Background
Globally, the older adult population has increased substantially, and it is estimated to reach approximately 22% of the world’s population by 2050 [
1,
2]. The risk of non-communicable diseases and disability increases with age, providing a challenge for health and social care resources [
3]. The World Health Organization has created many recommendations for behavior change to reduce the burden of non-communicable diseases and disabilities among the elderly [
4]. It is well established that physical activity plays a key role in the prevention of such diseases due to its close relationship with many of the chronic diseases and disabilities that largely affect the elderly, such as cardiovascular disease, cancer, type 2 diabetes, accidental falls, obesity, metabolic syndrome, mental disorders, and musculoskeletal diseases [
5,
6].
However, in the last decade, sedentary behavior has emerged as a new risk factor for health [
7‐
9]. Sedentary behaviors are characterized by any waking activity that requires an energy expenditure ranging from 1.0 to 1.5 basal metabolic rate and a sitting or reclining posture [
10]. Typical sedentary behaviors are television viewing, computer use, and sitting time [
10]. Epidemiological studies on different age groups show that a considerable amount of a human’s waking hours are spent in sedentary activities, creating a new public health challenge that must be tackled [
11,
12]. The scientific study of sedentary behavior has become popular in recent years. In fact, several systematic reviews of sedentary behaviors and health outcomes among children, adolescents, [
13‐
15] and adults [
11,
16‐
19] have recently been published. However, insights from these systematic reviews are limited for several reasons. Firstly, some of these systematic reviews did not evaluate the quality of evidence of the reviewed articles [
17,
16]. Secondly, some reviews included subjects with a wide age range (i.e., >18 years) [
16,
17]. Therefore, it is currently assumed that the deleterious health effects attributed to sedentary behaviors are similar among both adults (>18 years) and the elderly (>60 years). However, it has been observed that some cardiovascular risk factors (i.e., smoking, obesity, and consumption of alcohol) are less predictive of mortality in a large sample of Scandinavians aged 75 years or older [
20].
Furthermore, compared with other age groups, older adults are the most sedentary. Findings from studies in the US and Europe reported that objectively measured sedentary time was higher in those who were older than 50 years [
12] and 65 years, [
21] respectively. In addition, it has been reported that adults older than 60 years spend approximately 80% of their awake time in sedentary activities which represents 8 to 12 hours per day [
12,
21,
22]. Similarly, Hallal et al. conducted a global assessment in more than 60 countries and found that the elderly had the highest prevalence of reporting a minimum of 4 hours of sitting time daily [
23]. Despite this high exposure in the elderly, the health effects of sedentary behavior in this population have not yet been reviewed. Due to this knowledge gap, we systematically reviewed evidence to look for associations between sedentary behavior and multiple health outcomes in adults over 60 years of age.
Methods
Identification and selection of the literature
In May 2013, we searched the following databases: Medline, Excerpta Medica (EMBASE), Web of Science, SPORTDiscus, PsycINFO, Cumulative Index to Nursing and Allied Health Literature (CINAHL), Literatura Latino-Americana e do Caribe em Ciências da Saúde (LILLACS), and the Sedentary Behavior Research Database (SBRD).
The key-words used were as follows: exposure (sedentary behavior, sedentary lifestyles, sitting time, television viewing, driving, screen-time, video game, and computer); primary outcome (mortality, cardiovascular disease, cancer, type 2 diabetes mellitus); and secondary outcome (accidental falls, frail elderly, obesity, metabolic syndrome, mental disorders, musculoskeletal diseases). Further information regarding the search strategy is included in Additional file
1. According to the purpose of this systematic review, observational studies (cross-sectional, case–control, or cohort) involving older adults (all participants >60 years), with no restriction of language or date, were selected in the screening step.
In addition, we contacted the Sedentary Behaviour Research Network (SBRN) members in July 2013 to request references related to sedentary behavior in older adults. The SBRN is a non-profit organization focused on the scientific network of sedentary behavior and health outcomes. Additional information about the SBRN can be found elsewhere (
http://www.sedentarybehaviour.org/).
The studies retrieved were imported into the EndNote Web® reference management software to remove any duplicates. Initially, titles and abstracts were screened by two independent reviewers (LFMR and JPRL). Relevant articles were selected for a full read of the article. Disagreements between the two reviewers were settled by a third reviewer. In addition, the reference lists of the relevant articles were reviewed to detect additional articles that were not identified in the previous search strategy.
Studies were excluded if they met the following criteria: 1) Included adults <60 years of age; 2) did not include physical activity as a covariate; or 3) presented only a descriptive analysis of sedentary behavior.
Data extraction and quality assessment
The data from all of the eligible articles were extracted independently by two reviewers (LFMR and JPRL). The extracted data included the following information: author(s), year, country, age group, number of participants, type of population (general or patient), type of sedentary behavior, type of measurement tool, sedentary definition, adjusted confounders, and outcome (Additional file
2: Table S1).
The quality assessment was performed by two independent reviewers (LFMR, JPRL) and discussed during a consensus meeting. The quality of articles was assessed using the
Grades of Recommendation, Assessment, Development and Evaluation (GRADE) tool (Table
1). Briefly, the GRADE quality assessment tool begins with the design of the study. Studies with an observational design start with a low quality (2 points). The studies then lose points based on the presence of the following topics: risk of bias (−1 or −2 points), imprecision (−1 or −2 points), inconsistency (−1 or −2 points), and indirectness (surrogate outcome) (−1 or −2 points). However, studies gain points if the following criteria are met: a high magnitude of effect (RR 2–5 or 0.5 – 0.2) (+ 1 or 2 points), adequate confounding adjustment (+1 point), and a dose–response relationship (+1 point). Finally, the quality of the articles is categorized as follows: high (4 points), moderate (3 points), low (2 points), or very low (1 point). Further information about GRADE has been published elsewhere [
24].
Table 1
Quality of articles assessed using the Grades of Recommendation, Assessment, Development and Evaluation (GRADE)
| 2011 | Cross-sectional | −1 | −1 | −1 | −1 | 0 | 1 | 0 | 1 |
| 2011 | Cross-sectional | −1 | −1 | −1 | 0 | 0 | 1 | 0 | 1 |
| 2011 | Cross-sectional | 2 | −1 | −1 | 0 | 0 | 1 | 0 | 3 |
| 2012 | Cross-sectional | −1 | 0 | −1 | 0 | 0 | 1 | 0 | 1 |
| 2010 | Cross-sectional | −2 | −1 | −1 | 0 | 0 | 1 | 0 | 1 |
Gomez-Cabello et al. [ 30] | 2012 | Cross-sectional | −1 | −1 | −1 | 0 | 1 | 0 | 0 | 1 |
Gomez-Cabello et al. [ 31] | 2012 | Cross-sectional | −1 | −1 | −1 | 0 | 0 | 0 | 0 | 1 |
| 2010 | Cross-sectional | −2 | 0 | −1 | 0 | 0 | 1 | 0 | 1 |
| 2012 | Cross-sectional | −2 | 0 | −1 | 0 | 0 | 1 | 0 | 1 |
| 2012 | Cross-sectional | −2 | −1 | −1 | 0 | 0 | 1 | 0 | 1 |
| 2011 | Cross-sectional | 0 | −1 | −1 | 0 | 0 | 1 | 0 | 1 |
| 2007 | Cross-sectional | 0 | −1 | −1 | 0 | 1 | 1 | 1 | 3 |
| 2012 | Cross-sectional | −2 | −1 | −1 | 0 | 0 | 1 | 0 | 1 |
| 2012 | Cross-sectional | −1 | −1 | 0 | 0 | 1 | 1 | 0 | 2 |
| 2013 | Cross-sectional | 0 | −1 | −1 | 0 | 1 | 1 | 1 | 3 |
| 2011 | Cross-sectional | −1 | 0 | −1 | 0 | 1 | 1 | 0 | 2 |
| 2012 | Case–control | −1 | −1 | −1 | 0 | 1 | 0 | 0 | 1 |
Balboa-Castillo et al. [ 42] | 2011 | Prospective Cohort | −2 | −1 | 0 | 0 | 1 | 1 | 1 | 2 |
| 2013 | Prospective Cohort | −1 | 0 | 0 | 0 | 0 | 1 | 0 | 2 |
Martinez-Gomez et al. [ 44] | 2013 | Prospective Cohort | −1 | 0 | 0 | 0 | 1 | 1 | 0 | 3 |
| 2012 | Prospective Cohort | −1 | 0 | 0 | 0 | 1 | 1 | 1 | 4 |
| 2013 | Prospective Cohort | −1 | 0 | 0 | 0 | 1 | 1 | 1 | 4 |
| 2003 | Prospective Cohort | −2 | 0 | 0 | 0 | 1 | 1 | 0 | 2 |
Discussion
To the best of our knowledge, this is the first systematic review to examine the association between sedentary behavior and health outcomes in older people while considering the methodological quality of the reviewed studies. Similar to previous reviews in adults, [
16‐
19,
48] the present review shows observational evidence that greater time spent in sedentary activities is related to an increase risk of all-cause mortality in the elderly. However, in these studies, sedentary behavior was measured through self-reported questionnaires (e.g., hours/day of sitting time), which have moderate criterion validity [
49]. Studies with a moderate quality of evidence showed a relationship between sedentary behavior and metabolic syndrome, waist circumference, and overweight/obesity. The findings for other outcomes, such as mental health, renal cancer cells, and falls, remain insufficient to draw conclusions.
However, some sedentary activities (e.g., playing board games, craft activities, reading, computer use) were associated with a lower risk of dementia [
47]. Thus, future studies should take into account not only the amount of time spent in sedentary behavior but the social and cognitive context in which the activities takes place [
50]. To illustrate this point, some studies have shown that video game and computer use, even though classified as sedentary by energy expenditure criteria, may reduce the risk of mental health disorders [
51‐
53].
Methodological issues
To overcome the limitations of the observational studies available, future longitudinal studies with a high methodological quality are required. Moreover, the primary limitations found in the reviewed articles should be taken into account in future studies (Additional file
4). Based on these limitations, we offer several recommendations for future studies.
Selection bias
In nearly half of the reviewed articles (10 articles: 42%), the following selection biases were found: a low response rate; the use of independent and non-institutionalized volunteer participants; and an underrepresentation of some population subgroups [
25,
27‐
29,
31‐
34,
37,
41,
47].
To date, the use of accelerometers is the most valid and reliable method for evaluating sedentary behavior, although some devices are not able to distinguish sitting and standing posture [
54]. In studies of the elderly, 5 days of accelerometer use seems to be sufficient to evaluate the pattern of sedentary behavior [
55]. When using accelerometers, future studies should clearly specify the criteria established for non-wear time [
56] and use the most accurate sedentary cut-points (150 counts/min) [
57] to avoid misclassification. In the current review, all studies used at least 7 days of accelerometry, with a non-wear time criteria of 60 minutes without counts and sedentary cut-points of <100 counts/minute [
26,
28,
32,
37,
39] or <199 counts/minute [
33,
34].
Although subjective measurements present a low to moderate reliability, they allow for the evaluation of the contextual dimension of the sedentary activities [
49]. In the present review, information bias attributable to self-reported instruments was found in 20 articles (83%) [
25‐
27,
29‐
34,
37,
38,
42‐
44,
40,
41,
45‐
47]. In this sense, emergent objective methods (e.g., combination of geolocation data combined with acceleration signals in mobile phone) have been developed to obtain a precise and meaningful characteristic of the patterns of sedentary behavior [
49].
In addition, most of the studies in this review used different categorization criteria when measuring sedentary behavior [
43‐
46]. This variation in categorization criteria could limit future synthesis of the evidence. We recommend that future studies on the elderly use existing categorizations of sedentary behavior.
Imprecision
To reduce random error, future epidemiological studies, especially with longitudinal designs, should use an adequate sample size. In the present review, 14 (58%) studies presented imprecise results [
25‐
27,
29‐
31,
34‐
39,
41,
42].
Inconsistency
Subgroup and heterogeneity analysis should be performed and reported in future studies to evaluate the consistency of the findings. In the current study, only one article presented the consistency of the findings between subgroups [
25].
Indirectness: In the current review, indirectness (surrogate outcomes) was present in 17 articles (71%) [
25‐
37,
39‐
41]. Importantly, conclusions obtained with surrogate markers only allow a better understanding of the sedentary behavior physiology. However, researchers should not consider these surrogate markers as synonymous with the endpoint outcomes [
58].
Thus, endpoint outcomes (e.g., cardiovascular events, cancer and mortality) should be addressed in future studies.
Confounding adjustment
The confusion of effects (confounding) is a central issue in epidemiology. Although all of the studies in the present review included some covariates, such as moderate to vigorous physical activity, some residual confounding may be present [
59]. Moreover, health status should be measured and included as a covariate, especially in studies of the elderly to avoid confounding [
59]. Although most of the articles received better quality scores when they adjusted for potential confounders, [
25‐
29,
32‐
40,
42‐
47] only 3 studies included health status as a covariate [
25,
45,
46]. Future observational studies should include these important covariates in their statistical analysis.
Dose–response
Although sedentary behavior is a continuous variable, most of the studies categorized it as either an ordinal or a dummy variable. Such categorization could be an important limitation [
60,
61]. However, if future studies opt to categorize, they should use small intervals with more homogeneous groups that may allow for the observation of a dose–response gradient between sedentary behavior and health outcomes. In the present review, a dose–response was detected in 5 articles [
36,
39,
42,
45,
46].
Conclusion
This review confirms previous evidence of the relationship between sedentary behavior and all-cause mortality among adults. Due to the moderate quality of the studies, weak evidence exists regarding other health outcomes (metabolic syndrome, cardiometabolic biomarkers, obesity, and waist circumference). However, of note, some sedentary activities (e.g., playing board games, craft activities, reading, and computer use) had a protective relationship with mental health status (dementia). Future studies should consider the main methodological limitations summarized in this review to improve the current state of the art. Finally, intervention trials that support the observational knowledge are needed to create informed guidelines for sedentary behavior in the elderly.
Acknowledgements
We would like to thank the Sedentary Behaviour Research Network members for sending us titles of studies that were potentially eligible for inclusion in our systematic review. This study received financial support from the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP, São Paulo Research Foundation; Grant no. 2012/07314-8).
Competing interests
No financial disclosures were reported by the authors of this paper.
Authors’ contributions
Study Concept and design: LFMR, OCL; Search Strategy: LFMR; Identification and Selection of the Literature: LFMR, JPRL; Data Extraction and Quality Assessment: LFMR, JPRL; Narrative Synthesis: LFMR, JPRL; Drafting of the Manuscript: LFMR, JPRL; Study Supervision: OCL, VKRM. All authors read and approved the final manuscript.