Abstract
Cross-sectional evidence exists on the beneficial effects of breaks in sedentary time (BST) on frailty in older adults. Nonetheless, the longitudinal nature of these associations is unknown. This study aimed to investigate the direction and temporal order of the association between accelerometer-derived BST and frailty over time in older adults.
This longitudinal study analyzed a total of 186 older adults aged 67–90 (76.7 ± 3.9 years; 52.7% females) from the Toledo Study for Healthy Aging over a 4-year period. Number of daily BST was measured by accelerometry. Frailty was assessed with the Frailty Trait Scale. Multiple cross-lagged panel models were used to test the temporal and reciprocal relationship between BST and frailty.
For those physically inactive (n = 126), our analyses revealed a reciprocal inverse relationship between BST and frailty, such as higher initial BST predicted lower levels of later frailty (standardized regression coefficient [β] = −0.150, 95% confidence interval [CI] = −0.281, −0.018; p < .05); as well as initial lower frailty levels predicted higher future BST (β = −0.161, 95% CI = −0.310, −0.011; p < .05). Conversely, no significant pathway was found in the active participants (n = 60).
In physically inactive older adults, the relationship between BST and frailty is bidirectional, while in active individuals no associations were found. This investigation provides preliminary longitudinal evidence that breaking-up sedentary time more often reduces frailty in those older adults who do not meet physical activity recommendations. Targeting frequent BST may bring a feasible approach to decrease the burden of frailty among more at-risk inactive older adults.
Frailty is a condition of increased vulnerability associated with aging that leads to a number of adverse health outcomes, including disability, falls, hospitalization, and death (1). During the past decade, robust evidence has continued to accumulate on the health benefits of optimal levels of moderate-to-vigorous physical activity (MVPA) (2–4). Several of these studies have observed that higher MVPA was linked to a lower frailty status (5–8). Consequently, physical activity, particularly of moderate-to-vigorous intensity is considered as one of the keystones to prevent, delay, or even reverse the frailty syndrome in older adults (9). Unfortunately, MVPA occupies 2%–3% of the waking day in older adults (5,10–12). The majority of waking time is spent in sedentary behaviors (ie, sitting or reclining). Sedentary time (ST) is increasingly recognized as a novel risk factor for the wider health of individuals (13,14), also comprising frailty (15–17). For instance, a cross-sectional study with 3146 older people in the United States showed that total time spent sitting was associated with frailty (7). Other studies seem to confirm this ST throughout the day (ie, frequency of breaks in ST [BST]) have been demonstrated to be relevant for a variety of health-related outcomes including physical function, performance of activities of daily living, and disability (18–21). Two studies in the field found that breaking-up ST was associated with better physical function in older adults (20,21). However, in a sample of high-functioning and high-active older adults, more time spent in sedentary behavior and lower BST were associated with improved lower limb extensor muscle quality (22). These findings suggest that the relationship between BST and health outcomes may vary depending on whether the sample is active or inactive.
The number of studies assessing the role of BST on frailty is scarce (15). A recent cross-sectional study with 519 participants ≥65 years old assessed with accelerometers concluded that increasing daily BST was associated with less frailty in the population studied, regardless of meeting physical activity guidelines (23). In contrast, in another cross-sectional study with 2317 people from the National Health and Nutrition Examination Survey (NHANES), it was found no association between the frequency of BST and frailty (8). Nonetheless, the longitudinal association of BST and frailty outcomes remains unknown. In addition, there is the possibility of reverse causality, such as a high level of frailty being associated with lower number of BST in the future. Understanding the temporal order underpinning the relationship between BST and frailty has the potential to inform future public health interventions aimed to reduce the frailty burden among older adults.
In this study, we used a cross-lagged panel model to investigate the longitudinal and temporal order of the association between the number of daily BST and frailty in a sample of community-dwelling older adults assessed with accelerometers over a 4-year period. Cross-lagged panel analysis allows to explore the autoregressive and cross-lagged pathways, trying to identify the causal predominance in longitudinal panel data. The models are considered “crossed” because they estimate relationships from one variable to another and vice versa. They are considered “lagged” because they estimate relationships between variables across different time points (24). Based on some literature findings (22), we conducted the analyses in the whole sample as well as separately for physically inactive and active participants, to determine the potential impact of BST on these two groups. We hypothesized that there will be no relationship between daily BST and frailty in the whole sample or in physically active individuals, while there will be an inverse association in both crossed pathways between those physically inactive individuals.
Method
Study Design and Participants
This is a longitudinal study involving two data collection time points separated by 4 years (3.8 ± 0.8 years). This investigation used data from the second and third waves of the Toledo Study for Healthy Aging (TSHA). Details of the protocol of the TSHA have been described elsewhere (25). In the present study, a subsample of the TSHA with accelerometer data was included. The baseline assessment for this study started in July 2012 and lasted until June 2014. A total of 628 participants older than 65 years were assessed at baseline. However, 494 participants finally provided valid data for the analyses (78.7%). Those 494 participants were contacted again in 2015 and invited to participate in a follow-up study, conducted between May 2015 and July 2017 (26). Second evaluation was completed by 200 participants (59.5% missing). Of these, 186 participants (52.7% females) with complete data on all exposures, outcomes, and ≥80% covariates were included in the final analyses of this study. The flow of participants in the study has been described elsewhere (27). Signed informed consent was obtained from all participants prior to their involvement in the study. The study was approved by the Clinical Research Ethics Committee of the Toledo Hospital (approval code: 2010/93).
