Accelerometer-Measured Sedentary Behavior, Future Disease, and Cardiovascular-Kidney-Metabolic Health

Abstract

Background

Effects of sedentary behavior across the spectrum of future disease, including cardiovascular-kidney-metabolic (CKM) syndrome-related conditions, are poorly understood.

Objectives

The objective of the study was to examine associations between accelerometer-measured sedentary behavior and incidence of >700 conditions.

Methods

Among UK Biobank participants with accelerometer-measured sedentary time, we fit multivariable-adjusted Cox models with P value thresholds targeting a false discovery rate of 1%. To estimate the impact of sedentary behavior on population-level disease burden, we calculated population-attributable fractions (PAFs) for key CKM conditions and compared them to the American Heart Association Life’s Essential 8.

Results

Among 89,537 individuals (age 63 ± 8 years, 56.3% women) undergoing accelerometry with a median follow-up 8.0 years (quartile-1: 7.5, quartile-3: 8.5), sedentary time was associated with 81/761 (10.6%) incident diseases at false discovery rate 1%. Seventy-five associations (92.6%) indicated higher disease risk with greater sedentary time. Strongest associations were observed for CKM conditions, where using the second quartile (8.2-9.4 hours/day) as a referent, sedentary time in the top quartile (>10.6 hours/day) was associated with substantially higher risks of hypertension (HR: 1.13; 95% CI: 1.08-1.19), diabetes (HR: 1.18; 95% CI: 1.07-1.31), chronic kidney disease (HR: 1.14; 95% CI: 1.02-1.27), and obstructive sleep apnea (HR: 1.19; 95% CI: 0.99-1.42). At the >10.6 hours/day threshold, excess sedentary behavior was associated with population-level risks of hypertension (PAF: 1.6%; 95% CI: 1.2-1.9), diabetes (PAF: 1.3%; 0.5-2.7), chronic kidney disease (PAF: 1.9%; 1.2-2.5) and obstructive sleep apnea (PAF: 4.2%; 2.3-6.6) exceeding certain components of Life’s Essential 8 (eg, adequate sleep).

Conclusions

Excess sedentary behavior is broadly associated with future disease risk. Effects are prominent for CKM conditions, where population-level burden rivals consensus lifestyle factors. Avoiding excess sedentary behavior should be a key public health target.

Central Illustration

Current guidelines emphasize the importance of achieving ≥150 minutes of moderate-to-vigorous physical activity (MVPA) per week to promote cardiovascular (CV) health.1, 2, 3, 4, 5 Sedentary behavior, defined as energy expenditure ≤1.5 metabolic equivalents of task (METs) while in a seated, reclined, or lying posture,⁶ is increasingly recognized as harmful. However, available evidence has been insufficient to support guideline recommendations beyond general statements such as “sit less”.⁵

Physical activity has well-established dose-dependent protective effects against CV diseases, cancer, and premature mortality.^7,8 Although sedentary behavior has been associated with incident disease in multiple systems relevant to CV and non-CV health,^7,9, 10, 11 most prior studies have relied on self-reported activity metrics and focused on only 1 or a few conditions. In contrast, a robust and comprehensive assessment of measured sedentary behavior across the full spectrum of disease is needed to estimate the contemporary public health burden of excess sedentary time and inform future interventions focused on optimizing sedentary behavior. A holistic approach would further facilitate assessment of the effects of sedentary behavior on the aggregate of metabolic syndrome, chronic kidney disease (CKD), and CV disease, a growing public health problem termed CV-kidney-metabolic (CKM) syndrome.^12,13 The cornerstone of CKM management is early and aggressive promotion of a healthy lifestyle, as enumerated in the American Heart Association’s (AHA’s) Life’s Essential 8 and including healthy diet, sufficient sleep, normal blood glucose, blood cholesterol level, blood pressure and body mass index, participation in physical activity, and avoidance of nicotine.² Although avoidance of excess sedentary behavior is not currently included as a cardinal element of the Essential 8, it has important effects on CKM health.²

Here, we examine nearly 90,000 UK Biobank participants who underwent prospective accelerometer-based physical activity measurement and behavior classification using a validated method to systematically examine associations between sedentary time and future risk of >700 incident diseases, with a special emphasis on key CKM conditions, while accounting for total physical activity volume.

Methods

Study population

The UK Biobank is a prospective cohort of approximately 500,000 adults aged 40 to 69 years at recruitment.¹⁴ Briefly, between 2006 to 2010, 9.2 million individuals living within 25 miles of 22 assessment centers in the United Kingdom were invited, and 5.4% participated in the baseline assessment. Questionnaires and physical measurements were collected at recruitment, and all participants were followed for outcomes through linkage to national health-related data sets. Participants provided written informed consent. The UK Biobank was approved by the UK Biobank Research Ethics Committee (reference 11/NW/0382). The use of UK Biobank data (application 17,488) was approved by the local Mass General Brigham Institutional Review Board.

Accelerometer-derived physical activity

Between February 2013-December 2015, 236,519 UK Biobank participants were invited to participate in the accelerometer substudy, and 103,695 (45%) ultimately submitted data from an Axivity AX3 (Newcastle upon Tyne) wrist-worn triaxial accelerometer.¹⁵ Participants were instructed to wear the accelerometer consistently for 1 week while performing their usual activities. The sensor captured acceleration over 7 days at 100 Hz with dynamic range ± 8g. Signals were calibrated to gravity and analyzed as previously described (Supplemental Methods).¹⁸ We excluded individuals with insufficient wear-time to support imputation (<72 h of wear-time or no wear data in each 1 h period of the 24 h cycle), whose signals were insufficient for calibration or movement classification, whose mean acceleration values were nonphysiologic, and who did not contribute a full week (Supplemental Figure 1).¹⁶

Raw accelerometer signals were categorized into one of 4 activity classes (sleep [nonwaking behavior], sedentary [waking behavior at ≤1.5 METs such as sitting work at a computer, watching television, eating indoors or outdoors], light activity [waking behavior at <3 METs not meeting the sedentary behavior definition such as cooking or self-care], or MVPA [waking behavior at ≥3 METs, such as walking the dog or jogging]) using a machine learning method developed with manually labeled behaviors based on simultaneously captured accelerometer, video and time-diary logs ,and previously validated in a UK-based sample (overall accuracy 88%).¹⁶ The primary exposure in the present analysis was overall sedentary time accumulated over the week and averaged as a daily rate (hours/day).

Life’s Essential 8 elements and additional exposures

The AHA’s Life’s Essential 8 includes smoking status, blood pressure, body weight, total cholesterol, blood glucose/glycosylated hemoglobin (HbA1c), physical activity, diet, and sleep. Variable categorizations (eg, for the calculation of population attributable fractions [PAFs]) and definitions are provided in Supplemental Tables 1 and 2.² As categorized previously, we note a distinction between insufficient physical activity (ie, <150 min/week of MVPA) and total sedentary time, in which individuals may meet the criteria for guideline-adherent physical activity (ie, >150 min/week of MVPA), yet exhibit high sedentary time.¹⁷

Outcomes

We defined diseases using v1.2 of the Phecode Map, a set of 1867 disease definitions arranged into meaningful groups and identified using standardized sets of International Classification of Disease, 9th and 10th revision codes.18, 19, 20 Diagnostic code sources included hospital data through linkage to national health-related data sets, as well as outpatient general practitioner visit data through linkage to electronic health records. Follow-up began at the conclusion of accelerometry and ended at the earliest of an outcome event, death, or the last follow-up (Supplemental Methods).

Statistical analyses

Associations between accelerometer-derived sedentary time (hours/day) and incident disease were assessed using a phenome-wide association study approach with Cox proportional hazards regression. Our aim was to broadly identify plausible associations between sedentary behavior and disease at scale, so we selected a uniform set of covariates for adjustment using prior literature.^16,21 Specifically, we adjusted for age, sex, racial and ethnic background, tobacco use, Townsend Deprivation Index, alcohol intake, educational attainment, employment status, self-reported health, and diet quality. To estimate the effect of sedentary behavior independent of physical activity, all models were also adjusted for MVPA (modeled as a spline term to allow for disease-specific nonlinear relations).²² The effect of sedentary time was estimated as the HR per hourly increment of sedentary time and associated P value. To quantify the robustness of identified associations to residual confounding, we calculated E-values. E-values represent the minimum strength of association on the risk ratio scale that potential confounders would need to have with the exposure and outcome to nullify the observed association.^21,22 To prevent model instability, only diseases with ≥120 events were tested (ie ≥10 events per variable in the primary model). Given our interest in associations between sedentary time and incident disease, we excluded conditions related to pregnancy and congenital anomalies. As performed previously in prior physical activity studies, significance thresholds were corrected for multiple comparisons to target a false discovery rate (FDR) of 0.01, which would be expected to result in only 1 false discovery per 100 associations detected.22, 23, 24

Given particularly strong associations between sedentary time and CKM conditions (ie, endocrine/metabolic, circulatory, and genitourinary/renal Phecode categories), we selected 4 key CKM conditions for more detailed assessment: hypertension, diabetes, CKD, and sleep apnea. These exemplar conditions were chosen based on prior associations with physical activity as well as established characterization as CKM-related conditions.^22,25 We assessed the functional form of the relationship between sedentary time and the 4 exemplar conditions by fitting multivariable-adjusted B-spline basis functions with 3 degrees of freedom with the reference value set at the median. Linearity was assessed by comparing the Akaike Information Criterion for models fitting sedentary time as a restricted cubic spline, quadratic-transformed, or linear term. We also quantified associations across quartiles of sedentary time with the second quartile (sedentary time 8.2-9.4 hours/day) as the referent to allow better visualization of nonlinear effects.^17,26 The proportional hazards assumption was validated by inspecting Schoenfeld residuals.

Given evidence of inflection of risk at 10.6 hours/day, we conducted several analyses with sedentary time stratified at a binary threshold of 10.6 hours/day, including quantifying the associations with continuous sedentary time (hours/day) above and below this threshold, while testing for interaction. Disease incidence rates and cumulative incidence curves were generated across strata of both MVPA and sedentary time (at the 10.6 hours/day threshold). To estimate the public health burden of excess sedentary behavior on CKM disease,²⁷ we calculated multivariable-adjusted average PAFs using the approach of Eide and Gefeller,²⁸ implemented in the “averisk” package (version 4.1.1.) in R.

Multiple secondary analyses were performed (Online Methods). All analyses were performed using R (version 4.1.1). Tail-area based FDR thresholds were derived using a modified Grenander distribution-based algorithm as described previously using the “fdrtool” (version 1.2.17) package in R. For all other analyses, a 2-tailed P < 0.05 was considered statistically significant.

Results

Baseline characteristics

Among 89,537 UK Biobank participants who underwent activity measurement for 1 week, the mean age was 63.4 years (SD: 7.8 years), and 56% were women. The median sedentary time was 9.4 hours/day (quartile-1:8.2, quartile-3:10.6). Baseline characteristics across quartiles of sedentary time are presented in Table 1.

Table 1.

Sample Characteristics

	Overall (N = 89,537)	Quartile 1 (8.2-9.4 h/d) (n = 22,042)	Quartile 2 (<8.2 h/d) (n = 22,825)	Quartile 3 (9.4-10.6 h/d) (n = 22,304)	Quartile 4 (>10.6 h/d) (n = 22,366)
Age (y)	62 (8)	63 (8)	62 (8)	62 (8)	62 (8)
Sex
Female	50,451 (56%)	13,561 (62%)	15,118 (66%)	12,194 (55%)	9,578 (43%)
Male	39,086 (44%)	8,481 (38%)	7,707 (34%)	10,110 (45%)	12,788 (57%)
Racea
Asian	1,024 (1.1%)	213 (1.0%)	284 (1.2%)	239 (1.1%)	288 (1.3%)
Black	729 (0.8%)	162 (0.7%)	179 (0.8%)	177 (0.8%)	211 (0.9%)
Other	949 (1.1%)	219 (1.0%)	259 (1.1%)	237 (1.1%)	234 (1.0%)
White	86,835 (97%)	21,448 (97%)	22,103 (97%)	21,651 (97%)	21,633 (97%)
Body mass index, kg/m2	26.7 (4.5)	26.2 (4.2)	25.6 (4.0)	26.9 (4.4)	28.1 (5.1)
Systolic blood pressure	137 (18)	136 (18)	136 (18)	137 (18)	138 (18)
Diastolic blood pressure	82 (10)	81 (10)	81 (10)	82 (10)	83 (10)
Antihypertensive	15,582 (17%)	3,611 (16%)	3,233 (14%)	3,949 (18%)	4,789 (21%)
Diabetesb	4,085 (4.6%)	832 (3.8%)	639 (2.8%)	1,028 (4.6%)	1,586 (7.1%)
Obesec	17,350 (19%)	3,510 (16%)	2,929 (13%)	4,449 (20%)	6,462 (29%)
Tobacco use
Current	5,960 (6.7%)	1,351 (6.1%)	1,375 (6.0%)	1,410 (6.3%)	1,824 (8.2%)
Former	32,185 (36%)	7,986 (36%)	7,997 (35%)	8,176 (37%)	8,026 (36%)
Never	51,392 (57%)	12,705 (58%)	13,453 (59%)	12,718 (57%)	12,516 (56%)
Alcohol intake (g/wk)d	122 (132)	120 (128)	115 (125)	124 (129)	130 (143)
Townsend Deprivation Index	−1.74 (2.81)	−1.86 (2.74)	−1.85 (2.73)	−1.75 (2.79)	−1.50 (2.96)
Self-reported health
Excellent	19,328 (22%)	4,953 (22%)	5,238 (23%)	4,900 (22%)	4,237 (19%)
Fair	14,159 (16%)	3,120 (14%)	3,163 (14%)	3,467 (16%)	4,409 (20%)
Good	53,717 (60%)	13,534 (61%)	13,975 (61%)	13,450 (60%)	12,758 (57%)
Poor	2,333 (2.6%)	435 (2.0%)	449 (2.0%)	487 (2.2%)	962 (4.3%)
Educational attainment	15 (5)	15 (5)	15 (5)	16 (5)	16 (5)
Employment status
Employed	54,269 (61%)	12,767 (58%)	13,399 (59%)	13,649 (61%)	14,454 (65%)
Unemployed/retired	35,268 (39%)	9,275 (42%)	9,426 (41%)	8,655 (39%)	7,912 (35%)
Diet
Healthy	15,875 (18%)	4,046 (18%)	4,565 (20%)	3,892 (17%)	3,372 (15%)
Intermediate	44,874 (50%)	11,213 (51%)	11,681 (51%)	11,112 (50%)	10,868 (49%)
Unhealthy	28,788 (32%)	6,783 (31%)	6,579 (29%)	7,300 (33%)	8,126 (36%)
Total weekly MVPA, min	289 (243)	308 (240)	365 (298)	272 (213)	210 (178)

