https://orcid.org/0000-0001-8948-5419
Abstract
This study examines the relationship between various domains of sedentary behavior and subsequent cognitive function to evaluate whether different sedentary activities have specific associations with future cognitive performance.
Data were from 1 261 older adults participating in the Health, Aging, and Body Composition (Health ABC) Study between 1999/2000 and 2006/2007. Total sitting time (hours/day), reading time (hours/week), and TV time (≤27/≥28 h/wk) were self-reported at baseline and 3 years later. At follow-up, cognitive function was evaluated using the Teng Mini-Mental State Examination (3MS) and the Digit Symbol Substitution Test (DSST). Multivariable linear regression modeling examined the independent associations of baseline sedentary behaviors and 3-year change in those behaviors with cognitive function scores at follow-up, adjusting for important covariables.
Baseline total sitting time was positively associated with 3MS (β = 0.14 ± 0.07; p < .05) and DSST (β = 0.20 ± 0.10; p < .05) scores at follow-up, as was reading time (β = 0.09 ± 0.03; p < .05 for 3MS score and β = 0.14 ± 0.04; p < 0.01 for DSST score). Participants who increased their TV watching time over 3 years had a significantly lower 3MS score (β = −1.45 ± 0.71; p < .05) at follow-up, compared with those who maintained a low level of TV time (referent). These findings were independent of age, sex, race, education level, health status, depressive symptoms, and physical activity.
Some types of sedentary behavior may have benefits for cognitive function in older age, thus highlighting the importance of measuring different domains of sitting time.
Aging is the primary risk factor for cognitive decline and most types of dementia (1). Thus, due to aging demographics, the global prevalence of dementia is expected to nearly double from 36 million cases in 2010 to 66 million cases in 2030 (2,3). Dementia represents the most severe form of cognitive decline; however, older adults frequently experience deficits in cognitive function that do not rise to the level of a dementia diagnosis. Recent estimates indicate that cognitive impairment occurs in 11.2% of adults aged 55–69 and 27.3% of adults aged 65 or older (4,5). These estimates are expected to increase by 2050 (6,7) and have significant implications for the health and function of older adults, as well as for their families, and the healthcare system (8). Currently, there is no cure for dementia, which underscores the importance of examining the modifiable risk factors that can drive both the development and progression of this disease (9). One such modifiable factor is sedentary behavior, which to date has been understudied as a predictor of cognitive decline (10).
In 2017, the Sedentary Behaviour Research Network proposed a standardized definition of sedentary behavior as “any waking behaviour characterized by an energy expenditure ≤1.5 metabolic equivalents of task (METs) while in a sitting or reclining posture” (11). For reference, a MET is the ratio of the energy cost of performing a given activity relative to the resting metabolic rate = 1 MET. Sedentary behavior increases with age especially after the age of 60 years (12). In fact, recent estimates from the United States indicate that older adults report up to 11 h/d of sedentary time (13). This is concerning, as prolonged sedentary time has demonstrated a consistent adverse relationship with cardiometabolic risk, as well as with all-cause, cardiovascular, and cancer mortality (10).
The etiologic link between sedentary behavior and cognitive decline is complex, as is the time line for the development of cognitive impairments attributable to a sedentary lifestyle. Although research indicates that reducing sedentary time is beneficial for brain health, studies in older adults have been largely inconclusive, partially due to the cognitive status of the cohorts studied, the varying follow-up times, or inconsistent methods of assessing both sedentary behavior and cognitive function (14–19). Furthermore, the negative outcomes associated with being sedentary are exacerbated among those with low levels of physical activity, suggesting that the pattern of interaction between the 2 behaviors may be essential for understanding the association between activity levels and cognitive health (20).
Not all sedentary behaviors negatively affect cognitive function which may also explain some of the equivocal findings across studies. Indeed, while TV watching is associated with impaired cognitive function in older adults (21–23), reading and other cognitively stimulating activities (eg, computer use, puzzles, card games, or playing a musical instrument) performed while sedentary behaviors have been positively associated with cognition (24–28). To date, few studies have considered the role of different domains of sedentary behavior in relation to cognitive decline. Thus this study aimed to examine the distinct associations of baseline and 3-year changes in total sitting time, reading time, and TV watching with cognitive function at follow-up in a cohort of initially healthy participants in the Health, Aging, and Body Composition (Health ABC) Study. We hypothesized that a higher total sitting and TV time would be associated with lower cognitive performance, while greater reading time would show a positive association with cognitive performance, independent of physical activity and other known drivers of cognitive decline.
Method
Study Participants
Health ABC was conducted by the National Institute on Aging as a prospective cohort study of adults aged 70–79 years, who were Medicare beneficiaries living in Memphis, Tennessee, and Pittsburgh, Pennsylvania, at the time of recruitment in 1997–1998 (29). Participants were followed for 16 years, with data collection occurring using both questionnaires and clinical examinations. In order to be eligible for recruitment, participants had to report no difficulty walking ¼ mile, climbing 10 steps or performing basic activities of daily living, and walking without assistive equipment (30). In addition, participants could not have been treated for cancer in the 3 years prior to recruitment or have plans to leave the area within 3 years. By design, at least one-third of the participants were Black/African American. Data on sedentary behavior were first collected in Year 3 of the study (1999/2000), which thus served as our baseline for the current analysis. We also included data on sedentary behavior from 2002/2003 in order to assess 3-year changes in this behavior with respect to cognitive function at follow-up in 2006/2007 (Figure 1). The original Health ABC cohort in 1996/1997 comprised 3 075 participants, which dropped to 2 921 participants at the time of our baseline in 1999/2000. We then excluded 653 participants with missing data on sedentary behavior in both 1999/2000 and 2002/2003, and another 915 people with missing data on cognitive function at follow-up. Finally, to ensure that participants were not sedentary due to some functional limitation, the analysis was limited to those who report no more than a little difficulty walking ¼ mile. Thus, the final analytic cohort comprised 1 261 participants (Figure 2). A sensitivity analysis indicated that this analytic cohort had more white participants (71% vs 59%) and reported better health (53% vs 45% reporting excellent or very good health) compared with the full baseline cohort. Time spent in the domains of sedentary behavior appeared similar (Supplementary Table 1).
Study timeline to examine the longitudinal associations between domains of sedentary behaviors and cognitive function. 3MS = Teng Mini-Mental State Examination; CES-D = Centers for Epidemiologic Studies—Depression; DSST = Digit Symbol Substitution Test; Health ABC = Health, Aging, and Body Composition Study.
Participant selection from the Health ABC cohort. Those respondents who provided complete data on sedentary behavior and cognitive function or dementia and who were free of dementia and had limited difficult walking in 1999–2000 were eligible to be included in the analysis. 3MS = Teng Mini-Mental State Examination; CES-D = Centers for Epidemiologic Studies—Depression; DSST = Digit Symbol Substitution Test; Health ABC = Health, Aging, and Body Composition Study.
Sedentary Behavior
To assess sitting time, participants were asked about the number of hours they typically spend sitting upright over a 24-hour period. Reading time was assessed by asking participants about the hours they typically spend reading per week, which included books, newspapers, and magazines. Television time was assessed by asking participants about the hours per week that they typically spend watching TV, using a question with categorical responses (0–7; 8–14; 15–21; 22–28; 29–34; and ≥35 h/wk). For the analysis, sitting time and reading time were used as continuous variables, while categories of TV time were dichotomized (≤27/≥28 h/wk), based on recent findings suggesting a threshold of 3.5 h/d of TV watching before cognitive function becomes adversely affected (23).
Change in sitting and reading time between Years 3 and 6 was defined by creating a 3-level categorical variable for each (decreased = 2; increased = 1; maintained = 0 [referent]). Change in TV watching was categorized by a 4-level variable (maintaining a low level [0–27 h/wk) = 0 [referent]; maintaining a high level [28+ h/wk] = 1; decreasing from high to low = 2; or increasing from low to high = 3).
Cognitive Function Measures
Global cognition was assessed using the Teng Mini-Mental State Examination (3MS). The 3MS measures cognition across a range of domains, including orientation, attention, and recall, among others (31). Possible scores range from 0 to 100, with higher scores indicating better cognitive function. The Digit Symbol Substitution Test (DSST), a polyfactorial measure of cognition, assesses motor skills, attention, and executive function (32). One strength of the DSST is its high sensitivity to cognitive impairment, thus identifying deficits in cognitive function that the 3MS may not (32). Scores range from 0 to 100 and reflect the number of correctly matched numbers and symbols within 90 seconds, with higher scores indicating better performance.
Covariables
Covariables were included in the analysis based on their relationship to sedentary behavior or to cognitive decline in the literature (1,4,25,33). Sociodemographic variables included age, sex, race (white, black), and education level (<high school, high school grad, postsecondary), while health-related variables included smoking status (never, former, current smoker), body mass index (BMI: kg/m2), self-rated health (excellent, very good, good, fair, poor), physical activity (0–499, 500–1 999, 2 000+ kcal/wk; assessed using a modified version of the Minnesota Leisure-Time Physical Activity Questionnaire (34)), and depressive symptoms (assessed using the Centers for Epidemiologic Studies—Depression short-form scale [CES-D 10]) (35).
Statistical Analysis
Univariate statistics (mean ± SD; frequencies [%]) first were generated on all-study variables to determine their distributions within the study population. Pearson’s product moment correlation coefficients, Students t tests, and analysis of variance tested the unadjusted associations between each independent variable at baseline and cognitive function scores at follow-up. Multivariable linear regression modeled the associations of each domain of baseline sedentary behavior (sitting time, TV time, and reading time), as well as 3-year changes in these behaviors with 3MS and DSST scores at follow-up, while adjusting for important covariables. To achieve the most parsimonious models possible, final regression models included only those covariables achieving a p value of <.10 in the backward stepwise procedure, as well as physical activity. Finally, because sex and race may serve as important effect modifiers of the relation between sedentary time and cognitive function, first-order interaction terms between sex or race and each of the domains of sedentary time were entered into separate regression models. Parameter estimates, standard errors, and p values are reported for each domain of sedentary behavior. All analyses were performed in SAS 9.4 (SAS Institute, Cary, NC), at an alpha level of 0.05.
Results
Participants had an average score of 90.5 ± 9.1 on the 3MS and 35.7 ± 13.3 on the DSST. On average, participants were 75.1 ± 2.8 years of age (Table 1). Approximately 52% of participants were women and 71% were white. The majority of the study population was fairly well educated, as about 53% reported some postsecondary level of educational attainment. On average, participants were overweight (BMI = 27.2 ± 4.4 kg/m2), but rated their health as good or very good. About 75% were in the high or medium physical activity level category and reported current smoking was low. The mean score from the CES-D 10 was 3.9 ± 3.6, indicating low levels of depressive symptomology. Participants reported 6.7 ± 3.4 h/d of sitting and 12.5 ± 8.5 h/wk of reading, and about 80% of them reported watching TV between 0 and 27 h/wk.