Measurements
Frailty status
Frailty was assessed using the Frailty Trait Scale (FTS) (28). The FTS has been suggested as a measurement of frailty with superior predictive validity and sensitivity than previously validated scales such as the Frailty Phenotype (29) and the Frailty Index (30). The FTS includes seven domains to complete a continuous frailty scale in which the more decrease in the biological reserve, the closer it is to the threshold of presenting adverse effects derived from this reserve decline (functional deterioration, hospitalization, mortality, etc.). These domains become operational through 12 items (28):
Body mass index (BMI), waist circumference, unintentional weight loss, and serum albumin level were used to assess energy balance and nutrition.
Activity was assessed using the total score of the Physical Activity Scale for the Elderly.
The nervous system was calculated by considering verbal fluency and balance. Verbal fluency was estimated by asking the participants to give names of animals during 1 minute. Balance was measured by Romberg test.
The vascular system was measured by the brachial-ankle index done with Doppler ultrasound.
Weakness was estimated assessing grip strength in the dominant arm and the knee extension strength.
Endurance was assessed by the chair stand test, which measures the number of times that a person stands up in 30 seconds.
Slowness was estimated by calculating the time to walk 3 m at a “normal pace” according to a standard protocol.
Each item score was obtained as recommended elsewhere (28). The total score was calculated according to the formula: Total score = (Σ items score/total score possible by individual) * 100, standardizing the measure to a range from 0 (less frailty) to 100 (more frailty).
Physical activity and sedentary behavior assessment
Physical activity and ST were measured by accelerometry as previously described (5). Due to logistical reasons, ActiTrainer (ActiGraph, LLC, FortWalton Beach, FL) was used in the baseline, and wGT3X-BT (ActiGraph, LLC, Pensacola, FL) was used in the follow-up assessments. To make sure that the two accelerometers collected the same accelerations with the same movement, we decided to do a pilot study, not published, where both accelerometers were compared. The reliability test indicated a very high intraclass correlation coefficient (0.997) between the accelerations of axis 1 (used in this study) of both accelerometers. This method of comparison between ActiGraph models has been previously validated (31,32).
Briefly, all participants were asked to wear the accelerometer on the left hip during waking hours of a whole week, with exception for water activities. A valid day was defined as having ≥480 minutes (≥8 hours) of monitor wear, and the study included the results from participants with at least four valid days (33,34).
Each minute during which the accelerometer counts were below 100 cpm was defined as ST (35,36). Breaks in sedentary time was defined as at least 1 minute where the accelerometer registers ≥100 cpm following a sedentary period (8,20). Number of daily BST was then calculated to be used in the analyses.
Accelerometer counts more than or equal to 100 cpm were classified as physical activity with additional separation into light-intensity (100–1951 cpm), moderate-intensity (1952–5724 cpm), and vigorous-intensity (≥5725 cpm) physical activity (35).
In addition, compliance with physical activity recommendations was calculated in order to classify individuals as active or inactive at the baseline moment. To be active, at least one of these three premises had to be met (37): accumulate 150 minutes of moderate physical activity per week; accumulate 75 minutes of vigorous physical activity per week; or accumulate 150 minutes per week of an equivalent combination of MVPA.
Anthropometrics and confounding variables
Height and weight were measured using a balance-stadiometer (Seca 711 scales, Hamburg, Germany). Body mass index was determined as weight (kg) divided by height squared (m2).
Participants were asked about their age, sex, and ethnicity. Other sociodemographic variables such as education (no studies, primary school completed, secondary school completed or more) and marital status (single, married/living together, widowed, divorced/separated) were also self-reported in face-to-face interviews. The Charlson Comorbidity Index was used to account for comorbidity status of participants in the study (38). Mini-Mental State Examination (MMSE) was also collected in order to evaluate objective cognitive function (39).
Statistical Analysis
All analyses were conducted with the R software (R project version 3.5.1) and significance level was set at usual p < .05. In order to check whether the participants missing the follow-up assessment were systematically different from those retained in the study, demographics and other characteristics were compared through independent t test or chi-square test for continuous and categorical variables, respectively. The mean ± SD and frequencies (percentages) were used to describe continuous variables and categorical variables, respectively. A paired t test was used to compare participants retained in the study across time points of assessment. An independent t test was conducted to compare active and inactive participants.
We addressed our main objective using a structural equation modeling framework using functions from the R package Lavaan (40). The full information maximum likelihood was used to provide unbiased and efficient estimates of the parameters of interest using the complete available information (41). A cross-lagged panel model was designed to test the relationships between number of daily BST and frailty status between the initial assessment and the 4-year follow-up. Among the strengths of using a cross-lagged panel approach is that it tolerates simultaneous analysis of the two dependent outcomes, thereby allowing the identification of possible bidirectional associations over time. Covariates included sex as time-invariant variable, in addition to age, education, marital status, BMI, MMSE, MVPA, and accelerometer wear time as time-variant. Subsequently, two more cross-lagged panel models were made by stratifying the sample in those active and inactive individuals. Model fit was considered using a selection of fit indices and criteria as previously published (42): root mean square error of approximation (RMSEA) (≤0.06), standardized root mean square residual (SRMR) (≤0.08), confirmatory fit index (CFI) (≥0.95), and Tucker-Lewis index (TLI) (≥0.95).
Results
Attrition Data Across Time Points
Participants decreased from 494 with complete data at baseline to 186 with complete data at follow-up assessment (27). The causes and numbers who were lost to the follow-up assessment were death (n = 42), withdrawing (n = 225), and could not be located (n = 27). Additional missing data were lost by insufficient accelerometer wear time data (n = 9), missing frailty data (n = 2), or losing ≥80% of the covariates (n = 3). Compared with the retained sample, individuals who dropped the study at 4-year follow-up were significantly older, less educated, and spent more time on ST (Supplementary Table 1).
Descriptive Statistics
Table 1 displays the descriptive statistics for frailty, accelerometry variables as well as confounders at each of the two measurement time points divided into the whole sample, active individuals, and inactive individuals. At baseline, the whole sample had a mean age of 76.68 (SD = 3.90), a mean FTS of 35.35 (SD = 13.94), and a mean daily BST (n/day) of 71.66 (SD = 18.54). Frailty Trait Scale score and daily BST declined significantly from baseline to follow-up (p < .05). Body mass index, MMSE, and MVPA also decreased significantly between both time points (all p < .05). According to the current physical activity guidelines (37), one-third of the sample (n = 60) was categorized as physically active and the remaining as physically inactive (n = 126). There were significant differences between baseline and follow-up in BMI, FTS, and MVPA for physically active individuals (all p < .05). For those physically inactive, significant differences between baseline and follow-up were found in MMSE, FTS, and daily BST. Compared to physically active participants, those physically inactive were older, had a higher BMI, greater FTS, lower accelerometer wear time, and spent more time in ST and lower time in light physical activity and MVPA. There were not significant differences in daily BST between the two groups.
Sociodemographic and Descriptive Data
| Variables | Total Sample (n = 186) | Active Individuals (n = 60) | Inactive Individuals (n = 126) | |||
|---|---|---|---|---|---|---|
| Baseline | Follow-up | Baseline | Follow-up | Baseline | Follow-up | |
| Age (y)† | 76.68 ± 3.90 | 80.44 ± 4.24* | 75.42 ± 3.05 | 79.08 ± 3.37* | 77.29 ± 4.13θ | 81.08 ± 4.47* |
| Sex‡ | ||||||
| Men | 88 (47.3) | 88 (47.3) | 39 (65.0) | 39 (65.0) | 49 (38.9) | 49 (38.9) |
| Women | 98 (52.7) | 98 (52.7) | 21 (35.0) | 21 (35.0) | 77 (61.1) | 77 (61.1) |
| BMI (kg/m2)† | 30.82 ± 4.62 | 30.33 ± 4.40* | 29.11 ± 3.60 | 28.39 ± 3.62* | 31.63 ± 4.84 θ | 31.26 ± 4.45 Ŧ |
| Education‡ | ||||||
| None | 139 (74.7) | 116 (62.4) | 50 (83.3) | 38 (63.3) | 89 (70.6) | 78 (61.9) |
| Primary school | 30 (16.1) | 51 (27.4) | 4 (6.7) | 14 (23.3) | 26 (20.6) | 37 (29.4) |
| Secondary or more | 14 (7.5) | 19 (10.2) | 6 (10.0) | 8 (13.3) | 8 (6.3) | 11 (8.7) |
| Missing§ | 3 (1.6) | NA | NA | NA | 3 (2.4) | NA |
| Marital status‡ | ||||||
| Single | 7 (3.8) | 7 (3.8) | 1 (1.7) | 1 (1.7) | 6 (4.8) | 6 (4.8) |
| Married | 136 (73.1) | 125 (67.2) | 51 (85.0) | 47 (78.3) | 85 (67.5) | 78 (61.9) |
| Widower | 40 (21.5) | 51 (27.4) | 8 (13.3) | 11 (18.3) | 32 (25.4) | 40 (31.7) |
| Separated/divorced | 1 (0.5) | 2 (1.1) | NA | NA | 1 (0.8) | 2 (1.6) |
| Missing§ | 2 (1.1) | 1 (0.5) | NA | 1 (1.7) | 2 (1.6) | NA |
| MMSE† | 24.02 ± 3.73 | 23.32 ± 3.54* | 24.49 ± 3.52 | 24.27 ± 2.92 | 23.79 ± 3.82 | 22.89 ± 3.74* |
| Missing§ | 15 (8.1) | 15 (8.1) | 5 (8.3) | 5 (8.3) | 10 (7.9) | 10 (7.9) |
| Charlson Index, points† | 0.73 ± 1.20 | 0.65 ± 0.96 | 0.67 ± 1.37 | 0.58 ± 0.91 | 0.76 ± 1.11 | 0.67 ± 0.98 |
