerobic exercises, and 3) carrying out more strength exercises”. Response categories ranged from “absolutely disagree” to “absolutely agree”.

Physical activity behavior

Participants were asked to report: 1) how often they exercised per week and 2) how many hours they spent exercising every time they exercised. From these reports, we calculated how many hours per week were spent exercising.

Analyses

IBM SPSS statistics 20 was used for data preparation and R (R Core team, 2014) was used for data analysis. To measure performance on the implicit association tests, d-600 scores were calculated using the d-600 algorithm (Greenwald et al., 2003; see also the SPSS syntax file in Additional File 1). First, as a crude approximation of matching, we examined the difference between participants who would be classified as overweight/obese (BMI z-score ≥1) in the general and clinical samples (World Health Organization, 2015). Subsequently, to examine whether determinants of physical activity differ depending on BMI z-score, correlations (and 80% confidence intervals) were computed between BMI z-scores and attitudes, subjective norms, perceived behavioral controls, intentions, and impulsive determinants. We report both 80% confidence intervals of effect sizes (following Cohen’s (1990) recommendation as well as to prevent reading of the confidence intervals as ‘significance testing boundaries’, a problem of 95% confidence intervals) and Null Hypothesis Significance Testing (NHST) outcomes, based on the problems reported with NHST (e.g., Masicampo & Lalande, 2012)) and pleas to remedy this problematic approach (e.g., Denis, 2003).

Results

Youngsters participating in the outpatient lifestyle intervention program (COACH) might have been positively influenced by their program. Therefore (as a crude approximation of matching) before we executed correlational analyses for the total group, we split up the participants in three groups: the overweight and obese participants from COACH, the overweight and obese participants from the online panel (BMI z-score >1), and the rest of the general sample (BMI z-score > = 1). After that, we tested the differences between the three groups with regard to the demographic, reflective and impulsive determinants (see Table 1). The highest difference among the three subgroups has a p-value = .02 (uncorrected for multiple testing; see Additional File 1). We therefore concluded that there does not seem to be a systematic difference between the three subsamples.

Combining the COACH sample with the participants from the general sample (from Flycatcher) resulted in a final sample where the BMI z-score was bimodally distributed (see Figure 2). No associations with BMI z-scores were found for gender (Cohen’s d = .19 [–.17; .55], p = .30). Because the (mainly overweight and obese) participants from the COACH clinic were younger than the participants from the online panel, BMI z-score decreased with age. Therefore, when computing the correlation coefficients with BMI z-score, all variables were first corrected for age (besides the already for age corrected BMI z-score). Interested readers can refer to the uncorrected scores and the analyses per subgroup in Appendix 1.

BMI z-scores were not correlated with attitudes towards physical activity (r = – .05 [–.22, .12], p = .75), physical inactivity (r = .06 [–.11,.23], p = .71), strength exercises (r = –.02 [–.19,.15], p = .84), or aerobic exercises (r = –.09 [–.26,.08], p = .48) (see Figure 3). Although BMI z-scores were unrelated to subjective norms of parents (r = .16 [–.01, .32], p = .16), they were related to subjective norms of friends (r = .29 [.13, .44], p = .01): the higher the BMI z-score, the higher the perceived expectations of friends that the youngster should exercise. BMI z-scores were not related to perceived behavioral control regarding strength exercise (r = .11 [–.06, .27], p = .41) or aerobic exercise (r = .14 [–.03, .30], p = .25), but a positive correlation with perceived behavioral control related to exercise in general was found (r = .23 [.06, .38], p = .03): As BMI z-scores increased, youngsters were more confident that they could be more physically active if they wanted to. BMI z-scores were positively correlated with the three intention measures: the higher the BMI z-score, the higher the intention to be more physically active (r = .30 [.14, .45], p = .01), more engaged in aerobic exercises (r = .25 [.08, .40], p = .02), and more engaged in strength exercises (r = .23 [.06, .38], p = .03). There were no differences in BMI z-scores between the self-image groups (physical inactivity vs. physical activity: Cohen’s d = .01 [–.34, .35], p = .98; aerobic vs. strength: Cohen’s d = 26 [–0.12, .63], p = .28, note, these are not displayed in Figure 3 due to the different metrics). BMI z-scores were not correlated to whether participants’ reaction times implied stronger associations with either physical inactivity or physical activity (IAT1: r = .09 [–.21, .13], p = .75), or aerobic or strength exercise (IAT2: r = .12 [–.13, .21], p = .75). No correlation between BMI z-score and self-reported physical activity was found (r = –.01[–.18, .16], p = .91) (See Figure 3).

