Forest plots show pooled estimates of the hazard ratio (HR) for cardiovascular disease risk. Size of the data marker indicates the percentage weight of each study in the pooled analysis.
Dose-response associations are reported in the included studies and observed on pooled analysis. The pooled analysis estimates are shown by spline (smoothed fit) and 95% CIs of the pooled hazard ratio (HR) for adverse cardiovascular events by duration of sedentary time. The size of each data marker indicates study size.
Continuous dose-response association between total sedentary duration and risk for CVD after adjustment for all potential confounders including physical activity (9 studies); after adjustment for potential confounders excluding physical activity (4 studies); and after adjustment for potential confounders excluding body mass index (5 studies). Spline (smoothed fit) and 95% CI of pooled hazard ratio (HR) for CVD incidence by sedentary hours are shown.
eMethods. Data Collection, Exposure Assessment, and Sedentary Time Categories for Pooled Analysis
eTable 1. Reported Range of Time and Estimated Median Durations (in Hours/Day) for Sedentary Time Categories Across Studies
eTable 2. Physical Activity Assessment and Adjustment Across Included Studies
eTable 3. Quality Assessment of the Included Studies
eTable 4. Association Between Total Sedentary Time and Risk for Adverse Cardiovascular Events Among Different Subgroups and Sensitivity Analysis
eTable 5. Joint Effects of Physical Activity and Sedentary Time on Cardiovascular Disease Incidence
eFigure 1. Flowchart of Study Selection for Meta-analysis
eFigure 2. Funnel Plot for Assessment of Publication Bias
Customize your JAMA Network experience by selecting one or more topics from the list below.
Pandey A, Salahuddin U, Garg S, et al. Continuous Dose-Response Association Between Sedentary Time and Risk for Cardiovascular Disease: A Meta-analysis. JAMA Cardiol. 2016;1(5):575–583. doi:10.1001/jamacardio.2016.1567
Prior studies suggest that higher sedentary time is associated with a greater risk for cardiovascular disease (CVD). However, the quantitative, dose-response association between sedentary time and CVD risk is not known.
To determine the categorical and quantitative dose-response association between sedentary time and CVD risk.
Two independent investigators searched the MEDLINE and EMBASE databases for all studies published before July 6, 2015, that evaluated the association between sedentary time and incident CVD.
Prospective cohort studies with participants 18 years or older that reported the association between sedentary time and incident CVD were included.
Data Extraction and Synthesis
Two independent investigators performed the data extraction and collection using a standardized form. The study quality was assessed using the Newcastle-Ottawa Scale. The categorical dose-response association was evaluated by comparing the pooled hazard ratio (HR) for incident CVD associated with different levels of sedentary time (vs lowest sedentary time) across studies. The continuous dose-response association was assessed using random-effects generalized least squares spline models. Data were collected from April 5 to July 6, 2015.
Main Outcomes and Measures
Incident CVD (coronary heart disease, including nonfatal myocardial infarction, stroke, and cardiovascular mortality).
Nine prospective cohort studies with 720 425 unique participants (57.1% women; 42.9% men; mean age, 54.5 years) and 25 769 unique cardiovascular events and a median follow-up of 11 years were included. In categorical analyses, compared with the lowest sedentary time category (median, 2.5 h/d), participants in the highest sedentary time category (median, 12.5 h/d) had an increased risk for CVD (HR, 1.14; 95% CI, 1.09-1.19). However, no apparent risk associated with intermediate levels of sedentary time (HR for 7.5 h/d, 1.02; 95% CI, 0.96-1.08) was found. In continuous analyses, a nonlinear association between sedentary time and incident CVD was found (P for nonlinearity < .001), with an increased risk observed for more than 10 hours of sedentary time per day (pooled HR, 1.08; 95% CI, 1.00-1.14).
Conclusions and Relevance
The association between sedentary time and the risk for CVD is nonlinear with an increased risk only at very high levels. These findings could have implications for guideline recommendations regarding the risks related to sedentary behavior.
Physical inactivity has been identified as 1 of the 4 major modifiable risk factors for cardiovascular disease (CVD).1 The current guidelines recommend a minimum of 150 min/wk of moderate-intensity physical activity or 75 min/wk of vigorous-intensity physical activity in sessions lasting 10 minutes or more.2 Despite these recommendations and aggressive public health promotion, less than 4% of adults (aged 20-59 years) meet these guideline recommendations for exercise.3,4 Thus, novel preventive approaches aimed at other modifiable facets of the inactive lifestyle, such as excess sedentary or sitting time, are needed to reduce the burden of CVD.
Sedentary time is defined as time spent in activities involving low levels of energy expenditure (1.0-1.5 metabolic equivalent tasks, such as sitting, watching television, driving).5 Several epidemiologic studies have demonstrated that prolonged sedentary time is associated with an increased risk for CVD.6-17 Furthermore, recent studies suggest that the contributions of prolonged sedentary time toward increased CVD risk appear to be independent of the level of physical activity.7,12,17,18 However, the quantitative risk for CVD associated with different levels of sedentary time, independent of physical activity and other risk factors for cardiovascular (CV) events, is not known, and therefore, no consensus recommendations have been issued regarding limits on the amounts of sedentary time to optimize CVD prevention. To address this knowledge gap, we performed a pooled analysis of prospective cohort studies using a well-established meta-analytical approach19,20 to determine the categorical and quantitative dose-response association between sedentary time and CVD risk, independent of physical activity. We hypothesized that a dose-dependent association exists between sedentary time and the risk for adverse CV events independent of physical activity, particularly at higher levels of sedentary time.
Questions What is the quantitative association between sedentary time and the risk for cardiovascular disease?
Findings In this meta-analysis of 9 prospective cohort studies with 720 425 unique participants and 25 769 unique cardiovascular events, a nonlinear association between sedentary time and incident cardiovascular events was found, with an increased risk observed for more than 10 h/d of sedentary time.
Meaning The association of sedentary time with the risk for adverse cardiovascular events is nonlinear with an increased risk only at very high levels, which could have implications for guideline recommendations regarding the risks related to sedentary behavior.
We followed the protocol for Meta-analysis of Observational Studies in Epidemiology (MOOSE) for performing and reporting the present meta-analysis.21 We searched for all prospective cohort studies that examined the association between sedentary times and the risk for adverse CV events, including coronary artery disease, stroke, and CV-related mortality among adult participants (18 years or older at baseline). Systematic searches through the electronic databases (MEDLINE [OVID] and EMBASE [Elsevier Inc]) were performed with assistance from the study librarian (H.M.). Special search features, including EMTREE, explosion technique, and subheadings, were also used. We also performed additional manual searches through the reference lists of original publications and review articles. We used various combinations of the following key words and MeSH terms: sedentary time, sitting, cardiovascular risk, coronary artery disease, stroke, and cardiovascular disease mortality. The search was restricted to published English language articles that focused on human participants.
Prospective cohort studies that reported the association between baseline sedentary times and the risk for CVD incidence were included. Studies with all types of sedentary activity (as assessed by sitting time) were included in the initial study selection process. Because our primary aim is to quantify the risk for CVD associated with total sedentary time independent of physical activity, only studies that reported measures of total sedentary time and the associated risk for CVD adjusted for physical activity were included in the meta-analysis. If multiple articles were published from the same cohort, we included data from the study with the most detailed report of sedentary time and/or the larger sample size. The primary outcome of interest for this meta-analysis was incident atherosclerotic CVD (coronary heart disease, including nonfatal myocardial infarction, stroke, and CV-related mortality). Two independent investigators (A.P. and U.S.) conducted the initial screening of all titles or abstracts and then evaluated all potentially relevant articles based on full-text reviews. Studies were excluded if they failed to meet all the criteria detailed above. All discrepancies regarding study inclusion were adjudicated by the senior author (J.D.B.). For one study,6 the data on association between sedentary time categories and CVD risk were obtained from a related thesis publication by the same author group archived in the University of Tampere, Finland electronic database (https://tampub.uta.fi/bitstream/handle/10024/76853/gradu06623.pdf?sequence=1). The study quality was assessed using the Newcastle-Ottawa Scale, which allowed a total score of 9 points or fewer (9 indicates the highest quality) summarizing 8 aspects of each study.22
Two of us (A.P. and U.S.) independently performed the data collection from April 5 to July 6, 2015, using a standardized form. Information recorded for each study is detailed in the eMethods in the Supplement. Disagreements between the 2 reviewers regarding extracted data were resolved by consensus, and if needed, in consultation with the senior author (J.D.B.). No personal communication with study authors was required to obtain pertinent data from the selected studies.
