Modifications of 24-h movement behaviors to prevent obesity in retirement: a natural experiment using compositional data analysis

The study addresses how within-individual changes in the composition of 24-h movement behaviors—sleep, sedentary behavior (SED), light physical activity (LPA), and moderate-to-vigorous physical activity (MVPA)—relate to changes in obesity indicators during the transition from work to retirement. Obesity is a major health burden, and while MVPA is known to prevent obesity, time in MVPA is limited and codependent with other daily behaviors. Prior research often examined behaviors in isolation, suggesting benefits of LPA and harms of high SED, and mixed findings for short and long sleep. However, most studies did not account for the codependency of behaviors within the 24-h time budget. The 24-h time-use paradigm and compositional data analysis (CoDA) provide a framework to examine relative time reallocations. Retirement is a natural experiment that changes time use, often increasing sleep, altering SED, and modifying physical activity, and has been linked to increases in BMI. The research question is whether one-year within-individual changes in the composition of 24-h movement behaviors around retirement are associated with concurrent changes in BMI and waist circumference.

Emerging time-use epidemiology emphasizes integrating sleep, SED, LPA, and MVPA as co-dependent parts of a 24-h day. Cross-sectional CoDA studies typically show lower BMI, fat mass, and waist circumference with higher proportions of MVPA relative to other behaviors. Some studies suggest benefits of LPA relative to SED/sleep and of sleep relative to SED, but limitations include cross-sectional design and omission of sleep in some analyses. Only one prior longitudinal CoDA study in older adults showed that reallocating time from SED to MVPA was associated with decreases in BMI and adiposity over seven years but did not include sleep. Retirement transitions have been associated with increased sleep duration, decreased physical activity and increased sedentary time among manual workers, and slight increases in physical activity among non-manual workers, alongside increases in BMI. It remains unclear how longitudinal changes across all 24-h behaviors, including sleep, relate to changes in BMI and waist circumference during retirement.

Design: Longitudinal natural experiment within the Finnish Retirement and Aging Study (FIREA), with repeated annual assessments before and after transition to full-time statutory retirement. Participants: 213 public sector workers (mean age 63.5 years; 82% women; 69% non-manual), who had at least three valid accelerometer days before and after retirement. Measurements: 24-h movement behaviors assessed with thigh-worn triaxial Axivity AX3 accelerometer and daily logs. Participants wore devices continuously for at least four days (pre-retirement: ≥2 workdays and ≥2 days off; post-retirement: ≥4 days), including water-based activities (removed for sauna). Daily logs recorded dates, bed/wake times, reference times, and workdays. Data processing used Open Movement software and ActiPASS (automated Acti4). Non-wear: ≥60 min without movement. Valid day: ≥10 h wear during waking hours and log-determined sleep times. Behaviors: Sleep from log bed/wake times; SED = sitting + lying; LPA = standing, moving, slow walking (<100 steps/min); MVPA = fast walking (≥100 steps/min), stairs, running, cycling, other PA. Behavioral times averaged across valid days. Outcomes: Height, weight (light clothing) measured; BMI calculated; waist circumference measured twice at the midpoint between lowest rib and iliac crest during light exhalation; mean used. Covariates: Sex, date of birth, occupational status (manual vs non-manual from ISCO), pre-retirement smoking, chronic diseases, mobility limitations, self-reported sleep duration, sitting time, and non-occupational PA collected via questionnaire; main models adjusted for baseline BMI/waist circumference, baseline 24-h composition, age, sex, and occupation. Statistical analysis: Treated behavior times as compositional data; calculated compositional means and pre/post compositional differences; visualized with ternary plots by change groups (decrease/no change/increase). Applied isometric logratio (ilr) transformation using pivot coordinates to compare each behavior relative to the remaining behaviors. Fitted linear regression models with change in BMI or waist circumference over one year as outcomes, and changes in 24-h behaviors (ilr coordinates) as predictors, adjusted for baseline composition and covariates. Conducted compositional isotemporal substitution to estimate outcome changes for specified one-to-one reallocations based on mean pre-retirement composition: up to 60 min for reallocations involving MVPA and up to 120 min between sleep, SED, and LPA. Sensitivity analyses: additionally adjusted for pre-retirement measurement season and follow-up length; excluded participants with self-reported sleep >9 h/night pre-retirement to assess potential nonlinearity by sleep sufficiency.

- Baseline (pre-retirement) averages: sleep 8.3 h/day; SED 9.7 h/day; LPA 4.7 h/day; MVPA 77 min/day; BMI 26.3 kg/m² (SD 4.8); waist circumference 91.4 cm (SD 13.0). Post-retirement compositional mean: sleep 520 min, SED 572 min, LPA 272 min, MVPA 76 min. - Longitudinal associations (ilr-based, adjusted): - Increasing MVPA relative to sleep, SED, and LPA was associated with decreases in BMI (βilr = -0.60, 95% CI -1.16 to -0.04, p=0.04) and waist circumference (βilr = -2.14, 95% CI -4.25 to -0.03, p=0.05). - Increasing sleep relative to SED, LPA, and MVPA was associated with an increase in BMI (βilr = 1.34, 95% CI 0.21 to 2.47, p=0.02); no significant association with waist circumference (βilr = 1.51, 95% CI -2.71 to 5.73, p=0.48). - Increasing LPA relative to remaining behaviors showed no statistically significant associations (BMI βilr = -0.66, p=0.06; waist βilr = -1.76, p=0.17). - Increasing SED relative to remaining behaviors was not significantly associated with BMI (βilr = -0.09, p=0.85) or waist circumference (βilr = 2.39, p=0.15). - Isotemporal substitution estimates (one-to-one reallocations over one year): - 60 min reallocated from MVPA to sleep or SED increased BMI by approximately 0.80–0.91 kg/m² and waist circumference by about 3.0 cm; 30 min reallocation increased BMI by ~0.25–0.32 kg/m² and waist by ~1.0 cm; 10 min reallocation increased BMI by ~0.07–0.09 kg/m² and waist by ~0.3 cm. - Reallocating MVPA to LPA increased BMI and waist circumference to a smaller extent than reallocations to sleep or SED. - Reallocations among sleep, SED, and LPA yielded much smaller changes; approximately 90 min increase in LPA at the expense of SED was needed to approximate the effect of a 30 min increase in MVPA at the expense of SED. - Additional results: Effects of decreasing MVPA were larger than effects of increasing MVPA. Sensitivity analyses adjusting for season and follow-up length did not materially change results. Excluding long sleepers (>9 h) attenuated the sleep–BMI association and rendered the association between increasing SED and increasing waist circumference statistically significant.

The study demonstrates that during the transition from work to retirement, within-person increases in MVPA relative to other behaviors are associated with reductions in BMI and waist circumference, whereas increases in sleep relative to activity relate to increases in BMI. These findings align with the energy expenditure continuum: replacing low-expenditure behaviors (sleep, SED) with high-expenditure MVPA reduces adiposity metrics if energy intake remains constant. The larger adverse impact of decreasing MVPA compared to the benefit of increasing MVPA underscores the importance of maintaining MVPA levels during retirement, especially given the potential loss of work-related physical activity. While epidemiology often shows a U-shaped sleep-health relationship, the current findings suggest that increasing sleep from already sufficient levels, if it displaces physical activity, may reduce energy expenditure and contribute to weight gain. Reallocations between sleep and SED had negligible effects on adiposity, likely because both are very low energy expenditure states and sleep time was based on time in bed rather than sleep efficiency. Public health recommendations for retirees should consider preserving MVPA and, where feasible, reallocating time from SED to LPA and MVPA through leisure-time activities, active commuting, and breaking up sedentary time.

Increasing MVPA relative to sleep, SED, and LPA over one year around retirement was associated with reductions in BMI and waist circumference, while increasing sleep was associated with higher BMI. Maintaining MVPA during retirement appears particularly important, as decreases have larger adverse effects than the benefits of increases. Replacing SED with LPA may confer smaller benefits, requiring larger time reallocations. Future research should examine longer follow-up periods, include comprehensive assessments of sleep (e.g., sleep efficiency), body composition changes, contexts of physical activity, and potential dietary changes to clarify mechanisms and generalizability across diverse populations and occupational groups.

- Short-term concurrent change design precludes causal direction inference. - Relatively small sample size (n=213) increases risk of type II error. - No assessment of body composition changes or contexts/types of physical activity; removal of strenuous work tasks vs increased strength training could have differential effects. - Potential unmeasured confounding from changes in energy intake and diet. - Sleep measured via self-reported bed and wake times (time in bed), lacking sleep efficiency; may underestimate sleep benefits. - Study population predominantly women and non-manual workers and generally healthier/more active than survey-only participants, limiting generalizability. - Retirement-related seasonal and follow-up timing were considered, but residual confounding cannot be ruled out.