Health Reports
Reallocating time between sleep, sedentary and active behaviours: Associations with obesity and health in Canadian adults
Archived Content
Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.
by Rachel C. Colley, Isabelle Michaud and Didier Garriguet
Moderate-to-vigorous physical activity (MVPA) and sleep are positively associated with adults’ health,Note 1Note 2 whereas the association with sedentary behaviour (SED) is negative.Note 3 In addition, light-intensity physical activity (LPA) is emerging as an independent predictor of improved cardiovascular health.Note 4Note 5
While associations between each of these factors and health have been examined separately, interest in relationships between health and movement behaviours collectively has increased.Note 6Note 7 This reflects an acknowledgement that participating in a specific intensity of movement means not participating in another. As noted by Mekary et al., “The benefits of different activities depend not only on the specific activity, but also on the activity it displaces.”Note 8
Isotemporal substitution is a statistical technique that allows associations between increasing a given movement intensity and a health outcome to be estimated while considering the intensity of movement being replaced.Note 8 Research using isotemporal substitution has demonstrated that replacing SED with more intense movement is associated with improvements in cardiovascular disease riskNote 9Note 10Note 11Note 12Note 13Note 14Note 15 and with reduced mortality.Note 16Note 17 Stronger effects have been found for MVPA than for LPA.Note 10Note 11Note 14Note 15Note 16Note 18 In two studies of healthy adults, no benefit was observed when substituting LPA for SED.Note 12Note 14 However, in an older population, substituting LPA for SED was associated with better health, which suggests that LPA may be more beneficial for specific populations.Note 19
Most studies using isotemporal substitution have examined physical health outcomes; only a few examined perception of health, which is an indicator of quality of life and an independent predictor of morbidity and mortality.Note 20 Generally, inactive people are more likely to report poor or fair health,Note 21Note 22 and reallocation of time from SED to LPA has been associated with better general health in older adults.Note 19
Associations between various movement intensities and mental health are less consistent. A strong association between SED and psychological distress was observed by Hamer et al.,Note 23 who also found a strong association between high LPA (but not MVPA) and reduced psychological distress.Note 23 Increasing MVPA and decreasing SED have been associated with lower odds of depression,Note 24 and in another study, replacing TV watching with brisk (but not slow) walking was associated with a lower risk of depression in women.Note 18
This analysis uses isotemporal substitution regression models to examine how reallocating time among various movement intensities was related to indicators of obesity and self-rated general and mental health in a nationally representative sample of Canadian adults.
Data and methods
Data source
The Canadian Health Measures Survey (CHMS), an ongoing survey conducted by Statistics Canada, collects self-reported and measured health information from the household population aged 3 to 79. Residents of Indian Reserves, institutions and certain remote regions, and full-time members of the Canadian Forces are excluded. This analysis used data for adults aged 18 to 79 from the first four CHMS cycles: Cycle 1 (2007-to-2009), Cycle 2 (2009-to-2011), Cycle 3 (2011-to-2013), and Cycle 4 (2013-to-2015) (n = 10,621 respondents). Respondents answered an interviewer-administered questionnaire in their home and visited a mobile examination centre (MEC) to undergo physical measurements. Ethics approval for the CHMS was obtained from Health Canada’s Research Ethics Board. Details about the CHMS are available elsewhere.Note 25Note 26Note 27Note 28
Preparation of the dataset
Data from Cycles 1 to 4 were combined. Respondents aged 18 to 79 were included if they had valid accelerometer data and complete data for the outcome variables; 245 were deleted because of incomplete data for key variables.
Measurement of sleep, sedentary behaviour and physical activity
Sleep duration was self-reported as the number of hours respondents usually spend sleeping in a 24-hour period, excluding time spent resting.