Discussion

In this study, associations between weight status and physical activity, psychosocial impulsive, and reflective determinants of exercise behavior) in youngsters (8–18 years old) (distinguishing between strength exercises and aerobic exercises were examined. Despite finding significant associations between social norms of friends, perceived behavioral control for exercise, and all three intentions, our first finding is that there are not many significant correlations between BMI z-scores and either the measured reflective determinants, or impulsive determinants (IAT scores). This may imply that a) there really are very few associations between physical activity determinants and weight status (Craeynest et al., 2005), or that b) explicitly expressed cognitions are influenced by social desirability effects (Larson, 2000). Regardless of whether the participant was recruited via the online panel or COACH, participants with high BMI z-scores may well have been aware of the need for exercise and what is expected of them. However, the implicit measures which are not thought to be influenced by social desirability (Greenwald et al., 1998) did not show differences either, in line with the results of Craeynest et al. (2005). Moreover, inspection of the means for the three subpopulations (Table 1) does not suggest a systematic difference in terms of the youngsters participating in a lifestyle program giving more socially desirable answers than overweight/obese youngsters not in treatment.

What we do find is that youngsters with a higher BMI z-score are more in favor of exercise than youngsters with lower BMI z-scores: higher BMI z-scores were correlated with more positive norms for exercise (friends). They were also associated with higher perceived behavioral control for exercise and higher intentions to exercise (both aerobic and strength exercises). Youngsters with a lower BMI z-score do not have more negative correlations with exercise than youngsters with a higher z-scores in terms of concepts from current theories on reflective determinants (e.g. Montaño & Kasprzyk, 2015). This is contrary to the findings of De Bourdeaudhuij, et al. (2005) who reported that adolescents who are overweight had less favorable psychosocial correlates related to physical activity as compared to adolescents of a normal weight (BMI cut off point = 25). However, they also reported that the strength of associations between psychosocial variables and total physical activity were comparable in overweight/obese and normal weight adolescents. When we analyzed our data without correcting the BMI for age and gender, we again found few differences, but they are in the same direction as those found by De Bourdeaudhuij et al. (2005). In their sample, BMI scores were not corrected for age and gender, which makes it harder to compare the results of the two samples (Flegal & Ogden, 2011). Moreover, in another study from the same research group (Craeynest et al., 2005) no differences were found in reflective or impulsive determinants between children with obesity and a matched control group. An alternative explanation is that the lifestyle program could reasonably be expected to raise the participants’ self-efficacy for physical activity (i.e., through experience of physical activity) and their perception that their friends (e.g., from the program) think they should engage in physical activity. Resultantly, an increase in intention might also be expected. These effects could be driving the observed relationships between BMI z-score and subjective norm, perceived behavioral control, and intention. Inspection of the patterns of means in Table 1 suggests this is not the case for subjective norm but could well be the case for perceived behavioral control and intention. In both cases, it seems that the program effect might be driving most of these observed effects, but that some association might remain in the absence of this program effect.

We did not find significant correlations between BMI z-scores and either of the two implicit measures, and only a few between BMI z-scores and reflective determinants. We expected to find a difference between strength and aerobic exercises for overweight youngsters, but we did not. One explanation for this is that in order for these youngsters to find out that they are better suited to strength exercises, they need to experience strength exercises more often (Ten Hoor et al., 2016, 2017b). Targeting strength exercises might be more effective in the long-term, as this could reinforce the biological and psychological strengths of youngsters who are overweight/obese (Ten Hoor et al., 2014, 2016). Recently, other researchers have argued along the same lines (Schranz, Tomkison & Olds, 2013). The review by Schranz et al. (2013) about children and adolescents who are overweight provided empirical support for the (large) effects of strength training on muscle strength, in addition to the small effects of strength training on body composition. However, the effect of strength training on psychosocial outcomes remains unclear because of the limited number of studies addressing this topic. At least, our study does not show negative correlations with strength exercises in youngsters with a higher weight status.

Conclusion

Based on the reflective determinants, differences between normal weight youngsters and youngsters who are overweight or obese are small. Exercise intentions, including those for aerobic and strength exercises, were positively correlated with BMI z-scores. Youngsters with a higher BMI z-score often reported that their friends think they should exercise, and were more confident that they could exercise more if they wanted to. No other determinants, including impulsive determinants, were correlated with BMI z-scores, suggesting that children and adolescents with a different weight status can be targeted by the same intervention.

Additional File

The additional file for this article can be found as follows: