Sedentary behaviours and cognitive function Associations between sedentary behaviours and cognitive function: crosssectional and prospective findings from the UK Biobank IP T Authors: Kishan Bakrania, Charlotte L. Edwardson, Kamlesh Khunti, Stephan U SC R Bandelow, Melanie J. Davies, and Thomas Yates AN Correspondence address: Mr. Kishan Bakrania, Diabetes Research Centre, University of Leicester, Leicester General Hospital, Gwendolen Road, Leicester, M Leicestershire, LE5 4PW, United Kingdom. Phone: +44(0)116 258 4874 (e-mail: O RI G IN AL U N ED IT ED kb318@le.ac.uk). © The Author(s) 2017. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. 1 Sedentary behaviours and cognitive function Author affiliations: Department of Health Sciences, University of Leicester, Leicester General Hospital, Leicester, Leicestershire, LE5 4PW, United Kingdom (Kishan Bakrania); Diabetes Research Centre, University of Leicester, Leicester IP T General Hospital, Leicester, Leicestershire, LE5 4PW, United Kingdom (Kishan Bakrania, Charlotte L. Edwardson, Kamlesh Khunti, Melanie J. Davies, and Thomas SC R Yates); Leicester Diabetes Centre, University Hospitals of Leicester, Leicester General Hospital, Leicester, Leicestershire, LE5 4PW, United Kingdom (Kishan U Bakrania, Charlotte L. Edwardson, Kamlesh Khunti, Melanie J. Davies, and Thomas AN Yates); National Institute for Health Research (NIHR) Leicester Biomedical Research M Centre, Leicester General Hospital, Leicester, Leicestershire, LE5 4PW, United ED Kingdom (Kishan Bakrania, Charlotte L. Edwardson, Melanie J. Davies, and Thomas Yates); National Institute for Health Research (NIHR) Collaboration for Leadership in U N ED IT Applied Health Research and Care – East Midlands (CLAHRC – EM), Diabetes Research Centre, Leicester General Hospital, Leicester, Leicestershire, LE5 4PW, United Kingdom (Kishan Bakrania, and Kamlesh Khunti); and School of Sport, Exercise and Health Sciences, Loughborough University, Loughborough, IN AL Leicestershire, LE11 3TU, United Kingdom (Stephan Bandelow). O RI G Funding: The authors declare that there was no funding. Conflict of interest: The authors declare that they have no conflict of interest. Running head: Sedentary behaviours and cognitive function. 2 Sedentary behaviours and cognitive function IP T Abbreviations: Television (TV); Confidence Interval (CI). SC R Text Word Count: 3,418 Number of Tables: 4 U N ED IT ABSTRACT ED M AN U Number of Web Material Files: 1 We investigate the cross-sectional and prospective associations between different sedentary behaviours and cognitive function in a large sample of UK Biobank adults. Baseline data were available on 502,643 participants (years 2006-2010). Cognitive AL tests included prospective memory [n=171,585 (baseline only)], visual-spatial IN memory [round 1 (n=483,832); round 2 (n=482,762)], fluid intelligence [n=165,492], G and short-term numeric memory [n=50,370]. After a mean period of 5.3-years, RI between 12,091 and 114,373 participants also provided follow-up cognitive data. O Sedentary behaviours [Television (TV) viewing, driving, and non-occupational computer use time] were measured at baseline. At baseline, both TV viewing and driving time were inversely associated with cognitive function across all outcomes [e.g. for each additional hour spent watching TV, the total number of correct answers in the fluid intelligence test was 0.15 (99% confidence interval: 0.14, 0.16) lower]. 3 Sedentary behaviours and cognitive function Computer use time was positively associated with cognitive function across all outcomes. Both TV viewing and driving time at baseline were positively associated with the odds of having cognitive decline at follow-up across most outcomes. IP T Conversely, computer use time at baseline was inversely associated with the odds of having cognitive decline at follow-up across most outcomes. This study supports U SC R health policies designed to reduce TV viewing and driving in adults. AN Keywords: Cognitive decline; Cognitive function; Computer use; Driving; M Epidemiology; Sedentary behaviour; Television (TV) viewing; United Kingdom U N ED IT ED Biobank Currently, there are no effective long-term pharmacological therapies for the treatment or prevention of dementia. Therefore, identifying potentially modifiable risk factors of cognitive decline, a major characteristic of dementia, is a key priority. Engaging in healthy lifestyle practices, including physical activity, has been AL associated with a reduced risk of dementia and its symptoms, such as cognitive IN impairment (1, 2); suggesting a potential role for lifestyle therapies. Indeed, physical G activity intervention studies have shown changes to the structure and function of the O RI brain (3-7), supporting the observational associations. Along with physical activity, engaging in sedentary behaviour, defined as sitting or reclining with low energy expenditure (8), could also be an important determinant of poor cognitive function. There is cumulative evidence indicating that sedentary time is associated with poor cardiometabolic health, chronic disease, and mortality (9-12). 4 Sedentary behaviours and cognitive function A recent systematic review also suggested that sedentary behaviour is negatively associated with cognitive function; although the relationship between the two is complex, and recommend that future studies should focus on determining how IP T different sedentary behaviours are associated with cognitive function (13). Limited observational research has indicated that television (TV) viewing is inversely SC R associated with cognition (14-17). However, different sedentary behaviours may have different associations, with some evidence of computer/internet use linked to U cognitive improvement (15-18). Furthermore, most of the existing data have AN emerged from relatively limited cross-sectional findings (16-18). Therefore, this M warrants investigation in large-scale studies with prospective data. ED The aim of this paper was to use the nationally representative UK Biobank cohort to U N ED IT examine the cross-sectional and prospective associations between domains of sedentary behaviour (TV viewing, driving, and computer use) and cognitive function (prospective memory, visual-spatial memory, fluid intelligence and short-term AL numeric memory). IN METHODS G Design and population O RI The UK Biobank is a large prospective study of the middle-aged population (19-21). Approximately 500,000 adults (aged 37-73 years) were recruited between 20062010 via mailing out invitations to those registered with the National Health Service (NHS) and living within 25 miles of one of the 22 study assessment centres. Participants provided comprehensive baseline data on a broad range of biological, 5 Sedentary behaviours and cognitive function cognition, demographic, health, lifestyle, mental, social, and well-being outcomes. Approximately 300,000 participants also provided an email address to allow for the remote follow-up of cognitive function in the future. From 2014 to 2015, around IP T 125,000 participants provided some online follow-up cognitive function data. For the present study, baseline data were available on 502,643 individuals. Of these, SC R depending on the cognitive test, between 50,370 and 483,832 participants provided baseline cognitive function data (see Web Figure 1). Of these, after a mean period of U 5.3 years and depending on the cognitive test, between 12,091 and 114,373 AN participants