Upon completion of the MEC visit, ambulatory respondents were asked to wear an Actical accelerometer (Phillips – Respironics, Oregon, USA) over their right hip on an elasticized belt during their waking hours for 7 consecutive days. A valid day was defined as 10 or more hours of wear time; a valid respondent was defined as a minimum of 4 valid days.Note 30 Wear time was determined by subtracting nonwear time from 24 hours. Nonwear time was defined as at least 60 consecutive minutes of zero counts, with allowance for 1 to 2 minutes of counts between 0 and 100.Note 30 Published movement intensity thresholds were applied to the data to derive time spent in SED,Note 31 LPA, and MVPA.Note 32 A complete description of the accelerometer data reduction procedures is available elsewhere.Note 25Note 26Note 27Note 28Note 30Note 33
Health outcome variables
Body mass index (BMI) was calculated as measured weight in kilograms divided by measured height in metres squared (kg·m-2). Because of a change in the waist circumference (WC) measurement protocol after the first CHMS cycle, a correction factorNote 34 was applied to the Cycle 1 WC data to ensure comparability with subsequent cycles.
Perceived general and mental health were included as categorical variables and were based on the following questions:
- “In general, would you say your health is ...” (poor/fair/good/very good/excellent)
- “In general, would you say your mental health is ...” (poor/fair/good/very good/excellent)
Covariates
Age, sex, highest individual education, and household income were used as covariates in the regression models. Highest individual education was coded as: less than secondary school graduation, secondary school graduation, some postsecondary, and postsecondary graduation. Household income was imputed in all cases where it was not reported in the interview.Note 25Note 26Note 27Note 28 Accelerometer wear time (sum of SED, LPA, MVPA) was included as a covariate in the single models, but not in the partition or isotemporal substitution models because of multicollinearity.
Regression model descriptions
For continuous outcome variables (BMI, WC), linear regression was used. For multi-categorical outcome variables (self-rated general and mental health), generalized multi-logistic regression was used. All movement variables (time spent on sleep, SED, LPA, MVPA) were divided by 30 before analysis to facilitate interpretation of results. All beta coefficients represent the effect size of 30 minutes of a movement variable.
Single regression models were used to estimate the association between time spent at a given movement intensity with the outcome variables, while controlling for covariates (including wear time), but not for time spent at other movement intensities.
Partition regression models were used to estimate associations between time spent at a given movement intensity and the outcome variables, while controlling for the other movement variables and covariates (excluding wear time due to multicollinearity). Although the movement data were not complete (did not sum to 24 hours), inclusion of all movement variables in the same model raised multicollinearity concerns. Multicollinearity risk was assessed using correlation analysis and verifying variance inflation factors. MVPA was correlated with LPA (Rho = 0.19, p < .0001) and SED (Rho = -0.12, p < .0001). LPA was correlated with SED (Rho = -0.22, p < .0001) and sleep (Rho = -0.064, p < .0001). SED was correlated with sleep (Rho = -0.051, p < .0001). MVPA and sleep were not correlated. The variance inflation factors were all less than 2 using the “VIF TOL COLLINOINT” option in PROC REG in SAS. Despite the low variance inflation factors, four partition models were completed with one movement variable dropped each time to mitigate the multicollinearity risk. A final model including all variables was also completed:
- Partition Model 1: LPA – MVPA – Sleep
- Partition Model 2: MVPA – Sleep – SED
- Partition Model 3: Sleep – SED - LPA
- Partition Model 4: SED – LPA - MVPA
- Partition Model 5: SED – LPA – MVPA – Sleep
Isotemporal substitution models were used to estimate the effect of substituting a specified amount of time in one movement intensity for another, while controlling for total time (SED + LPA + MVPA + sleep and covariates except wear time). For example, in the isotemporal model examining reallocation of SED to LPA or to MVPA, the model drops SED but includes LPA, MVPA, total time, and the other covariates. The beta coefficients for LPA and MVPA, therefore, represent the result of substituting 30 minutes of SED with LPA or MVPA, respectively. Additional isotemporal substitution models were run on a sample split by age [18 to 49 (n = 5,990) versus 50 to 79 (n = 4,631)] and by obesity status [underweight or healthy weight (n = 3,985) versus overweight or obese (n = 6,636)].
