Sedentary Behaviors, Light-Intensity Physical Activity, and Healthy Aging

Key Points Question Besides moderate to vigorous physical activity and sleep, are sedentary behavior and light physical activity independently associated with healthy aging, and how could they be reallocated to promote healthy aging? Findings In this cohort study among 45 176 female participants in the Nurses’ Health Study, sedentary behavior was associated with reduced odds of healthy aging, while light physical activity was associated with increased odds of healthy aging. Replacing television time with light physical activity, moderate to vigorous physical activity, or sleep (in participants with inadequate sleep) were associated with better odds of healthy aging. Meaning These findings expand on the literature reporting that replacing sedentary behavior with light or moderate to vigorous physical activity is associated with decreased mortality by suggesting that this increased lifespan might be accompanied by better overall health.


Odds ratio
Odds ratio Odds ratio eAppendix 1. Assessment of healthy aging In order to comprehensively assess the health status of the respondents, based on the concept of successful aging proposed by Rowe and Kahn 1 and other related studies [2][3][4][5] , we defined healthy aging as surviving to at least 70 years old with maintenance of 4 health domains including being free of 11 main chronic diseases, no impairment of physical, memory and mental health.Participants who did not meet these 4 domains or died during the 20 years' follow-up were classified as usual agers 6 .
Information of main chronic diseases was reported on the biennial questionnaires (including cancer, diabetes, myocardial infarction, coronary artery bypass graft surgery or percutaneous transluminal coronary angioplasty, congestive heart failure, stroke, kidney failure, chronic obstructive pulmonary disease, Parkinson's disease, multiple sclerosis, and amyotrophic lateral sclerosis).Study physicians blinded to the participants' exposure status confirmed self-reported diagnosis through medical record review, pathology report review, telephone interview, or supplementary questionnaire inquiries, and self-reported diagnoses were shown to have high validity in this cohort 2 .Women who did not report these 11 diseases at the end of follow-up were considered as being free of chronic diseases.
Physical function was assessed by 10 questions in the Medical Outcomes Study Short-Form Health Survey (SF-36).We inquired about physical limitations in performing different kinds of activities.Each activity had three response choices: "Yes, limited a lot," "Yes, limited a little," or "No, not limited at all." Impairment of physical function was defined as limited at least "a little" on 1 moderate activities (such as moving a table, bowling, or pushing a vacuum cleaner; climbing one flight of stairs; walking more than a mile; walking several blocks; bathing or dressing); or limited "a lot" on 1 more difficult activities (vigorous activities, like running, lifting heavy objects; lifting or carrying groceries; climbing several flights of stairs; bending, kneeling, or stooping); otherwise, participants were considered as having no impairment of physical function 3 .The validity and reliability of the SF-36 and its components have been previously established 7 .
Mental status was evaluated by using the Geriatric Depression Scale-15 (GDS-15).The score on this scale ranges from 1 to 15, with lower scores representing better mental health 8 .No mental health limitation was defined as a GDS-15 score less than or equal to 1 (the median value in this cohort) 3,4 .
Lastly, based on the Structured Telephone Interview for Dementia Assessment 9 , memory function was assessed through seven questions including general memory; remembering short lists; remembering things from one second to the next; remembering recent events; understanding or following spoken instructions; following a group conversation or a plot in a TV program; and navigating familiar streets.These items were strongly associated with objective cognitive function 10 , and have been used to identify individuals with possible cognitive impairment 9 .No impairment in memory was defined as no more than one memory complaint 3,4 .