| Frailty Trait Scale, points† | 35.35 ± 13.94 | 43.79 ± 13.86* | 27.45 ± 12.54 | 34.56 ± 10.11* | 39.11 ± 13.00 θ | 48.19 ± 13.25* |
| Accelerometer wear time, min/ valid day† | 781.36 ± 83.14 | 777.61 ± 74.45 | 807.95 ± 79.31 | 800.40 ± 69.19 | 768.69 ± 82.22 θ | 766.77 ± 74.67 |
| Sedentary time, min/valid day† | 530.18 ± 84.86 | 542.61 ± 75.91Ŧ | 505.52 ± 86.18 | 525.44 ± 67.62 | 541.92 ± 81.98 θ | 550.78 ± 78.49 |
| LPA, min/valid day† | 231.05 ± 86.50 | 221.79 ± 84.83 | 253.61 ± 84.94 | 245.42 ± 22.81 | 220.31 ± 85.48 θ | 210.31 ± 89.18 |
| MVPA, min/valid day† | 20.12 ± 23.30 | 13.21 ± 18.73* | 48.82 ± 19.43 | 29.54 ± 22.81* | 6.45 ± 6.45 θ | 5.44 ± 9.20 |
| Daily BST, n/day† | 71.66 ± 18.54 | 65.94 ± 18.11* | 69.19 ± 16.08 | 66.39 ± 15.57 | 72.83 ± 19.56 | 65.73 ± 19.26* |
| Variables | Total Sample (n = 186) | Active Individuals (n = 60) | Inactive Individuals (n = 126) | |||
|---|---|---|---|---|---|---|
| Baseline | Follow-up | Baseline | Follow-up | Baseline | Follow-up | |
| Age (y)† | 76.68 ± 3.90 | 80.44 ± 4.24* | 75.42 ± 3.05 | 79.08 ± 3.37* | 77.29 ± 4.13θ | 81.08 ± 4.47* |
| Sex‡ | ||||||
| Men | 88 (47.3) | 88 (47.3) | 39 (65.0) | 39 (65.0) | 49 (38.9) | 49 (38.9) |
| Women | 98 (52.7) | 98 (52.7) | 21 (35.0) | 21 (35.0) | 77 (61.1) | 77 (61.1) |
| BMI (kg/m2)† | 30.82 ± 4.62 | 30.33 ± 4.40* | 29.11 ± 3.60 | 28.39 ± 3.62* | 31.63 ± 4.84 θ | 31.26 ± 4.45 Ŧ |
| Education‡ | ||||||
| None | 139 (74.7) | 116 (62.4) | 50 (83.3) | 38 (63.3) | 89 (70.6) | 78 (61.9) |
| Primary school | 30 (16.1) | 51 (27.4) | 4 (6.7) | 14 (23.3) | 26 (20.6) | 37 (29.4) |
| Secondary or more | 14 (7.5) | 19 (10.2) | 6 (10.0) | 8 (13.3) | 8 (6.3) | 11 (8.7) |
| Missing§ | 3 (1.6) | NA | NA | NA | 3 (2.4) | NA |
| Marital status‡ | ||||||
| Single | 7 (3.8) | 7 (3.8) | 1 (1.7) | 1 (1.7) | 6 (4.8) | 6 (4.8) |
| Married | 136 (73.1) | 125 (67.2) | 51 (85.0) | 47 (78.3) | 85 (67.5) | 78 (61.9) |
| Widower | 40 (21.5) | 51 (27.4) | 8 (13.3) | 11 (18.3) | 32 (25.4) | 40 (31.7) |
| Separated/divorced | 1 (0.5) | 2 (1.1) | NA | NA | 1 (0.8) | 2 (1.6) |
| Missing§ | 2 (1.1) | 1 (0.5) | NA | 1 (1.7) | 2 (1.6) | NA |
| MMSE† | 24.02 ± 3.73 | 23.32 ± 3.54* | 24.49 ± 3.52 | 24.27 ± 2.92 | 23.79 ± 3.82 | 22.89 ± 3.74* |
| Missing§ | 15 (8.1) | 15 (8.1) | 5 (8.3) | 5 (8.3) | 10 (7.9) | 10 (7.9) |
| Charlson Index, points† | 0.73 ± 1.20 | 0.65 ± 0.96 | 0.67 ± 1.37 | 0.58 ± 0.91 | 0.76 ± 1.11 | 0.67 ± 0.98 |
| Frailty Trait Scale, points† | 35.35 ± 13.94 | 43.79 ± 13.86* | 27.45 ± 12.54 | 34.56 ± 10.11* | 39.11 ± 13.00 θ | 48.19 ± 13.25* |
| Accelerometer wear time, min/ valid day† | 781.36 ± 83.14 | 777.61 ± 74.45 | 807.95 ± 79.31 | 800.40 ± 69.19 | 768.69 ± 82.22 θ | 766.77 ± 74.67 |
| Sedentary time, min/valid day† | 530.18 ± 84.86 | 542.61 ± 75.91Ŧ | 505.52 ± 86.18 | 525.44 ± 67.62 | 541.92 ± 81.98 θ | 550.78 ± 78.49 |
| LPA, min/valid day† | 231.05 ± 86.50 | 221.79 ± 84.83 | 253.61 ± 84.94 | 245.42 ± 22.81 | 220.31 ± 85.48 θ | 210.31 ± 89.18 |
| MVPA, min/valid day† | 20.12 ± 23.30 | 13.21 ± 18.73* | 48.82 ± 19.43 | 29.54 ± 22.81* | 6.45 ± 6.45 θ | 5.44 ± 9.20 |
| Daily BST, n/day† | 71.66 ± 18.54 | 65.94 ± 18.11* | 69.19 ± 16.08 | 66.39 ± 15.57 | 72.83 ± 19.56 | 65.73 ± 19.26* |
Notes: BMI = body mass index; BST = number of breaks in sedentary time; LPA = light physical activity; MMSE = Mini-Mental State Examination; MVPA = moderate-to-vigorous physical activity; NA = not available.
†Continuous variable; mean ± SD.
‡Categorical variable; n (%).
§Missing data; n (%).
*Significant differences between baseline versus follow-up (p < .05).
θ Significant differences between active versus inactive individuals at baseline (p < .05).
ŦTrend toward significance between baseline versus follow-up (p < .08 > .05).
Sociodemographic and Descriptive Data
| Variables | Total Sample (n = 186) | Active Individuals (n = 60) | Inactive Individuals (n = 126) | |||
|---|---|---|---|---|---|---|
| Baseline | Follow-up | Baseline | Follow-up | Baseline | Follow-up | |
| Age (y)† | 76.68 ± 3.90 | 80.44 ± 4.24* | 75.42 ± 3.05 | 79.08 ± 3.37* | 77.29 ± 4.13θ | 81.08 ± 4.47* |
| Sex‡ | ||||||
| Men | 88 (47.3) | 88 (47.3) | 39 (65.0) | 39 (65.0) | 49 (38.9) | 49 (38.9) |
| Women | 98 (52.7) | 98 (52.7) | 21 (35.0) | 21 (35.0) | 77 (61.1) | 77 (61.1) |
| BMI (kg/m2)† | 30.82 ± 4.62 | 30.33 ± 4.40* | 29.11 ± 3.60 | 28.39 ± 3.62* | 31.63 ± 4.84 θ | 31.26 ± 4.45 Ŧ |
| Education‡ | ||||||
| None | 139 (74.7) | 116 (62.4) | 50 (83.3) | 38 (63.3) | 89 (70.6) | 78 (61.9) |
| Primary school | 30 (16.1) | 51 (27.4) | 4 (6.7) | 14 (23.3) | 26 (20.6) | 37 (29.4) |
| Secondary or more | 14 (7.5) | 19 (10.2) | 6 (10.0) | 8 (13.3) | 8 (6.3) | 11 (8.7) |
| Missing§ | 3 (1.6) | NA | NA | NA | 3 (2.4) | NA |
| Marital status‡ | ||||||
| Single | 7 (3.8) | 7 (3.8) | 1 (1.7) | 1 (1.7) | 6 (4.8) | 6 (4.8) |
| Married | 136 (73.1) | 125 (67.2) | 51 (85.0) | 47 (78.3) | 85 (67.5) | 78 (61.9) |
| Widower | 40 (21.5) | 51 (27.4) | 8 (13.3) | 11 (18.3) | 32 (25.4) | 40 (31.7) |
| Separated/divorced | 1 (0.5) | 2 (1.1) | NA | NA | 1 (0.8) | 2 (1.6) |
| Missing§ | 2 (1.1) | 1 (0.5) | NA | 1 (1.7) | 2 (1.6) | NA |
| MMSE† | 24.02 ± 3.73 | 23.32 ± 3.54* | 24.49 ± 3.52 | 24.27 ± 2.92 | 23.79 ± 3.82 | 22.89 ± 3.74* |
| Missing§ | 15 (8.1) | 15 (8.1) | 5 (8.3) | 5 (8.3) | 10 (7.9) | 10 (7.9) |
| Charlson Index, points† | 0.73 ± 1.20 | 0.65 ± 0.96 | 0.67 ± 1.37 | 0.58 ± 0.91 | 0.76 ± 1.11 | 0.67 ± 0.98 |
| Frailty Trait Scale, points† | 35.35 ± 13.94 | 43.79 ± 13.86* | 27.45 ± 12.54 | 34.56 ± 10.11* | 39.11 ± 13.00 θ | 48.19 ± 13.25* |
| Accelerometer wear time, min/ valid day† | 781.36 ± 83.14 | 777.61 ± 74.45 | 807.95 ± 79.31 | 800.40 ± 69.19 | 768.69 ± 82.22 θ | 766.77 ± 74.67 |
| Sedentary time, min/valid day† | 530.18 ± 84.86 | 542.61 ± 75.91Ŧ | 505.52 ± 86.18 | 525.44 ± 67.62 | 541.92 ± 81.98 θ | 550.78 ± 78.49 |
| LPA, min/valid day† | 231.05 ± 86.50 | 221.79 ± 84.83 | 253.61 ± 84.94 | 245.42 ± 22.81 | 220.31 ± 85.48 θ | 210.31 ± 89.18 |
| MVPA, min/valid day† | 20.12 ± 23.30 | 13.21 ± 18.73* | 48.82 ± 19.43 | 29.54 ± 22.81* | 6.45 ± 6.45 θ | 5.44 ± 9.20 |
| Daily BST, n/day† | 71.66 ± 18.54 | 65.94 ± 18.11* | 69.19 ± 16.08 | 66.39 ± 15.57 | 72.83 ± 19.56 | 65.73 ± 19.26* |
| Variables | Total Sample (n = 186) | Active Individuals (n = 60) | Inactive Individuals (n = 126) | |||
|---|---|---|---|---|---|---|
| Baseline | Follow-up | Baseline | Follow-up | Baseline | Follow-up | |
| Age (y)† | 76.68 ± 3.90 | 80.44 ± 4.24* | 75.42 ± 3.05 | 79.08 ± 3.37* | 77.29 ± 4.13θ | 81.08 ± 4.47* |
| Sex‡ | ||||||
| Men | 88 (47.3) | 88 (47.3) | 39 (65.0) | 39 (65.0) | 49 (38.9) | 49 (38.9) |
| Women | 98 (52.7) | 98 (52.7) | 21 (35.0) | 21 (35.0) | 77 (61.1) | 77 (61.1) |
| BMI (kg/m2)† | 30.82 ± 4.62 | 30.33 ± 4.40* | 29.11 ± 3.60 | 28.39 ± 3.62* | 31.63 ± 4.84 θ | 31.26 ± 4.45 Ŧ |
| Education‡ | ||||||
| None | 139 (74.7) | 116 (62.4) | 50 (83.3) | 38 (63.3) | 89 (70.6) | 78 (61.9) |
| Primary school | 30 (16.1) | 51 (27.4) | 4 (6.7) | 14 (23.3) | 26 (20.6) | 37 (29.4) |
| Secondary or more | 14 (7.5) | 19 (10.2) | 6 (10.0) | 8 (13.3) | 8 (6.3) | 11 (8.7) |
| Missing§ | 3 (1.6) | NA | NA | NA | 3 (2.4) | NA |
| Marital status‡ | ||||||
| Single | 7 (3.8) | 7 (3.8) | 1 (1.7) | 1 (1.7) | 6 (4.8) | 6 (4.8) |
| Married | 136 (73.1) | 125 (67.2) | 51 (85.0) | 47 (78.3) | 85 (67.5) | 78 (61.9) |
| Widower | 40 (21.5) | 51 (27.4) | 8 (13.3) | 11 (18.3) | 32 (25.4) | 40 (31.7) |
| Separated/divorced | 1 (0.5) | 2 (1.1) | NA | NA | 1 (0.8) | 2 (1.6) |
| Missing§ | 2 (1.1) | 1 (0.5) | NA | 1 (1.7) | 2 (1.6) | NA |
| MMSE† | 24.02 ± 3.73 | 23.32 ± 3.54* | 24.49 ± 3.52 | 24.27 ± 2.92 | 23.79 ± 3.82 | 22.89 ± 3.74* |
| Missing§ | 15 (8.1) | 15 (8.1) | 5 (8.3) | 5 (8.3) | 10 (7.9) | 10 (7.9) |
| Charlson Index, points† | 0.73 ± 1.20 | 0.65 ± 0.96 | 0.67 ± 1.37 | 0.58 ± 0.91 | 0.76 ± 1.11 | 0.67 ± 0.98 |
| Frailty Trait Scale, points† | 35.35 ± 13.94 | 43.79 ± 13.86* | 27.45 ± 12.54 | 34.56 ± 10.11* | 39.11 ± 13.00 θ | 48.19 ± 13.25* |
| Accelerometer wear time, min/ valid day† | 781.36 ± 83.14 | 777.61 ± 74.45 | 807.95 ± 79.31 | 800.40 ± 69.19 | 768.69 ± 82.22 θ | 766.77 ± 74.67 |
| Sedentary time, min/valid day† | 530.18 ± 84.86 | 542.61 ± 75.91Ŧ | 505.52 ± 86.18 | 525.44 ± 67.62 | 541.92 ± 81.98 θ | 550.78 ± 78.49 |
| LPA, min/valid day† | 231.05 ± 86.50 | 221.79 ± 84.83 | 253.61 ± 84.94 | 245.42 ± 22.81 | 220.31 ± 85.48 θ | 210.31 ± 89.18 |
| MVPA, min/valid day† | 20.12 ± 23.30 | 13.21 ± 18.73* | 48.82 ± 19.43 | 29.54 ± 22.81* | 6.45 ± 6.45 θ | 5.44 ± 9.20 |
| Daily BST, n/day† | 71.66 ± 18.54 | 65.94 ± 18.11* | 69.19 ± 16.08 | 66.39 ± 15.57 | 72.83 ± 19.56 | 65.73 ± 19.26* |
Notes: BMI = body mass index; BST = number of breaks in sedentary time; LPA = light physical activity; MMSE = Mini-Mental State Examination; MVPA = moderate-to-vigorous physical activity; NA = not available.
†Continuous variable; mean ± SD.
‡Categorical variable; n (%).
§Missing data; n (%).
*Significant differences between baseline versus follow-up (p < .05).
θ Significant differences between active versus inactive individuals at baseline (p < .05).
ŦTrend toward significance between baseline versus follow-up (p < .08 > .05).
Cross-Lagged Panel Model 1: Total Sample
Figure 1 shows the final cross-lagged model for the whole sample. The model fits the data well (RMSEA = 0.000; SRMR = 0.012; CFI = 1.000; TLI = 1.010). Significant associations were only detected in the autoregressive pathways. That is, initial daily BST predicted future daily BST (standardized regression coefficient [β] = 0.366, 95% confidence interval [CI] = 0.253, 0.480; p < .01) and baseline frailty scores predicted future frailty scores, respectively (β = 0.320, 95% CI = 0.204, 0.436; p < .01). The cross-lagged effect from baseline daily BST to follow-up frailty status was not statistically significant (β = −0.081, 95% CI = −0.182, 0.020; p = .12). Similarly, the cross-lagged effect from initial frailty status to later daily BST was also not statistically significant (β = −0.130, 95% CI = −0.268, 0.008; p = .07).
Cross-lagged panel model 1: total sample.BST = number of daily breaks in sedentary time. Model adjusted for age, sex, body mass index (BMI), education, marital status, Mini-Mental State Examination, moderate-to-vigorous physical activity (MVPA), and accelerometer wear time. Bold indicates statistical significance (p < .05).
Cross-Lagged Panel Model 2: Physically Active Individuals
Figure 2 shows the final cross-lagged model for the physically active individuals. The model fits the data well (RMSEA = 0.025; SRMR = 0.025; CFI = 0.992; TLI = 0.980). As in the previous model, the autoregressive pathways were the only significant ones. Baseline daily BST predicted daily BST in the follow-up (β = 0.376, 95% CI = 0.176, 0.577; p < .01) as well as baseline frailty status predicted later frailty status (β = 0.226, 95% CI = 0.010, 0.442; p < .05). The cross-lagged effects from initial daily BST to subsequent frailty status and initial frailty status to later daily BST were not statistically significant (β = −0.056, 95% CI = −0.267, 0.156; p = .61; and β = 0.043, 95% CI = −0.171, 0.256; p = .70, respectively).
Cross-lagged panel model 2: physically active individuals. BST = daily number of breaks in sedentary time. Model adjusted for age, sex, body mass index (BMI), education, marital status, Mini-Mental State Examination, moderate-to-vigorous physical activity (MVPA), and accelerometer wear time. Bold indicates statistical significance (p < .05).
Cross-Lagged Panel Model 3: Physically Inactive Individuals
Figure 3 shows the final cross-lagged model for the inactive individuals. The model fits the data well (RMSEA = 0.021; SRMR = 0.020; CFI = 0.994; TLI = 0.986). The autoregressive pathways were again significant for both BTS and FTS (β = 0.371, 95% CI = 0.235, 0.507; and β = 0.339, 95% CI = 0.203, 0.475, respectively; p < .01). The cross-lagged pathways were also significant. Specifically, initial daily BST predicted future frailty (β = −0.150, 95% CI = −0.281, −0.018; p < .05), indicating that lower daily BST at baseline predicted higher frailty score 4 years later, adjusting for baseline frailty status. Likewise, initial FTS predicted future daily BST (β = −0.161, 95% CI = −0.310, −0.011; p < .05), adjusting for baseline daily BST, pointing out that higher frailty at baseline predicted lower daily BST 4 years later.
Cross-lagged panel model 3: physically inactive individuals. BST = daily number of breaks in sedentary time. Model adjusted for age, sex, body mass index (BMI), education, marital status, Mini-Mental State Examination, moderate-to-vigorous physical activity (MVPA), and accelerometer wear time. Bold indicates statistical significance (p < .05).
Discussion
To our knowledge, the present study tested for the first time the temporal and bidirectional associations between the number of daily BST and frailty status in a sample of community-dwelling older adults. The main findings were that, in physically inactive individuals there was an inverse relationship between BST and frailty status 4 years later. The reverse was also true (ie, frailty status at baseline was inversely associated with BST 4 years later). There was no evidence of longitudinal association between BST and frailty in older adults who were considered physically active. Consequently, increasing daily BST could be a promising strategy to reduce the burden of frailty syndrome in physically inactive older adults.
Different cross-sectional studies have explored the associations of BST with physical function (20), disability (18), and frailty (23) in older adults. A previous study found a positive association between BST and physical function, after adjusting for total ST and MVPA (20). In another study conducted in older adults assessed with accelerometers, daily BST was associated with lower frailty in older adults (23). Our study extends these previous findings and investigated the longitudinal and bidirectional associations between daily BST and frailty in a sample of community-dwelling participants aged 65 years and older.
Mechanisms underlying these findings are likely to be complex. Experimental studies have provided evidence related to the physiologic and cardiometabolic benefits of breaking-up and reducing sitting time (43). The evidence suggests that beyond the increase in energy expenditure that requires a transition from sitting to standing position (44), the benefits of frequent BST can be explained by the muscular contractions derived from such transitions (45). These muscular contractions, which will mostly be provided in light physical intensity activities, can lead to important functional adaptations through different physiological and molecular pathways (45–49) that ultimately affect human health (50). Therefore, it seems evident that physically inactive individuals who break their ST more often may reduce their frailty status.
Likewise, less frail individuals will find it easier to break ST more often compared to other more structured physical task, thus becoming a positive circle (negative if the question is raised backwards). This is in accordance with a previous study of our group that demonstrated that older adults with comorbidities may benefit more from replacing ST with light-intensity physical activity compared to healthier counterparts (5). A recent meta-analysis also reported the benefits of replacing sitting time with light-intensity physical activity for cardiometabolic health and mortality (51). Furthermore, it should be noted that although those who continued the study were younger, more active, less sedentary, and more educated, the association between daily BST and frailty remained significant, even in this healthier population of physically inactive individuals. This means that the results are likely understated and that the true magnitude of the association may be even greater. Together, these findings suggest the potential of targeting reductions in ST to improve the health of older adults (43,52).
According to our estimates, the increase of 20 daily BST could reduce frailty by 2 points in the FTS (0–100) 4 years later. Other cross-sectional studies found greater effects for other health outcomes (8,18,20). Factors such as our population is considered a healthier, nonclinical population and that in this study the exposure is predicting the outcome longitudinally may partially explain this observation. Despite the observed small effect, it is worth noting that a sedentary break can be as short as 1 minute, so continuous and/or more vigorous exposure may have greater effects. Taking together, these preliminary findings extend those from previous cross-sectional evidence and remain clinically relevant: ST fragmentation may result in the prevention of future frailty among more at-risk inactive older adults.
In contrast, we did not detect a statistical association between baseline daily BST and follow-up frailty status among physically active older adults. In a previous study, it was suggested that lower BST was linked with better muscle quality in a sample of highly functioning and highly active older adults (22). It is plausible that active participants are also fit (53) and that a stimulus stronger than muscle contractions resulting from breaking-up ST is necessary to evoke reductions in the frailty status among these individuals. This suggests that the effects of BST on frailty may be moderated by fitness. Similarly, we speculate a higher physiological reserve (ie, less frailty, as described in Table 1) present in active participants may have accounted for the lack of association between initial frailty status and daily BST at follow-up. It is possible that active participants are less likely to become frail, particularly due to the relatively short follow-up period. Future studies with longer follow-up should revisit this issue.
The lack of significant results for the total sample is likely to reflect the heterogeneity in the estimations for active and inactive individuals in this study (ie, the positive findings for inactive individuals are canceled out by the null findings among active individuals).
These findings are likely to be policy-relevant: current physical activity guidelines for older adults focused mainly on MVPA (37,54). Further to increase MVPA, our results also stress the relevance of breaking-up ST, particularly in those physically inactive individuals. From a health-promotion perspective, encouraging small bouts of activity into otherwise sedentary periods may be a more feasible and less challenging approach for older adults than taking part in more intense activities. Breaking-up ST can occur trivially in a variety of daily living activities (at home, during transport, or leisure time) because:
i) it does not require a high degree of commitment or planning,
ii) it can be achieved with a physically lower load, and
iii) it does not require a high level of fitness or complex motor skills.
Altogether, our observations indicate the possibility of targeting BST to reduce the frailty levels in physically inactive older adults (43) and provide preliminary evidence that may inform the development of lifestyle strategies related to sedentary behavior to maintain functional capacity and prevent frailty in older adults.
Strengths and Limitations
A key strength of our work is that it comprises a relatively large sample of community-dwelling older adults with data follow-up of 4 years. The use of accelerometer-derived sedentary and physical activity behavior in this study is also a strength. There are some inherent issues with the selected, commonly used threshold to determine sedentary behavior and activity intensity in our sample. Despite some studies suggest different thresholds may classify better sedentary behaviors and physical activity in older adults, consensus is yet to be reached. At the very least, our results are comparable with other international studies (36). Future studies should be conducted to determine optimal accelerometer cutoff points to classify activity in older adults, especially in those more frail and with less functional performance individuals. Another strength is the robustness of the FTS to assess the frailty level of participants. The FTS has demonstrated superior predictive validity and responsiveness than previously validated constructs such as the Frailty Phenotype (29) and the Frailty Index (30). Furthermore, another key strength of this study is the use of a statistical method (ie, cross-lagged panel model) that allowed us to investigate the temporal order and bidirectional, longitudinal associations of BST with frailty over time in the study participants.
Despite the methodological rigor of this study, some limitations must to be acknowledged. First, there was a significant loss to the follow-up in our study. This could have influenced our results, mainly underestimating the magnitude of the association in those most vulnerable. Our study did not look at the incidence of frailty; therefore, we cannot rule out that reverse causality may still be present. Future research should address this issue. The WHO physical activity guidelines for older adults may not be appropriate for frail individuals (55). Therefore, some participants in this study may have been misclassified as inactive, potentially biasing our results. Future studies should consider the development of guidelines relative to the functional capacity of individuals. An additional limitation is the limited ability of accelerometers to discriminate between sitting and standing compared with posture-based devices (56). Importantly, in our study, a break in ST reflects a modification in acceleration instead of a change in posture, corresponding to a transition from none or slight movement (<100 cpm) to some movement (≥100 cpm). Further studies should replicate our analysis with posture-based devices. Future research should investigate optimal strategies to increase the number of BST throughout the day in older adults, particularly among those not achieving the recommended level of physical activity.
Conclusion
According to our hypothesis, the longitudinal relationship between the daily number of sedentary breaks and the frailty levels is bidirectional in physically inactive older individuals. We found no evidence of a longitudinal association between BST and frailty in older adults considered physically active. Our study provides preliminary longitudinal evidence that breaking-up ST more often reduces frailty in older adults that do not meet physical activity recommendations. Pending on experimental confirmation, targeting frequent reductions in ST may provide with a feasible approach to reduce the burden of frailty among more at-risk inactive older adults.
Acknowledgments
The authors would like to thank the cohort members, investigators, research associates, and team members.
Funding
This work was supported by the Biomedical Research Networking Center on Frailty and Healthy Aging (CIBERFES) and FEDER funds from the European Union (CB16/10/00477), (CB16/10/00456), and (CB16/10/00464). It was further funded by grants from the Government of Castilla-La Mancha (PI2010/020; Institute of Health Sciences, Ministry of Health of Castilla-La Mancha, 03031-00), Spanish Government (Spanish Ministry of Economy, “Ministerio de Economía y Competitividad,” Instituto de Salud Carlos III, PI10/01532, PI031558, PI11/01068), and by European Grants (Seventh Framework Programme: FRAILOMIC). P.B.J. is supported by the Portuguese Foundation for Science and Technology (SFRH/BPD/115977/2016). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Conflict of Interest
None declared.
Author Contributions
Conceptualization: A.M., B.d.P.-C., and I.A. Formal analysis: A.M. and B.d.P.-C. Methodology: A.M., I.R.-G., J.L.-R., L.R.-M., F.J.G.-G., and I.A. Investigation: A.M., B.d.P.-C., I.R.-G., J.L.-R., P.B.J., L.B.S., L.R.-M., F.J.G.-G., and I.A. Data curation: A.M., B.d.P.-C., I.R.-G., J.L.-R., L.R.-M., F.J.G.-G, and I.A. Original draft preparation: A.M. Review and editing: A.M., B.d.P.-C., I.R.-G., J.L.-R., P.B.J., L.B.S., L.R.-M., F.J.G.-G., and I.A. All authors have read and approved the final version of the manuscript.
References
Author notes
These authors contributed equally to this work.
These authors contributed equally to this work.
