The Dose-Response Associations of Sedentary Time with Chronic Diseases and the Risk for All-Cause Mortality Affected by Different Health Status: A Systematic Review and Meta-Analysis

Abstract

Purpose

To determine the dose-response associations of sedentary behaviour with cardiovascular diseases (CVD), cancer, and all-cause mortality, and to examine whether the sedentary-associated all-cause mortality risk was affected by appearance of diabetes and hypertension, physical activity, and body mass index (BMI).

Design

We carried out a systematic review and meta-analysis to search Medline, SportDiscus, and Web of Science for eligible studies.

Settings

Prospective cohort studies that reported sedentary time and CVD, cancer, and mortality incidents.

Measurements

Two authors independently extracted data based on predefined criteria. The effect estimates were evaluated by hazard ratios (HRs) with 95% confidences (CIs).

Results

Twenty-four studies met the inclusion criteria. Sitting time showed dose-response associations with CVD, cancer, and all-cause mortality, with each 1-hour increment of sitting time daily accounting for HRs 1.04 (95% CIs 1.02–1.07), 1.01 (1.00–1.02), and 1.03 (1.02–1.03), respectively. The link between sitting time and CVD and all-cause mortality was non-linear (p_non-linear < 0.0001). The relationship between TV viewing and CVD and all-cause mortality was dose-dependent, with HRs 1.07 (1.06–1.09) and 1.04 (1.01–1.06) for per 1-hour increment of TV time every day, respectively. The regression was curved (p_non-linear < 0.0001). When the analysis was stratified by the percentage of diabetes and hypertension, BMI values, and physical activity levels, we found that higher BMI and a greater percentage of diabetes and hypertension further increased all-cause mortality risk in the most sedentary populations, whereas higher physical activity levels decreased it.

Conclusion

Sitting time and TV viewing significantly increased cardiovascular, cancer, and mortality risk; the associations were dose-dependent. More importantly, sedentary behaviour in combination with chronic diseases or high BMI increased all-cause mortality risk whereas physical activity was likely to alleviate the adverse associations.

Introduction

Sedentary behaviour is frequently defined as any waking sitting or lying behaviour with a low energy expenditure—usually less than 1.5 metabolic equivalent units (1, 2). Therefore, sedentary behaviour typically refers to sitting or lying behaviour rather than a simple absence of moderate-to-vigorous intensity physical activity (MVPA) (1). Sedentary behaviour is popular in modern society and constitutes a major health problem for adults' health.

Compelling evidence has demonstrated that both sitting time and watching TV pose an obvious risk for chronic diseases; e.g. type two diabetes mellitus (3, 4), cardiovascular diseases (CVD) (5, 6), cancer (7, 8), and increased mortality risk (9, 10, 11). Although the qualitative and quantitative aspects of the relationship between sedentary behaviour and mortality have been investigated in previous meta-analyses (12, 13, 14), considerable questions remain unaddressed. A major concern is that it is still unknown whether sedentary-associated mortality increases or decreases in association with diabetes and hypertension, body mass index (BMI), and physical activity. Elucidation of these issues is critical for adults to manage their mortality risk by modifying sedentary behaviour, lifestyles, and health status. Consequently, we performed a systematic review of studies with similar interests and applied dose-response meta-analysis methods to examine the associations between sitting time and TV viewing with cardiovascular, cancer, and all-cause mortality. We further determined the effect of the appearance of diabetes and hypertension, BMI, and physical activity levels on the associations between sedentary time and mortality in the most sedentary of adults.

Methods

Search Strategy

This meta-analysis was conducted in accordance with PRISMA recommendations and the criteria of the reporting of meta-analysis guideline (15). Electronic database searches were conducted in PubMed, SportDiscus, and Web of Science for articles published up to 1 May 2019. Terms used for database searching included (1) sedentary, sitting time, TV viewing/time, screen time; AND (2) mortality, death, fatal; AND (3) prospective study, follow-up study, longitudinal study (Supplementary materials, Text S1). No language restriction was applied. We also checked the reference lists of articles and conducted a forward search.

Inclusion Criteria

We only included prospective cohort studies that conducted a survey of sedentary behaviour and reported cardiovascular, cancer, and mortality incidents. The included studies had to meet the following criteria: (1) prospective or follow-up studies; (2) participants aged ≥ 18 years; (3) sitting time and TV viewing measures ≥ 3 categories; and (4) the number of CVD, cancer, and mortalities presented according to the categories of sitting time or TV viewing.

Data Extraction

Two review authors (WB and YC) independently extracted data from each included study and resolved any disagreement through discussion with a senior expert in the field. We contacted the corresponding authors to acquire the relevant information if necessary. The extracted details included the following: name of the first author, study location, the source and the number of participants, mean or median age of participants, the respective proportions of women and men, follow-up time, cardiovascular, cancer and all-cause mortality, the methods of ascertainment, assessment details for sitting time, adjustment, and adjusted hazard ratios (HRs). The data extraction followed the methods recommended by the Cochrane Reviewers' Handbook (16).

Risk of Bias Assessment

The quality of the included studies was evaluated independently by two authors (WB and YC) using the New Castle-Ottawa Quality Assessment Scale assessment tools (17). Each included study was ranked as high, moderate, and low quality according to the score values ≥ 7, ≥ 5–6, and < 5, respectively.

Statistical Analyses

The primary outcome of our study was risk of CVD, cancer, and mortality, and the effect estimates were evaluated by HRs with 95% confidences (CIs). Because the studies were conducted in different countries, heterogeneity between studies might exist; therefore, random-effects models were used to evaluate effect sizes. The adjusted HRs was used to conduct the meta-analysis. If only adjusted risk ratios (RRs) were given, we designated the RRs as the HRs. If studies did not report the person-years for each category of sedentary time, we calculated the separate person-years by multiplying the total person-years by the percent of populations in each category or by multiplying the number of participants in each category by follow-up years. For each of the included studies, the reported median or mean value of sitting time or TV viewing in each category was designated as the sitting time or TV viewing category.

We performed the dose-response meta-analysis employing methods described by Greenland and Longnecker (18), and the study-specific slopes (linear trends), and 95% CIs were calculated from the natural logs of the HR and 95% CIs across categories of sitting time and TV viewing. The potential non-linear dose-response associations of mortality with sitting time and TV viewing were estimated by using fractional polynomial models.

We further determined whether the mortality risk increased or decreased with changes in the percentage of diabetes or hypertension, BMI, or physical activity levels. To do this, we extracted the baseline data of diabetes, hypertension, BMI measures, and physical activity levels from the most sedentary group in each included study, and categorized those data into three groups, low, moderate, and high, respectively. We then determined HRs and 95% CIs across various levels of the percentage of diabetes (≤ 7%, 7.1–11.9%, and > 12%) and hypertension (≤ 40%, 41–44%, and ≥ 45%), BMI (Sitting time: ≤25.29 kg/m², 25.6–26.99 kg/m², and ≥27 kg/m²; TV: ≤ 26.9 kg/m², 27–28.9 kg/m², and ≥ 29 kg/m²), and physical activity levels (≤ 10 MET•h/w, 10.1–11.9 MET•h/w, and ≥ 20 MET•h/w). Subgroup analyses were only conducted for associations of sitting time and TV viewing with all-cause mortality.

The heterogeneity of results between the studies was determined using Chi² test (with a significant level at p < 0.10) and I² (low: I² < 30%; moderate: 30%≤ I² <60%; and high: I² ≥ 60%). A two-tailed p-value < 0.05 was considered significant. All statistical analyses were performed using STATA (Version 15, StataCorp LP) and MATLAB (R2015b, The Mathworks, Inc) software.

Sensitivity and Publication Bias Analyses

To conduct the sensitivity analysis, we tested the associations of sitting time and TV viewing with cardiovascular, cancer, and mortality risk by dropping the studies with extreme outcomes (lying outside the 95% CIs limit of summary estimated intervention effects). We performed Begg's test for all outcomes to explore the small-study effects.

Results

Characteristics of the Included Studies

Database searches identified and screened 5025 potential abstracts; of these, 4938 were excluded because they were either unrelated to the specific topic or duplicate studies from different databases (Figure 1). Eighty-seven full-text articles were then reviewed for eligibility; 63 studies were excluded for the following reasons: (1) categories of sitting time or TV viewing ≤ 3; (2) no mortality data; (3) no sedentary-stratified physical activity or BMI data; and (4) other reasons: e.g. cross-sectional studies, data on deaths not according to the categories of sedentary time, and reviews. Finally, twenty-four prospective studies met the inclusion criteria, of which 16 studies measured summary sitting time and 10 reported TV viewing (Table 1). These studies included 1156400 participants aged 18–99 years, who were followed-up for 2–15.7 years. During this period, 151810 died (13.1%). Included studies were conducted in the US, UK, Denmark, Norway, Canada, Spain, Japan, and Australia (Table 1). Other information regarding the participants in each study is listed in Table 1.

Figure 1.

Flow chart of the study selection

Table 1.

Characteristics of included studies

Author, year and country	Study	Participants	Mortality outcomes (N)	Variables adjusted for	QS
Bjork Petersen et al., 2014(5), Denmark	Danish Health Examination Survey, 5.4-year follow-up	71363 adults, aged 18–99 years	ST: All-cause, 1074	Age, sex, educational level, smoking, BMI, alcohol consumption, and hypertension	8
Campbell et al., 2013(39), US	Cancer Prevention Study-II, follow-up for 6.8 years	2293 men and women	ST: CVD, 145; cancer, 369; all-cause, 816	Age, sex, smoking status, BMI, red meat intake, surveillance, physical activity, and education	6
Cao et al., 2015(40), US	The HPFS cohort study, follow-up for 22 years	926 adults, aged 40–75 years	TV: cancer, 169; all-cause, 471	Age, stage of disease, grade of differentiation, smoking status, alcohol, folate, calcium, red meat intake, and energy intake	7
Chau et al., 2015(10), Norway	The Nord-Trøndelag Health Study, follow-up for 3.3 years	50817 adults, aged ≥20 years	ST: all-cause, 640. TV: all-cause, 684	Sex, BMI, education level, PA, smoking status, general health status, cardiometabolic disease status	7
Dunstan et al., 2010(41), Australia	The Australian Diabetes, Obesity and Lifestyle Study, 6.6-year follow-up	8800 men and women aged ≥25 years	TV: CVD, 87; cancer, 125; all-cause, 284	Age, sex, education, BMI, smoking, energy intake, alcohol, hypertension, total cholesterol, triglycerides, glucose tolerance, and diabetes	8
Ford et al., 2012(42), US	The NHANES study, follow-up 5.8 years	7350 men and women, aged ≥20 years	TV: All-cause, 542	Age, sex, race education, smoking, and Healthy Eating Index Score	7
George et al., 2013,(11) US	The HEAL cohort study, follow-up for 7 years	687 women, aged ≥18 years	TV: all-cause, 89	MVPA, race, menopausal status, treatment, tamoxifen, number of comorbidities, and BMI	7
Hagger-Johnson et al., 2016(43), UK	The UK Women's Cohort Study, follow-up for 12 years	12778 women, aged 37–78 years	ST: all-cause, 577	Age, chronic disease, PA, smoking, alcohol use, daily fruit consumption, sleep, education, occupation	8
Ikehara et al., 2015(44), Japan	The JACC Study, follow-up for 19.2 yeas	35959 men and 49940 women, aged 40–79 years	TV: CVD, 5403	Age, BMI, smoking, ethanol intake, education, PA, sleep duration, mental stress, presence of job, histories of hypertension and diabetes	8
Inoue et al., 2008,(45) Japan	Japan Public Health Centre-based Prospective Study, 8.7-year follow-up	83034 men and women, aged 45–74 years	ST: CVD, 974; cancer, 2044; All-cause, 4564;	Age, sex, geographical area, occupation, history of diabetes, smoking, alcohol consumption, BMI, and total energy intake	7
Katzmarzyk et al., 2009(46), Canada	Canada Fitness Survey, follow-up for 8 years	17013 adults aged 18–90 years	ST: CVD, 759; cancer, 547; All-cause, 1832	Age, sex, smoking, and alcohol consumption	8
Kim et al., 2013(47), US	Multiethnic Cohort Study, follow-up for 13.7 years	134596 men and women aged 45–75 years	ST: CVD, 6535; cancer, 6697; All-cause, 19143.	Age, sex, race or ethnic origin, education, smoking history, history of diabetes or hypertension, energy intake, and alcohol consumption	7
Leon-Munoz et al., 2013(48), Spain	A 2-year follow-up prospective study	2635 adults aged ≥60 years	ST: All-cause, 846	Age, sex, education, BMI, smoking, alcohol consumption, weight, BMI, chronic disease, morbidity limitations	7
Martinez-Gomez et al., 2016(49), Spain	The UAM cohort study, follow-up for 8.7 years	2470 people, aged ≥60 years	ST: all-cause, 982	Age, sex, education, BMI, waist circumference, systolic blood pressure, smoking, alcohol, CVD, cancer, diabetes, depression, and PA	7
Matthews et al., 2012(50), US	NIH-AARP Diet and Health Study, follow-up for 8.5 years	240814 adults aged 50–71 years	ST: CVD, 4684; cancer, 7652; all-cause, 17044. TV: CVD, 4684; cancer, 7652; all-cause, 17044;	Age, sex, race, education, BMI, smoking, and diet	8
Matthews et al., 2014(51), US	Southern Community Cohort Study, follow-up for 6.4 years	63308 adults aged 40–79 years	ST: CVD, 1376; all-cause, 5007;	Age, sex, race, education, income, marital status, occupation, comorbidity, alcohol intake, smoking, and BMI	8
Patel et al., 2010(52), US	The CPS-II Nutrition Cohort study, follow-up for 14 years	123216 adults aged 50–74 years	ST: CVD 6369; cancer, 6989; all-cause, 19230;	Age, sex, race, education, BMI, alcohol consumption, smoking status, marital status, total energy intake, and comorbidity	8
Pavey et al., 2015(53), Australia	The Australian Longitudinal Study on Women's Health, follow-up for 6 years	6656 participants aged 76–81 years	ST: all-cause mortality, 1364	Age, education, marital status, area, smoking, alcohol consumption, BMI, PA, chronic conditions, health and assistance	6
Pulsford et al., 2015(54), UK	The Whitehall II study, follow-up for 15.7 years	5132 participants aged 35–55 years	ST: all-cause, 443	Age, gender, employment grade, ethnicity, smoking, alcohol consumption, food, BMI, physical functioning, walking time and MVPA	7
Seguin et al., 2014(55), US	The WHI OS&ES study, follow-up for 12 years	92234 women aged 50–79 years	ST: CVD, 3878; cancer, 4759; all-cause, 10243	Age, BMI, race/ethnicity, PA, physical function, and history of chronic diseases	8
Stamatakis et al., 2011(56), UK	The 2003 Scottish Health Survey, 4.3-year follow-up	4512 adults aged ≥ 35 years	TV: all-cause, 325	Age, sex, BMI, smoking, marital status, ethnicity, social class, illness, PA, diabetes and hypertension, MVPA	8
Ukawa et al., 2014(57), Japan	The JACC Study, follow-up for 19.2 years	69752 adults, aged	TV, cancer, 267	Age, study area, smoking, alcohol and coffee consumption, BMI, education, marital status, and a history of diseases	7
van der Ploeg et al., 2012, Australia	The 45 and Up Study, follow-up for 2.8 years	222497 adults, aged ≥45 years	ST: all-cause, 5405	Age, sex, education, urban or rural residence, BMI, marital status, smoking, self-rated health	7
Warren et al., 2010(58), US	The ACLS study, follow-up for 21 years	7744 men, aged 20–89 years	TV: CVD, 377	Age, PA, current smoker, alcohol intake, BMI, family history of CVD, hypertension, diabetes	7

Note. ST: sitting time; CVD: cardiovascular disease; N: number of mortalities; MV(PA): moderate-to-vigorous intensity (physical activity); BMI: body mass index; QS: quality score.

Meta-analysis

We first examined the quantified aspects of the associations between sedentary time and cardiovascular, cancer, and mortality risk. Sitting time showed a dose-response relationship with CVD, cancer and all-cause mortality, each 1-hour increment of sitting time daily accounted for HRs of 1.04 (95% CIs 1.02–1.07), 1.01 (1.00–1.02), and 1.03 (1.02–1.03), respectively. Further analysis indicated that the relationship between sitting time and CVD (p_non-linear < 0.0001) and all-cause mortality (p_non-linear < 0.0001) was non-linear (Figure 2A/B). The relationship between TV viewing and CVD and all-cause mortality was dose-dependent, with per 1-hour increment of TV time each day increasing HRs of 1.07 (1.06–1.09) and 1.04 (1.01–1.06), respectively. The regression was curved, p_non-linear < 0.0001 and p_non-linear < 0.0001, respectively (Figure 2C/D).

Figure 2.

Non-linear regression of the associations of sitting time (ST) and TV viewing with hazard ratio (HR) of mortality and 95% confidence interval (CIs)

h: hour; CVD: cardiovascular diseases

To determine the effect of participants' characteristics on the relationship between sitting time and mortality risk, we analysed the effects sizes stratified by ranking the percentage of diabetic and hypertensive populations, BMI values, and physical activity levels. The mortality risk increased proportionally to the increment of the percentage of diabetes and hypertension in populations with the longest time of sitting and TV viewing (Figure 3A/B; Supplementary materials, Table S1). TV viewing-related mortality increased with the increment of BMI values and decreased with enhancement of physical activity levels in parallel (Figure 3B; Supplementary material, Table S1). Additionally, the increment of BMI promoted the sitting time-associated mortality risk from ≤ 25.5 kg/m² to 25.6–26.9 kg/m², and the increase of physical activity levels decreased the mortality risk from ≤ 10 MET•h/w to 10.1–19.9 MET•h/w. However, further increment of BMI (≥ 27 kg/m²) or physical activity (≥ 20 MET•h/w) seemed to no longer change the estimated effects of sitting time-associated mortality risk (Figure 3A; Supplementary material, Table S1).

Figure 3.

Meta-analysis of all-cause mortality stratified by diabetes, hypertension, physical activity (PA) level, and BMI measures in the most sedentary groups

ST: sitting time; HR: hazard ratio; h: hours; w: week; MET: metabolic equivalent.

Risk of Bias, Sensitivity, and Publication Bias Analyses

Methodological quality assessment of the included studies indicated that most studies had a good quality score with score values ranging from 7 to 8 points (Table 1). Sensitivity analysis was conducted by dropping the studies with extreme outcomes, suggesting our results were robust (Supplementary materials, Table S2). We also explored the possibility of the appearance of small-study effects. The Begg's test showed there was weak evidence of the presence of small-study effects (Supplementary materials, Table S2).

Discussion

Our study provided quantified evidence concerning the relationship between sedentary time and cardiovascular, cancer, and mortality risk. The findings indicated that the associations of sitting time and TV viewing with CVD, cancer, and all-cause mortality were dose-dependent. Our results further demonstrated when sedentary behaviour was combined with diabetes, hypertension, and high BMI, the mortality risk was remarkedly upgraded. However, high physical activity levels tended to decrease the sedentary-associated mortality risk. The sensitivity analysis suggested our results were robust.

Most prior reviews only explored the qualitative aspects of the relationship between sedentary time and mortality (9, 14, 19). Additionally, many studies (9, 14) mixed the mortality data induced by sitting time and TV viewing together. Our study revealed that the relationship between sedentary time and mortality was dose-dependent, indicating that mortality risk increased proportionally to the increment of sedentary behaviour. Our results were similar to the findings of one recent meta-analysis (12) that also determined the quantitative aspects of the relationship between sedentary time and mortality risk. However, this review was conducted only at a summary level, and aspects of the adverse associations with appearance of chronic diseases, BMI, and physical activity remain unclear.

Two meta-analyses (13, 20) recently examined the role of physical activity in modifying the adverse associations of sedentary behaviour and mortality risk; however, considerable questions remained about the effects of chronic diseases and BMI status on these adverse associations. Our study indicated that participants with a higher percent of diabetes and hypertension at baseline further increased the mortality risk in the most sedentary populations. This suggests that sedentary behaviour in combination with chronic diseases is more likely to damage one's health as compared to sedentary behaviour alone. Given the fact that, with age accrual, some older people predestine to have chronic diseases, the findings alert those elderly adults that modifying existing chronic diseases is also crucial for managing sedentary-associated mortality risk. Diabetes and hypertension are one of the major sources of mortality (21, 22, 23); therefore, we preferred using baseline disease conditions to stratify the subgroup analysis instead of employing new disease occurrences, which may be too novel to cause mortality.

We also found that high physical activity levels remarkably reduced and a high BMI increased the risk of all-cause mortality in the most sedentary groups. Several studies (24, 25, 26) have reported associations between physical activity levels and BMI status with mortality risk in the general populations. We further confirmed the relationship in sedentary populations. Furthermore, distinct from previous meta-analyses (9, 13, 20, 27, 28, 29), our study extended the analysis to determine the effects of the combination of physical activity and BMI status with sedentary time on mortality risk. Our findings suggested that modification of physical activity levels and BMI might help manage mortality risk in the most sedentary of adults.

Several mechanisms may explain the protective effects of high physical activity levels and low BMI against mortality in sedentary people. Increased physical activity has potential metabolic effects, including reducing BMI, sex hormones, adiposity, insulin resistance and c-peptide levels, and it may impact immune system inflammation (28, 30, 31, 32), all of which are risk factors for sedentary-induced mortality. High BMI is frequently regarded as linked with increased adiposity, insulin resistance, elevated inflammation levels, and so on (33, 34, 35, 36, 37, 38); therefore, low BMI might help improve those deleterious conditions and thus reduce the mortality risk.

Simultaneously, apart from modifying sedentary behaviour to reduce mortality risk, our findings further demonstrated that improving chronic disease conditions, increasing physical activity levels, and reducing BMI might also constitute key components of strategies for the management of mortality risk. However, well-designed and population-based prospective studies are still needed to confirm our findings and elucidate the underling mechanism.

We used the New Castle-Ottawa Quality Assessment Scale assessment tools to evaluate the study quality. The quality of the included studies was relatively good, with scores ranging from 7 to 8 points. A sensitivity analysis also suggested that our results were robust. The Begg's test showed there was weak evidence of the presence of small-study effects.

Nonetheless, our study had some limitations. Because the noted studies were conducted across diverse countries, heterogeneity across studies was unavoidable. Furthermore, as the included studies did not present sufficient data for examining the effects of diabetes, hypertension, physical activity levels, and BMI status on cardiovascular and cancer mortality, the subgroup analyses were thus confined to all-cause mortality.

Conclusion

In sum, our study provided quantified evidence on the associations of sitting time and TV viewing with the risk of CVD, cancer, and mortality. The dose-response relationship suggested that the risk increased proportional to the increment in sedentary time. Apart from modifying sedentary behaviour, managing existing chronic diseases, increasing physical activity levels, and reducing BMI should also be considered for prevention of sedentary-associated mortality risk. Future prospective studies with large sample sizes are however needed to confirm our findings. Nevertheless, our study may help adults maintain their health by promoting them to modify their daily behaviour.

Funding

This work was supported in part by the Talent Project of Yangzhou University (No. 5011/137011159).

Conflict of Interest

The authors declare that no competing interest.

Ethical Standards

This study did not include any animal or human experiments.

Electronic supplementary material

Supplementary material is available for this article at https://doi.org/10.1007/s12603-019-1298-3and is accessible for authorized users.

Supplementary Materials