Skip to main content
NCBI home page
The Journals of Gerontology Series B: Psychological Sciences and Social Sciences logoLink to The Journals of Gerontology Series B: Psychological Sciences and Social Sciences
. 2025 Apr 18;80(7):gbaf071. doi: 10.1093/geronb/gbaf071

Nonlinear Associations of Accelerometer-Based Sedentary Time With Cognitive Functions in the UK Biobank

Silvio Maltagliati 1,✉, Daniel H Aslan 2, M Katherine Sayre 3, Pradyumna K Bharadwaj 4, Madeline Ally 5, Mark H C Lai 6, Rand R Wilcox 7, Yann C Klimentidis 8,9, Gene E Alexander 10,11, David A Raichlen 12,13
Editor: Martina Luchetti14
PMCID: PMC12150772  PMID: 40247819

Abstract

Objectives

Device-based sedentary time shows a nonlinear association with incident dementia among older adults. However, associations between sedentary time and cognitive performance have been inconsistent. We examined potential nonlinear associations between sedentary time and performance on cognitive tests among older adults.

Methods

We used data from the UK Biobank and included 32,875 adults aged 60–79. Sedentary time was estimated from a machine learning–based analysis of 1 week of wrist-worn accelerometer data. The primary outcomes were performance on 6 cognitive tests completed online (fluid intelligence test, short-term numeric memory test, symbol substitution test, visual-spatial memory test, alphanumeric, and numeric trail making tests), as well as a composite cognitive score.

Results

Except for the visual-spatial memory test, nonlinear approaches provided a better fit than linear methods to model the associations of sedentary time with other cognitive outcomes. For these outcomes, segmented regression models showed that, although effect sizes were small, higher sedentary time was associated with better cognitive performance up to a threshold of sedentary time that varied from 9.7 to 12.3 hr per day. Above this threshold, the association between sedentary time and cognitive performance was attenuated toward the null or became negative (for the symbol substitution test only).

Discussion

As accounted by our nonlinear approach, the association between sedentary time and cognitive performance may shift from positive to null or negative above a 10–12-hr threshold among older adults. A combination of device-based and self-report assessments of sedentary behavior is needed to better understand these nonlinear associations.

Keywords: Aging, Brain, Cognition, Epidemiology, Sitting time

Dementia affects around 50 million individuals worldwide—with prevalence having doubled since 1990 and expected to increase to 152 million by 2050 (Livingston et al., 2020). Promoting cognitive health across the lifespan constitutes a public health priority, and nonpharmacological lifestyle interventions are thought to be central to this goal as an estimated 40% of worldwide dementia cases may be potentially prevented through changes in modifiable risk factors (Livingston et al., 2020). Among those factors, some evidence suggests adopting a physically active lifestyle may protect against cognitive decline in older adults. Physical activity is not only associated with reduced risks of dementia (Iso-Markku et al., 2022) and structural indicators of brain health in older age (e.g., greater hippocampal volumes, Erickson et al., 2011), but it also favors the maintenance of cognitive functions across aging (Cheval et al., 2021; Hamer et al., 2018). For example, a Mendelian randomization study drawing on large-scale genome-wide association studies has recently supported the beneficial role of physical activity on cognitive functioning (Cheval et al., 2023). Although the factors attenuating (e.g., air pollution, Raichlen, Furlong, et al., 2022) or reinforcing (e.g., exposure to nature, Rhee et al., 2023) the benefits of this behavior remain to be fully identified, previous research has collectively demonstrated the potential importance of physical activity as a modifiable lifestyle factor that can support and enhance cognitive functioning across aging.

As opposed to physical activity, sedentary behaviors (i.e., any behavior defined by an energy expenditure of 1.5 Metabolic Equivalent of Task or less, while sitting, reclining, or lying down, Tremblay et al., 2017) occupy a significant portion of adults’ waking time (Diaz et al., 2016). While high amounts of sedentary time constitute a risk factor for several health outcomes even after accounting for physical activity (e.g., for all-cause mortality, diabetes), links between sedentary behavior and cognitive health have received less attention and current evidence remains mixed (Dillon et al., 2022; Falck et al., 2017; Olanrewaju et al., 2020). Results are particularly heterogeneous when focusing on performance on cognitive tests. Some observational studies have shown that sedentary behaviors were associated with poorer cognitive performance among older adults (e.g., Hamer & Stamatakis, 2014; Kesse-Guyot et al., 2014), while others reported nonsignificant associations (e.g., Maasakkers et al., 2020). Greater total sedentary time was even shown to predict better performances on cognitive tests in some studies (Major et al., 2023; Wanders et al., 2021). These inconsistencies may partly result from the fact that previous studies have not examined the potential nonlinear nature of the relationships between sedentary time and cognition (Falck et al., 2023). Recently, a large study of older adults indeed found that sedentary time, as measured by accelerometry, was not significantly associated with dementia risk below approximately 10 hr of sitting per day (Raichlen et al., 2023). Above this threshold, higher sedentary time was linked with increased risk for all-cause dementia (Raichlen et al., 2023). Such findings call for a more comprehensive investigation of the potential nonlinear nature of relationships between sedentary time and cognition. Determining the threshold above which sedentary time shows the greatest risks for cognitive performance would help clarify contradictory results from prior work and help refine public health recommendations in terms of identifying what “too much sitting” means.

In the UK Biobank—the largest cohort of community-dwelling adults—we used accelerometer-based measures to test the hypothesis that sedentary time shows nonlinear relationships with cognitive functions (Hypothesis 1). Notably, we hypothesized that similar to results for incident dementia (Raichlen et al., 2023), higher sedentary time would be associated with poorer cognitive functions above an expected 10-hr per day threshold (Hypothesis 2).

Method

Data from the UK Biobank were used to determine the relationship between sedentary behavior and cognition. This large-scale study cohort was composed of community-dwelling adults aged 40–69 years, living in England, Scotland, or Wales, registered with the National Health Service, and living within 40 km of one of the 22 study assessment centers (Sudlow et al., 2015). A baseline visit was conducted between 2006 and 2010 with about 500,000 participants who provided information on a wide range of demographic, health, and lifestyle variables. Later, between 2013 and 2015, 103,684 adults agreed to wear a 3-axis logging accelerometer (AX3; Axivity, York, UK) for 24 hr per day for 7 days on their dominant wrist. Cognitive performance was assessed during an online follow-up (from 2014 to 2015), for which all participants were invited by email. We focused on data collected at the online follow-up, which provided the largest sample to date with both accelerometer and cognitive test data.

Participants were included in the present analyses if they had valid accelerometer data (i.e., at least 3 valid days [> 16 hr/day] of wear time), had completed at least one cognitive test on the online follow-up, were at least 60 years old at the time of wearing the accelerometer, were free of all-cause dementia prior to participating, and had complete covariate data (measured at baseline, between 2006 and 2010). The number of participants included in the analyses differs across cognitive tests as some participants did not complete some parts of the survey, and completed some cognitive tests, but not others.

Exposure

Methods to determine sedentary time followed Raichlen et al. (2023). Accelerometer-based sedentary time was measured in 2013 in a subsample of the UK Biobank. Participants wore an Axivity AX3 (Axivity) accelerometer on their dominant wrist for seven consecutive days (Doherty et al., 2017). The Axivity AX3 is a small (23 × 32.5 × 8.9 mm) and unobtrusive (11 g) triaxial monitor, with a sampling frequency set to 100 Hz. Sedentary time was estimated from raw accelerometer data using a validated machine learning algorithm, developed for use with the UK Biobank (Walmsley et al., 2022). The algorithm was implemented from a cohort of 152 adults (aged 18–91 years) who wore the AX3 accelerometer and a wearable camera and kept a time-use diary during daily life. The researchers annotated accelerometer data with activities from the Compendium of Physical Activities and trained machine-learning models to classify behaviors in 30-s time windows of accelerometer data. This machine-learning based estimation of waking sedentary time had acceptable precision (Walmsley et al., 2022) and was shown to predict cases of dementia in the UK Biobank among older adults (Raichlen et al., 2023).

Outcomes

Cognitive functions were tested in an online follow-up assessment from 2014 to 2015. Cognitive tests were presented without supervision on the participants’ personal computers. Five cognitive tests were administered at this timepoint: a fluid intelligence test, a short-term numeric memory test, a symbol substitution test, a visual-spatial memory test, and two trail making tests—an alphanumeric path and a numeric path. More details on these cognitive tests are provided in Supplementary Material Section 1, in previous literature (Fawns-Ritchie & Deary, 2020; Lyall et al., 2016), and are also described on the UK Biobank website: https://biobank.ndph.ox.ac.uk/showcase/label.cgi?id=116.

Covariates

Covariates included: age at the time of wearing the accelerometer (in years), sex (female vs. male), education (higher education vs. no higher education), townsend deprivation index (a higher score is associated with higher socio-economic deprivation), body mass index (BMI, scored from measured height and weight), chronic health conditions (received vs. not received a diagnosis of cardiovascular disease, diabetes, or cancer), depression (yes vs. no, self-report or doctor diagnosed), smoking status (never smoker vs. former smoker vs. or current smoker), alcohol use (none vs. moderate alcohol consumption vs. high alcohol consumption following, derived from Lourida et al., 2019), adherence to a healthy diet (yes vs. no, derived from Lourida et al., 2019), accelerometer-based physical activity level (measured in 2013, determined using the machine learning algorithm described above, and expressed in hours per day), and APOE status (presence vs. absence of the APOE ε4 allele, a genetic risk factor for dementia). In a first set of sensitivity analyses, we controlled for the time interval between the acquisition of cognitive performance and accelerometer-based measures. In a second set of sensitivity analyses, we additionally controlled for participants’ job status at baseline (i.e., in paid employment or self-employment, retired or other situations—Data-Field 6142).

Statistical Analyses

After computing descriptive statistics, all continuous variables were standardized and scaled (i.e., mean = 0, standard deviation = 1). The associations between sedentary time and scores on each cognitive test were first analyzed separately (i.e., one set of models for each cognitive test). We began by testing our first hypothesis of a nonlinear association between sedentary time and scores on the cognitive tests. The Kolmogorov test and the Cramér–von Mises test were computed to determine whether the null hypothesis H0 of a linear association could be rejected at p < .05, using the lintest function of the WRS package (Wilcox & Rousselet, 2023).

Provided that the null hypothesis of linear associations was rejected, we tested our second hypothesis using segmented regression analyses, using the segmented package (Muggeo, 2008). In short, these models provide a finer-grained analysis on the nonlinear nature of a relation between an exposure and an outcome by estimating different slopes across segments in the exposure. These segments were partitioned by breakpoints that are iteratively estimated along the exposure range, until significant changes in slope parameters were detected (Muggeo, 2008). We notably expected to detect breakpoints around a 10-hr threshold as found by Raichlen et al. (2023) regarding dementia cases, suggesting that the association between sedentary time and cognitive scores would shift around this threshold.

To further strengthen these findings, we computed ordinary least squares (OLS) models that estimated slopes of sedentary time on cognitive scores below and above the breakpoint identified in segmented regressions. The coefficients of these two slopes were compared below and above the threshold using the ols2ci function of the WRS package (Wilcox & Rousselet, 2023). In OLS models, multivariate outliers were excluded using the MAD-median rule (Wilcox & Rousselet, 2023).

Following previous work (Fawns-Ritchie & Deary, 2020; Lyall et al., 2016), we then used principal component analyses (PCA) to compute a composite cognitive score, from loadings of the different cognitive tests on the first unrotated dimension. We replicated previous analyses using this score as a global measure of cognition.

Results

Following inclusion criteria, the final analytic sample was composed of 32,875 older adults (age = 67.38 ± 4.14 years; 54% female), with a mean sedentary time of 9 hr and 23 min per day. Descriptive statistics are provided in Table 1, and bivariate correlations between sedentary time and cognitive scores are reported in Figure 1.

Table 1.

Sample Characteristics and Descriptive Statistics for Cognitive Scores

Characteristics Mean ± (SD) Median (Q1, Q3) Min; max %
Sedentary time (hours/day) 9.38 ± 1.75 9.35 (8.21; 10.52) 2.37; 19.47
MVPA (hours/day) 0.69 ± 0.59 0.55 (0.25, 0.96) 0; 7.11
Sex
 Women (%) 54.18
 Men (%) 45.82
Age (at the accelerometer-wearing period) 67.38 ± 4.14 67.28 (64.07, 70.49) 60.00; 79.17
BMI (at baseline) 26.72 ± 4.36 26.10 (23.73, 28.96) 14.08; 59.37
Education
 High education 43.57
 Low education 56.43
Depression (at baseline)
 Depression 36.18
 No depression 63.82
Townsend deprivation index (at baseline) −1.98 ± 2.67 −2.66 (−3.91, −0.63) −6.25; 9.41
Smoking status (at baseline)
 Smoker 40.40
 Not a smoker 54.12
Healthy diet (at baseline)
 Healthy diet 58.03
 Unhealthy diet 41.97
Alcohol (at baseline)
 Low alcohol consumption 27.05
 Moderate alcohol consumption 67.51
 High alcohol consumption 5.44
Health chronic conditions (at baseline)
 No chronic condition 37.56
 At least one chronic condition 62.44
APOE allele
 Carrier of APOE ɛ4 allele 25.67
 Not a carrier of APOE ɛ4 allele 74.33
Cognitive scores on the online follow-up
Visual-spatial memory test (number of incorrect answers; n = 31,269) 4.45 ± 3.21 4 (2.00, 6.00) 0; 43
Short-term numeric memory test (number of correct answers; n = 29,118) 6.85 ± 1.48 7.00 (6.00, 8.00) 2; 11
Symbol substitution test (number of correct answers; n = 31,259) 18.39 ± 4.87 19 (16.00, 22.00) 0; 72
Fluid intelligence (number of correct answers; n = 32,374) 5.38 ± 1.97 5.00 (4.00, 7.00) 0; 13
Trail making test—alphanumeric (duration in seconds; n = 27,071) 71.59 ± 26.44 65.82 (53.85, 82.42) 21.78; 477.73
Trail making test—numeric (duration in seconds; n = 27,072) 41.63 ± 16.09 37.81 (31.29, 47.54) 17.28; 733.97
Open in a new tab

Notes: BMI = body mass index; MVPA = accelerometer-based moderate-to-vigorous physical activity; Q1 = first quartile; Q3 = third quartile; SD = Standard deviation.

Figure 1.

Alt Text: Bivariate correlations between sedentary time and cognitive scores. Cells in blue indicate significant and positive correlations. The crossed cell indicates a non-significant correlation (p < .05).

Open in a new tab

Bivariate correlations between sedentary time and cognitive scores. Notes: All correlations are significant, with p < .05, with the exception of the crossed cell representing the bivariate correlation between sedentary time and scores in the visual-spatial memory test (r = 0.01). Incorrect answers in the visual memory test and duration to complete the trail making test (alphanumeric and numeric versions) were reversed so that higher values corresponded to better performances in these tests. Significance of correlations remains unchanged after further adjustment for age, sex, and education.

Is Sedentary Time Nonlinearly Associated With the Cognitive Scores (Hypothesis 1)?

For scores on the UK Biobank fluid intelligence test, the short-term numeric memory test, the symbol substitution test, and the trail making tests (alphanumeric and numeric paths), the Kolmogorov test and the Cramér–von Mises test indicated that the H0 hypothesis of a linear relation between sedentary time and cognitive scores could be rejected (p < .040). These results suggested a nonlinear modeling approach would be needed to examine the association between sedentary time and the scores in these five cognitive tests. Only for the UK Biobank visual-spatial memory test, the Kolmogorov test and the Cramér–von Mises test indicated that the H0 hypothesis of a linear relationship between sedentary time and this score could not be rejected at p < .05. As such, an OLS model was computed to examine the association between sedentary time and scores on the visual-spatial memory test.

Is Sedentary Time Associated With Poorer Cognitive Scores Above a 10-hr Threshold (Hypothesis 2)?

Segmented regression indicated that for scores in the fluid intelligence test, the breakpoint was 9.7 hr (standard error [SE]: 0.4 hr) of sedentary time (Figure 2A). Sedentary time was positively associated with scores in the fluid intelligence test below this threshold (b [95% confidence intervals, CI] = 0.09 [0.08, 0.10]). This association remained positive but was significantly weaker after this threshold (b [95% CI] = 0.05 [0.04, 0.06]). The difference in slope coefficients was confirmed in robust OLS models, with an estimated difference in slope coefficients of Δb [95% CI] = 0.04 [0.02; 0.06], p < .001.

Figure 2.

Alt Text: Plots from segmented regression models illustrating the non-linear associations of sedentary time with cognitive scores. Significant breakpoints are identifiable above a 10-12-hour threshold of sedentary time per day.

Open in a new tab

Segmented regression models with scores in the fluid intelligence test (A), the short-term numeric memory test (B), the symbol substitution test (C), the alphanumeric trail making test (D), the numeric trail making test (E), and the five-test composite cognitive score (F) as outcomes. Notes: A rug plot for observations of sedentary time is displayed on the x-axis of the graph. Dotted lines represent 95% confidence intervals around the estimated slopes for sedentary time on the cognitive scores. Models were adjusted for all abovementioned covariates.

For the short-term numeric memory test, the breakpoint was 10.9 hr (SE: 0.6 hr) of sedentary time (Figure 2B). Sedentary time was positively associated with scores in the short-term numeric memory test below this threshold (b [95% CI] = 0.05 [0.04; 0.06]). This association became nonsignificant after this threshold (b [95% CI] = 0.01 [−0.01; 0.04]. The difference in slope coefficients was confirmed in robust OLS models, with an estimated difference in slope coefficients of Δb = [95% CI] = 0.04 [0.01; 0.07], p = .007.

For the symbol substitution test, the breakpoint was 12.0 hr (SE: 0.3 hr) of sedentary time (Figure 2C). Sedentary time was positively associated with scores in the symbol substitution test below this threshold (b [95% CI] = 0.07 [0.06; 0.08]). This association became negative after this threshold (b [95% CI] = −0.05 [−0.10; −0.01]). The difference in slopes coefficients was confirmed in robust OLS models, with an estimated difference in slope coefficients of Δb [95% CI] = 0.13 [0.08; 0.18], p < .001.

For the alphanumeric trail making test, the breakpoint was 12.3 hr (SE: 0.3 hr) of sedentary time (Figure 2D). Sedentary time was negatively associated with reaction times in the alphanumeric trail making test, indicating better cognitive performances, below this threshold (b [95% CI] = −0.08 [−0.09; −0.07]). This association became nonsignificant after this threshold (b [95% CI] = 0.05 [−0.01; 0.10]). The difference in slope coefficients was confirmed in robust OLS models, with an estimated difference in slope coefficients of Δb [95% CI] = 0.14 [0.07; 0.20], p < 001.

For the numeric trail making test, the breakpoint was 12.3 hr (SE: 0.3 hr) of sedentary time (Figure 2E). Sedentary time was negatively associated with reaction times in the numeric trail making test, indicating better cognitive functions, below this threshold (b [95% CI] = −0.07 [−0.08; −0.06]). This association was nonsignificant after this threshold (b [95% CI] = 0.05 [−0.01; 0.11]). The difference in slope coefficients was confirmed in robust OLS models, with an estimated difference in slope coefficients of Δb [95% CI] = 0.12 [0.06; 0.19], p < 001.

Regarding the visual-spatial memory test, results from the OLS model showed that a higher sedentary time was linearly associated with a lower number of incorrect answers in this test, reflecting better cognitive functions (b [95% CI] = −0.05 [−0.10; −0.01], p < .001).

Association Between Sedentary Time and the Composite Global Cognitive Score

Results from PCA using the six cognitive tests suggested that a single component could fit the data, with an eigenvalue value = 2.47 for the first component and 41.3% of variance explained by the first component. However, the loading for scores of the visual-spatial memory test on this first component was low (loading = 0.23). Therefore, we performed another PCA, removing scores of the visual-spatial memory test. Results for the five cognitive tests showed that a single component could fit the data, with an eigenvalue value = 2.45 for the first component and 48.9% of variance explained by the first component. All loadings of the scores for the four cognitive tests were >0.48, suggesting that a composite cognitive score could be extracted from the loadings of these five cognitive tests on the first component. A higher composite cognitive score reflected better cognitive functions. Above mentioned analyses (i.e., test of the linearity of the association, segmented regression, and robust OLS models) were repeated using this composite cognitive score as the outcome. For completeness, we also computed these analyses with the composite cognitive score created from the six cognitive tests (i.e., including the visual-spatial memory test), and we report the results in Supplementary Material Section 2.

The Kolmogorov test and the Cramér–von Mises test indicated that the H0 hypothesis of a linear relation between sedentary time and the composite cognitive scores could be rejected (p < .001). These results suggested that a nonlinear modeling approach was appropriate to examine the association between sedentary time and the composite cognitive score.

Segmented regression indicated that the breakpoint was 11.0 hours (SE: 0.3 hours) of sedentary time (Figure 2F). The association of sedentary time with the compositive cognitive scores was positive below this threshold (b [95% CI] = 0.10 [0.09; 0.11]. This association became nonsignificant after this threshold (b [95% CI] = 0.03 [−0.00; 0.06]. The difference in slope coefficients was confirmed in robust OLS models, with an estimated difference in slope coefficients of Δb [95% CI] = 0.11 [0.05; 0.16], p < .001. Similar results were obtained using the six-test composite cognitive score (see Supplementary Material Section 2).

Sensitivity analyses

Results remained consistent when controlling for the time interval between the exposure and the cognitive outcomes (Supplementary Material Section 3). Results also remained unchanged when adjusting for participants’ employment status at baseline (Supplementary Material Section 4).

Discussion

Main Findings

Drawing on a large-scale sample of community-dwelling older adults, this study extends previous literature by providing evidence that, consistent with our first hypothesis, for most of the available cognitive tests, accelerometer-based sedentary time shows nonlinear associations with cognitive functions among community-dwelling older adults. Notably, we observed that below a 10–12-hr threshold, daily sedentary time showed positive associations with cognitive scores. However, above this 10–12-hr threshold of daily sedentary time, in partial support of our second hypothesis, the associations between sedentary time and cognitive scores either attenuated toward the null, were nonsignificant, or became negative.

Comparison With Previous Studies

The present work addresses an important limitation of previous literature that may explain the equivocal or contradictory associations between sedentary time and cognition. Previous work mostly relied on self-reported measures of sedentary time to examine its association with cognitive functions. Beyond the social desirability and recall biases that are inherent to self-reported measures of movement-based behaviors (Prince et al., 2008), television watching during leisure time was often used as the proxy for sedentary time (Falck et al., 2017). Television time not only provides a crude estimation of total sedentary time (Clark et al., 2011), but, more importantly, using this metric may have blurred the association between sedentary time and cognition. Sedentary television time is generally considered a mentally passive leisure-time activity and does not account for sedentary time in other cognitive contexts. Similar to associations with incident dementia (Raichlen, Klimentidis, et al., 2022), self-reported mentally passive sedentary time (e.g., watching TV), but not self-reported mentally active sedentary time (e.g., using a computer), was negatively associated with cognitive performance (Bakrania et al., 2018; Ringin et al., 2023). Relying on device-based methods was therefore warranted to better reflect the full spectrum of sedentary time and its association with cognition. As such, the present work represents the largest effort to date to estimate the associations between accelerometer-based sedentary time and cognitive performance in healthy older adults.

There are some key differences in the results presented here and findings from previous work that may help us better understand the links between sedentary time and cognition. Most notably, the observation that, until a certain threshold, accelerometer-based sedentary time was associated with better scores in all cognitive tests stands in contrast with studies that used television time as a proxy for sedentary time (e.g., Hamer & Stamatakis, 2014; Kesse-Guyot et al., 2014). However, our findings also align with previous work that found positive associations between total sitting time and cognitive performances (Major et al., 2023; Wanders et al., 2021), or another study that showed that the detrimental effect of sedentary time on cognitive performances was only observed in individuals with long accelerometer-based sedentary time (i.e., above a 10-hr threshold; Han et al., 2023). Although the effect sizes were small, the positive association between sedentary time and cognitive performance that was observed up to a 10–12-hr threshold in our study may be driven by several factors. Notably, participants with lower total accelerometer-based sedentary time (i.e., below < 11 hr) were also more physically active, had a healthier diet at baseline and reported fewer chronic conditions, compared to those with higher total accelerometer-based sedentary time (i.e., above 11 hr; see Supplementary Material Section 5 for stratified descriptive statistics). Another important characteristic that could explain the small positive effect of sedentary time in participants with lower sedentary time could be driven by the type of sedentary behavior. Mentally passive and mentally active sedentary activities appear to have opposite effects on cognitive health (Bakrania et al., 2018; Raichlen, Klimentidis, et al., 2022). As such, not only some participants could report a limited amount of time spent sitting/lying down (i.e., below the 10–12-hr threshold), but they could also engage in less mentally passive sedentary behaviors (e.g., watching television) and more mentally active sedentary behaviors (e.g., using a computer). This assumption may notably apply to participants who were still working at the time of the study and who may hence engage in a long time sitting over the day, but also potentially a higher engagement in mentally active behaviors. Combining in situ self-reported measures with device-based estimates of sedentary time is needed to better understand how sedentary time, the type of behaviors performed in a sitting/lying position, as well as the domain in which they occur (i.e., workplace vs. at home), interact to predict cognition-related outcomes.

Above a threshold that varied from 9.7 to 12.3 hr of sedentary time per day, the association between sedentary time and cognitive performance was attenuated toward the null or was negative. These results echo our previous study showing that accelerometer-based sedentary time was associated with higher risks for all-cause dementia only above a ~10-hour threshold (Raichlen et al., 2023). While underlying neurophysiological mechanisms remain to be identified (e.g., alteration of cerebral blood flow, e.g., Zlatar et al., 2019), these findings collectively demonstrate that the negative effects of sedentary time on cognition may only emerge at higher levels than previously thought. Moreover, we found a certain heterogeneity in the (nonlinear) nature of the association between sedentary time and scores in the six different cognitive tests. Future research is needed to identify whether, and why, the detrimental effects of sedentary time appear earlier on certain cognitive measures than on others. Conjointly, an avenue for future research could be to explore whether the threshold above which the detrimental cognitive effects of sedentary time emerges depends on types of sedentary behavior (e.g., watching television vs. using a computer).

Limitations and Strengths

Among the strengths of the current study are its large sample size, the reliance on multiple tests to assess cognitive functions, and the accelerometer-based estimation of sedentary time. This study also has several limitations. First, as mentioned above, the type of sedentary behaviors being performed while sitting/lying down (i.e., reading a book vs. watching television) remains unknown with accelerometer-based measures. Moreover, despite using a validated machine-learning algorithm to predict sedentary time, our results should be replicated by combining different devices, including those that can better dissociate different positions across the day (e.g., lying down vs. sitting) and/or polysomnography to better identify waking versus sleeping time. In addition, we observed very high (≈ 19 hr) and very low values of daily sedentary time (≈ 2 hr) in our sample. Although outliers were detected and removed from analyses in OLS models, our findings should thus be interpreted in light of the limitations of exposure measures and of its relative accuracy in classifying movement-based behaviors. Second, despite the adjustment for a wide range of covariates, we cannot rule out reversed causality. For example, higher cognitive functions may increase the odds of engaging in more sedentary behaviors, especially the more mentally active ones. Third, given the selection bias inherent to the UK Biobank procedure (i.e., healthy, racially, and ethnically homogeneous cohort), our findings may not be generalizable to other populations. Fourth, covariates were measured at baseline and, as some of them are likely to change across time (e.g., health conditions, self-reported depression), the timeframe between the measure of these variables and of the exposure may have resulted in an inadequate adjustment of our models. Fifth, although our robust smoothing approach aimed to minimize the risk of forcing models to detect breakpoints (Breit et al., 2023), further research is needed to consolidate our knowledge on the nonlinear nature of the association between sedentary time and cognition in order to better inform public health policies on the meaning of “too much sitting.”

Conclusion

Among older adults, accelerometer-based sedentary time is associated with better scores in cognitive tests, with a small effect size, until reaching a threshold of between 10 and 12 hr, depending on the cognitive task. Above this threshold, in four of the five cognitive tests available and on a global measure of cognitive functions, the association between sedentary time and cognitive performance was attenuated toward the null or became negative. Future research is needed to further clarify the nonlinear nature of the association between accelerometer-based sedentary time and cognition and should better characterize the degree of engagement in mentally active versus passive activity during sedentary behavior.

Supplementary Material

gbaf071_suppl_Supplementary_Materials
gbaf071_suppl_supplementary_materials.docx (692.7KB, docx)

Acknowledgments

This research was conducted using the UK Biobank Resource under Application Number 15678. We thank the participants and organizers of the UK Biobank.

Contributor Information

Silvio Maltagliati, Human and Evolutionary Biology Section, Department of Biological Sciences, University of Southern California, Los Angeles, California, USA.

Daniel H Aslan, Human and Evolutionary Biology Section, Department of Biological Sciences, University of Southern California, Los Angeles, California, USA.

M Katherine Sayre, Department of Anthropology, University of California Santa Barbara, Santa Barbara, California, USA.

Pradyumna K Bharadwaj, Department of Psychology, University of Arizona, Tucson, Arizona, USA.

Madeline Ally, Department of Psychology, University of Arizona, Tucson, Arizona, USA.

Mark H C Lai, Department of Psychology, University of Southern California, Los Angeles, California, USA.

Rand R Wilcox, Department of Psychology, University of Southern California, Los Angeles, California, USA.

Yann C Klimentidis, Department of Epidemiology and Biostatistics, Mel and Enid Zuckerman College of Public Health, University of Arizona, Tucson, Arizona, USA; BIO5 Institute, University of Arizona, Tucson, Arizona, USA.

Gene E Alexander, Department of Psychology, University of Arizona, Tucson, Arizona, USA; Evelyn F. McKnight Brain Institute, University of Arizona, Tucson, Arizona, USA.

David A Raichlen, Human and Evolutionary Biology Section, Department of Biological Sciences, University of Southern California, Los Angeles, California, USA; Department of Anthropology, University of Southern California, Los Angeles, California, USA.

Martina Luchetti, (Psychological Sciences Section).

Funding

This work was supported by the National Institute on Aging of the National Institutes of Health (P30AG072980, P30AG019610, R56AG067200, R01AG064587, R01AG072445), the state of Arizona and Arizona Department of Health Services, and the McKnight Brain Research Foundation.

Conflict of Interest

None.

Data Availability

Data are available through the UK Biobank (https://www.ukbiobank.ac.uk). Researchers must apply for access to the dataset through the UK Biobank. The study was not preregistered.