Subject Characteristics: Health ABC Study (N = 1 261)
| Age (y) | 75.1± 2.7 |
| Female (%) | 52.2 |
| Race (%) | |
| Black | 29.0 |
| White | 71.1 |
| Educational attainment (%) | |
| Less than high school | 14.5 |
| High school graduate | 32.4 |
| Postsecondary | 53.1 |
| Self-rated health (%) | |
| Excellent | 15.5 |
| Very good | 37.6 |
| Good | 36.2 |
| Fair | 10.3 |
| Poor | 0.5 |
| Smoking status (%) | |
| Never | 49.1 |
| Current | 5.2 |
| Former | 45.8 |
| Physical activity, n (%) | |
| 2 000 + kcal/wk | 41.6 |
| 500–1 999 kcal/wk | 34.7 |
| 0–499 kcal/wk | 23.8 |
| Body mass index (kg/m2) | 27.2 ± 4.4 |
| CES-D 10 score | 3.9 ± 3.6 |
| Total sitting time (h/d) | 6.7 ± 3.3 |
| Reading time (h/wk) | 12.5 ± 8.6 |
| TV watching (%) | |
| 0–27 h/wk | 80.7 |
| 28+ h/wk | 19.3 |
| Age (y) | 75.1± 2.7 |
| Female (%) | 52.2 |
| Race (%) | |
| Black | 29.0 |
| White | 71.1 |
| Educational attainment (%) | |
| Less than high school | 14.5 |
| High school graduate | 32.4 |
| Postsecondary | 53.1 |
| Self-rated health (%) | |
| Excellent | 15.5 |
| Very good | 37.6 |
| Good | 36.2 |
| Fair | 10.3 |
| Poor | 0.5 |
| Smoking status (%) | |
| Never | 49.1 |
| Current | 5.2 |
| Former | 45.8 |
| Physical activity, n (%) | |
| 2 000 + kcal/wk | 41.6 |
| 500–1 999 kcal/wk | 34.7 |
| 0–499 kcal/wk | 23.8 |
| Body mass index (kg/m2) | 27.2 ± 4.4 |
| CES-D 10 score | 3.9 ± 3.6 |
| Total sitting time (h/d) | 6.7 ± 3.3 |
| Reading time (h/wk) | 12.5 ± 8.6 |
| TV watching (%) | |
| 0–27 h/wk | 80.7 |
| 28+ h/wk | 19.3 |
Note: CES-D = Centers for Epidemiologic Studies—Depression.
Subject Characteristics: Health ABC Study (N = 1 261)
| Age (y) | 75.1± 2.7 |
| Female (%) | 52.2 |
| Race (%) | |
| Black | 29.0 |
| White | 71.1 |
| Educational attainment (%) | |
| Less than high school | 14.5 |
| High school graduate | 32.4 |
| Postsecondary | 53.1 |
| Self-rated health (%) | |
| Excellent | 15.5 |
| Very good | 37.6 |
| Good | 36.2 |
| Fair | 10.3 |
| Poor | 0.5 |
| Smoking status (%) | |
| Never | 49.1 |
| Current | 5.2 |
| Former | 45.8 |
| Physical activity, n (%) | |
| 2 000 + kcal/wk | 41.6 |
| 500–1 999 kcal/wk | 34.7 |
| 0–499 kcal/wk | 23.8 |
| Body mass index (kg/m2) | 27.2 ± 4.4 |
| CES-D 10 score | 3.9 ± 3.6 |
| Total sitting time (h/d) | 6.7 ± 3.3 |
| Reading time (h/wk) | 12.5 ± 8.6 |
| TV watching (%) | |
| 0–27 h/wk | 80.7 |
| 28+ h/wk | 19.3 |
| Age (y) | 75.1± 2.7 |
| Female (%) | 52.2 |
| Race (%) | |
| Black | 29.0 |
| White | 71.1 |
| Educational attainment (%) | |
| Less than high school | 14.5 |
| High school graduate | 32.4 |
| Postsecondary | 53.1 |
| Self-rated health (%) | |
| Excellent | 15.5 |
| Very good | 37.6 |
| Good | 36.2 |
| Fair | 10.3 |
| Poor | 0.5 |
| Smoking status (%) | |
| Never | 49.1 |
| Current | 5.2 |
| Former | 45.8 |
| Physical activity, n (%) | |
| 2 000 + kcal/wk | 41.6 |
| 500–1 999 kcal/wk | 34.7 |
| 0–499 kcal/wk | 23.8 |
| Body mass index (kg/m2) | 27.2 ± 4.4 |
| CES-D 10 score | 3.9 ± 3.6 |
| Total sitting time (h/d) | 6.7 ± 3.3 |
| Reading time (h/wk) | 12.5 ± 8.6 |
| TV watching (%) | |
| 0–27 h/wk | 80.7 |
| 28+ h/wk | 19.3 |
Note: CES-D = Centers for Epidemiologic Studies—Depression.
In the simple (unadjusted) analysis, the total sitting time at baseline was positively correlated with 3MS (r = 0.07; p < .05) and DSST (r = 0.06; p < .05) scores at follow-up, as was reading time (r = 0.17; p < .001 and r = 0.18; p < .001 for the 3MS and DSST scores, respectively). Scores on the 3MS were slightly higher in participants reporting 0–27 h/wk of TV watching (90.7 ± 9.0), compared with those who watched more (89.4 ± 9.2; p < .06). There was no association between baseline TV watching and DSST score at follow-up. We also observed significant associations between age, sex, race, education, and self-rated health with cognitive function scores in the simple analysis.
After adjustment for the covariables age, sex, race, education, self-rated health, CES-D score, and physical activity in the multivariable analysis, the total sitting time remained positively associated with both 3MS (β = 0.14 ± 0.07; p < .05) and DSST (β = 0.20 ± 0.10; p < .05) scores at follow-up, which was contrary to the proposed hypothesis (Table 2a). As indicated by the parameter estimates, each hour of sitting per day increased 3MS scores by 0.14 points and DSST scores by 0.20 points. Reading time also remained positively associated with both 3MS (β = 0.09 ± 0.03; p < .01) and DSST (β = 0.14 ± 0.04; p < .01) scores (Table 2b). Baseline TV watching demonstrated no independent associations with any of the cognitive function scores at follow-up in the presence of several consistently powerful covariables—namely, age, sex, race, education, and health status. We did not observe a statistically significant interaction between either sex or race and any of the domains of sedentary time.