Dose-response meta-analysis was performed using the statistical analysis approach as previously described.19,20 Briefly, we estimated the median sedentary duration for each sedentary time category as described in the eMethods in the Supplement and assigned it to the corresponding hazard ratio (HR) for each study. The reported ranges of sedentary time categories and estimated medians are detailed in eTable 1 in the Supplement.
For the present meta-analysis we used the HR or relative risk as available with their 95% CIs as a measure of the effect size associated with each category of sedentary time for all studies. In articles that studied more than 1 type of sedentary activity (eg, sitting time, television time), overall sedentary time was preferentially included for analysis. We used HRs from multivariable-adjusted models with the most complete adjustment for potential baseline confounders, including the presence of baseline CVD risk factors and physical activity levels for primary analysis.23
Categorical and continuous dose-response analysis was performed in the present study. The categorical dose-response analysis was performed with STATA software (version 10.0; StataCorp). Because most of the studies had 3 or fewer sedentary time categories (6 of 9), we used a previously described approach19 to pool data across studies and generated 3 pooled categories of sedentary time (ie, highest, intermediate, and lowest) as described in the eMethods in the Supplement. The median durations of the pooled highest, intermediate, and lowest sedentary time categories were 2.5 (interquartile range, 1.5- 2.9), 7.5 (interquartile range, 6.6-7.6), and 12.5 (interquartile range, 9.5-13.8) hours, respectively. The pooled HRs and 95% CI for CVD associated with different categories of sedentary time were calculated by comparing the highest and intermediate with the lowest sedentary time categories using the random-effects modeling technique as described by DerSimonian and Laird.24 Maximally adjusted HRs, when reported, were used for the primary analysis to account for confounding variables. Pooled analysis comparing highest vs lowest sedentary times included all available studies (n = 9), whereas comparisons of intermediate vs lowest sedentary time categories included 8 studies. We assessed heterogeneity using the I2 test (I2 > 50% was assumed to be a result of significant heterogeneity). We tested the robustness of the observed associations by performing subgroup analyses based on age of participants (>55 vs ≤55 years), geographic location (US vs non-US studies), physical activity levels (above vs below the pooled median level), multivariable adjustment strategy used in the analysis (HR associated with models with vs without adjustment for physical activity and body mass index [BMI]), and duration of follow-up (<10 vs ≥10 years). Publication bias was assessed using contour-enhanced funnel plots and the Begg’s rank correlation test. All P values were-2 tailed. For all tests, a probability level less than .05 was considered statistically significant.
The continuous dose-response association between sedentary time and risk for CVD was assessed using a generalized least squares regression model with the maximum likelihood method using SAS software (version 9.2; SAS Corporation), as previously described in the literature.19,20,25 This method accounts for appropriate variance-covariance associations between and within studies. It uses the multiple data points available in all studies simultaneously to provide the best overall pooled dose-response estimate in a single estimation. Nonlinearity in the association between sedentary time and CVD risk was assessed by modeling sedentary time duration with the use of restricted cubic splines with 3 knots at fixed centiles (5%, 50%, and 95%) of the distribution. We first estimated a restricted cubic spline model with a generalized least squares regression, considering the correlation within each set of reported HRs. We then combined the study-specific estimates, using the restricted maximum likelihood method in a multivariable random-effects meta-analysis. We also performed sensitivity analyses using HRs from models without adjustment for physical activity and BMI levels to determine whether the observed dose-response association was different with vs without adjustment for these potential confounders.
The study selection process and results from the literature search are shown in eFigure 1 in the Supplement. We included 9 cohort studies with 720 425 unique participants (57.1% women; 42.9% men; mean age, 54.5 years) and 25 769 unique incident CVD events during a median follow-up of 11 years. Baseline characteristics of the included studies are shown in Table 1. Of the 9 studies that were included,6,8,9,11-16 2 studies9,11 included only women and 7 studies6,8,12-16 included men and women. Sedentary behavior was assessed by self-reported questionnaire in the included studies. Objective criteria (administrative database or national mortality and/or death index) were used to identify the incident CVD or associated mortality events in most studies (Table 2). Most studies adjusted for covariates such as age (n = 9), sex (n = 9), BMI (n = 7), smoking (n = 9), physical activity (n = 9), and CVD risk factors (diabetes and/or hypertension) (n = 6) in the most adjusted model evaluating the association between sedentary time and CVD risk (Table 2). Across most of the included studies, higher sedentary time was associated with lower levels of physical activity (eTable 2 in the Supplement).
Assessment of study quality yielded a mean score of 8.7, and all studies had a score of 8 or above (eTable 3 in the Supplement). We did not observe a significant publication bias in the present meta-analysis (P for Egger line regression test = .26; P for Begg rank correlation test = .45 [eFigure 2 in the Supplement]).
Figure 1 shows the pooled estimates the risk for CVD associated with different sedentary time categories. No significant heterogeneity was observed in the pooled analysis comparing the highest vs lowest sedentary time categories across studies (I2 = 39.7%; P = .08), whereas moderate heterogeneity was observed in the pooled comparison between intermediate vs lowest sedentary time categories (I2 = 61.8%; P = .004). Compared with the lowest sedentary time category (median duration, 2.5 h/d), participants in the highest sedentary time category (median duration, 12.5 h/d) had an increased risk for CVD (pooled HR, 1.14; 95% CI, 1.09-1.19). However, the risk associated with the intermediate sedentary time category (median duration, 7.5 h/d) compared with the lowest sedentary time category was not statistically significant (pooled HR, 1.02; 95% CI, 0.96-1.08). In continuous analyses, we found a nonlinear association between sedentary time and CVD risk (P < .001 for nonlinearity), with a nonsignificant increased risk observed only at sedentary times more than 6.8 h/d (pooled HR, 1.01; 95% CI, 0.95-1.08), and that became statistically significant at times more than 10.04 h/d (pooled HR, 1.08; 95% CI, 1.00-1.14) (Figure 2).
To confirm the robustness of our study findings, we conducted additional sensitivity and subgroup analysis evaluating the association between the highest levels of sedentary time and the risk for CVD. We did not observe a significant difference in magnitude or direction of the association between high levels of sedentary time and CVD risk in subgroup analyses by age, geographic location, follow-up duration, and levels of physical activity at baseline (eTable 4 in the Supplement).
To characterize better the effect of physical activity on the observed associations between sedentary time and CVD risk, we conducted additional sensitivity analyses by pooling HRs from multivariable-adjusted models without adjustment for physical activity (4 studies11,13-15). We observed similar associations between the highest levels of sedentary time and CVD risk using HRs unadjusted for physical activity (eTable 4 in the Supplement). Similar findings were also observed in continuous dose-response analyses with a statistically significant risk for CVD events observed at a modestly lower sedentary time threshold of 9.2 hours compared with the primary analysis (10.04 h/d) (Figure 3A and B).
We also assessed the role of BMI as a potential contributor to the observed association between sedentary time and CVD risk by performing separate pooled analysis using HRs from models that evaluated the association between sedentary time and CVD risk with vs without adjustment for BMI. In categorical analyses, the pooled estimate was not different in studies with vs without adjustment for BMI (eTable 4 in the Supplement). Similar findings were also observed in continuous dose-response analysis using hazard ratios unadjusted for BMI (Figure 3C).
The joint effect of physical activity and sedentary time was reported quantitatively in 3 studies.8,9,15 We did not perform a meta-analysis to evaluate the joint effects of physical activity and sedentary time on CVD risk owing to the small number of available studies. However, we compared the relative hazard associated with higher levels of sedentary time among physically inactive vs active groups in each of the 3 studies (eTable 5 in the Supplement). Five studies8,9,11,13,15 performed interaction testing between physical activity levels and sedentary time for CVD risk. Of these, 4 studies8,9,11,15 observed no significant interaction between the 2 variables, whereas 1 study13 reported a significant interaction between television viewing time and physical activity level among men for CVD risk (eTable 5 in the Supplement).
The findings from this meta-analysis, based on 720 425 participants from 9 cohort studies and including 25 769 CVD events, demonstrate a nonlinear association between sedentary time levels and CVD risk. After adjustment for physical activity and other CVD risk factors, significant risk for CVD was observed with very high levels of sedentary time (>10 h/d), with no apparent risk associated with intermediate levels of sedentary time. To our knowledge, the present study represents the largest and most comprehensive evaluation of the dose-response association between sedentary time and CVD risk in the general population.
Recent meta-analyses have evaluated the dichotomous association between sedentary time (high vs low) and the risk for adverse clinical outcomes and demonstrated an increased risk for diabetes, CVD, and incident cancer with greatest levels of sedentary time.7,26 However, these studies were limited by lack of adjustment for physical activity in the included studies26 or the lack of assessment of the graded dose-response association between sedentary time and CVD risk.7 Thus, the independent contributions of different levels of sedentary time toward CVD risk were not assessed. The present study adds significantly to the existing body of literature by providing a comprehensive evaluation of the dose-response association between sedentary time and CVD risk.
We observed a nonlinear association between total sedentary duration and CVD risk, with increased risk observed only at a sedentary duration greater than 10 hours. The biological mechanism underlying this nonlinear association is not completely understood, but it appears to reflect an apparent threshold effect of sedentary time on cardiometabolic risk factors. Along these lines a recent study by Qi et al27 demonstrated deleterious associations between objectively measured sedentary time and cardiometabolic biomarkers, with more pronounced abnormalities in triglyceride levels, glucose intolerance, insulin sensitivity, and C-reactive protein levels observed only at the highest levels of sedentary time.
The nonlinear association between total sedentary time and CVD risk observed in this study is in contrast to the linear association between television viewing time and CVD risk that was previously reported by Grøntved and Hu in 2012.10 This discrepancy between the study findings could be owing to differences in the measures of sedentary behavior and pooled HRs for CVD events used in the 2 meta-analyses. Although overall sedentary time was used as a measure of sedentary behavior in the 9 cohort studies included in the present meta-analysis, Grøntved and Hu10 included fewer studies and evaluated the association between television viewing, a less informative measure of overall sedentary behavior,28 and the risk for CVD. As a result, the observed differences in the associations could have been related to the difference in the exposure variable of interest. Furthermore, Grøntved and Hu10 observed a nonlinear association between television viewing and mortality in a pooled analysis, similar to our study findings.