References

  1. Bakrania, K., Edwardson, C. L., Khunti, K., Bandelow, S., Davies, M. J., & Yates, T. (2018). Associations between sedentary behaviors and cognitive function: Cross-sectional and prospective findings From the UK Biobank. American Journal of Epidemiology, 187(3), 441–454. https://doi.org/ 10.1093/aje/kwx273 [DOI] [PubMed] [Google Scholar]
  2. Breit, M., Preuß, J., Scherrer, V., & Preckel, F. (2023). Why the use of segmented regression analysis to explore change in relations between variables is problematic: A simulation study. Psychological Methods. Advance online publication. https://doi.org/ 10.1037/met0000576 [DOI] [PubMed] [Google Scholar]
  3. Cheval, B., Csajbók, Z., Formánek, T., Sieber, S., Boisgontier, M. P., Cullati, S., & Cermakova, P. (2021). Association between physical-activity trajectories and cognitive decline in adults 50 years of age or older. Epidemiology and Psychiatric Sciences, 30, e79. https://doi.org/ 10.1017/S2045796021000688 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Cheval, B., Darrous, L., Choi, K. W., Klimentidis, Y. C., Raichlen, D. A., Alexander, G. E., Cullati, S., Kutalik, Z., & Boisgontier, M. P. (2023). Genetic insights into the causal relationship between physical activity and cognitive functioning. Scientific Reports, 13(1), 5310. https://doi.org/ 10.1038/s41598-023-32150-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Clark, B., Healy, G. N., Winkler, E. A. H., Gardiner, P. A., Sugiyama, T., Dunstan, D. W., Matthews, C. E., & Owen, N. (2011). Relationship of television time with accelerometer-derived sedentary time. Medicine & Science in Sports & Exercise, 43(5), 822–828. https://doi.org/ 10.1249/MSS.0b013e3182019510 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Diaz, K. M., Howard, V. J., Hutto, B., Colabianchi, N., Vena, J. E., Blair, S. N., & Hooker, S. P. (2016). Patterns of sedentary behavior in US middle-age and older adults: The REGARDS study. Medicine and Science in Sports and Exercise, 48(3), 430–438. https://doi.org/ 10.1249/MSS.0000000000000792 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Dillon, K., Morava, A., Prapavessis, H., Grigsby-Duffy, L., Novic, A., & Gardiner, P. A. (2022). Total sedentary time and cognitive function in middle-aged and older adults: A systematic review and meta-analysis. Sports Medicine - Open, 8(1), 127. https://doi.org/ 10.1186/s40798-022-00507-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Doherty, A., Jackson, D., Hammerla, N., Plötz, T., Olivier, P., Granat, M. H., White, T., van Hees, V. T., Trenell, M. I., Owen, C. G., Preece, S. J., Gillions, R., Sheard, S., Peakman, T., Brage, S., & Wareham, N. J. (2017). Large scale population assessment of physical activity using wrist worn accelerometers: The UK Biobank study. PLoS One, 12(2), e0169649. https://doi.org/ 10.1371/journal.pone.0169649 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Erickson, K. I., Voss, M. W., Prakash, R. S., Basak, C., Szabo, A., Chaddock, L., Kim, J. S., Heo, S., Alves, H., White, S. M., Wojcicki, T. R., Mailey, E., Vieira, V. J., Martin, S. A., Pence, B. D., Woods, J. A., McAuley, E., & Kramer, A. F. (2011). Exercise training increases size of hippocampus and improves memory. Proceedings of the National Academy of Sciences of the United States of America, 108(7), 3017–3022. https://doi.org/ 10.1073/pnas.1015950108 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Falck, R. S., Davis, J. C., Khan, K. M., Handy, T. C., & Liu-Ambrose, T. (2023). A Wrinkle in measuring time use for cognitive health: How should we measure physical activity, sedentary behaviour and sleep? American Journal of Lifestyle Medicine, 17(2), 258–275. https://doi.org/ 10.1177/15598276211031495 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Falck, R. S., Davis, J. C., & Liu-Ambrose, T. (2017). What is the association between sedentary behaviour and cognitive function? A systematic review. British Journal of Sports Medicine, 51(10), 800–811. https://doi.org/ 10.1136/bjsports-2015-095551 [DOI] [PubMed] [Google Scholar]
  12. Fawns-Ritchie, C., & Deary, I. J. (2020). Reliability and validity of the UK Biobank cognitive tests. PLoS One, 15(4), e0231627. https://doi.org/ 10.1371/journal.pone.0231627 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Hamer, M., Muniz Terrera, G., & Demakakos, P. (2018). Physical activity and trajectories in cognitive function: English longitudinal study of ageing. Journal of Epidemiology and Community Health, 72(6), 477–483. https://doi.org/ 10.1136/jech-2017-210228 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Hamer, M., & Stamatakis, E. (2014). Prospective study of sedentary behavior, risk of depression, and cognitive impairment. Medicine & Science in Sports & Exercise, 46(4), 718–723. https://doi.org/ 10.1249/MSS.0000000000000156 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Han, X., Song, L., Li, Y., Dong, Y., Liu, R., Han, Q., Wang, X., Mao, M., Cong, L., Tang, S., Hou, T., Zhang, Q., Liu, C., Han, X., Shi, L., Nyberg, L., Launer, L. J., Wang, Y., Du, Y., & Qiu, C. (2023). Accelerometer-measured sedentary behavior patterns, brain structure, and cognitive function in dementia-free older adults: A population-based study. Journal of Alzheimer's Disease: JAD, 96(2), 657–668. https://doi.org/ 10.3233/JAD-230575 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Iso-Markku, P., Kujala, U. M., Knittle, K., Polet, J., Vuoksimaa, E., & Waller, K. (2022). Physical activity as a protective factor for dementia and Alzheimer’s disease: Systematic review, meta-analysis and quality assessment of cohort and case–control studies. British Journal of Sports Medicine, 56(12), 701–709. https://doi.org/ 10.1136/bjsports-2021-104981 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Kesse-Guyot, E., Andreeva, V. A., Lassale, C., Hercberg, S., & Galan, P. (2014). Clustering of midlife lifestyle behaviors and subsequent cognitive function: A longitudinal study. American Journal of Public Health, 104(11), e170–e177. https://doi.org/ 10.2105/AJPH.2014.302121 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Livingston, G., Huntley, J., Sommerlad, A., Ames, D., Ballard, C., Banerjee, S., Brayne, C., Burns, A., Cohen-Mansfield, J., Cooper, C., Costafreda, S. G., Dias, A., Fox, N., Gitlin, L. N., Howard, R., Kales, H. C., Kivimäki, M., Larson, E. B., Ogunniyi, A., … Mukadam, N. (2020). Dementia prevention, intervention, and care: 2020 report of the lancet commission. Lancet (London, England), 396(10248), 413–446. https://doi.org/ 10.1016/S0140-6736(20)30367-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Lourida, I., Hannon, E., Littlejohns, T. J., Langa, K. M., Hyppönen, E., Kuzma, E., & Llewellyn, D. J. (2019). Association of lifestyle and genetic risk with incidence of dementia. JAMA, 322(5), 430–437. https://doi.org/ 10.1001/jama.2019.9879 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Lyall, D. M., Cullen, B., Allerhand, M., Smith, D. J., Mackay, D., Evans, J., Anderson, J., Fawns-Ritchie, C., McIntosh, A. M., Deary, I. J., & Pell, J. P. (2016). Cognitive test scores in UK Biobank: Data reduction in 480,416 participants and longitudinal stability in 20,346 participants. PLoS One, 11(4), e0154222. https://doi.org/ 10.1371/journal.pone.0154222 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Maasakkers, C. M., Claassen, J. A. H. R., Gardiner, P. A., Olde Rikkert, M. G. M., Lipnicki, D. M., Scarmeas, N., Dardiotis, E., Yannakoulia, M., Anstey, K. J., Cherbuin, N., Haan, M. N., Kumagai, S., Narazaki, K., Chen, T., Ng, T. P., Gao, Q., Nyunt, M. S. Z., Crawford, J. D., Kochan, N. A., … Melis, R. J. F.; COSMIC Collaborators. (2020). The association of sedentary behaviour and cognitive function in people without dementia: A coordinated analysis across five cohort studies from COSMIC. Sports Medicine (Auckland, N.Z.), 50(2), 403–413. https://doi.org/ 10.1007/s40279-019-01186-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Major, L., Simonsick, E. M., Napolitano, M. A., & DiPietro, L. (2023). Domains of sedentary behavior and cognitive function: The health, aging, and body composition study, 1999/2000 to 2006/2007. The Journals of Gerontology. Series A, Biological Sciences and Medical Sciences, 78(11), 2035–2041. https://doi.org/ 10.1093/gerona/glad020 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Muggeo, V. M. R. (2008). Segmented: an R package to fit regression models with broken-line relationships. RNews, 8(1), 20–25. https://journal.r-project.org/articles/RN-2008-004/RN-2008-004.pdf [Google Scholar]
  24. Olanrewaju, O., Stockwell, S., Stubbs, B., & Smith, L. (2020). Sedentary behaviours, cognitive function, and possible mechanisms in older adults: A systematic review. Aging Clinical and Experimental Research, 32(6), 969–984. https://doi.org/ 10.1007/s40520-019-01457-3 [DOI] [PubMed] [Google Scholar]
  25. Prince, S. A., Adamo, K. B., Hamel, M., Hardt, J., Connor Gorber, S., & Tremblay, M. (2008). A comparison of direct versus self-report measures for assessing physical activity in adults: A systematic review. International Journal of Behavioral Nutrition and Physical Activity, 5(1), 56. https://doi.org/ 10.1186/1479-5868-5-56 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Raichlen, D. A., Aslan, D. H., Sayre, M. K., Bharadwaj, P. K., Ally, M., Maltagliati, S., Lai, M. H. C., Wilcox, R. R., Klimentidis, Y. C., & Alexander, G. E. (2023). Sedentary behavior and incident dementia among older adults. JAMA, 330(10), 934–940. https://doi.org/ 10.1001/jama.2023.15231 [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Raichlen, D. A., Furlong, M., Klimentidis, Y. C., Sayre, M. K., Parra, K. L., Bharadwaj, P. K., Wilcox, R. R., & Alexander, G. E. (2022). Association of physical activity with incidence of dementia is attenuated by air pollution. Medicine & Science in Sports & Exercise, 54(7), 1131–1138. https://doi.org/ 10.1249/MSS.0000000000002888 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Raichlen, D. A., Klimentidis, Y. C., Sayre, M. K., Bharadwaj, P. K., Lai, M. H. C., Wilcox, R. R., & Alexander, G. E. (2022). Leisure-time sedentary behaviors are differentially associated with all-cause dementia regardless of engagement in physical activity. Proceedings of the National Academy of Sciences, 119(35), e2206931119. https://doi.org/ 10.1073/pnas.2206931119 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Rhee, J. H., Schermer, B., Han, G., Park, S. Y., & Lee, K. H. (2023). Effects of nature on restorative and cognitive benefits in indoor environment. Scientific Reports, 13(1), 13199. https://doi.org/ 10.1038/s41598-023-40408-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Ringin, E., Dunstan, D. W., McIntyre, R. S., Owen, N., Berk, M., Rossell, S. L., Hallgren, M., & Van Rheenen, T. E. (2023). Differential associations of mentally-active and passive sedentary behaviours and physical activity with putative cognitive decline in healthy individuals and those with bipolar disorder: Findings from the UK Biobank Cohort. Mental Health and Physical Activity, 24, 100514. https://doi.org/ 10.1016/j.mhpa.2023.100514 [DOI] [Google Scholar]
  31. Sudlow, C., Gallacher, J., Allen, N., Beral, V., Burton, P., Danesh, J., Downey, P., Elliott, P., Green, J., Landray, M., Liu, B., Matthews, P., Ong, G., Pell, J., Silman, A., Young, A., Sprosen, T., Peakman, T., & Collins, R. (2015). UK Biobank: An Open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Medicine, 12(3), e1001779. https://doi.org/ 10.1371/journal.pmed.1001779 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Tremblay, M. S., Aubert, S., Barnes, J. D., Saunders, T. J., Carson, V., Latimer-Cheung, A. E., Chastin, S. F. M., Altenburg, T. M., & Chinapaw, M. J. M. (2017). Sedentary Behavior Research Network (SBRN)–Terminology consensus project process and outcome. International Journal of Behavioral Nutrition and Physical Activity, 14(1), 75. https://doi.org/ 10.1186/s12966-017-0525-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Walmsley, R., Chan, S., Smith-Byrne, K., Ramakrishnan, R., Woodward, M., Rahimi, K., Dwyer, T., Bennett, D., & Doherty, A. (2022). Reallocation of time between device-measured movement behaviours and risk of incident cardiovascular disease. British Journal of Sports Medicine, 56(18), 1008–1017. https://doi.org/ 10.1136/bjsports-2021-104050 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Wanders, L., Bakker, E. A., van Hout, H. P. J., Eijsvogels, T. M. H., Hopman, M. T. E., Visser, L. N. C., Wouters, H., & Thijssen, D. H. J. (2021). Association between sedentary time and cognitive function: A focus on different domains of sedentary behavior. Preventive Medicine, 153, 106731. https://doi.org/ 10.1016/j.ypmed.2021.106731 [DOI] [PubMed] [Google Scholar]
  35. Wilcox, R. R., & Rousselet, G. A. (2023). An updated guide to robust statistical methods in neuroscience. Current Protocols, 3(3), e719. https://doi.org/ 10.1002/cpz1.719 [DOI] [PubMed] [Google Scholar]
  36. Zlatar, Z. Z., Hays, C. C., Mestre, Z., Campbell, L. M., Meloy, M. J., Bangen, K. J., Liu, T. T., Kerr, J., & Wierenga, C. E. (2019). Dose-dependent association of accelerometer-measured physical activity and sedentary time with brain perfusion in aging. Experimental Gerontology, 125, 110679. https://doi.org/ 10.1016/j.exger.2019.110679 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

gbaf071_suppl_Supplementary_Materials
gbaf071_suppl_supplementary_materials.docx (692.7KB, docx)

Data Availability Statement

Data are available through the UK Biobank (https://www.ukbiobank.ac.uk). Researchers must apply for access to the dataset through the UK Biobank. The study was not preregistered.

Articles from The Journals of Gerontology Series B: Psychological Sciences and Social Sciences are provided here courtesy of Oxford University Press

RESOURCES