Values are mean ± SD or n (%).

MVPA = moderate-to-vigorous physical activity.

^aRepresents self-reported racial and ethnic background. Race classification of other defined as self-report of a race other than Asian, Black, multiracial, multiethnic, or White.

^bIndicates covariates that likely lie in the causal pathway between sedentary behavior and cardiovascular events and are included in secondary adjusted models (see Supplemental Tables 10 and 11).

^cDefined as BMI ≥30 kg/m².

^dCan be converted to standard U.S. drinks per week by dividing by 14 g.

Associations of sedentary behavior with incident disease

After a median follow-up of 8.0 years (quartile-1: 7.5, quartile-3: 8.5), sedentary time was associated with risk of 81 of 761 incident diseases (10.6% of all conditions tested) at an FDR of 1% (Figure 1). Among significant associations, 75 (92.6%) indicated a higher risk of disease with greater sedentary time (HR: range 1.02 to 1.29 per 1-hour increase in sedentary time) (Supplemental Table 3). Associations spanned 15 of the 16 categories evaluated. Among associations indicating higher risk, endocrine/metabolic conditions were most common (n = 17 [22.7% of significant associations]), followed by circulatory (n = 12 [16.0%]), respiratory (n = 7 [9.3%]), and genitourinary/renal (n = 5 [6.7%]) (Figure 1). At the same time, we observed strong associations for a breadth of conditions including Parkinson Disease (HR: 1.29; 95% CI: 1.21-1.36; P = 1.9 × 10⁻¹⁷) and bacterial infection (HR: 1.07; 95% CI: 1.05-1.10; P = 1.3 × 10⁻⁷). A modest number of associations with lower disease risk (n = 6) primarily represented selected circulatory diseases (eg varicose veins) (Supplemental Table 3). The top 5 strongest statistical associations (ie, lowest P values) within each category, including individual effect sizes and E-values, are provided in Supplemental Table 4.

Figure 1.

Associations Between Sedentary Time and Incident Disease

(A) Depicted is the negative log₁₀P value for the association between accelerometer-measured daily sedentary time and incident disease (grouped by category on the x-axis in Cox proportional hazards models adjusted for age, sex, racial and ethnic background, tobacco use, Townsend Deprivation Index, alcohol intake, educational attainment, employment status, self-reported health, diet quality, and weekly moderate-to-vigorous physical activity). Darker shaded points represent associations meeting significance at a false discovery rate (FDR) of 1% (horizontal red line). Upward facing triangles represent higher risk (HRs >1), whereas downward-facing triangles represent lower risk (HR <1). P values smaller than 1 × 10⁻²⁰ are displayed as 1 × 10⁻²⁰ for graphical purposes. Only log₁₀P values > 15 or the highest log₁₀P value within each category are annotated. (B) Depicted is the proportion of diseases within each category with a significant association indicating higher risk (ie, HRs greater than 1) with increasing sedentary time. Only significant associations at FDR of 1% are depicted.

Associations between sedentary behavior and cardiovascular-kidney-metabolic conditions

Across >700 diseases tested, associations between sedentary time and CKM conditions were particularly prominent (n = 36 [44% of significant associations]) (Figure 1 and Central Illustration, Supplemental Tables 3 to 6). For example, sedentary time was associated with a higher risk of hypertension (HR: 1.04 per 1-hour increase; 95% CI: 1.03-1.05; P = 6.9 × 10⁻¹²), diabetes (HR: 1.09; 95% CI: 1.07-1.12; P = 6.7 × 10⁻¹⁵), CKD (HR: 1.05; 95% CI: 1.02-1.07; P = 1.2 × 10⁻⁴), and sleep apnea (HR: 1.13; 95% CI: 1.09-1.17; P = 2.8 × 10⁻¹⁰). Using the second quartile (8.2-9.6 hours/day) as a referent, sedentary time in the top quartile (>10.6 hours/day) was associated with substantially higher risks of incident hypertension (HR: 1.13; 95% CI: 1.08-1.19), diabetes (HR: 1.18; 95% CI: 1.07-1.31), CKD (HR: 1.14; 95% CI: 1.02-1.27), and sleep apnea (HR: 1.19; 95% CI: 0.99-1.42) (Figure 2, Supplemental Figure 2). Associations between sedentary time and key CKM conditions did not deviate statistically from nonlinearity (P linearity >0.05 for each), although the effects of additional sedentary time did appear consistently greater above the >10.6 hour/day threshold (HR range: 1.05-1.21) vs below (HR range: 1.03-1.09) (Supplemental Figure 3). There was also a dose-response relationship between the number of days the >10.6-hour sedentary threshold was exceeded and risks of incident hypertension (HR: 1.04 per 1-day increase; 95% CI: 1.03-1.05), diabetes (HR: 1.08; 95% CI: 1.06-1.10), CKD (HR: 1.06; 95% CI: 1.04-1.08), and sleep apnea (HR: 1.10; 95% CI: 1.07-1.13) (Supplemental Figure 4).

Central Illustration.

Device-Measured Sedentary Behavior and Future Health Risk

Depicted is an overview of the study design and the key study findings. BMI = body mass index; BP = blood pressure; CKM = cardiovascular-kidney-metabolic; HDL = high-density lipoprotein; NOS = not otherwise specified.

Figure 2.

Risk of Cardiovascular-Kidney-Metabolic Diseases According to Quartiles of Sedentary Time

Depicted are multivariable-adjusted HRs for associations between sedentary time and key cardiovascular-kidney-metabolic conditions. Quartiles 1 (<8.2 hours/day), 3 (9.4-10.6 hours/day), and 4 (>10.6 hours/day) are compared to quartile 2 (8.2-9.4 hours/day). The models were adjusted for age, sex, racial and ethnic background, tobacco use, Townsend Deprivation Index, alcohol intake, educational attainment, employment status, self-reported health, and diet quality. E-values estimate the minimum strength of association on the risk ratio scale that potential confounders would need to have with the exposure and outcome to nullify the observed association. E_point denotes E-values for the point estimate and E_null denotes the E-value for the confidence interval bound closest to the null.

Public health burden of sedentary behavior on cardiovascular-kidney-metabolic health

Using multivariable-adjusted PAFs, the estimated burden of sedentary behavior was greatest for sleep apnea (PAF: 4.2%; 95% CI 2.3%-6.6%), followed by CKD (PAF: 1.9%; 95% CI: 1.2%-2.5%), hypertension (PAF: 1.6%; 95% CI: 1.2%-1.9%), and diabetes (PAF: 1.3%; 95% CI: 0.5%-2.7%) (Figure 3). When considered alongside the components of Life’s Essential 8, excess sedentary behavior was consistently in the top 6 contributors, generally demonstrating higher PAF compared to sleep or diet.

Figure 3.

Population-Attributable Fractions for Sedentary Behavior, Life’s Essential-8 for Cardiovascular-Kidney-Metabolic Conditions

Average population-attributable fractions (PAFs) were estimated using the method of Eide and Gefeller (see text). Here, the average PAF is an estimate of the proportion of incident disease that may be prevented by achieving a desirable value of a given risk factor (eg, limiting sedentary time to ≤10.6 hours/day). Estimates for individual risk factors were truncated at a lower limit of 0, as negative PAFs are not possible. PAF point estimates are provided above each bar. BMI = body mass index; BP = blood pressure; HDL = high-density lipoprotein.

Relations between sedentary behavior, physical activity, and cardiovascular-kidney-metabolic health

Among individuals with sedentary time >10.6 hours/day, event rates were lower in those achieving guideline-recommended levels of MVPA (Figure 4). However, associations between sedentary time and future risk of CKM conditions persisted even in the presence of guideline-recommended MVPA levels (Supplemental Figure 5). Specifically, the effects of >10.6 hours/day of sedentary time persisted whether MVPA was ≥150 min/week (CKD: HR: 1.10 [0.96-1.27]; diabetes: HR: 1.19 [1.06-1.34]; hypertension: HR: 1.15 [1.08-1.21]; and sleep apnea: HR: 1.41 [1.16-1.71]) vs <150 min/week (CKD: HR: 1.26 [1.12-1.41]; diabetes: HR: 1.30 [1.17-1.44]; hypertension: HR: 1.20 [1.13-1.28]; and sleep apnea: HR: 1.35 [1.13-1.63]).

Figure 4.

Incidence of Cardiovascular-Kidney-Metabolic Conditions by Sedentary Behavior and Moderate-to-Vigorous Physical Activity

Depicted are plots of crude cumulative risk of 4 exemplar cardiovascular-kidney-metabolic conditions, stratified by MVPA guideline recommendations (active, ≥150 minutes MVPA per week; inactive, <150 minutes MVPA per week) and daily sedentary time (sedentary, >10.6 hours per day; not sedentary, ≤10.6 hours per day). The sedentary group is depicted in red, the nonsedentary group in green.

Effects of reallocating sedentary time on cardiovascular-kidney-metabolic health

We estimated the effects of reallocating sedentary to other activities on the risk of developing key CKM conditions. Reallocation of sedentary time to all activities proportionally was associated with lower risk of hypertension (eg, 1-hour decrease in sedentary time HR: 0.95; 95% CI: 0.94-0.97), diabetes (HR: 0.91; 95% CI: 0.89-0.93), CKD (HR: 0.97; 95% CI: 0.95-0.99), and sleep apnea (HR: 0.86; 95% CI: 0.83-0.89) (Supplemental Figure 6).

Secondary analyses

There were numerically fewer associations: 1) using the more stringent Bonferroni correction threshold of P = 6.57 × 10⁻⁵ (43 total associations, 38 [88%] with higher risk of disease) (Supplemental Figure 7); 2) with the institution of a 2-year blanking period (65 total associations, 61 [93.8%] with higher risk of disease) (Supplemental Figure 8, Supplemental Table 7); 3) among only individuals with “good to excellent” self-reported health (37 total associations) (Supplemental Figure 9, Supplemental Table 8); and 4) among only individuals without prevalent musculoskeletal conditions (51 total associations) (Supplemental Figure 10, Supplemental Table 9), but associations with key CKM conditions persisted in each case and with similar effect sizes (Supplemental Table 10).

We assessed models with additional adjustment for potential mediators (ie, body mass index, blood pressure, and antihypertensive use). Results were similar with adjustment for blood pressure and antihypertensive use (73 total associations) (Supplemental Table 10, Supplemental Figure 10), whereas there were numerically fewer associations with adjustment for body mass index (37 associations) (Supplemental Table 11, Supplemental Figure 11). Although associations remained significant, effect sizes for CKM conditions were smaller after adjustment for body mass index (eg, hypertension HR: 1.01 per 1-hour increase in sedentary time; 95% CI: 1.00-1.02; P = 0.03; diabetes HR: 1.04; 95% CI: 1.02-1.06; P = 9.4 × 10⁻⁴; CKD HR: 1.02; 95% CI: 1.00-1.05; P = 0.05; sleep apnea HR: 1.06; 95% CI: 1.02-1.10; P = 3.0 × 10⁻³) which also showed evidence of substantial mediation (Supplemental Table 12).

Discussion

Among 89,537 community-based middle-aged and older adults wearing a wrist-worn activity tracker for 1 week, we examined associations between device-measured sedentary behavior and future risk of >700 diseases. Sedentary behavior was associated with a higher risk of 75 conditions spanning the full spectrum of human disease. Associations with CKM disease were particularly prominent, accounting for nearly half of the associations identified. For such conditions, sedentary behavior >10.6 hours/day, observed in one-quarter of the sample, was consistently and strongly associated with higher risk, even after adjustment for MVPA. Using PAFs, we estimate that ∼1 to 4% of population-level CKM disease may be attributable to excess sedentary time, representing disease burden estimates rivaling key components of the AHA’s Life’s Essential 8.

Our findings extend prior assessments of the role of sedentary behavior in the development of disease.^6,7,9,10,26,29, 30, 31 The relatively narrow focus of prior efforts (ie, 1 or a few conditions) has precluded a comprehensive understanding of the composite risks associated with sedentary behavior, the diseases for which optimization may be expected to have the greatest impact, and the volumes of sedentary behavior that may be most relevant across varying conditions, when assessed comparatively in a single large sample. Furthermore, most studies have relied on self-reported behavior, which is prone to misclassification and may overestimate sedentary behavior by up to 60%.32, 33, 34 Here, we leverage a unique resource of prospectively collected device-measured movement data classified using contemporary machine-learning algorithms with sufficient statistical power to enable systematic association testing with >700 future conditions spanning the full spectrum of health and disease, with a special emphasis on key CKM conditions. We highlight 3 key implications relevant for future clinical and public health efforts.

First, optimization of sedentary behavior may have broad-ranging benefits on disease risk extending far beyond the common cardiometabolic diseases emphasized in current guidelines.^9,35 After controlling for potential confounders including health status, socioeconomic factors, and total MVPA, we found that sedentary behavior was associated with a greater risk of >70 conditions spanning 15 of 16 categories tested. Importantly, many associations are supported by preclinical, observational, and experimental evidence. For example, associations between sedentary behavior and central nervous system diseases (eg, Parkinson Disease)^26,30 have been linked to reductions in cerebral blood flow, altered glucose utilization, inflammation, and impaired autoregulation.¹⁰ Therefore, our findings add support for the key role for movement behaviors in promoting brain health,³⁶ highlighting the specific importance of sedentary behavior. Although we acknowledge that some associations may be susceptible to reverse causation, most associations remained robust with institution of a 2-year blanking period, including all associations with key CKM conditions. Overall, our findings demonstrate the potential wide-ranging adverse impact of sedentary behavior on disease risk and prioritize key conditions meriting further study to elucidate potential mechanisms.

Second, efforts to optimize sedentary time may have greatest impact on CKM health, for which a 10.6-hour threshold may be relevant. In the present study, diseases related to CKM health were over-represented (36/81) among conditions associated with sedentary behavior. We consistently observed substantially higher risk of key CKM conditions with sedentary time in excess of 10.6 hours per day (4th quartile), which aligns with prior work reporting increasing disease risk with sedentary time over ∼10 hours/day.^11,17,37 Although our study is observational and cannot establish a causal relation between sedentary behavior and CKM disease, prior findings support the plausibility of a causal effect,⁷ including temporality and dose-responsiveness,^11,17 supportive human experimental evidence,^10,38 and biological plausibility.^10,39 Future mechanistic evaluation is needed to clarify potential pathways by which sedentary behavior may affect the risk of disease and CKM-related pathology, which may include potentiation of inflammation and oxidative stress, increased blood lipids, higher sympathetic nervous system tone, reduced insulin sensitivity, and altered fatty acid and glucose utilization.^10,38 If a causal relationship exists, our results imply the public health effects on CKM may be appreciable, with excess sedentary time accounting for 1% to 4% of the total burden of hypertension, diabetes, CKD, and sleep apnea, estimates rivaling certain components of the AHA’s Life’s Essential 8 rubric. Overall, our findings suggest that a greater emphasis should be placed on the role of sedentary behavior in the development of CKM diseases, and identify avoidance of sedentary time in excess of 10.6-hours per day as a potentially reasonable minimal target. Optimization of sedentary behavior merits consideration as a cardinal element of the AHA’s Life’s Essential prevention framework.²

Third, avoidance of excess sedentary behavior appears relevant for CKM health even in the context of adequate MVPA. We observed associations between sedentary time and >70 future diseases despite adjustment for total MVPA. Furthermore, associations with CKM disease were substantial whether or not individuals met guideline-based MVPA levels (ie, ≥150 min/week). Our findings align with prior studies demonstrating that reallocating time from sedentary behavior to other activities was estimated to translate to significantly lower body mass index, waist circumference, blood cholesterol, triglyceride levels, HbA1c, systolic, and diastolic blood pressure.^40,41 Several small trials focused on reallocating sedentary time to other activities, including MVPA, have also demonstrated improvements in HbA1c levels and blood pressure.^42,43 On balance, our results should compel future investigation focused on the formulation and assessment of combined public health interventions emphasizing the importance of optimizing sedentary behavior and achieving guideline-recommended levels of MVPA as separate important targets.