Analytical parameters
The data were analyzed with SAS 9.3 (SAS Institute, Cary, NC) and SUDAAN 11.0 using appropriate denominator degrees of freedom (46) for the full sample in the SUDAAN procedure statements. Survey and bootstrap weightsNote 25Note 26Note 27Note 28 were used in the variance estimations and calculation of confidence intervals to account for the survey design and to adjust for non-response (average response rate is about 40%).Note 29
Results
Descriptive characteristics
The analysis was based on 10,621 adults aged 18 to 79 (48% male), with relatively equal sample sizes by age group (Table 1). Average daily MVPA decreased with increasing BMI (Table 2) and was lower among individuals whose WC exceeded the threshold for metabolic syndrome (Table 2). Average daily MVPA rose as self-rated general health moved from poor/fair to excellent (Table 2). The average number of daily minutes of MVPA was stable across response categories of self-rated mental health (Table 2).
Single regression models
SED and MVPA were significantly associated with BMI and WC in the single regression models (Table 3). The effect size for MVPA was negative and of greater magnitude for BMI (-1.36 kg·m-2) and WC (-1.57 inches) than for SED (0.11 kg·m-2 and 0.17 inches). A 30-minute increment in SED was associated with an increased likelihood of reporting poor/fair rather than excellent general and mental health. A 30-minute increment in LPA or sleep was associated with a decreased likelihood of reporting poor/fair rather than excellent general and mental health. A 30-minute increment in MVPA was associated with a decreased likelihood of reporting poor/fair rather than excellent general, but not mental, health.
Partition regression models
MVPA was associated with BMI and WC in all partition models (p < .001) except Model 3 (sleep-SED-LPA), where it was the dropped variable (Table 3). The effect size remained stable across the various models (BMI range: -1.36 to -1.37 kg·m-2; WC range: -1.56 to -1.58 inches). All movement variables were associated with a decreased likelihood of reporting poor/fair rather than excellent general health, except for SED, which was significant (in the opposite direction) only when LPA was excluded from the model (Model 2). SED was associated with an increased likelihood of reporting poor/fair rather than excellent mental health in all models. LPA and sleep were associated with a decreased likelihood of reporting poor/fair rather than excellent in all models. MVPA was not associated with mental health in any model.
Isotemporal substitution: BMI and WC
The isotemporal substitution models showed that reallocation of 30 minutes from SED, LPA or sleep to MVPA (increasing movement) was associated with a 1.28 to 1.38 kg·m-2 lower BMI (p < .001) (Figure 1). Results were similar for WC: a 30-minute reallocation from SED, LPA or sleep to MVPA was associated with a 1.49- to 1.56-inch smaller WC (p < .001). Time reallocations between sleep, SED and LPA yielded modest, non-significant results for BMI and WC (Figure 1).
Isotemporal substitution: Self-rated general and mental health
Reallocation of 30 minutes from SED, LPA or sleep to MVPA decreased the odds of reporting poor/fair rather than excellent general health (OR = 0.35 to 0.42, p < .001) (Figure 2). The odds were also decreased if SED was transferred to LPA or to sleep, but not as much as a transfer to MVPA (OR = 0.84 to 0.86, p < .001). Time reallocations to MVPA were not significantly related to self-reported mental health. However, a 30-minute transfer from SED to LPA or to sleep was associated with decreased odds of reporting poor/fair rather than excellent mental health (OR = 0.85).
Isotemporal substitution by age: BMI and WC
Thirty-minute time reallocations to MVPA were associated with lower BMI and smaller WC, regardless of age or whether the time came from SED, LPA or sleep (p < .001) (Figure 3). In all time reallocations to MVPA, the effect size was 1.2 to 1.5 times greater for older (ages 50 to 79) than for younger (ages 18 to 49) adults.