eAppendix 2. Statistical methods
Because the outcome of this study was whether healthy aging is achieved, and the assessment of all four dimensions was only available after 20 years of follow-up, and the specific occurrence time of the outcome cannot be obtained, we used logistic regression models (instead of Cox regression models) to evaluate the independent association between various behaviors and healthy aging.To be consistent with previous studies 11 , all exposures were categorized into five groups: 0-1, 2-5, 6-20, 21-40, and 41 hours/week.Tests for linear trend were conducted by assigning the median score for each category and treating the median score as a continuous variable.Secondly, to compare the independent role of all five exposures and MVPA as in a prior study 11 , we fit a mutually adjusted model (with these behaviors included simultaneously); in this model, all exposures were treated as continuous variables (per 2 hours/day).Because the major type of MVPA was walking and the most common walking-pace was within 2-2.9mph in this cohort 12 , we standardized the total MVPA as walking time at this pace by using the formula: total MVPA(MET-hours per week)/3(MET per one hour of walking at 2-2.9mph pace) /5(days per week).
In the above analyses, we adjusted for age, education, marital status, annual household income (estimated from the census tract of participant's residence, geocoded to the 1994 U.S. Census); family history of cancer, myocardial infarction and diabetes; baseline hypertension and high cholesterol; menopausal status and postmenopausal hormone use, aspirin use; smoking history, alcohol intake, total energy intake and diet quality assessed by a validated semi-quantitative food frequency questionnaire [13][14][15] , and sleep duration.Considering that BMI could be in the causal pathway between 24-hour behaviors and healthy aging, and to explore the degree to which the exposure-outcome relationships are explained by BMI, we ran an additional model further adjusting for BMI.Considering the important influence of age on healthy ager's evaluation (especially surviving to the end of follow-up), we examined the association between TV time and healthy aging stratified by age (50-, 55-, ≥60 years at baseline, corresponding to 70-, 75-, ≥80 years at follow-up).Moreover, we calculated the population-attributable risk (PAR), an estimate of the percentage of healthy agers during follow-up that could have been achieved if they engaged in the low risk category for these exposures, assuming that the observed associations were causal 16 .
We fitted the ISM to quantify the associations of replacing one hour of one behavior for equal amount of another behavior on healthy aging while the total amount of time of all behaviors was kept constant.Due to the ISM requirement for a linear association between each exposure and the outcome, we modeled sleep as a piecewise variable with a cutoff at 7 h (≤7 h/day and >7 h/day) 17,18 to account for the potential nonlinear, U-shaped association between sleep duration and healthy aging 5 .The Isotemporal Substitution Model is originally expressed as a basic simple model as follows: Healthy aging risk = (b0) sitting watching TV + (b1) sitting at work + (b2) other sitting at home +(b3) LPA at work +(b4) LPA at home + (b5) MVPA + (b6) sleep +(b7) total time* + (b8) covariates By eliminating one behavior from the model (for example, sitting watching TV), the coefficient (b3) represents the effect of substituting 60 minutes/day of LPA at work for 60 minutes/day of sitting watching TV, while holding total time constant; the coefficient (b4) represents the effect of substituting 60 minutes/day of LPA at home for 60 minutes/day of TV-watching while holding total time constant; and so on.The coefficient (b7) for total time represents the omitted behavior (herein, time spent on sitting watching TV).Similar interpretation for the remaining substitution models can be applied when other behaviors are omitted from the model.For comparison, we also fitted the partition model, which partitions "total time" into its components, can be expressed as follows: Healthy aging risk = (b0) sitting watching TV + (b1) sitting at work + (b2) other sitting at home +(b3) LPA at work +(b4) LPA at home + (b5) MVPA + (b6) sleep + (b7) covariates; Because "total time" is not included in the partition model, it is therefore not held constant.Therefore, each coefficient for a certain behavior represents the effect of adding, rather than substituting, that behavior, which indicates that time per day is infinite.
Because the proportion of missing values for each exposure was less than 5% (the highest was only 3.1% for sitting at work or away from home or while driving), in the main ISM analysis, imputation using the median values was used.In addition, we performed the complete cases analysis.To further verify the robustness of our results, for women who had at least 1 of the time use variables missing, we used multiple imputations (10 imputations) and the Expectation-Maximisation algorithm 17,18 .The multiple imputation model included age, and non-missing time use variables as covariates.
Similarly, we analyzed the independent and replacement association of all exposures with the four domains of healthy aging.Potential heterogeneity in the relationship was also explored by stratified analysis by MVPA (being physically active, inactive).Participants were dichotomized into physically active and inactive groups using a threshold of 7.5 MET-hours/week, corresponding to the minimum physical activity recommendations 19 (moderate-intensity activity, 3 MET for 0.5 hour on 5 days/week = 3*0.5*5=7.5 MET-hours/week).The significance of interaction between the stratification variable and exposures was tested using the likelihood ratio test.In secondary analyses, to further evaluate the association between these exposures and healthy aging among survivors, we excluded participants who died before 2012 from usual agers, and then repeated all analyses.
All statistical tests were two-sided and P values <0.05 were considered statistically significant.Data management and statistical analyses were performed using SAS 9.4 (SAS Institute, Cary, NC).eReferences