also provided online follow-up cognitive function data (see Web Figure M 2). All participants provided written informed consent and the study was approved by U N ED IT available elsewhere (19-21). ED the NHS National Research Ethics Service (Ref: 11/NW/0382). Further details are Cognitive function tests Questionnaires administered through a computerised touchscreen interface AL assessed cognitive function at baseline. Using the same methodology minus the IN touchscreen ability, follow-up measurements were obtained via online questionnaires that were completed remotely. To ensure effortless application on a large scale and RI G wide response distributions, the cognitive function tests, which were refined over O piloting, were designed comprehensively and specifically for UK Biobank. Prospective memory (available at baseline only), visual-spatial memory, fluid intelligence, and short-term numeric memory tests were included in this analysis. At baseline, there were variations between the numbers of individuals who completed each cognitive assessment due to tests being: abandoned or skipped by 6 Sedentary behaviours and cognitive function participants, incorporated towards the end of recruitment (e.g. fluid intelligence), and/or phased out during the early stages of recruitment (e.g. short-term numeric IP T memory). For more details on the cognitive function tests, see Web Appendix 1. SC R Sedentary behaviours Data on sedentary behaviours were self-reported and collected at baseline using a U computerised questionnaire. Domains of sedentary behaviour included: TV viewing AN time (<1, 1, 2, 3, ≥4 hours/day), driving time (<1, 1, 2, ≥3 hours/day), and non- M occupational computer use time (<1, 1, 2, ≥3 hours/day). For more details, see Web Covariate data U N ED IT ED Appendix 2. Covariate data included: anthropometric (body mass index), demographic (age, sex, ethnicity, social deprivation index, employment status, education level), health AL (number of cancers, number of non-cancer illnesses, number of IN medications/treatments), and lifestyle (smoking status, alcohol drinking status, sleep duration, fruit and vegetable consumption, physical activity) variables. For more O RI G details, see Web Appendix 3. Statistical analysis Statistical analyses were executed using Stata/MP V14.0 (Stata Corporation, College Station, Texas, USA). Data were analysed in February 2017. With the 7 Sedentary behaviours and cognitive function intention of maximising the use of the data, pairwise deletion was used to handle missing data (see Web Figure 1 and Web Figure 2). Participant characteristics were tabulated. Categorical variables were presented as numbers and proportions, IP T whereas continuous variables were summarised as means and standard deviations SC R (SD); and presented with their minimum and maximum values. AN U Cross-sectional analysis Regression analysis was used to examine the cross-sectional associations between M the three domains of sedentary behaviour and cognitive function at baseline. Multiple ED logistic regression models were fitted for each binary cognitive outcome variable (prospective memory, visual-spatial memory (round 1), and visual-spatial memory U N ED IT (round 2)). Multiple linear regression models were fitted for each continuous cognitive outcome variable (fluid intelligence and short-term numeric memory). For more details on the nature of the cognitive outcome variables used in the crosssectional analysis, see Web Appendix 1. Model 1 was mutually adjusted for the other AL sedentary behaviours and for age and sex. Model 2 was further adjusted for body IN mass index, ethnicity, social deprivation index, employment status, education level, smoking status, alcohol drinking status, fruit and vegetable consumption, sleep G duration, physical activity (frequency of ≥10 minutes of walking (days/week), O RI frequency of ≥10 minutes of moderate physical activity (days/week), frequency of ≥10 minutes of vigorous physical activity (days/week)), number of cancers, number of non-cancer illnesses, and number of medications/treatments. For each sedentary behaviour, the ‘<1 hour/day’ category was selected as the reference group. Linear trends (linear terms) across the categories of each sedentary behaviour were 8 Sedentary behaviours and cognitive function reported. Interaction terms were separately added to the fully adjusted model (Model 2) to observe whether the associations between the sedentary behaviours and cognitive function were modified by age or sex. Significant results for age were SC R IP T stratified at 60 years. Prospective analysis AN U Multiple logistic regression models investigated the prospective associations between the three domains of sedentary behaviour at baseline and cognitive function M at follow-up. These models estimated the odds of having cognitive decline (i.e. a ED poor outcome) at follow-up. Cognitive outcomes included: visual-spatial memory (round 1), visual-spatial memory (round 2), fluid intelligence, and short-term numeric U N ED IT memory. For full details on the definitions and nature of the cognitive outcome variables used in the prospective analysis, see Web Appendix 1. As well as controlling for the baseline result/score of the cognitive test under consideration, models were adjusted for all the covariates mentioned previously (see list of AL confounders in Models 1 and 2 of the cross-sectional analyses). Linear trends across IN the categories of each sedentary behaviour were reported. Interactions by age and O RI G sex were also investigated. Sensitivity analysis To assess the generalizability of our findings, the cross-sectional and prospective analyses investigating the associations between sedentary behaviours and cognitive function (Model 1 and Model 2) were repeated across the sample of participants 9 Sedentary behaviours and cognitive function without a medical history of cancer, cardiovascular disease, and/or cognitive/psychiatric illnesses as sensitivity analyses. For more details on the IP T specific diseases/illnesses, see Web Appendix 4. SC R Statistical reporting For each variable of interest (sedentary behaviours), the beta coefficient (linear AN U regression) or odds ratio (logistic regression) with 99% confidence intervals (99% CIs) and p-values are reported. All analyses employed robust standard errors and all M reported p-values are two-sided. To account for multiple comparisons, p<0.01 was ED considered to be statistically significant for the main analyses. For the interaction RESULTS U N ED IT analyses, p<0.05 was considered to be statistically significant. Cross-sectional findings AL Table 1 presents the characteristics of the 502,643 participants with baseline data. IN The mean (SD) age of these individuals was 56.5 (8.1) years and 273,467 (54.4%) G were female. O RI Table 2 presents the associations between the sedentary behaviours and cognitive function. In the fully adjusted models (Model 2), the cross-sectional data showed that TV viewing time was inversely associated with cognitive function across all outcomes apart from visual-spatial memory (round 2). For example, for each additional hour spent watching TV up to ≥4 hours/day, the fluid intelligence and short-term numeric 10 Sedentary behaviours and cognitive function memory scores were 0.15 (99% CI: 0.14, 0.16) and 0.09 (0.07, 0.10) units lower, respectively. Correspondingly, the odds of a poor result in the prospective memory and visual-spatial memory (round 1) tests