Time reallocations from SED to LPA were significantly associated with a lower BMI and smaller WC among older (p < .05), but not younger, adults. Also, reallocations from sleep to LPA were significantly associated with a lower BMI and smaller WC only for older respondents (p < .05). Reallocations from SED, LPA or sleep to MVPA were associated with decreased odds of reporting poor/fair rather than excellent general health for both age groups; however, the magnitude was greater for older than for younger adults [ORSED to MVPA: 0.487 (p < .0001) versus 0.203 (p < .0001); ORLPA to MVPA: 0.564 (p < .05) versus 0.244 (p < .0001)] (data not shown). The effect size was relatively similar for the SED-to-LPA reallocation [ORSED to LPA: 0.864 (p < .05) versus 0.831 (p < .0001)] (data not shown). No notable differences by age were apparent for self-rated mental health.
Isotemporal substitution by obesity status: BMI and WC
Thirty-minute time reallocations to MVPA were all associated with a lower BMI and smaller WC, regardless of obesity status or whether the time came from SED, LPA or sleep (Figure 4). In time reallocations to MVPA, the effect size for BMI was 2.7 to 5.2 times greater for overweight/obese adults than for those who were underweight/healthy weight. The difference for WC was more modest (1.6 to 2.3 times greater among those who were overweight/obese).
Reallocating time from SED to LPA was significantly associated with a lower BMI and smaller WC for overweight/obese adults (p < .001). Reallocations from sleep to LPA were significantly associated with a smaller WC for overweight/obese individuals (p < .05). Reallocations from sleep and SED to LPA exhibited a modest (0.06 to 0.08 kg·m-2), but statistically significant, increase in BMI among those who were underweight/healthy weight. Very little difference was observed between those who were overweight/obese and those who were not in their odds of reporting poor/fair rather than excellent general and mental health (data not shown).
Discussion
Using isotemporal substitution, this study found that time reallocation from SED to MVPA was associated with improved obesity markers and a decreased likelihood of reporting poor/fair health, especially among older and overweight/obese individuals. Time reallocation from SED to LPA was significantly associated with improved obesity markers only for older and overweight/obese individuals.
A number of studies that employed isotemporal substitution have demonstrated an association between MVPA and a decrease in BMI and WC. In the present study, the result for BMI of replacing 30 minutes of SED with MVPA (-1.4 kg·m-2) is similar to that observed in a group of healthy adults (-1.2 kg·m-2),Note 12 more than that observed in a group of breast cancer survivors (-0.5 to -0.93 kg·m-2),Note 34 and less than that observed (-2.2 kg·m-2) in a group of adults with newly diagnosed type 2 diabetes.Note 35 The analysis of CHMS data found that reallocation of 30 minutes to MVPA was associated with a 1.6-inch smaller waist circumference, regardless of from where the time was reallocated, a finding similar to previous studies.Note 34Note 35 The effect sizes observed in the present study differed little between single regression and partition models with various combinations of movement variables. The association with MVPA appears to be consistent regardless of whether other movement variables were included in the models.
This analysis applied isotemporal substitution models to different age groups to test whether MVPA becomes less important with age,Note 14 and the finding that substituting LPA for SED was beneficial to health in a sample of older adults.Note 19 A greater effect size was found for reallocation of time to MVPA in older individuals and in those who were overweight/obese.
Time reallocation from SED or sleep to LPA was beneficial for older individuals and for those who were overweight/obese, but not for adults younger than 50 or for those who were not overweight/obese. These age-related differences may reflect decreased efficiency of movement at older ages.Note 36Note 37 Similarly, obesity is associated with decreased gait efficiency and increased energy expenditure for a given task because of greater mass displacement.Note 38 Time reallocation from SED to LPA likely resulted in greater total daily energy expenditure among older and overweight/obese individuals, thus explaining why the SED-to-LPA transition was significantly associated with less obesity among these two subgroups.