eFigure 2 |
11 910 women excluded for missing healthy aging including those who were lost to follow-up 57 086 participants were followed up for 20 years 164 women excluded for implausible exposures, including MVPA time > 4h/day, or at least one behavior time (sleep, total sitting, LPA, or MVPA) reported as 0 57 250 participants remained 10 824 women excluded for missing TV time 68 074 participants aged 50 years or over and without major chronic diseases in 1992 21 614 women excluded for having: Design and analysis framework of this study 20 years' follow-up Continuum spectrum of different intensity behaviors eFigure 3 | Association between TV time, light-intensity physical activities and healthy aging, stratified analysis by MVPA a 19 20 a MVPA: moderate or vigorous-intensity physical activity; TV: sitting while watching television.Adjusted for age (years), education (registered nurse, bachelor, or graduate), marital 21 status (married, widowed, or separated/divorced), household income (quintiles); family history of cancer, myocardial infarction and diabetes (yes or no); baseline hypertension and 22 high cholesterol (yes or no); menopausal status and postmenopausal hormone use (pre-menopausal, post-menopausal and never user, post-menopausal and past user, post-menopausal 23 P interaction =0.34 MVPA<7.5 METs/wk MVPA>=7.5 METs/wk TV time (hours/wk) P interaction =0.17

MVPA<7. 5 eFigure 4 |
METs/wk MVPA>=7.5 METs/wk Standing or walking around at home (hours/wk) Odds ratio of heathy aging and its 4 domains associated with sedentary behaviors and physical activities a a All sedentary behavior variables, LPA and MVPA are included simultaneously in the model.Other sitting at home includes reading, mealtime, and at desk.LPA-H: standing or walking around at home; LPA-W: standing or walking around at work.Error bars indicate 95% confidence intervals.Adjusted for age (years), education (registered nurse, bachelor, or graduate), marital status (married, widowed, or separated/divorced), household income (quintiles); family history of cancer, myocardial infarction and diabetes (yes or no); baseline hypertension and high cholesterol (yes or no); menopausal status and postmenopausal hormone use (pre-menopausal, post-menopausal and never user, post-menopausal and past user, post-menopausal and current user), aspirin use (regular use or not); smoking history (never, former smoker, current smoker), alcohol intake (none, 1-14.9, ≥15 g/d), total energy intake (quintiles), diet quality (Alternate Healthy Eating Index score, quintiles), sleep duration (5, 6, 7, 8, 9h), body mass index (BMI <18.5, 18.5-24.9,25-29.9,≥30 kg/m 2 ).b MVPA (hours per day) = total MET-hours per week /3 MET (for 1 hour of normal-pace walking) /5 days.

eTable 2 .
Odds ratios of healthy aging among women according to sedentary behaviors and light-intensity physical activities in hours per week, stratified by age groups a a Adjusted for age ). Percentage of usual agers that can be potentially prevented by adopting active lifestyles