We found that the 24-hour movement behaviour composition was associated with BMI z-score among Australian preschoolers. When examining the association between each movement behaviour relative to the others and BMI z-scores, we observed an association for MVPA (p = 0.023). There were no significant association for the other behaviours, however for sleep the p-value was 0.056. In the overall regression model an interaction effect for weight group was found. In stratified analyses, the association between high BMI z-scores and the overall 24-hour movement behaviour composition did not apply for the group with overweight and obesity.
Consistent with this finding, two previous studies have also reported statistically significant associations between the 24-hour movement behaviour composition and BMI z-scores in pre-schoolers (17, 18). This suggests that the distribution of time spent on each of the 24-hour movement behaviours within a day makes a difference when it comes to healthy weight.
Our finding that more MVPA relative to the other behaviours was related to a higher BMI z-score may appear counterintuitive, considering the more robust evidence among other populations (school-aged children and adults) indicating that more MVPA is associated with lower adiposity (2). In particular, our prediction of higher BMI z-scores when more time is spent in MVPA relative to the remaining behaviours contradicts previous studies in a school-aged population (2). The study of Fairclough et al. (2017), showed higher BMI z-scores when less time was spent in MVPA, relative to either sleep, SB or LPA (15-minute reallocations were associated with + 0.88, + 0.83 and + 0.89 BMI z-score, respectively) (36). However, the effect sizes observed in our current study are notably small, casting doubt on the clinical significance of our findings (a 10-minute reallocation away from MVPA to sleep, SB and LPA collectively was only associated with − 0.073 BMI z-score). The direction of our findings align with an 8 year longitudinal study by Moore et al. (2003) (37). In the latter study, children classified as “high active” at four years old had higher BMI values compared those classified as “low active”. However, the “high active” children had lower BMI gains compared to the “low active” throughout the study. These seemingly contradictory relationships found among preschool populations may be attributed to the onset of adiposity rebound (37, 38). In the current study we were not able to account for adiposity rebound due to the cross-sectional nature. Additionally, we should be aware that the favourable effect of MVPA on healthy weight might need some time to manifest and may become more evident as the child ages. We should also consider that other confounding factors, including muscle mass, dietary intake and recruitment bias, might have influenced the results. Recruitment bias might be induced as participation in the study was voluntary, meaning that it is possible that parents with a particular interest in movement behaviours were more likely to participate, while those who are less concerned about movement behaviours and health may be opted out of the study (39). This could have led to an overrepresentation of preschoolers with parents who prioritize MVPA (39). However, we were not able to control for this. To have a better understanding of the relationship between 24-hour movement behaviour compositions and preschoolers’ adiposity, future studies should consider these possible confounders and include other measures of adiposity (e.g., waist circumference, body fat percentage). In addition, longitudinal studies are recommended and may provide insight in how 24-hour movement behaviours affect adiposity levels as children age. Although the association between 24-hour movement behaviour composition and adiposity in preschoolers is currently uncertain, promoting a healthy amount of time spent in each of the movement behaviours is recommended due to the strong evidence that exists for school-aged children (2).
In contrast to our results, Carson et al. (2017) did not find significant associations with the individual ILR pivot coordinates representing each behaviour, relative to the remaining behaviours among a group of 552 three- to four-year-olds. In line with our results for sleep (although statistically non-significant), Taylor et al. (2018) found significant associations for more sleep, relative to the remaining behaviours, and lower BMI z-scores in 231 three-and-a-half-year-olds. However, they also found an association for SB and LPA, but not for MVPA, each relative to the remaining behaviours. The association between sleep and adiposity, however not significant (p = 0.056), is in line with a systematic review of Chaput et al. (2017) showing that shorter sleep duration was associated with higher adiposity in 20 out of 30 studies in young children from zero to four years old. Nonetheless, these findings are more mixed compared to studies in school-aged children, showing a more robust relationship between shorter sleep duration and higher adiposity (40). This is likely because it might be more challenging to identify associations with adverse health indicators in younger children. The effects might become apparent over time if short sleep duration persists. This could be confirmed with longitudinal studies.
We found an interaction effect for weight group in the overall composition regression model. The association between the overall 24-hour movement behaviour composition and BMI z-score was statistically significant only for the non-overweight group. It could be that the sample size of preschoolers with overweight and obesity (n = 24) was too small to provide enough statistical power to observe significant differences. Therefore, this result should be interpreted with caution. To have a better understanding of the association between the 24-hour movement behaviour composition and BMI z-scores in weight groups, higher number of participants would be beneficial.
Strengths and limitations
The strengths of this study were the use of a compositional analytical approach and the device-based measurement of SB, LPA and MVPA which are less susceptible to recall bias. The study has several limitations. First, there was a large drop-out of participants. The missingness of data was shown to be not at random, limiting the generalizability of our findings. Participants with higher parental education were less likely to have missing accelerometry data. Parents with a lower educational background might have found it more difficult to understand the study instructions, leading to invalid data and questionable generalizability of the study. However, data of a more disadvantaged population is crucial within public health research as this population is more vulnerable for non-communicable diseases (41). Therefore, more effort should be made to make these studies more inclusive. Second, participants only needed one day of accelerometry data to be included, whereas standard procedures recommend minimally four days to obtain high reliability including both week- and weekend days (22). The minimum of one day was decided based on the large drop-out. Third, sleep was derived from a parent reported questionnaire, which might have caused an over-or underestimation of sleep duration due to social desirability or recall bias (42). Fourth, only BMI-z score was used as adiposity indicator, which does not distinguish between excess body fat and bone/muscle mass (43). Additional studies using body composition measures are needed to elucidate such associations. Nonetheless, BMI-z scores take into account standards for age and sex, which is important due to natural developmental changes in weight and height with age and sex, as well as their relation to body fatness among children (44). Previous research using other adiposity indicators, for instance waist-to-height ratio, that do not consider what is normal for a specific age and sex among preschoolers (2, 45). For this kind of indicators, this raises the question to what extent we can rely on the results for a sample of children without taking age and sexes into account. Future research should address the need to use sex and age specific adiposity indicators other than BMI z-scores among preschoolers. Fifth, the study has a cross-sectional design, which prevents interpretation of how adiposity in association with 24-hour movement behaviour compositions manifest through child development. Sixth, we were not able to control the analyses for potential confounding factors such as adiposity rebound or dietary intake.