Independent Association of Baseline Sitting Time and Baseline Reading Time With 3MS and DSST Scores at Follow-up.
| (a) Baseline Sitting Time | ||||||
|---|---|---|---|---|---|---|
| 3MS Score | DSST Score | |||||
| Coefficient | Std. Err | p | Coefficient | Std. Err | p | |
| Total sitting (h/d) | 0.14 | 0.07 | * | 0.20 | 0.10 | * |
| Physical activity | ||||||
| 2 000 + kcal/wk | REF | REF | ||||
| 500–1 999 kcal/wk | 0.16 | 0.55 | 0.41 | 0.77 | ||
| 0-499 kcal/wk | −1.17 | 0.62 | −0.43 | 0.88 | ||
| Age (y) | −0.48 | 0.09 | ** | −0.82 | 0.12 | ** |
| Female | 1.64 | 0.48 | ** | 4.42 | 0.68 | ** |
| Black race | −3.18 | 0.56 | ** | −6.33 | 0.78 | ** |
| Educational attainment | ||||||
| Postsecondary | REF | REF | ||||
| High school graduate | −2.58 | 0.53 | ** | −3.34 | 0.75 | ** |
| Less than high school | −6.87 | 0.73 | ** | −11.08 | 1.04 | ** |
| Self-rated health | ||||||
| Excellent | REF | REF | ||||
| Very good | −0.02 | 0.71 | −2.03 | 1.00 | * | |
| Good | −1.35 | 0.74 | −4.57 | 1.04 | ** | |
| Fair | −1.76 | 1.02 | −5.85 | 1.44 | ** | |
| Poor | −1.19 | 3.47 | −9.02 | 4.85 | ||
| CES-D 10 score | −0.13 | 0.07 | −0.17 | 0.10 | ||
| (b) Baseline Reading Time | ||||||
| Reading time (h/wk) | 0.09 | 0.03 | ** | 0.14 | 0.04 | ** |
| Physical activity | ||||||
| 2 000 + kcal/wk | REF | REF | ||||
| 500–1 999 kcal/wk | 0.09 | 0.54 | 0.31 | 0.77 | ||
| 0–499 kcal/wk | −1.15 | 0.62 | −0.39 | 0.88 | ||
| Age (y) | −0.48 | 0.09 | ** | −0.83 | 0.12 | ** |
| Female | 1.58 | 0.48 | ** | 4.33 | 0.68 | ** |
| Black race | −3.14 | 0.56 | ** | −6.26 | 0.78 | ** |
| Educational attainment | ||||||
| Postsecondary | REF | REF | ||||
| High school graduate | −2.34 | 0.54 | ** | −2.98 | 0.76 | ** |
| Less than high school | −6.55 | 0.74 | ** | −10.54 | 1.05 | ** |
| Self-rated health | ||||||
| Excellent | REF | REF | ||||
| Very good | −0.04 | 0.71 | −2.05 | 1.00 | * | |
| Good | −1.25 | 0.74 | −4.42 | 1.04 | ** | |
| Fair | −1.65 | 1.02 | −5.69 | 1.44 | ** | |
| Poor | −1.17 | 3.46 | −8.97 | 4.83 | ||
| CES-D 10 score | −0.12 | 0.07 | −0.16 | 0.10 |
| (a) Baseline Sitting Time | ||||||
|---|---|---|---|---|---|---|
| 3MS Score | DSST Score | |||||
| Coefficient | Std. Err | p | Coefficient | Std. Err | p | |
| Total sitting (h/d) | 0.14 | 0.07 | * | 0.20 | 0.10 | * |
| Physical activity | ||||||
| 2 000 + kcal/wk | REF | REF | ||||
| 500–1 999 kcal/wk | 0.16 | 0.55 | 0.41 | 0.77 | ||
| 0-499 kcal/wk | −1.17 | 0.62 | −0.43 | 0.88 | ||
| Age (y) | −0.48 | 0.09 | ** | −0.82 | 0.12 | ** |
| Female | 1.64 | 0.48 | ** | 4.42 | 0.68 | ** |
| Black race | −3.18 | 0.56 | ** | −6.33 | 0.78 | ** |
| Educational attainment | ||||||
| Postsecondary | REF | REF | ||||
| High school graduate | −2.58 | 0.53 | ** | −3.34 | 0.75 | ** |
| Less than high school | −6.87 | 0.73 | ** | −11.08 | 1.04 | ** |
| Self-rated health | ||||||
| Excellent | REF | REF | ||||
| Very good | −0.02 | 0.71 | −2.03 | 1.00 | * | |
| Good | −1.35 | 0.74 | −4.57 | 1.04 | ** | |
| Fair | −1.76 | 1.02 | −5.85 | 1.44 | ** | |
| Poor | −1.19 | 3.47 | −9.02 | 4.85 | ||
| CES-D 10 score | −0.13 | 0.07 | −0.17 | 0.10 | ||
| (b) Baseline Reading Time | ||||||
| Reading time (h/wk) | 0.09 | 0.03 | ** | 0.14 | 0.04 | ** |
| Physical activity | ||||||
| 2 000 + kcal/wk | REF | REF | ||||
| 500–1 999 kcal/wk | 0.09 | 0.54 | 0.31 | 0.77 | ||
| 0–499 kcal/wk | −1.15 | 0.62 | −0.39 | 0.88 | ||
| Age (y) | −0.48 | 0.09 | ** | −0.83 | 0.12 | ** |
| Female | 1.58 | 0.48 | ** | 4.33 | 0.68 | ** |
| Black race | −3.14 | 0.56 | ** | −6.26 | 0.78 | ** |
| Educational attainment | ||||||
| Postsecondary | REF | REF | ||||
| High school graduate | −2.34 | 0.54 | ** | −2.98 | 0.76 | ** |
| Less than high school | −6.55 | 0.74 | ** | −10.54 | 1.05 | ** |
| Self-rated health | ||||||
| Excellent | REF | REF | ||||
| Very good | −0.04 | 0.71 | −2.05 | 1.00 | * | |
| Good | −1.25 | 0.74 | −4.42 | 1.04 | ** | |
| Fair | −1.65 | 1.02 | −5.69 | 1.44 | ** | |
| Poor | −1.17 | 3.46 | −8.97 | 4.83 | ||
| CES-D 10 score | −0.12 | 0.07 | −0.16 | 0.10 |
Note: 3MS = Teng Mini-Mental State Examination; CES-D = Centers for Epidemiologic Studies—Depression; DSST = Digit Symbol Substitution Test.
*p < .05. **p < .01.
Independent Association of Baseline Sitting Time and Baseline Reading Time With 3MS and DSST Scores at Follow-up.
| (a) Baseline Sitting Time | ||||||
|---|---|---|---|---|---|---|
| 3MS Score | DSST Score | |||||
| Coefficient | Std. Err | p | Coefficient | Std. Err | p | |
| Total sitting (h/d) | 0.14 | 0.07 | * | 0.20 | 0.10 | * |
| Physical activity | ||||||
| 2 000 + kcal/wk | REF | REF | ||||
| 500–1 999 kcal/wk | 0.16 | 0.55 | 0.41 | 0.77 | ||
| 0-499 kcal/wk | −1.17 | 0.62 | −0.43 | 0.88 | ||
| Age (y) | −0.48 | 0.09 | ** | −0.82 | 0.12 | ** |
| Female | 1.64 | 0.48 | ** | 4.42 | 0.68 | ** |
| Black race | −3.18 | 0.56 | ** | −6.33 | 0.78 | ** |
| Educational attainment | ||||||
| Postsecondary | REF | REF | ||||
| High school graduate | −2.58 | 0.53 | ** | −3.34 | 0.75 | ** |
| Less than high school | −6.87 | 0.73 | ** | −11.08 | 1.04 | ** |
| Self-rated health | ||||||
| Excellent | REF | REF | ||||
| Very good | −0.02 | 0.71 | −2.03 | 1.00 | * | |
| Good | −1.35 | 0.74 | −4.57 | 1.04 | ** | |
| Fair | −1.76 | 1.02 | −5.85 | 1.44 | ** | |
| Poor | −1.19 | 3.47 | −9.02 | 4.85 | ||
| CES-D 10 score | −0.13 | 0.07 | −0.17 | 0.10 | ||
| (b) Baseline Reading Time | ||||||
| Reading time (h/wk) | 0.09 | 0.03 | ** | 0.14 | 0.04 | ** |
| Physical activity | ||||||
| 2 000 + kcal/wk | REF | REF | ||||
| 500–1 999 kcal/wk | 0.09 | 0.54 | 0.31 | 0.77 | ||
| 0–499 kcal/wk | −1.15 | 0.62 | −0.39 | 0.88 | ||
| Age (y) | −0.48 | 0.09 | ** | −0.83 | 0.12 | ** |
| Female | 1.58 | 0.48 | ** | 4.33 | 0.68 | ** |
| Black race | −3.14 | 0.56 | ** | −6.26 | 0.78 | ** |
| Educational attainment | ||||||
| Postsecondary | REF | REF | ||||
| High school graduate | −2.34 | 0.54 | ** | −2.98 | 0.76 | ** |
| Less than high school | −6.55 | 0.74 | ** | −10.54 | 1.05 | ** |
| Self-rated health | ||||||
| Excellent | REF | REF | ||||
| Very good | −0.04 | 0.71 | −2.05 | 1.00 | * | |
| Good | −1.25 | 0.74 | −4.42 | 1.04 | ** | |
| Fair | −1.65 | 1.02 | −5.69 | 1.44 | ** | |
| Poor | −1.17 | 3.46 | −8.97 | 4.83 | ||
| CES-D 10 score | −0.12 | 0.07 | −0.16 | 0.10 |
| (a) Baseline Sitting Time | ||||||
|---|---|---|---|---|---|---|
| 3MS Score | DSST Score | |||||
| Coefficient | Std. Err | p | Coefficient | Std. Err | p | |
| Total sitting (h/d) | 0.14 | 0.07 | * | 0.20 | 0.10 | * |
| Physical activity | ||||||
| 2 000 + kcal/wk | REF | REF | ||||
| 500–1 999 kcal/wk | 0.16 | 0.55 | 0.41 | 0.77 | ||
| 0-499 kcal/wk | −1.17 | 0.62 | −0.43 | 0.88 | ||
| Age (y) | −0.48 | 0.09 | ** | −0.82 | 0.12 | ** |
| Female | 1.64 | 0.48 | ** | 4.42 | 0.68 | ** |
| Black race | −3.18 | 0.56 | ** | −6.33 | 0.78 | ** |
| Educational attainment | ||||||
| Postsecondary | REF | REF | ||||
| High school graduate | −2.58 | 0.53 | ** | −3.34 | 0.75 | ** |
| Less than high school | −6.87 | 0.73 | ** | −11.08 | 1.04 | ** |
| Self-rated health | ||||||
| Excellent | REF | REF | ||||
| Very good | −0.02 | 0.71 | −2.03 | 1.00 | * | |
| Good | −1.35 | 0.74 | −4.57 | 1.04 | ** | |
| Fair | −1.76 | 1.02 | −5.85 | 1.44 | ** | |
| Poor | −1.19 | 3.47 | −9.02 | 4.85 | ||
| CES-D 10 score | −0.13 | 0.07 | −0.17 | 0.10 | ||
| (b) Baseline Reading Time | ||||||
| Reading time (h/wk) | 0.09 | 0.03 | ** | 0.14 | 0.04 | ** |
| Physical activity | ||||||
| 2 000 + kcal/wk | REF | REF | ||||
| 500–1 999 kcal/wk | 0.09 | 0.54 | 0.31 | 0.77 | ||
| 0–499 kcal/wk | −1.15 | 0.62 | −0.39 | 0.88 | ||
| Age (y) | −0.48 | 0.09 | ** | −0.83 | 0.12 | ** |
| Female | 1.58 | 0.48 | ** | 4.33 | 0.68 | ** |
| Black race | −3.14 | 0.56 | ** | −6.26 | 0.78 | ** |
| Educational attainment | ||||||
| Postsecondary | REF | REF | ||||
| High school graduate | −2.34 | 0.54 | ** | −2.98 | 0.76 | ** |
| Less than high school | −6.55 | 0.74 | ** | −10.54 | 1.05 | ** |
| Self-rated health | ||||||
| Excellent | REF | REF | ||||
| Very good | −0.04 | 0.71 | −2.05 | 1.00 | * | |
| Good | −1.25 | 0.74 | −4.42 | 1.04 | ** | |
| Fair | −1.65 | 1.02 | −5.69 | 1.44 | ** | |
| Poor | −1.17 | 3.46 | −8.97 | 4.83 | ||
| CES-D 10 score | −0.12 | 0.07 | −0.16 | 0.10 |
Note: 3MS = Teng Mini-Mental State Examination; CES-D = Centers for Epidemiologic Studies—Depression; DSST = Digit Symbol Substitution Test.
*p < .05. **p < .01.
Neither a 3-year change (increase or decrease) in sitting time nor reading time was independently associated with cognitive function at follow-up, compared with those who maintained the same amount of sitting or reading. In contrast, however, a 3-year categorical increase in TV watching (0–27 to ≥28 h/wk) was associated with a significantly lower 3MS score (β = −1.45 ± 0.71; p < .05) at follow-up, compared with participants who maintained a low level of TV watching (Table 3).
Independent Associations of 3-Year Change in TV Watching Time With 3MS and DSST Scores at Follow-up.
| 3MS Score | DSST Score | |||||
|---|---|---|---|---|---|---|
| Coefficient | Std. Err | p | Coefficient | Std. Err | p | |
| Change in TV watching | ||||||
| 0–27 h/wk both years | REF | REF | ||||
| 28+ h/wk both years | −0.12 | 0.79 | 1.42 | 1.11 | ||
| 3-year increase | −1.45 | 0.71 | * | −0.80 | 1.01 | |
| 3-year decrease | −1.10 | 0.85 | −0.98 | 1.20 | ||
| Physical activity | ||||||
| 2 000 + kcal/wk | REF | REF | ||||
| 500–1 999 kcal/wk | 0.15 | 0.55 | 0.31 | 0.77 | ||
| 0–499 kcal/wk | −1.10 | 0.63 | −0.46 | 0.88 | ||
| Age (y) | −0.48 | 0.09 | ** | −0.82 | 0.12 | ** |
| Female | 1.61 | 0.49 | ** | 4.41 | 0.69 | ** |
| Black race | −3.12 | 0.56 | ** | −6.31 | 0.79 | ** |
| Educational attainment | ||||||
| Postsecondary | REF | REF | ||||
| High school graduate | −2.54 | 0.54 | ** | −3.50 | 0.75 | ** |
| Less than high school | −6.87 | 0.74 | ** | −11.19 | 1.04 | ** |
| Self-rated health | ||||||
| Excellent | REF | REF | ||||
| Very good | −0.03 | 0.71 | −2.03 | 1.00 | * | |
| Good | −1.33 | 0.74 | −4.57 | 1.04 | ** | |
| Fair | −1.74 | 1.02 | −5.86 | 1.44 | ** | |
| Poor | −1.51 | 3.47 | −9.50 | 4.85 | * | |
| CES-D 10 score | −0.11 | 0.07 | −0.16 | 0.10 |
| 3MS Score | DSST Score | |||||
|---|---|---|---|---|---|---|
| Coefficient | Std. Err | p | Coefficient | Std. Err | p | |
| Change in TV watching | ||||||
| 0–27 h/wk both years | REF | REF | ||||
| 28+ h/wk both years | −0.12 | 0.79 | 1.42 | 1.11 | ||
| 3-year increase | −1.45 | 0.71 | * | −0.80 | 1.01 | |
| 3-year decrease | −1.10 | 0.85 | −0.98 | 1.20 | ||
| Physical activity | ||||||
| 2 000 + kcal/wk | REF | REF | ||||
| 500–1 999 kcal/wk | 0.15 | 0.55 | 0.31 | 0.77 | ||
| 0–499 kcal/wk | −1.10 | 0.63 | −0.46 | 0.88 | ||
| Age (y) | −0.48 | 0.09 | ** | −0.82 | 0.12 | ** |
| Female | 1.61 | 0.49 | ** | 4.41 | 0.69 | ** |
| Black race | −3.12 | 0.56 | ** | −6.31 | 0.79 | ** |
| Educational attainment | ||||||
| Postsecondary | REF | REF | ||||
| High school graduate | −2.54 | 0.54 | ** | −3.50 | 0.75 | ** |
| Less than high school | −6.87 | 0.74 | ** | −11.19 | 1.04 | ** |
| Self-rated health | ||||||
| Excellent | REF | REF | ||||
| Very good | −0.03 | 0.71 | −2.03 | 1.00 | * | |
| Good | −1.33 | 0.74 | −4.57 | 1.04 | ** | |
| Fair | −1.74 | 1.02 | −5.86 | 1.44 | ** | |
| Poor | −1.51 | 3.47 | −9.50 | 4.85 | * | |
| CES-D 10 score | −0.11 | 0.07 | −0.16 | 0.10 |
Note: 3MS = Teng Mini-Mental State Examination; CES-D = Centers for Epidemiologic Studies—Depression; DSST = Digit Symbol Substitution Test.
**p < .01. *p < .05.
Independent Associations of 3-Year Change in TV Watching Time With 3MS and DSST Scores at Follow-up.
| 3MS Score | DSST Score | |||||
|---|---|---|---|---|---|---|
| Coefficient | Std. Err | p | Coefficient | Std. Err | p | |
| Change in TV watching | ||||||
| 0–27 h/wk both years | REF | REF | ||||
| 28+ h/wk both years | −0.12 | 0.79 | 1.42 | 1.11 | ||
| 3-year increase | −1.45 | 0.71 | * | −0.80 | 1.01 | |
| 3-year decrease | −1.10 | 0.85 | −0.98 | 1.20 | ||
| Physical activity | ||||||
| 2 000 + kcal/wk | REF | REF | ||||
| 500–1 999 kcal/wk | 0.15 | 0.55 | 0.31 | 0.77 | ||
| 0–499 kcal/wk | −1.10 | 0.63 | −0.46 | 0.88 | ||
| Age (y) | −0.48 | 0.09 | ** | −0.82 | 0.12 | ** |
| Female | 1.61 | 0.49 | ** | 4.41 | 0.69 | ** |
| Black race | −3.12 | 0.56 | ** | −6.31 | 0.79 | ** |
| Educational attainment | ||||||
| Postsecondary | REF | REF | ||||
| High school graduate | −2.54 | 0.54 | ** | −3.50 | 0.75 | ** |
| Less than high school | −6.87 | 0.74 | ** | −11.19 | 1.04 | ** |
| Self-rated health | ||||||
| Excellent | REF | REF | ||||
| Very good | −0.03 | 0.71 | −2.03 | 1.00 | * | |
| Good | −1.33 | 0.74 | −4.57 | 1.04 | ** | |
| Fair | −1.74 | 1.02 | −5.86 | 1.44 | ** | |
| Poor | −1.51 | 3.47 | −9.50 | 4.85 | * | |
| CES-D 10 score | −0.11 | 0.07 | −0.16 | 0.10 |
| 3MS Score | DSST Score | |||||
|---|---|---|---|---|---|---|
| Coefficient | Std. Err | p | Coefficient | Std. Err | p | |
| Change in TV watching | ||||||
| 0–27 h/wk both years | REF | REF | ||||
| 28+ h/wk both years | −0.12 | 0.79 | 1.42 | 1.11 | ||
| 3-year increase | −1.45 | 0.71 | * | −0.80 | 1.01 | |
| 3-year decrease | −1.10 | 0.85 | −0.98 | 1.20 | ||
| Physical activity | ||||||
| 2 000 + kcal/wk | REF | REF | ||||
| 500–1 999 kcal/wk | 0.15 | 0.55 | 0.31 | 0.77 | ||
| 0–499 kcal/wk | −1.10 | 0.63 | −0.46 | 0.88 | ||
| Age (y) | −0.48 | 0.09 | ** | −0.82 | 0.12 | ** |
| Female | 1.61 | 0.49 | ** | 4.41 | 0.69 | ** |
| Black race | −3.12 | 0.56 | ** | −6.31 | 0.79 | ** |
| Educational attainment | ||||||
| Postsecondary | REF | REF | ||||
| High school graduate | −2.54 | 0.54 | ** | −3.50 | 0.75 | ** |
| Less than high school | −6.87 | 0.74 | ** | −11.19 | 1.04 | ** |
| Self-rated health | ||||||
| Excellent | REF | REF | ||||
| Very good | −0.03 | 0.71 | −2.03 | 1.00 | * | |
| Good | −1.33 | 0.74 | −4.57 | 1.04 | ** | |
| Fair | −1.74 | 1.02 | −5.86 | 1.44 | ** | |
| Poor | −1.51 | 3.47 | −9.50 | 4.85 | * | |
| CES-D 10 score | −0.11 | 0.07 | −0.16 | 0.10 |
Note: 3MS = Teng Mini-Mental State Examination; CES-D = Centers for Epidemiologic Studies—Depression; DSST = Digit Symbol Substitution Test.
**p < .01. *p < .05.
Discussion
This study is among the first to examine how different domains of sedentary behavior, as well as changes in those behaviors, are longitudinally associated with cognitive function in older adults. Our results were inconsistent, thereby supporting only a few of our hypotheses. This suggests that sedentary behavior may have important contributions to cognitive function in older age, but that these relationships are complex, and may be especially so when controlling for several very powerful drivers of cognitive decline.
Previous research on the relationship between sedentary behavior and cognitive function has also yielded mixed results (15–17,36). One possible explanation for these equivocal findings is that not all types of sedentary behavior affect cognitive function similarly. Indeed, cognitively stimulating activities performed while seated could actually maintain or even improve cognition (24–28). The baseline measure of total sitting time in the Health ABC study was positively associated with higher 3MS (a measure of global cognitive function) and DSST scores (a polyfactorial measure of cognition) at follow-up. Other investigators (24,27,36) also recently reported higher cognitive function with greater amounts of self-reported and objectively measured total sedentary time. In a sub-group analysis, however, Wanders et al. (27) reported that work-related sitting (but not leisure-time sitting, TV time, or reading time) and nonoccupational computer time were the primary determinants of this positive association. Thus, survey questions and device-based measures of total sitting time may be unable to distinguish the specific contribution of different sitting domains to brain health.
In light of this complexity, we also investigated reading and TV time to examine the relationships between domains of sedentary activity that are and are not cognitively stimulating. Higher amounts of reading time at baseline were associated with higher scores on the 3MS and DSST at follow-up, which corroborates results from previous research (25,26). On the other hand, baseline levels of TV watching (an activity that is passive in nature) were not associated with cognitive function scores at follow-up, which is similar to other recent findings (23,27) using an outcome measure of global cognitive function similar to the 3MS. The best evidence to date comes from a prospective study of 146 651 older participants in the UK Biobank (28) from which the authors report that time spent watching TV increased the risk of dementia by nearly 25% (hazard ratio [HR] = 1.24; 95% CI: 1.15, 1.32), whereas time spent using a computer decreased dementia risk (HR = 0.85; 95% CI: 0.80, 0.90), independent of physical activity and other covariables. Nemoto and colleagues (37) recently reported on the independent and joint effects of sedentary time and self-reported physical activity with 5-year dementia risk in older Japanese adults. In this cohort, daily TV watching was not independently associated with dementia risk, while daily book-reading time was. The observed benefits of reading to cognitive health were especially pronounced at higher levels of physical activity. In fact, dementia risk for participants reporting ≥16.0 MET-h/wk of activity and at least 10 min/d of reading was about 60% lower compared with those reporting low physical activity (<2.5 MET-h/wk) and low reading time (<10 min/d). Results from previous other studies on different domains of sedentary behavior have been inconsistent, due to the age of the population studied (32), the study design (36), or the domains of cognitive function assessed (22,24,38,39).
Three-year change in TV watching was the only longitudinal domain of sedentary time that was significantly associated with cognitive function at follow-up. Those who increased their TV time from the low category (0–27 h/wk) to the high category (28+ h/wk) over 3 years performed significantly worse on the 3MS at follow-up than those who maintained the same level of TV viewing, although the effect size was somewhat small. To our knowledge, the only other study (33) that examined longitudinal changes between sedentary behavior and cognitive decline reported null findings, which may have been due to their study cohort being about 20 years younger than the Health ABC cohort (55 vs 75 years). Increased TV watching over 3 years may be an important risk factor for cognitive decline among people in their eighth decade of life, or it could be indicative of a prevailing decline in cognitive function, as people begin to disassociate from social and/or cognitively stimulating activities in favor of passive activities. Thus, a change in this particular domain of sedentary behavior could be a correlate, rather than the cause of the cognitive decline.
Importantly, the Health ABC cohort is a relatively robust group of older men and women. As such, selective survival is always a concern with regard to null findings. That is, the punitive risks of sedentary behavior to health and function may affect more vulnerable adults before age 70, and therefore, they are not represented in our study population. Health ABC study participants are well educated and the average score on the DSST was higher than national normative values for this age group (40). In fact, 53% of the final analytic cohort had some postsecondary level of educational attainment, which is substantially higher than that of the full Health ABC cohort at enrollment (41). It is possible that in these healthy, cognitively intact older people, a longer follow-up time would be necessary to detect more meaningful declines in cognitive performance. Also, we adjusted our analyses for level of physical activity. Edwards and Loprinzi (42) provide evidence that the association between sedentary behavior and cognitive function becomes attenuated when moderate-to-vigorous physical activity is included in the statistical modeling. In theory, adequate levels of physical activity could mitigate the harmful effects of sedentary behavior on cognitive performance. We performed a second sensitivity analysis in which we ran the multivariable modeling with and without adjustment for physical activity. The results indicated that the parameter estimates were only marginally attenuated by the addition of physical activity and thus, the results were unchanged.
A major strength of this analysis was the longitudinal design, which allowed the examination of the relationship between changes in sedentary behavior and cognitive function. We know of only 1 other study that has examined these associations longitudinally (33). Our study population also was balanced by sex, and we tried to reduce the possibility of reverse causality by excluding participants who screened positive for dementia at baseline, as well as those reporting mobility limitations. The validity and reliability of the sedentary domain questions have not been established, however, and information on sedentary behavior and physical activity was self-reported. There is ample evidence to indicate that people tend to underreport sedentary time, while overreporting physical activity (43). This was likely the case in our study, as participants reported an average of about 7 h/d of sitting and 42% reported ≥2 000 kcal/wk of physical activity, which does not match national age-specific normative values for these behaviors (13,44). Misreporting of sedentary behavior and physical activity also may have been nondifferential across cognitive function scores, further attenuating our findings toward the null.
Our findings and those of others indicate that the relationship between sedentary behavior and cognition remains complex. Some types of sedentary behavior may have benefits for cognitive function in older age, thus highlighting the importance of measuring different domains of sitting time and overall time use. Given global aging demographics and the fact that older adults may spend up to 60% of their waking hours sitting or reclining (45), there is a substantial public health benefit to understanding how sedentary behavior affects and/or reflects brain health in older age.
Acknowledgments
L.M. and L.D.P. conceptualized the hypotheses and designed the statistical analysis plan; L.M. performed all data analyses and was supported by E.M.S., M.A.N., and L.D.P.; all authors (L.M., E.M.S., M.A.N., and L.D.P.) contributed equally to the writing, editing, and revising of the manuscript.
Funding
This work was supported by National Institute of Aging (contracts N01AG62101, N01AG62103, and N01AG62106) to E.M.S.; National Institute of Aging (R01AG028050); and National Institute of Nursing Research (R01NR012459).
Conflict of Interest
None declared.