Study limitations

Several limitations should be considered when interpreting our work. First, sedentary behavior was assessed over a single 1-week period and, longitudinal follow-up (median 8 years), although favorable to most accelerometer studies,⁴⁴ may not have been sufficient to detect associations with less common conditions. However, prior studies support the use of 7-day accelerometer measurements as reasonable surrogates for longer-term habitual activity.⁴⁵ Second, our findings are observational, and causality cannot be inferred. However, we note that we accounted for multiple confounders and performed secondary analyses to assess for potential reverse causality (eg, 2-year blanking period), in which our findings remained robust. Reassuringly, E-values suggest that large unadjusted confounding effects would need to be present to nullify our main findings but we acknowledge that E-values do not account for selection bias or exposure measurement error. Third, as performed previously, we defined diseases using Phecodes,^22,23 which are a systematic framework for simultaneous assessment of hundreds of conditions. Given the breadth of our analysis, it would be impractical to have used curated disease definitions, although we acknowledge that Phecodes may be less accurate for specific conditions. Fourth, the UK Biobank accelerometer subsample is subject to selection bias, resulting in a predominantly White sample which is enriched generally for health and socioeconomic status. Selection effects are relevant to the inflection of risk we observed at 10.6 hours/day of sedentary time, as thresholds may vary in different populations. However, prior UK Biobank analyses have demonstrated disease incidence rates similar to other relatively healthy research samples,²² supported the validity of exposure-outcome associations,⁴⁶ and provided previous insights pertaining to measured physical activity which have replicated in independent cohorts.⁴⁷ Fifth, to further investigate the effect of substituting sedentary time for other activities, we performed compositional data analysis focused on our key CKM conditions. Given our goal of performing a broad survey of conditions associated with sedentary time, we did not extend this approach to our phenome-wide analysis. Sixth, the machine learning activity classification algorithm has been validated with good performance (overall accuracy 88%), but its accuracy for specific comparisons (eg, sedentary vs light activity) may be lower. However, these data have served as the basis for a number of subsequent phenotypic studies providing expected exposure-outcome relations,^23,26,44,48, 49, 50 as well as genotypic analyses identifying biologically plausible loci, providing complementary evidence of validity.⁵¹

Conclusions

Over 8 years of follow-up, greater measured sedentary behavior was associated with higher risk of 75 future diseases, especially CKM conditions. Sedentary time in excess of 10.6 hours/day was strongly associated with CKM disease risk regardless of whether or not individuals also achieved guideline-adherent MVPA levels (≥150 min/week). The estimated population-level burden of CKM disease attributable to excess sedentary behavior rivals those of consensus lifestyle measures (eg, adequate sleep, healthy diet). Optimization of sedentary behavior should be a key public health priority separate from adequate MVPA.

Perspectives.

COMPETENCY IN SYSTEMS-BASED PRACTICE: In a study of approximately 90,000 UK Biobank participants wearing a wrist-worn activity tracker for 1 week, greater sedentary time was associated with a higher risk of 75 conditions, especially CKM conditions, independent of MVPA levels. The estimated population-level burden of CKM disease attributable to excess sedentary behavior rivals those of consensus lifestyle measures (eg, adequate sleep, healthy diet).

TRANSLATIONAL OUTLOOK: Sedentary behavior is a risk factor for future adverse health outcomes independently of MVPA. Optimization of sedentary behavior should be a key public health priority separate from adequate MVPA.

Funding support and author disclosures

Dr Ajufo is supported by the John S. LaDue Memorial Fellowship in Cardiovascular Medicine or Vascular Biology grant. Dr Kany is supported by the Walter Benjamin Fellowship from the Deutsche Forschungsgemeinschaft (521832260). Dr Rämö is supported by a research fellowship from the Sigrid Jusélius Foundation. Dr Churchill is supported by the National Institutes of Health (K23HL159262-01A1). Dr Guseh is supported by the American Heart Association (19AMFDP34990046; SFRN awards 24SFRNCCN1276092, 25ISFRNCP1501445, 24SFRNPCN1284389). Dr Aragam is supported by grants from the National Institutes of Health (1K08HL153937 and R01HL174912) and the American Heart Association (862032). Dr Ellinor is supported by grants from the National Institutes of Health (RO1HL092577, R01HL157635), the American Heart Association (961045), the European Union (MAESTRIA 965286), and the Foundation Leducq (24CVD01). Dr Khurshid is supported by the NIH (K23HL169839-01) and the American Heart Association (2023CDA1050571). Dr Aragam receives sponsored research support from Sarepta Therapeutics, Bayer AG, and Foresite Labs. Dr Guseh is a co-inventor on a pending patent related to ECG-based deep learning methods to predict peak VO₂ and cardiovascular outcomes, and serves as team cardiologist for the New England Patriots and New England Revolution. Dr Ellinor receives sponsored research support from Bayer AG, IBM Research, Bristol Myers Squibb, Pfizer and Novo Nordisk; he has also served on advisory boards and/or consulted for Bayer AG. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.