The findings add evidence to previous research suggesting that LPA may be important for subpopulations who find exercise programs comprised largely of MVPA challenging to adopt and sustain.Note 19Note 39 The results also lend support to criticism of current threshold-based guidelines that ignore proven health benefits of very modest doses and intensities of physical activity, especially in older populations.Note 40Note 41Note 42
A shortcoming of population health surveys that use accelerometers is that the increased cost of movement for older and heavier people is not captured. This means that the energy expenditure of two people with equal minutes of MVPA could differ considerably. Physical activity levels of overweight/obese individuals tend to be lower than those of healthy weight people,Note 33 and the lack of information about actual energy expenditure at the individual level precludes a true understanding of how human movement relates to health.Note 38
While beyond the scope of the present analysis, future studies may be able to examine how much LPA would be required to equal the benefit of MVPA and determine how this differs by age and obesity status. Such information may contribute public health messaging in which MVPA may be the focus for some groups (for instance, young and healthy), while LPA or a combination of LPA and MVPA could be recommended for other groups (older and overweight/obese).
The prevalence of reporting poor/fair general health was significantly lower when 30 minutes were reallocated from SED to MVPA; however, this was not true for mental health. Previous studies have also had mixed or modest findings related to the association between physical activity and mental health in both cross-sectionalNote 23Note 24 and prospective randomized controlled trials.Note 43 As well, it is possible that a single question about mental health fails to capture the complexity of the issue, as “mental health” is more than the absence of mental illness.Note 44Note 45
Reallocation of time from SED to LPA or to sleep was associated with a decreased likelihood of reporting poor/fair rather than excellent mental health, which is consistent with research that supports the importance of sleep in maintaining mental healthNote 46Note 47 and the idea that reducing SED and engaging in LPA are also important.Note 23Note 24
Strengths and limitations
The study has several notable strengths. The sample is large and representative of Canadian adults. As well, SED, LPA, and MVPA were measured objectively by accelerometry.
However, accelerometers are limited in their ability to capture some types of movement such as swimming, cycling, and load-bearing. As well, CHMS respondents did not wear the accelerometers for 24 hours; therefore, sleep duration was derived from self-reported data.
Use of a single threshold for MVPA for adults of all ages in most population health surveys, including the CHMS, is a limitation because it assumes that the energy cost and health benefit of a given acceleration do not differ between people. Adjustment of intensity thresholds is a method of overcoming this limitation and appears to be more effective at identifying survey respondents at increased health riskNote 48 and reducing discrepancies between reported and measured physical activity in overweight/obese individuals.Note 49 However, in the context of large population health surveys, adjustment of cut-points may be unrealistic.
Isotemporal substitution takes other movement behaviours into account when examining the effect of a given movement behaviour on a health outcome. Inclusion of multiple movement variables that theoretically add up to 24 hours makes controlling for accelerometer wear time in the regression models challenging. Wear time was controlled in the single models, but it could not be included as a covariate in the partition and isotemporal substitution models. SED was significantly associated with BMI and WC in the single models when wear time was included, but not in any other models. This suggests that the true effect of SED is muted in the partition and isotemporal substitution models, which may be partially explained by wear time.
Another limitation of isotemporal substitution is its inability to provide an overall assessment of the association between the 24-hour movement profile and a health outcome. This has been challenging because of multicollinearity issues when all movement variables within a finite period (24 hours) are included in the same model. New techniques such as compositional data analysis are emerging to help address this limitation.Note 50Note 51
Finally, this analysis cannot assess cause-and-effect, because it was not a prospective intervention study in which outcomes were examined before and after time was purposefully reallocated from SED to other activity intensities.
Conclusion
These findings confirm previous studies that found a strong association between MVPA and obesity and health. The beneficial effect of MVPA was greater in older and overweight/obese individuals. This analysis adds evidence to the idea that LPA is an important contributor to health, particularly for older and overweight/obese individuals.
Isotemporal substitution allows for a more comprehensive perspective to be taken on how all intensities of movement relate to health. This more inclusive approach confirms that sedentary behaviour, sleep and light-intensity movement should be considered alongside volitional exercise when developing strategies to improve health.
- Date modified: