Introduction
Previous studies have found that physical activity (PA) and sedentary behavior (SB) are associated with adiposity [
1‐
4]. In prospective cohort studies, low PA and high SB at baseline have usually been associated with higher subsequent adiposity [
5,
6]. However, some studies provided evidence suggesting that adiposity may lead to subsequent changes in movement behaviors [
7]; specifically, to reduced PA [
8] and increased SB [
9]. It seems, therefore, that the association between adiposity and movement behaviors may be bidirectional.
Higher PA may plausibly lead to lower adiposity through higher energy expenditure, but conversely higher adiposity may lead to reductions in PA through increased discomfort and lower fitness. The same applies mutatis mutandis to SB.
Importantly, PA and SB are co-dependent. Since there are only 24 h in a day, any change to the amount of time we spend in one of the two behaviors must be compensated by an equal and opposite change in the other behavior and sleep collectively [
10,
11]. It therefore makes more sense to talk about activity compositions rather than individual movement behaviors, and to ask whether composition of activity is related to adiposity. For this purpose, it is recommended to use compositional data analysis [
10‐
14]. Some studies have found significant cross-sectional relationships, with compositions with high levels of moderate to vigorous PA (MVPA) being favorably associated with adiposity [
10]. This relationship has also been explored in several longitudinal studies [
15,
16]. However, there is a lack of compositional analyses of longitudinal data which focus on the association between baseline adiposity and subsequent changes in movement behaviors.
Using a compositional approach, this study, therefore, aimed to examine the prospective associations between baseline adiposity and 7-year changes in PA and SB among elderly women. We also explored the longitudinal associations between change in adiposity and change in movement-behavior composition.
Discussion
Our results suggest that among elderly women a longitudinal change in adiposity over 7 years is associated with a concurrent change in movement-behavior composition. We did not find a significant prospective association between baseline adiposity and 7-year changes in movement-behavior composition.
Longitudinal changes in adiposity and movement-behavior composition found in our study concur with previous findings of age-related increase in adiposity [
5,
6] and time spent in SB [
33‐
35] and age-related decrease PA [
36‐
38]. Furthermore, over the 7-year study period, the elderly women in our sample became more sedentary and less physically active. The increase of SB at the expense of MVPA is broadly considered to increase the risk of adverse health outcomes [
39,
40].
The results of this study did not provide clear evidence of the association between baseline adiposity and subsequent changes in movement-behavior composition. Some previous prospective studies suggest obesity might be a risk factor for decrease in PA [
8] or may lead to increase in SB [
9]. However, these studies included mainly middle-aged populations and did not consider the movement behaviors as compositional data. Given the lack of previous evidence for prospective associations between adiposity and subsequent changes in movement-behavior composition, it is not possible to make direct comparisons of our results with findings from previous studies. Some of the non-significant associations between baseline adiposity and subsequent changes in movement behaviors found in our study sample point in the same direction as the associations found in previous, non-CoDA based studies [
8,
9,
41]. For example, in our sample, higher baseline FM% was associated with an increase of SB at the expense of MVPA and with a decrease of MVPA in favor of SB and LPA. It is possible that these associations were not significant, because the sample in our study was smaller than in the two previous studies [
9,
41]. A post hoc analysis of statistical power revealed that our sample size was large enough to ensure > 0.99 probability of obtaining a significant beta coefficient, if the true effect size in the population is at least medium (
f2 ≥ 0.15) [
42]. The probability of obtaining a significant beta coefficient in case the true effect size in the population is small (
f2 ≥ 0.02) was 0.46, which is relatively low. However, Pulsford and colleagues [
41] did not find a prospective association between baseline adiposity and total sitting time either, despite the fact their study was conducted in a much larger sample. The lack of significant associations in our study and in the Pulsford et al. study [
41] might therefore also be an indication of a very low or no prospective association between adiposity and movement behaviors. Given the inconsistencies in findings between studies, this should be further investigated.
Although there were no prospective associations, we found significant longitudinal change-to-change associations. Our findings suggested that 7-year increases in both BMI and FM% were associated with a concurrent increase of SB at the expense of other movement behaviors. And conversely, longitudinal decrease in BMI was associated with a decrease of MVPA in favor of other movement behaviors. It is possible that the change in adiposity is more important in determining concurrent changes to movement behaviors than starting levels of adiposity. This may reflect older women’s decreasing ability to participate in MVPA and increasing propensity for SB and LPA as their adiposity increases. However, given that these were change-to-change associations, we are unable to determine the direction of this association and more research is needed to confirm these assumptions.
The main strengths of this study are its longitudinal design, the use of CoDA, and the assessment of movement behaviors using accelerometers. Additionally, the carefully selected adjustments for confounding strengthen the findings of this study. This study was subject to several limitations. The sample cannot be considered as fully representative of the population of elderly women in the Czech Republic, Poland and the Slovak Republic. The findings should, therefore, be generalized with caution. Furthermore, we did not find significant differences in baseline movement behaviours between the participants included in the analysis and those who were lost to follow-up. However, the latter group had a significantly higher age (by on average 2.03 years), BMI (by on average 1.1 kg/m
2), and fat mass percentage (by on average 2.3%) compared with those who remained in the study. This may also have reduced the generalisability of findings. There are also limitations related to the assessment of movement behaviors. We did not measure sleep duration, despite the fact that sleep has been found to be associated with adiposity [
43] and is co-dependent with SB, LPA and MVPA. A potential limitation may stem from our choice of the threshold for non-wear time. The large amount of time that older adults spend in sedentary behaviours, increases the likelihood of misclassifying sedentary time as non-wear time [
44]. Some authors have, therefore, suggested using longer thresholds for non-wear time for older adults [
45]. Despite that, the 60-min threshold that we used in the current paper remains the most commonly used threshold in epidemiological studies conducted in this age group [
22]. Furthermore, the accelerometer cut-points used to determine the time spent in movement behaviors were validated in samples with somewhat different characteristics compared with our study sample. We also did not consider other characteristics movement behaviors such as bout length, timing and consistency, some of which have been associated with adiposity in older women [
15]. Lastly, although the regression models were adjusted for illness or disability occurrence before baseline and during the 7-year follow-up, it may be that some aspects of illness and disability were not fully captured by this question. This may have led to residual confounding.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.