were 2% (0%, 3%) and 3% (2%, 4%) IP T higher, respectively. Driving time was inversely associated with cognitive function across all outcomes. In contrast, computer use time was positively associated with SC R cognitive function across all outcomes. U Interaction analyses showed that most findings were modified by age and sex AN (p<0.05). Stratification indicated that the associations were generally stronger in older adults (≥60 years) and in males (see Web Figure 3 (age) and Web Figure 4 U N ED IT Prospective findings ED M (sex)). Table 3 presents the cognitive function data of the participants with cognitive data at both baseline and follow-up. Cognitive decline over time was apparent since participants performed better in each cognitive test at baseline than at follow-up. For AL example, the mean (SD) fluid intelligence score (n=46,704) at baseline and follow-up IN was 6.7 (2.1) and 5.5 (2.0), respectively; with 15,384 (32.9%) individuals reporting a G good outcome at follow-up (baseline fluid intelligence score ≤ follow-up fluid RI intelligence score) and 31,320 (67.1) individuals reporting a poor outcome at follow- O up (baseline fluid intelligence score > follow-up fluid intelligence score). The other tests followed a similar pattern. 11 Sedentary behaviours and cognitive function Those with follow-up data had similar characteristics to the full UK Biobank cohort, although they were better educated and more likely to be employed (see Web Table 1). IP T Table 4 presents the associations between the sedentary behaviours at baseline and cognitive function at follow-up. In the fully adjusted models (Model 2), both TV SC R viewing and driving time at baseline were positively associated with the odds of U having cognitive decline at follow-up across most outcomes. For example, for each AN additional hour spent watching TV up to ≥4 hours/day at baseline, the odds of a lower fluid intelligence score at follow-up were 9% (6%, 11%) higher. Similarly, for M each additional hour spent driving up to ≥3 hours/day at baseline, the odds of a ED lower fluid intelligence score at follow-up were 11% (7%, 15%) higher. In contrast, U N ED IT computer use time at baseline was inversely associated with the odds of having cognitive decline at follow-up across most outcomes. Interaction analyses showed that only the associations between TV viewing time and visual-spatial memory (round 2) were modified by age (p<0.05) (see Web Figure 5). Findings were not IN AL modified by sex. G Sensitivity analyses RI The cross-sectional and prospective findings were generalizable across the sample O of participants without cancer, cardiovascular disease, and/or cognitive/psychiatric illnesses (see Web Figure 6 (cross-sectional associations) and Web Figure 7 (prospective associations)). 12 Sedentary behaviours and cognitive function DISCUSSION Key findings This is the first study to quantify the cross-sectional and prospective associations IP T between domains of sedentary behaviour and cognitive function in a large cohort of SC R UK adults. At baseline, both TV viewing and driving time were inversely associated with cognitive function. In contrast, computer use time was positively associated with U cognitive function. Most findings were modified by age and sex, with stronger AN relationships generally observed in older adults and in males. These novel results M suggest that the influence of sedentary behaviour on cognition is enhanced in older age and in men. Both TV viewing and driving time at baseline were positively ED associated with the odds of having cognitive decline at follow-up across most U N ED IT outcomes. In contrast, computer use time at baseline was inversely associated with the odds of having cognitive decline at follow-up across most outcomes. The crosssectional and prospective findings were robust and generalizable across the sample of participants without cancer, cardiovascular disease, and/or cognitive/psychiatric IN AL illnesses. G Interpretations O RI To our knowledge, only a few number of studies have attempted to examine the prospective associations between the different types of sedentary behaviours and cognitive function (14-17, 22-26). However, these studies have all been limited by a small sample size (N ranging between 469 and 8,462), populations that only involved children or older adults, analyses that only considered one domain or test of 13 Sedentary behaviours and cognitive function cognitive function, and/or cognitive data that were only collected at a single time point. Therefore, this novel study in a large sample of middle-aged adults representing the general population provides the most comprehensive observational IP T analysis to date. Our findings are consistent with the existing data in this research area. Observational SC R studies have previously demonstrated an inverse association between TV viewing U and cognition (14-17), and a positive association between computer/internet use and AN cognition (15-18). However, until this study, the interactions with age or the deleterious influence of driving on cognitive health were less clear. The inverse M associations of TV viewing and driving time with cognitive function could be due to ED several factors. Cognition has previously been linked to cardiometabolic health (27, U N ED IT 28), and numerous studies have demonstrated inverse associations of TV viewing and driving time with cardiometabolic health (9-12, 29-31). Therefore, it is possible that the observed associations act via pathways linked to the risk of vascular dysfunction and chronic diseases. As vascular dysfunction and chronic diseases are linked to aging, this mechanism would also help explain the observed interactions AL with age. Other mediating factors could also explain the results for driving; it is IN known that driving is related to stress and fatigue (32), and with several studies G previously showing the links between these factors and cognitive decline (33-35), it RI is plausible that the observed relationships are enhanced via this pathway. O Furthermore, some types of sedentary behaviours, such as TV viewing and driving, could possibly segregate individuals from social networks and restrict external collaborations, factors which are known to affect cognition (36-38); this again could be particularly important in older adults. In contrast, the positive relationship shared between computer use and cognitive function coincides with previous work where 14 Sedentary behaviours and cognitive function improved cognition or a lower risk of dementia were reported in those engaging in cognitively vitalising sedentary behaviours or leisure activities (15-18). Therefore, as computer use is likely to involve some level of cognitive challenge, stimulate social IP T interactions and reduce solitariness, it may compensate for the associated sedentary behaviour in relation to cognitive health. Some of the mechanisms mentioned above SC R are also linked to and vary by gender (39, 40); and therefore, they could help explain U the observed interactions with sex. AN The differences observed in cognitive function across the categories of sedentary behaviour in our analyses are likely to be clinically important beyond the risk of M cognitive decline. For example, higher fluid intelligence scores have previously been ED shown to be strongly associated with a lower risk of all-cause mortality (41, 42). In a U N ED IT sample of 5,572 middle-aged British adults, Sabia and colleagues observed that a higher fluid intelligence score by 1 SD was associated with a 14% lower risk of allcause mortality (41). Similarly, in a sample of 896 older Australian adults, Batterham and colleagues observed that a higher fluid intelligence score by 1 SD was associated with a 24% lower risk of all-cause mortality (42). In our analysis at AL baseline (Model 2), the SD of fluid intelligence score was 2.1. Regression analyses IN investigating the associations of sedentary behaviours with fluid intelligence G demonstrated that TV viewing and driving time were linearly associated with lower RI fluid intelligence scores of 0.15 and 0.24 units, respectively. In contrast, computer O use time was linearly associated with a higher fluid intelligence score of 0.12 units. Hence, using the data above, it can be estimated that lower fluid intelligence scores by 0.15 and 0.24 units would approximately equate to a 1.1%-3.2% higher risk of allcause mortality. In contrast, a higher fluid intelligence score by 0.12 units would 15 Sedentary behaviours and cognitive function approximately equate to a 0.9%-1.6% lower risk of all-cause mortality. For more Strengths and limitations SC R This study has several strengths and some limitations. Strengths include: IP T details on these calculations, see Web Appendix 5. exploitation of a large sample of adults representing the national population, follow- AN U up cognitive function data allowing for prospective associations to be investigated, evaluation of dose-response and linear relationships between mutually adjusted and M time quantified sedentary behaviours with a wide range of comprehensive cognitive ED outcomes, detailed covariate data enabling several important and relevant factors to be controlled for, interactions by age and sex, and robust sensitivity analyses U N ED IT investigating the associations in the healthy population. Although the UK Biobank is representative of the general population with respect to age, sex, ethnicity, and deprivation within the age range recruited, it may not be representative in other regards (43). While this limits the ability to generalize prevalence rates, estimates of AL the magnitude of associations in our study are unlikely to have been substantially IN affected by this due to the large and multifaceted base population (43, 44). Furthermore, the cognitive data from the UK Biobank cohort has recently been RI G shown to be an important and valid resource for investigating predictors and O modifiers of cognitive abilities and associated health outcomes in the general population (45). The sedentary behaviour data used in this study have both strengths and limitations. Only three sedentary domains included; thus, the findings are restricted and cannot be generalized to other types of sedentary behaviour. Self-reported assessments of 16 Sedentary behaviours and cognitive function sedentary behaviour are subjective and are influenced by recall and response issues (46, 47); hence, they tend to have low validity and increase the risk of regression dilution. However, although data that are more robust can be obtained using IP T objective measurement tools (e.g. accelerometers) (46, 47), they would not provide information on the specific type of sedentary behaviour performed. Furthermore, SC R since the reasons for using the computer outside work were unknown (e.g. utilised for activities such as: reading, watching videos, internet browsing, playing games, U etc.), it is not possible to accurately classify or infer the type of computer use AN undertaken, and it may have involved crossover into cognitively inert tasks. M Additionally, only those who provided an email address at baseline (~300,000) were ED contacted to participate in the online follow-up of cognitive function. Therefore, these participants all had computer access and presumably, some computer use U N ED IT experience. This may also have resulted in the small differences in characteristics (including level of education and employment status) in the follow-up sample (see Web Table 1). Consequently, the prospective analysis may be biased and lack generalizability. Moreover, at baseline, the cognitive function tests were implemented AL using questionnaires that were administered via a touchscreen interface. At followup, the measurements were obtained remotely via online questionnaires that were IN administered on a computer via a mouse interface. Therefore, this difference in the G mode of administration could possibly account for some of the variability in cognitive O RI performance and change over time. Nevertheless, the prospective analysis broadly supports and is consistent with the cross-sectional associations reported for the full cohort at baseline. Although we adjusted for a wide range of covariates, some unmeasured factors (e.g. type of employment/occupation) may have further confounded the reported associations. Our results may be subject to residual 17 Sedentary behaviours and cognitive function confounding or reverse causality. For example, it is possible that the positive association observed between computer use and cognitive function was simply reflecting greater familiarity for interacting with a computer rather than better IP T cognitive function as such. Correspondingly, individuals with better cognitive function are more likely to engage in healthy behaviours and abstain from unhealthy ones, a SC R concept known as neuroselection (48, 49). Whilst we investigated interactions by age and sex in our study, it must be highlighted that similar differences observed in U cognitive function across different groups (i.e. in younger adults vs. older adults, and AN females vs. males) may have different clinical meanings and should be interpreted M with caution. For example, a unit difference in a cognitive function test score in a ED younger adult may not have the same result or significance on cognitive health as a unit difference in an older adult. Lastly, due to large variations between the numbers U N ED IT of individuals who completed each cognitive assessment at both baseline and followup, analyses were based on different sample sizes. AL Conclusions IN Our analysis, conducted in a large national sample of adults, demonstrates that some sedentary domains, but not all, are associated with poor cognition. Watching RI G TV and driving are inversely associated with cognitive function, whereas computer O use is positively associated with cognitive function. Of note, the associations were consistently stronger in older adults. Intervention studies are required to confirm these findings. Nevertheless, these results provide robust observational data supporting public health policies aimed at reducing TV viewing and driving time in adults. 18 Sedentary behaviours and cognitive function ACKNOWLEDGEMENTS This research has been conducted using the United Kingdom Biobank resource under application number 10813. The United Kingdom Biobank was established IP T by the Wellcome Trust medical charity, Medical Research Council, Department of Health, Scottish Government and the Northwest Regional Development Agency. It SC R has also had funding from the Welsh Assembly Government and the British Heart U Foundation, and is supported by the National Health Service (NHS). This research AN was supported by the National Institute for Health Research (NIHR) Leicester Biomedical Research Centre, the NIHR Collaboration for Leadership in Applied M Health Research and Care – East Midlands (CLAHRC – EM) and the Leicester ED Clinical Trials Unit. The views expressed are those of the authors and not U N ED IT necessarily those of the NHS, the NIHR or the Department of Health. AUTHOR CONTRIBUTIONS Thomas Yates and Kishan Bakrania had the original idea for the analysis, which was AL further developed and refined by Charlotte L. Edwardson, Kamlesh Khunti, Stephan IN Bandelow and Melanie J. Davies. Kishan Bakrania had full access to the data in the G study and takes responsibility for the integrity of the data and the accuracy of the RI analysis. Kishan Bakrania wrote the first draft of the manuscript. Kishan Bakrania, O Charlotte L. Edwardson, Kamlesh Khunti, Stephan Bandelow and Melanie J. Davies and Thomas Yates edited/reviewed the manuscript and approved the final version of the manuscript. 19 Sedentary behaviours and cognitive function ETHICAL APPROVAL The United Kingdom Biobank study received ethical approval from the NHS National IP T Research Ethics Service (Ref: 11/NW/0382). SC R DATA SHARING U All the data reported in this study are fully available via application to the United M AN Kingdom Biobank. ED CONFLICT OF INTEREST O RI G IN AL U N ED IT The authors declare that they have no conflict of interest. 20 Sedentary behaviours and cognitive function REFERENCES 1. Lautenschlager NT, Cox KL, Flicker L, et al. Effect of physical activity on cognitive function in older adults at risk for Alzheimer disease: a randomized 2. IP T trial. JAMA 2008;300(9):1027-1037. Laurin D, Verreault R, Lindsay J, et al. Physical activity and risk of cognitive Erickson KI, Voss MW, Prakash RS, et al. Exercise training increases size of U 3. SC R impairment and dementia in elderly persons. Arch Neurol 2001;58(3):498-504. AN hippocampus and improves memory. Proc Natl Acad Sci U S A 2011;108(7):3017-3022. Colcombe SJ, Kramer AF, Erickson KI, et al. Cardiovascular fitness, cortical M 4. Chapman SB, Aslan S, Spence JS, et al. Shorter term aerobic exercise U N ED IT 5. ED plasticity, and aging. Proc Natl Acad Sci U S A 2004;101(9):3316-3321. improves brain, cognition, and cardiovascular fitness in aging. Front Aging Neurosci 2013;5:75. 6. Colcombe SJ, Erickson KI, Scalf PE, et al. Aerobic exercise training increases brain volume in aging humans. J Gerontol A Biol Sci Med Sci Smith JC, Nielson KA, Antuono P, et al. Semantic memory functional MRI and IN 7. AL 2006;61(11):1166-1170. G cognitive function after exercise intervention in mild cognitive impairment. J RI Alzheimers Dis 2013;37(1):197-215. O 8. Network SBR. Letter to the editor: standardized use of the terms "sedentary" and "sedentary behaviours". Applied physiology, nutrition, and metabolism = Physiologie appliquee, nutrition et metabolisme 2012;37(3):540-542. 21 Sedentary behaviours and cognitive function 9. Biswas A, Oh PI, Faulkner GE, et al. Sedentary time and its association with risk for disease incidence, mortality, and hospitalization in adults: a systematic review and meta-analysis. Ann Intern Med 2015;162(2):123-132. Edwardson CL, Gorely T, Davies MJ, et al. Association of sedentary behaviour with metabolic syndrome: a meta-analysis. PLoS One 11. SC R 2012;7(4):e34916. IP T 10. Grontved A, Hu FB. Television viewing and risk of type 2 diabetes, U cardiovascular disease, and all-cause mortality: a meta-analysis. JAMA Wilmot EG, Edwardson CL, Achana FA, et al. Sedentary time in adults and M 12. AN 2011;305(23):2448-2455. ED the association with diabetes, cardiovascular disease and death: systematic review and meta-analysis. Diabetologia 2012;55(11):2895-2905. Falck RS, Davis JC, Liu-Ambrose T. What is the association between U N ED IT 13. sedentary behaviour and cognitive function? A systematic review. Br J Sports Med 2017;51(10):800-811. 14. Hoang TD, Reis J, Zhu N, et al. Effect of Early Adult Patterns of Physical AL Activity and Television Viewing on Midlife Cognitive Function. JAMA Psychiatry 2016;73(1):73-79. Wang JY, Zhou DH, Li J, et al. Leisure activity and risk of cognitive IN 15. G impairment: the Chongqing aging study. Neurology 2006;66(6):911-913. Hamer M, Stamatakis E. Prospective study of sedentary behavior, risk of O RI 16. depression, and cognitive impairment. Med Sci Sports Exerc 2014;46(4):718723. 22 Sedentary behaviours and cognitive function 17. Kesse-Guyot E, Charreire H, Andreeva VA, et al. Cross-sectional and longitudinal associations of different sedentary behaviors with cognitive performance in older adults. PLoS One 2012;7(10):e47831. Tun PA, Lachman ME. The association between computer use and cognition IP T 18. across adulthood: use it so you won't lose it? Psychol Aging 2010;25(3):560- 19. SC R 568. Sudlow C, Gallacher J, Allen N, et al. UK biobank: an open access resource Allen N, Sudlow C, Downey P, et al. UK Biobank: Current status and what it M 20. AN old age. PLoS Med 2015;12(3):e1001779. U for identifying the causes of a wide range of complex diseases of middle and 21. ED means for epidemiology. Health Policy and Technology 2012;1(3):123-126. UK Biobank. UK Biobank: Protocol for a large-scale prospective U N ED IT epidemiological resource. 2016. (http://www.ukbiobank.ac.uk/wpcontent/uploads/2011/11/UK-Biobank-Protocol.pdf). (Accessed 10 March 2017). 22. Iwasa H, Yoshida Y, Kai I, et al. Leisure activities and cognitive function in AL elderly community-dwelling individuals in Japan: a 5-year prospective cohort study. J Psychosom Res 2012;72(2):159-164. Wilson RS, Mendes De Leon CF, Barnes LL, et al. Participation in cognitively IN 23. G stimulating activities and risk of incident Alzheimer disease. JAMA O RI 2002;287(6):742-748. 24. Verghese J, Lipton RB, Katz MJ, et al. Leisure activities and the risk of dementia in the elderly. N Engl J Med 2003;348(25):2508-2516. 23 Sedentary behaviours and cognitive function 25. Akbaraly TN, Portet F, Fustinoni S, et al. Leisure activities and the risk of dementia in the elderly: results from the Three-City Study. Neurology 2009;73(11):854-861. Aggio D, Smith L, Fisher A, et al. Context-Specific Associations of Physical IP T 26. Activity and Sedentary Behavior With Cognition in Children. Am J Epidemiol 27. SC R 2016;183(12):1075-1082. Elias MF, Sullivan LM, D'Agostino RB, et al. Framingham stroke risk profile Kaffashian S, Dugravot A, Nabi H, et al. Predictive utility of the Framingham AN 28. U and lowered cognitive performance. Stroke 2004;35(2):404-409. M general cardiovascular disease risk profile for cognitive function: evidence 29. ED from the Whitehall II study. Eur Heart J 2011;32(18):2326-2332. Belkic K, Savic C, Theorell T, et al. Mechanisms of cardiac risk among 30. U N ED IT professional drivers. Scand J Work Environ Health 1994;20(2):73-86. Morris JN, Heady JA, Raffle PAB, et al. Coronary heart-disease and physical activity of work. The Lancet 1953;262(6796):1111-1120. 31. Robinson CF, Burnett CA. Truck drivers and heart disease in the United 32. AL States, 1979-1990. Am J Ind Med 2005;47(2):113-119. Howard A, Joint M. Fatigue and Stress in Driving. Employee Counselling IN Today 1994;6(6):3-7. van der Linden D, Frese M, Meijman TF. Mental fatigue and the control of G 33. RI cognitive processes: effects on perseveration and planning. Acta Psychol O (Amst) 2003;113(1):45-65. 34. Abd-Elfattah HM, Abdelazeim FH, Elshennawy S. Physical and cognitive consequences of fatigue: A review. J Adv Res 2015;6(3):351-358. 24 Sedentary behaviours and cognitive function 35. Scott SB, Graham-Engeland JE, Engeland CG, et al. The Effects of Stress on Cognitive Aging, Physiology and Emotion (ESCAPE) Project. BMC Psychiatry 2015;15:146. Shankar A, Hamer M, McMunn A, et al. Social isolation and loneliness: IP T 36. relationships with cognitive function during 4 years of follow-up in the English 37. SC R Longitudinal Study of Ageing. Psychosom Med 2013;75(2):161-170. Bassuk SS, Glass TA, Berkman LF. Social disengagement and incident U cognitive decline in community-dwelling elderly persons. Ann Intern Med Golden J, Conroy RM, Bruce I, et al. Loneliness, social support networks, M 38. AN 1999;131(3):165-173. 2009;24(7):694-700. Mosca L, Barrett-Connor E, Wenger NK. Sex/gender differences in U N ED IT 39. ED mood and wellbeing in community-dwelling elderly. Int J Geriatr Psychiatry cardiovascular disease prevention: what a difference a decade makes. Circulation 2011;124(19):2145-2154. 40. McCarrey AC, An Y, Kitner-Triolo MH, et al. Sex differences in cognitive AL trajectories in clinically normal older adults. Psychol Aging 2016;31(2):166175. Sabia S, Guéguen A, Marmot MG, et al. Does cognition predict mortality in IN 41. G midlife? Results from the Whitehall II cohort study. Neurobiol Aging O RI 2010;31(4):688-695. 42. Batterham PJ, Christensen H, Mackinnon AJ. Fluid intelligence is independently associated with all-cause mortality over 17 years in an elderly community sample: An investigation of potential mechanisms. Intelligence 2009;37(6):551-560. 25 Sedentary behaviours and cognitive function 43. Collins R. What makes UK Biobank special? Lancet 2012;379(9822):11731174. 44. Manolio TA, Collins R. Enhancing the feasibility of large cohort studies. JAMA 45. IP T 2010;304(20):2290-2291. Lyall DM, Cullen B, Allerhand M, et al. Cognitive Test Scores in UK Biobank: SC R Data Reduction in 480,416 Participants and Longitudinal Stability in 20,346 Participants. PLoS One 2016;11(4):e0154222. Troiano RP, Berrigan D, Dodd KW, et al. Physical activity in the United States U 46. Prince SA, Adamo KB, Hamel ME, et al. A comparison of direct versus self- M 47. AN measured by accelerometer. Med Sci Sports Exerc 2008;40(1):181-188. ED report measures for assessing physical activity in adults: a systematic review. Int J Behav Nutr Phys Act 2008;5:56. Belsky DW. Invited Commentary: Lessons for Research on Cognitive Aging U N ED IT 48. From a Study of Children. Am J Epidemiol 2016;183(12):1083-1085. 49. Aggio D, Smith L, Fisher A, et al. Aggio et al. Respond to "Lessons for O RI G IN AL Research on Cognitive Aging". Am J Epidemiol 2016;183(12):1086-1087. 26 Sedentary behaviours and cognitive function TABLES Table 1 - Baseline characteristics of the UK Biobank participants (n = 502,643; years 2006 to 2010) Participant Characteristics Number % Mean (SD) Range IP T Anthropometrics Missing 3,105 12.1–74.7 37.0–73.0 0.6 Demographics Age, years b 0.0 White British 442,699 Other 57,166 Missing 2,778 Female Missing 0.5 54.4 229,176 45.6 0 0.0 273,467 U N ED IT Male 11.4 ED Sex c 88.1 M Ethnicity c U 0 AN 56.5 (8.1) Missing Social deprivation index b Missing -1.3 (3.1) 627 0.1 In paid employment or self-employed 287,234 57.1 Not in paid employment or self-employed 212,451 42.3 2,958 0.6 161,210 32.1 No college or university degree 331,291 65.9 Missing 10,142 2.0 Never 273,603 54.4 Previous 173,099 34.4 IN AL Employment status c Missing G Education level c RI College or university degree O 27.4 (4.8) SC R Body mass index a b Lifestyle Smoking status c 27 -6.3–11.0 Sedentary behaviours and cognitive function Current 52,989 10.6 Missing 2,952 0.6 Never 22,547 4.5 Previous 18,114 3.6 Current 460,479 91.6 Missing 1,503 0.3 <5 300,352 59.8 ≥5 189,979 37.8 Missing 12,132 2.4 4,218 0.8 ED 2.5 29,991 6.0 39,339 7.8 40,036 8.0 80,039 15.9 50,082 9.9 228,697 45.5 8,545 1.7 61,178 12.2 1 38,290 7.6 RI Frequency of ≥10 minutes of walking, days/week c 2 69,799 13.9 3 71,507 14.2 4 47,201 9.4 5 71,441 14.2 6 26,436 5.3 7 89,506 17.8 0 2 3 4 5 AL 6 7 12,455 13,459 2.7 U N ED IT 1 Missing IN Frequency of ≥10 minutes of moderate physical activity, days/week c O G 0 28 SC R AN Missing M Sleep duration, hours/day b U Fruit and vegetable consumption, portions/day c IP T Alcohol drinking status c 7.2 (1.1) 1.0–23.0 Sedentary behaviours and cognitive function Missing 27,285 5.4 0 178,275 35.5 1 66,853 13.3 2 75,055 14.9 3 65,276 13.0 4 30,705 6.1 5 32,452 6.5 6 9,430 1.9 7 17,005 3.4 Missing 27,592 5.5 SC R U M Number of cancers c AN Health 460,075 ED 0 ≥1 91.5 41,706 8.3 862 0.2 U N ED IT Missing 126,639 25.2 134,113 26.7 98,825 19.6 62,828 12.5 79,376 15.8 862 0.2 137,704 27.4 1 94,776 18.8 RI Number of non-cancer illnesses c 0 2 77,673 15.4 3 57,819 11.5 4 42,211 8.4 5 29,937 6.0 ≥6 61,661 12.3 862 0.2 1 2 ≥4 AL 3 Missing IN Number of medications/treatments c O G 0 Missing 29 IP T Frequency of ≥10 minutes of vigorous physical activity, days/week c Sedentary behaviours and cognitive function Medical history of cancer, cardiovascular disease, and/or cognitive/psychiatric illnesses c No 402,897 80.2 Yes 99,746 19.8 0 0.0 <1 39,456 7.8 1 62,503 12.4 2 132,780 26.4 3 116,940 23.3 ≥4 145,546 29.0 Missing IP T Sedentary behaviours 5,418 259,920 ED <1 1 ≥3 51.7 140,144 27.9 60,977 12.1 U N ED IT 2 U 1.1 M Driving time, hours/day c AN Missing 31,663 6.3 9,939 2.0 240,648 47.9 140,821 28.0 62,859 12.5 48,939 9.7 9,376 1.9 Good result 130,910 26.0 Poor result 40,675 8.1 Missing 331,058 65.9 Good result 345,685 68.8 Poor result 138,147 27.5 Missing Computer use time, hours/day c <1 2 IN ≥3 AL 1 Missing G Cognitive function at baseline O RI Prospective memory test c d Visual-spatial memory test (round 1) c e 30 SC R TV viewing time, hours/day c Sedentary behaviours and cognitive function Missing 18,811 3.7 Good result 82,130 16.3 Poor result 400,632 79.7 Missing 19,881 4.0 Visual-spatial memory test (round 2) c f Total number of correct answers Missing 337,151 67.1 Short-term numeric memory test b h Continuous variable. c Categorical variable. 90.0 AN b 452,273 M Weight (kg)/height (m)2. 2.0–12.0 6.7 (1.3) Missing a 0.0–13.0 U Maximum digits remembered correctly 6.0 (2.2) SC R IP T Fluid intelligence test b g d ED Prospective memory result: good result [correct recall on first attempt]; or poor result [incorrect recall on first attempt (i.e. correct recall on second attempt, instruction not recalled, skipped or incorrect)]. Pairs matching result (round 1): good result [<1 incorrect matches]; or poor result [≥1 incorrect matches]. f Pairs matching result (round 2): good result [<2 incorrect matches]; or poor result [≥2 incorrect matches]. g Fluid intelligence score: total number of correct answers. h Numeric memory score: maximum digits remembered correctly. O RI G IN AL U N ED IT e 31 PT Sedentary behaviours and cognitive function (range) = 44,097 to 471,474; years 2006 to 2010) Visual-Spatial Memory Test OR 99% CI Round 1 (Model 1: n = 471,474; Model 2: n = 422,731) d P Value OR 99% CI h P Value h Round 2 (Model 1: n = 470,433; Model 2: n = 421,851) e P Value OR 99% CI h Model 1 2 0.94 3 0.97 ≥4 1.23 Linear trend 1.07 (0.93, 1.07) (0.88, 1.00) (0.91, 1.04) (1.15, 1.30) (1.05, 1.08) - p=0.838 1.05 p=0.012 1.07 p=0.303 1.14 p<0.001 1.26 p<0.001 1.06 Driving time, hours/day i 1.05 2 1.12 ≥3 1.50 (1.02, 1.09) (1.07, 1.18) (1.41, 1.60) - - p<0.001 0.99 p<0.001 1.05 IN 1 - G - p<0.001 O RI <1 - - (1.01, 1.09) (1.03, 1.11) (1.10, 1.18) (1.22, 1.31) (1.06, 1.07) p=0.002 1.26 - 1.01 ED IT 0.99 - p<0.001 1.02 p<0.001 1.03 p<0.001 1.08 U N 1 - AL - ED TV viewing time, hours/day i <1 - (0.97, 1.01) (1.02, 1.08) (1.22, 1.31) Fluid Intelligence Test (Model 1: n = 161,348; Model 2: n = 145,124) f U Prospective Memory Test (Model 1: n = 166,401; Model 2: n = 148,327) c M AN Cross-sectional Analysis: Sedentary Behaviours and Cognitive Function (Model 1 and Model 2) a b SC RI Table 2 - Cross-sectional associations at baseline between sedentary behaviours and cognitive function within the UK Biobank participants (n p<0.001 1.02 - - p=0.270 1.01 p<0.001 1.03 p<0.001 1.12 32 - (0.96, 1.05) (0.98, 1.06) (0.99, 1.08) (1.03, 1.12) (1.01, 1.03) (0.98, 1.03) (1.00, 1.06) (1.07, 1.17) β 99% CI - - - p=0.712 -0.13 p=0.303 -0.30 p=0.036 -0.54 p<0.001 -0.99 p<0.001 -0.26 - - p=0.569 -0.19 p=0.024 -0.37 p<0.001 -0.88 (-0.19, 0.07) (-0.36, 0.24) (-0.60, 0.49) (-1.04, 0.93) (-0.27, 0.25) (-0.22, 0.16) (-0.41, 0.33) (-0.94, 0.82) P Value Short-term Numeric Memory Test (Model 1: n = 49,035; Model 2: n = 44,097) g h β 99% CI - - - p<0.001 -0.15 p<0.001 -0.25 p<0.001 -0.35 p<0.001 -0.55 p<0.001 -0.13 - - p<0.001 -0.02 p<0.001 -0.13 p<0.001 -0.28 (-0.22, 0.08) (-0.31, 0.19) (-0.41, 0.29) (-0.61, 0.48) (-0.15, 0.12) (-0.06, 0.01) (-0.18, 0.08) (-0.34, 0.21) P Value h p<0.001 p<0.001 p<0.001 p<0.001 p<0.001 p=0.139 p<0.001 p<0.001 (1.09, 1.12) p<0.001 1.05 (1.04, 1.06) p<0.001 1.03 (1.01, 1.04) p<0.001 <1 - - - - - - - - - 1 0.68 p<0.001 0.79 p<0.001 0.85 2 0.69 p<0.001 0.77 p<0.001 0.80 ≥3 0.86 p<0.001 0.81 p<0.001 0.82 Linear trend 0.91 p<0.001 0.91 p<0.001 0.92 Computer use time, hours/day i (0.83, 0.87) (0.78, 0.83) (0.79, 0.85) (0.91, 0.93) p<0.001 0.52 U (0.77, 0.80) (0.75, 0.79) (0.79, 0.84) (0.90, 0.92) - p<0.001 0.58 p<0.001 0.40 p<0.001 0.18 - - p=0.403 -0.12 p=0.128 -0.21 p=0.136 -0.33 p=0.053 -0.58 p=0.057 -0.15 - - p<0.001 -0.28 p<0.001 -0.43 M AN (0.66, 0.71) (0.66, 0.72) (0.82, 0.90) (0.89, 0.92) -0.24 (-0.26, 0.23) -0.08 (-0.09, 0.06) p<0.001 - - - - p<0.001 0.21 p<0.001 0.21 p<0.001 0.16 p<0.001 0.07 - - p<0.001 -0.13 p<0.001 -0.20 p<0.001 -0.25 p<0.001 -0.38 p<0.001 -0.09 - - p<0.001 -0.06 p<0.001 -0.18 p<0.001 SC RI 1.11 Linear trend PT Sedentary behaviours and cognitive function - (0.49, 0.55) (0.53, 0.62) (0.35, 0.44) (0.17, 0.20) (0.17, 0.25) (0.16, 0.26) (0.11, 0.22) (0.06, 0.09) p<0.001 p<0.001 p<0.001 p<0.001 TV viewing time, hours/day i 2 0.96 3 0.96 ≥4 1.09 Linear trend 1.02 (0.96, 1.12) (0.90, 1.03) (0.89, 1.03) (1.01, 1.17) (1.00, 1.03) 1.07 p=0.142 1.07 p=0.159 1.09 p=0.003 1.14 p=0.001 1.03 <1 - 1 1.21 2 1.27 - p=0.232 O RI Driving time, hours/day i - - - p<0.001 1.05 G (1.17, 1.27) (1.20, 1.34) - p<0.001 - 1.10 - ED IT 1.04 - (1.03, 1.12) (1.03, 1.11) (1.05, 1.14) (1.10, 1.19) (1.02, 1.04) p<0.001 1.01 p<0.001 1.02 U N 1 - AL - IN <1 ED Model 2 (1.03, 1.07) (1.06, 1.13) p<0.001 1.03 p<0.001 1.03 p<0.001 1.01 - - p<0.001 1.04 p<0.001 1.07 33 - (0.97, 1.06) (0.98, 1.07) (0.98, 1.07) (0.99, 1.08) (1.00, 1.02) (1.02, 1.07) (1.03, 1.10) (-0.18, 0.06) (-0.27, 0.16) (-0.39, 0.28) (-0.64, 0.53) (-0.16, 0.14) (-0.31, 0.25) (-0.48, 0.39) (-0.20, 0.06) (-0.26, 0.13) (-0.32, 0.19) (-0.44, 0.31) (-0.10, 0.07) (-0.10, 0.03) (-0.23, 0.13) p<0.001 p<0.001 p<0.001 p<0.001 p<0.001 p<0.001 p<0.001 1.54 Linear trend 1.15 (1.43, 1.66) (1.13, 1.17) p<0.001 1.23 p<0.001 1.06 - - p<0.001 0.85 p<0.001 0.81 p<0.001 0.84 p<0.001 0.92 (1.19, 1.28) (1.05, 1.07) p<0.001 1.11 p<0.001 1.04 - - p<0.001 0.88 p<0.001 0.83 p<0.001 0.84 p<0.001 0.93 (1.06, 1.16) (1.02, 1.05) p<0.001 p<0.001 1 0.77 2 0.74 ≥3 0.86 Linear trend 0.92 (0.74, 0.81) (0.70, 0.78) (0.81, 0.91) (0.90, 0.94) (0.83, 0.87) (0.79, 0.83) (0.81, 0.86) (0.91, 0.93) OR = odds ratio. 99% CI = 99% confidence interval. β = beta coefficient. Model 1 was mutually adjusted for the other sedentary behaviours and for age and sex. ED IT a b (0.86, 0.90) (0.80, 0.86) (0.81, 0.88) (0.92, 0.94) - - p<0.001 0.32 p<0.001 0.40 p<0.001 0.26 p<0.001 0.12 M AN - ED <1 -0.24 U Computer use time, hours/day i -0.73 (-0.79, 0.68) (-0.25, 0.22) p<0.001 SC RI ≥3 PT Sedentary behaviours and cognitive function - (0.29, 0.35) (0.36, 0.44) (0.22, 0.31) (0.11, 0.14) -0.27 p<0.001 -0.09 - - p<0.001 0.14 p<0.001 0.15 p<0.001 0.13 p<0.001 0.06 (-0.34, 0.19) (-0.11, 0.07) (0.10, 0.17) (0.10, 0.20) (0.07, 0.18) (0.04, 0.07) p<0.001 p<0.001 p<0.001 p<0.001 p<0.001 p<0.001 p<0.01 indicates statistical significance. i <1 hour/day = reference. O RI G h IN AL U N Model 2 was further adjusted for body mass index, ethnicity, social deprivation index, employment status, education level, smoking status, alcohol drinking status, fruit and vegetable consumption, sleep duration, frequency of ≥10 minutes of walking, frequency of ≥10 minutes of moderate physical activity, frequency of ≥10 minutes of vigorous physical activity, number of cancers, number of non-cancer illnesses, and number of medications/treatments. c Prospective memory result: categorical: good result [(reference) correct recall on first attempt]; or poor result [incorrect recall on first attempt (i.e. correct recall on second attempt, instruction not recalled, skipped or incorrect)]. An odds ratio of less than 1 indicates lower odds of a poor result; and an odds ratio of greater than 1 indicates higher odds of a poor result. d Pairs matching result (round 1): categorical: good result [(reference) <1 incorrect matches]; or poor result [≥1 incorrect matches]. An odds ratio of less than 1 indicates lower odds of a poor result; and an odds ratio of greater than 1 indicates higher odds of a poor result. e Pairs matching result (round 2): categorical: good result [(reference) <2 incorrect matches]; or poor result [≥2 incorrect matches]. An odds ratio of less than 1 indicates lower odds of a poor result; and an odds ratio of greater than 1 indicates higher odds of a poor result. f Fluid intelligence score: continuous: total number of correct answers. A beta coefficient of greater than 0 indicates a higher score; and a beta coefficient of less than 0 indicates a lower score. g Numeric memory score: continuous: maximum digits remembered correctly. A beta coefficient of greater than 0 indicates a higher score; and a beta coefficient of less than 0 indicates a lower score. 34 SC RI PT Sedentary behaviours and cognitive function Table 3 - Cognitive function data of the UK Biobank participants with cognitive data at both baseline and follow-up (n (range) = 12,091 to 114,373; Baseline Number % Good result 89,137 77.9 Poor result 25,236 22.1 Visual-spatial memory test (round 1) a b U N 23,262 90,217 % 70,761 61.9 43,612 38.1 70,761 61.9 43,612 38.1 20.5 14,886 13.1 79.5 98,593 86.9 14,886 13.1 98,593 86.9 IN O RI ED Mean (SD) Range 5.5 (2.0) 0.0–13.0 46,704 6.7 (2.1) G Total number of correct answers Good outcome at follow-up Poor outcome at follow-up Follow-up Number AL Poor result Range ED IT 113,479 Good result Fluid intelligence test d e Mean (SD) 114,373 Good outcome at follow-up Poor outcome at follow-up Visual-spatial memory test (round 2) a c Good outcome at follow-up Poor outcome at follow-up M AN Total Number Cognitive Function U mean follow-up period of 5.3 years) 35 0.0–13.0 15,384 32.9 31,320 67.1 Short-term numeric memory test d f Maximum digits remembered correctly Good outcome at follow-up Poor outcome at follow-up 2.0–12.0 6.9 (1.5) 7,791 64.4 4,300 35.6 U 7.0 (1.2) SC RI 12,091 Categorical variable. 2.0–11.0 M AN a PT Sedentary behaviours and cognitive function b Pairs matching result (round 1): good result [<1 incorrect matches]; or poor result [≥1 incorrect matches]. Good outcome at follow-up [<1 incorrect matches at follow-up]; or poor outcome at follow-up [≥1 incorrect matches at follow-up]. c Pairs matching result (round 2): good result [<2 incorrect matches]; or poor result [≥2 incorrect matches]. Good outcome at follow-up [<2 incorrect matches at follow-up]; or poor outcome at follow-up [≥2 incorrect matches at follow-up]. Continuous variable. ED d e ED IT Fluid intelligence score: total number of correct answers. Good outcome at follow-up [baseline fluid intelligence score ≤ follow-up fluid intelligence score]; or poor outcome at follow-up [baseline fluid intelligence score > follow-up fluid intelligence score]. f Numeric memory score: Maximum digits remembered correctly. Good outcome at follow-up [baseline numeric memory score ≤ follow-up numeric memory score]; or poor outcome at follow-up [baseline numeric memory score > follow-up numeric memory score]. U N Table 4 - Prospective associations between sedentary behaviours at baseline and cognitive function at follow-up within the UK Biobank participants (N (range) = 11,299 to 113,129; mean follow-up period of 5.3 years) Round 1 (Model 1: n = 113,129; Model 2: n = 106,665) c 99% CI P Value g IN OR O RI TV viewing time, hours/day h <1 AL Visual-Spatial Memory Test Round 2 (Model 1: n = 112,252; Model 2: n = 105,861) d OR 99% CI Fluid Intelligence Test (Model 1: n = 46,158; Model 2: n = 43,350) e Short-term Numeric Memory Test (Model 1: n = 11,957; Model 2: n = 11,299) f P Value g OR 99% CI P Value g OR 99% CI P Value g - - - - - - - Model 1 G Prospective Analysis: Sedentary Behaviours and Cognitive Function (Model 1 and Model 2) a b - - - - - 36 1.09 3 1.13 ≥4 1.17 Linear trend 1.04 p=0.154 1.02 p<0.001 1.00 p<0.001 1.03 p<0.001 1.01 p<0.001 1.00 - - p<0.001 1.01 p=0.002 1.00 p<0.001 1.01 (0.93, 1.11) (0.92, 1.08) (0.94, 1.12) (0.93, 1.10) (0.99, 1.02) p=0.623 1.15 p=0.961 1.24 p=0.439 1.37 p=0.672 1.66 p=0.612 1.13 - - p=0.480 1.15 p=0.903 1.10 p=0.831 1.44 p=0.709 1.10 - - p=0.068 0.93 p<0.001 0.94 p<0.001 0.96 p<0.001 0.98 Driving time, hours/day h - 1 1.06 2 1.07 ≥3 1.18 Linear trend 1.05 (1.02, 1.10) (1.01, 1.12) (1.09, 1.28) (1.03, 1.07) p<0.001 Computer use time, hours/day 0.96 2 0.90 ≥3 0.91 Linear trend 0.96 (0.93, 1.00) (0.86, 0.94) (0.86, 0.96) (0.95, 0.98) - - p=0.013 0.96 p<0.001 0.87 p<0.001 0.89 p<0.001 0.95 AL 1 O RI TV viewing time, hours/day h - IN - G <1 1.00 U N h (0.96, 1.07) (0.93, 1.08) (0.90, 1.13) (0.98, 1.03) ED - ED IT <1 (1.02, 1.28) (1.12, 1.37) (1.24, 1.52) (1.50, 1.84) (1.10, 1.15) (0.91, 1.02) (0.81, 0.93) (0.83, 0.96) (0.93, 0.97) Model 2 37 p=0.002 SC RI 2 (0.97, 1.11) (1.03, 1.15) (1.07, 1.20) (1.10, 1.25) (1.03, 1.06) U 1.04 M AN 1 PT Sedentary behaviours and cognitive function - (1.08, 1.22) (1.00, 1.21) (1.25, 1.66) (1.06, 1.14) (0.87, 1.00) (0.86, 1.02) (0.88, 1.05) (0.96, 1.01) 1.05 p<0.001 1.13 p<0.001 1.26 p<0.001 1.43 p<0.001 1.10 - - p<0.001 1.05 p=0.008 1.09 p<0.001 1.11 p<0.001 1.04 - - p=0.006 0.90 p=0.041 0.77 p=0.293 0.86 p=0.150 0.93 (0.85, 1.31) (0.93, 1.37) (1.03, 1.55) (1.17, 1.76) (1.05, 1.15) (0.93, 1.18) (0.92, 1.30) (0.85, 1.44) (0.98, 1.11) (0.79, 1.02) (0.65, 0.90) (0.72, 1.03) (0.88, 0.98) p=0.557 p=0.112 p=0.003 p<0.001 p<0.001 p=0.319 p=0.193 p=0.318 p=0.108 p=0.034 p<0.001 p=0.035 p=0.001 1 1.02 2 1.07 3 1.08 ≥4 1.09 Linear trend 1.02 (0.96, 1.09) (1.00, 1.13) (1.02, 1.15) (1.02, 1.17) (1.01, 1.04) - - p=0.348 1.03 p=0.006 1.01 p=0.001 1.03 p=0.001 1.00 p<0.001 1.00 - - p<0.001 1.02 p=0.001 1.01 (0.94, 1.12) (0.93, 1.09) (0.94, 1.12) (0.91, 1.10) (0.98, 1.02) - - p=0.470 1.16 p=0.815 1.21 p=0.416 1.29 p=0.993 1.45 p=0.955 1.09 - - p=0.294 1.19 p=0.624 1.15 p=0.895 1.43 p=0.552 1.11 - - p=0.250 0.94 p<0.001 0.94 p=0.001 0.97 p<0.001 0.99 1 1.07 2 1.08 ≥3 1.16 Linear trend 1.05 (1.03, 1.12) (1.02, 1.14) (1.06, 1.26) (1.03, 1.07) Computer use time, hours/day h - 0.97 2 0.91 ≥3 0.90 Linear trend 0.96 (0.93, 1.01) (0.86, 0.95) (0.85, 0.96) (0.95, 0.98) p<0.001 O RI 1.01 1.01 - - p=0.053 0.97 p<0.001 0.88 p<0.001 0.90 p<0.001 0.96 AL 1 IN - G <1 p<0.001 - (0.97, 1.08) (0.94, 1.10) (0.90, 1.13) (0.98, 1.04) ED IT - U N - ED Driving time, hours/day h <1 (1.03, 1.30) (1.09, 1.35) (1.15, 1.44) (1.29, 1.62) (1.06, 1.11) (0.92, 1.03) (0.82, 0.94) (0.83, 0.98) (0.93, 0.98) OR = odds ratio. 99% CI = 99% confidence interval. 38 - SC RI - U - M AN <1 PT Sedentary behaviours and cognitive function - (1.11, 1.27) (1.04, 1.27) (1.24, 1.66) (1.07, 1.15) (0.87, 1.01) (0.86, 1.03) (0.88, 1.06) (0.96, 1.02) - p=0.001 1.02 p<0.001 1.08 p<0.001 1.16 p<0.001 1.29 p<0.001 1.07 - - p<0.001 1.05 p<0.001 1.05 p<0.001 1.05 p<0.001 1.02 - - p=0.020 0.92 p=0.073 0.76 p=0.359 0.84 p=0.207 0.92 (0.82, 1.28) (0.88, 1.33) (0.94, 1.44) (1.04, 1.61) (1.02, 1.12) (0.92, 1.19) (0.88, 1.27) (0.80, 1.39) (0.96, 1.10) (0.80, 1.05) (0.64, 0.90) (0.69, 1.01) (0.87, 0.98) p=0.817 p=0.310 p=0.066 p=0.003 p<0.001 p=0.363 p=0.466 p=0.650 p=0.363 p=0.090 p<0.001 p=0.016 p<0.001 a PT Sedentary behaviours and cognitive function Model 1 was mutually adjusted for the other sedentary behaviours and for age, sex and the baseline result/score of the cognitive test under consideration. b p<0.01 indicates statistical significance. h <1 hour/day = reference. O RI G IN AL U N ED IT g ED M AN U SC RI Model 2 was further adjusted for body mass index, ethnicity, social deprivation index, employment status, education level, smoking status, alcohol drinking status, fruit and vegetable consumption, sleep duration, frequency of ≥10 minutes of walking, frequency of ≥10 minutes of moderate physical activity, frequency of ≥10 minutes of vigorous physical activity, number of cancers, number of non-cancer illnesses, and number of medications/treatments. c Pairs matching result (round 1): categorical: good outcome at follow-up [<1 incorrect matches at follow-up]; or poor outcome at follow-up [≥1 incorrect matches at follow-up]. An odds ratio of less than 1 indicates lower odds of having cognitive decline at follow-up (i.e. a good outcome at follow-up); and an odds ratio of greater than 1 indicates higher odds of having cognitive decline at follow-up (i.e. a poor outcome at follow-up). d Pairs matching result (round 2): categorical: good outcome at follow-up [<2 incorrect matches at follow-up]; or poor outcome at follow-up [≥2 incorrect matches at follow-up]. An odds ratio of less than 1 indicates lower odds of having cognitive decline at follow-up (i.e. a good outcome at follow-up); and an odds ratio of greater than 1 indicates higher odds of having cognitive decline at follow-up (i.e. a poor outcome at follow-up). e Fluid intelligence score: categorical: good outcome at follow-up [baseline fluid intelligence score ≤ follow-up fluid intelligence score]; or poor outcome at follow-up [baseline fluid intelligence score > follow-up fluid intelligence score]. An odds ratio of less than 1 indicates lower odds of having cognitive decline at follow-up (i.e. a good outcome at follow-up); and an odds ratio of greater than 1 indicates higher odds of having cognitive decline at follow-up (i.e. a poor outcome at follow-up). f Numeric memory score: categorical: good outcome at follow-up [baseline numeric memory score ≤ follow-up numeric memory score]; or poor outcome at follow-up [baseline numeric memory score > follow-up numeric memory score]. An odds ratio of less than 1 indicates lower odds of having cognitive decline at follow-up (i.e. a good outcome at follow-up); and an odds ratio of greater than 1 indicates higher odds of having cognitive decline at follow-up (i.e. a poor outcome at follow-up). 39