Our study findings have important clinical and public health implications. Current public health guidelines are focused on physical activity promotion in the general population and recommend at least 30 minutes of moderate- to vigorous-intensity physical activity on most days of the week among healthy adults.2 In contrast, no such guideline recommendations exist regarding the targets for reducing daily duration of sedentary times in the general population. This lack represents an important gap in the public health guidelines when we consider that as many as two-thirds of adult waking hours are sedentary. Our study findings provide important insights into the thresholds beyond which sedentary time may be detrimental to CV health, independent of physical activity levels. Reduction of sedentary times for CVD prevention may be particularly relevant among individuals who may not be able to tolerate guideline-recommended physical activity levels owing to comorbidities or other limitations.15,29 Nonambulatory interventions, such as sit-stand workstations and activity-permissive desks, aimed at reducing sedentary times among these individuals with very high levels of sedentary time (>10 h/d) may help to reduce CVD risk. This finding is also supported by recent observational data that have reported significant benefits of increased standing or stepping for cardiometabolic risk biomarkers.30
The main strengths of our meta-analysis are inclusion of large, well-established prospective studies from a range of geographic locations; a large number of included participants and follow-up CV events; and detailed adjustment for a wide range of potential confounding risk factors that allowed us to evaluate the independent contributions of different levels of sedentary time to CVD risk. The included studies were of high quality (mean score of 8.7 of 9), and we found no significant publication bias in our included studies. We observed low to moderate degrees of heterogeneity in categorical dose-response analyses, with moderate heterogeneity only for the intermediate vs lowest sedentary time comparison. Furthermore, to confirm the robustness of our study findings, we performed several sensitivity and subgroup analyses and observed no significant change in the magnitude of the direction of the pooled effect size for the association between total sedentary duration and CVD risk.
Our study also has several important limitations. First, the study protocol was not published a priori. Second, because we performed a meta-analysis of observational studies, the results are subject to unmeasured or residual confounding. Third, measurement errors in self-reported sedentary time and variability in the scale of sitting time across studies may have limited the analytical power of the study to determine the association between sedentary time and CVD risk, particularly at higher levels. However, owing to the large number of participants and events included in our dose-response pooled analysis, our study was well powered to detect even a very small risk for CVD associated with unit increases in sedentary time, even at the extreme levels. Future studies with objectively measured sedentary time levels using accelerometers and long-term follow-up are needed to confirm our study findings. Finally, we only included English language studies in our meta-analysis to negate culture-based heterogeneity in our study findings. Although this criterion may have led to potential language bias, previous studies have failed to demonstrate a systematic bias from use of language restriction.31
Our study findings suggest that CVD risk associated with total sedentary time is nonlinear, with an increased risk only at very high levels (>10 h/d). Furthermore, the CVD risk associated with very high levels of sedentary time is independent of the baseline CVD risk factor burden and physical activity levels. Future studies are needed to characterize the mechanisms through which high levels of sedentary time increase CVD risk and to determine whether nonambulatory sitting reduction interventions can potentially reduce the risk for adverse CVD events among individuals with very high levels of sedentary time.
Corresponding Author: Jarett D. Berry, MD, MS, Division of Cardiology, Department of Internal Medicine, University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd, Dallas, TX 75390 (firstname.lastname@example.org).
Accepted for Publication: April 21, 2016.
Published Online: July 13, 2016. doi:10.1001/jamacardio.2016.1567.
Author Contributions: Drs Pandey and Salahuddin contributed equally to this study. Dr Berry had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Pandey, Kulinski, Berry.
Acquisition, analysis, or interpretation of data: Pandey, Salahuddin, Garg, Ayers, Anand, Mayo, Kumbhani, de Lemos, Berry.
Drafting of the manuscript: Pandey, Salahuddin, Garg, Anand, Mayo, Berry.
Critical revision of the manuscript for important intellectual content: Pandey, Ayers, Kulinski, Kumbhani, de Lemos, Berry.
Statistical analysis: Pandey, Garg, Ayers, Berry.
Administrative, technical, or material support: Pandey, Salahuddin, Mayo.
Study supervision: Berry.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Funding/Support: This study was supported by the Dedman Family Scholar in Clinical Care Endowment at University of Texas Southwestern Medical Center and grant 14SFRN20740000 from the American Heart Association Prevention Network (Dr Berry).
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Create a personal account or sign in to: