Is the relationship between sedentary behaviour and cardiometabolic health in adolescents independent of dietary intake? A systematic review.

Screen time, but not overall sedentary behaviour, is consistently related to cardiometabolic health in adolescents. Because of the associations screen time has with dietary intake, diet may be an important factor in the screen time and health relationship; however, evidence has not previously been synthesized. Thus, the aim of this systematic review was to explore whether the associations between various sedentary behaviours and cardiometabolic risk markers are independent of dietary intake in adolescents. Online databases and personal libraries were searched for peer-reviewed original research articles published in English before March 2014. Included studies assessed associations between sedentary behaviour and cardiometabolic markers in 12- to 18-year-olds and adjusted for dietary intake. Twenty-five studies met the inclusion criteria. From the 21 studies examining sedentary behaviour and adiposity, the majority found significant positive associations between television viewing, screen time and self-reported overall sedentary behaviour with markers of adiposity, independent of dietary intake. No significant associations between screen time with blood pressure and cholesterol were reported. Sedentary behaviour appears to be associated with adiposity in adolescents, irrespective of dietary intake. However, the variability of dietary variables between studies suggests further work is needed to understand the role of dietary intake when examining these associations in youth.

Introduction

Cardiometabolic risk factors, specifically obesity, insulin resistance, hypertension and dyslipidaemia, are becoming more prevalent among adolescents 1, 2. In Western countries, 25–33% of adolescents are overweight or obese 3, 4, 5 and 5–10% of adolescents have one or more of the other cardiometabolic markers 1. Spending large amounts of time using screen‐based media, in particular time spent watching television, has been consistently associated with cardiometabolic health in children and adolescents 6, 7, 8, 9, 10. This is concerning since current figures indicate that 62–83% of adolescents from Westernized countries are exceeding the screen‐based recommendations of no more than 2 h of screen time per day 5, 11, 12.

In contrast to the consistent links shown for television viewing and screen time, there is inconsistent evidence of associations between overall sedentary time (usually measured by accelerometry) and cardiometabolic risk markers in youth 13, 14. This inconsistency suggests that some of the observed associations between screen time and health may be explained by other factors, rather than time spent sedentary or ‘sitting’ per se. There is a possibility that the differences may be simply due to different measures used, with self‐reported measures and objective measures reported to have low to moderate correlations (−0.71 to 0.96) 15.

Alternatively, there may be other lifestyle behaviours such as physical activity and dietary intake that may explain the stronger relationship seen between screen time and health. For example, there is some evidence indicating that adolescents who spend more time in screen time behaviours are less physically active 16, 17. However, findings from a recent systematic review reveal screen time and physical activity among youth are not strongly inversely associated (r = 0.080, 95% CI −0.101, −0.060) 18, suggesting they are independent behaviours. Additionally, most studies that report associations between screen time and health outcomes in this age group account for physical activity levels 9, 19.

In youth, there is consistent evidence that television viewing is linked to elements of a less healthy diet such as lower fruit and vegetable consumption, higher consumption of energy‐dense snacks, drinks, and fast foods, and higher total energy intake 20. This suggests that dietary intake may play a role in the relationship between television viewing and cardiometabolic health. However, the mediating role of dietary intake has rarely been examined in youth, while the few studies among adults have found no consistent relationship 21, 22, 23, 24. There is now an emerging amount of literature that has examined the relationship between sedentary behaviour and cardiometabolic risk markers in adolescents that has accounted for dietary intake in the analyses. However, there have been no systematic reviews to determine whether these associations are consistent, independent of dietary intake, or whether the studies are of sufficiently high quality. With the majority of Western adolescents not meeting the current screen time guidelines 5, 11, 12 and the increase in prevalence of cardiometabolic conditions such as obesity, type 2 diabetes and cardiovascular disease 1, 2, adolescence is an important time to encourage healthy behaviour changes. Therefore, the aim of this systematic review was to identify and synthesize evidence from studies that have accounted for dietary intake when examining the relationship between sedentary behaviour and cardiometabolic health in 12‐ to 18‐year‐old adolescents.

Methods

This review is registered with PROSPERO, CRD42014010359.

Search strategy

Online databases (Medline, Global Health, PsycInfo, Web of Science and Embase), reference lists and personal libraries were searched for peer‐reviewed original research articles published in English before 25 March 2014. The following keyword combinations were used for sedentary behaviour (sedentar*, sitting, indoor*, screen time, computer*, television, inactivit*, video game*, internet), dietary intake (diet*, nutrition, food*, snack*, drink*, beverage*, eating, energy intake*, meals), health outcome (overweight, obes*, adiposity, waist circumference*, waist‐hip ratio, body mass index, blood pressure, hypertens*, glucose intolerance, blood glucose, insulin, cholesterol, lipoprotein, triglycerides, lipid metabolism, cardiometabolic, metabolic syndrome, cardiovascular disease*) and age (youth, adolescen*, child*, young people). The search terms were restricted to title and abstract only.

Inclusion and exclusion criteria

To be included in the review, studies were required to meet the following criteria: (i) be published as a peer‐reviewed original research article, with full‐text availability; (ii) the study participants’ mean age was between 12 and 18 years, or for longitudinal studies the mean age was between 12 and 18 years at baseline or at follow‐up; (iii) the study included a measure of sedentary behaviour as an independent variable defined by ‘any waking behaviours characterized by low energy expenditure (≤1.5 METS) while in a sitting or reclining posture’ 25, and not the absence of sufficient levels of physical activity; (iv) the statistical analyses included an adjustment for dietary intake as defined by the intakes of energy, macronutrients, foods, and/or beverages) and (v) the study assessed at least one cardiometabolic risk marker as the main outcome (i.e. adiposity, blood pressure; insulin sensitivity; glucose tolerance or lipid levels) or included a chronic cardiometabolic condition (i.e. metabolic syndrome, cardiovascular disease or type 2 diabetes).

The exclusion criteria included the following: (i) the study included all three measures of interest but did not include them simultaneously in the statistical analyses; (ii) the study included only special populations (e.g. participants with type 1 diabetes); (iii) the study was not original research and/or (iv) the reviewers could not access the full text after contacting the corresponding author.

Identification of relevant studies

Two reviewers (EF, RL) independently reviewed all articles based on title and abstract initially, then assessed full text for eligibility. For articles that needed further clarification in order to assess their eligibility for the present review, the corresponding author was contacted. Any discrepancies between reviewers about article eligibility were discussed with all authors until a final consensus decision was reached.

Data extraction and coding

Data extraction was performed by one reviewer (EF) using a predetermined data extraction template. The following information was extracted from each article: (i) study design and length (for longitudinal studies only); (ii) participant characteristics (sample size, % males, age range, mean age, % overweight/obese and location); (iii) measures used to assess sedentary behaviour, dietary intake and cardiometabolic risk markers; (iv) statistical analysis methods (e.g. statistical models used, variables included in model) and (v) main findings in regard to sedentary behaviour, dietary intake and cardiometabolic risk markers. Where studies combined multiple domains of sedentary behaviours (e.g. television viewing, video watching and video playing), this was coded as ‘screen time’ in the results table. This method was also applied to dietary variables where multiple drinks were assessed and coded as ‘sugar sweetened beverages’. For this review, only dietary variables related to food and drinks consumed, energy intake or macronutrient intakes were reported.

Methodological quality assessment

The methodological quality of the included studies was independently examined by two reviewers (EF, RL) using an adapted 15‐item quality criteria checklist 26, 27 (Table 1). The 15‐item checklist consisted of assessing various methodological aspects (e.g. study design and sample, data sources and measurement of variables, statistical methods used) and involved a yes (0.5 or 1 point) or no/unclear (0 point) answer format. The change made to the original quality criteria checklist was mainly regarding the measurement questions, where the reliability and validity of the three measurements of interest were examined separately, as opposed to assessing the reliability and validity of one measurement. A quality score ranging from 0 to 15 points was calculated for each study. This score was divided by 15 points to calculate a percentage from 0 to 100% and a study was considered to be of high quality if the score was >50% 27.

Table 1

Criteria list for assessment of the methodological quality of observational studies (based on Tooth et al. 26 and Singh et al. 27)

Criteria (rating of criteria: 0.5 or 1 = yes, 0 = no or unclear)	Score (total 15)
Study design
1. Is the study design presented	__ /0.5
AND is the study design longitudinal and not cross‐sectional?	__ /0.5
Target population
2. Do the authors describe the target population they wanted to research?	__ / 1
Sample
3. Was a random sample of the target population taken/described	__ /0.5
AND/OR was the response rate 60% or more?	__ /0.5
4. Is participant selection described, or referred to?	__ / 1
5. Is participant recruitment described, or referred to?	__ / 1
6. Are the inclusion and/or exclusion criteria stated, or referred to?	__ / 1
7. Is the study sample described? (minimum description = size, gender, age, BMI)	__ / 1
8. Are the numbers of participants at each stage of the study reported? (authors should report at least numbers eligible, numbers recruited and numbers with complete data)	__ / 1
Variables
9. Are the measures of sedentary behaviour, dietary intake and the health outcome sufficiently described in detail?	__ / 1
Data sources and collection
10. Do authors describe the source of their data? (e.g. registry, health survey)	__ /0.5
AND did authors describe how the data were collected? (e.g. by mail, by survey)	__ /0.5
Measurements
11. Was the validity of sedentary behaviour mentioned or referred to?	__ /0.5
AND/OR was the reliability of sedentary behaviour mentioned or referred to?
(if measured by accelerometry, automatically full points)	__ /0.5
12. Was the validity of dietary behaviour mentioned or referred to?	__ /0.5
(if measured by 24‐h food recall, automatically full points)	__ /0.5
13. Was the health outcome measured objectively and not by self‐report?	__ / 1
Statistical methods
14. Were appropriate statistical methods used and adequately described (including taking into account number of participants and clustering effects)?	__ /0.5
AND/OR did the statistical methods address confounders?	__ /0.5
15. Were the numbers/percentages of participants with missing data for sitting and the health outcome indicated?	__ /0.5
AND if more than 20% of data in the primary analyses were missing, were methods used to address missing data?	__ /0.5

Synthesis of studies

All findings relating to sedentary behaviour and cardiometabolic health from each study were categorized and tabulated into two categories: (i) no association between sedentary behaviour and the cardiometabolic risk marker (denoted by ‘0’) or (ii) a significant positive or negative association between sedentary behaviour and cardiometabolic risk marker, independent of dietary intake (denoted by ‘+’ or ‘−’). Following the ‘quality rating system’ by Singh et al. 27, the overall findings were synthesized into four levels of scientific evidence. However, because of all studies in this review being considered to be of ‘high quality’, cut‐off percentages were given to differentiate the level of ‘strong’ and ‘moderate’ evidence. Thus, the four levels of scientific evidence used in this review are as follows: (i) strong evidence where ≥2 high‐quality studies have ≥70% consistent findings; (ii) moderate evidence where ≥2 high‐quality studies have between 50 and 69% of consistent findings; (iii) inconsistent evidence where ≥2 studies have <49% of consistent findings and (iv) insufficient evidence where only one study was available.

Results

Search results

The literature search identified a total of 539 records (Figure 1). After removing 12 duplicates and checking the titles, abstracts and full‐text articles, 18 articles met the inclusion criteria. A further seven articles were identified from other sources (e.g. reference lists and personal libraries), resulting in a total of 25 articles being included in this systematic review.

Figure 1

Flow diagram of the identification of literature for inclusion in this systematic review.

Study characteristics

All 25 articles reported on 25 unique study samples. Twenty‐one articles reported cross‐sectional results and four reported longitudinal study results (ranging from 1 to 4 years of follow‐up). The majority of studies originated from the United States (n = 9), Canada (n = 3) or China (n = 3). The number of participants ranged from 282 to 39,011 with most studies including an approximately even ratio of males to females, and only one study including females only 28. The age of participants ranged from 10 to 19 years (mean 14.4 years) and the majority were of healthy weight, with the body mass index (BMI) ranging from 18 to 24 kg m−2. Nearly all studies used some form of regression analyses (e.g. linear, logistic, longitudinal, multilevel) to examine the associations between sedentary behaviour and a cardiometabolic health outcome, adjusting for dietary intake in the model, with one study conducting a mediation analysis with dietary intake as the mediator 29. The characteristics and the main findings of studies are presented in the Supporting Information (Table S1).

Overview of measures

The most common sedentary behaviour measure was television viewing 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, followed by total screen time 28, 30, 42, 43, 44, 45, 46, 47, 48, 49, personal computer use 31, 32, 36, 37, 40, 41, and video game playing or video viewing 37, 40, 49. Three studies measured total sedentary behaviour 28, 50, 51 and only one measured a non‐screen time behaviour such as homework time (results not reported in the review) 31. The majority of sedentary behaviours were measured via a self‐reported questionnaire, with two studies using a proxy report for those aged 12 or below 45, 52, and one study each used a 24‐h activity recall 28 or interview‐administered questionnaire 47. The dietary variables that were included in the analyses varied widely, with the majority of studies (n = 19) simultaneously adjusting for multiple dietary variables. The most common variables included total energy intake 34, 35, 40, 41, 46, 47, 49, 50, 51 and the consumption of energy‐dense snacks 29, 30, 32, 36, 39, 42, 43, 44, 45, 48, sugar‐sweetened beverages 28, 30, 31, 32, 33, 36, 37, 44, 45, 50, fruits and vegetables 28, 31, 32, 36, 38, 42, 43, 50, and fast food 30, 42, 45, 50, with few studies measuring macronutrient intakes 28, 35, 40, 47 or specific food items (e.g. cereals, meat) 50, 52. Adiposity was the most commonly reported cardiometabolic risk marker, with 21 studies measuring adiposity mostly by weight status. Only four studies assessed other cardiometabolic markers such as blood pressure, insulin resistance, glucose tolerance and lipid levels 40, 46, 47, 48.

A summary of the methodological quality of the studies is presented in the Supporting Information (Table S2). The initial discrepancy between the two reviewers when scoring the 25 articles was 6.3%. The majority of disagreements were resolved after discussion, with 9 of the 33 discrepancies being discussed with all authors. The quality scores of the papers ranged from 60 to 84%. The most common reasons why studies did not get a quality score of 100% were the study ‘did not use methods to address missing data’ and ‘did not report the inclusion and exclusion criteria of the sample’. According to the quality criteria 27, all studies were considered to be of high quality (>50%) with eight studies scoring 80% or more 28, 30, 31, 41, 42, 46, 47, 49.

Fifteen of the 25 studies included a random sample of the target population 29, 30, 31, 32, 33, 34, 36, 38, 39, 41, 42, 43, 45, 46, 47, 48, 50 with 14 studies reporting a response rate of 60% or more 29, 30, 31, 33, 34, 38, 41, 42, 43, 44, 46, 47, 49, 50. All studies used either self‐administered questionnaires or interviews to measure sedentary behaviour with four studies reporting the validity 28, 29, 48, 49 and two studies reporting the reliability 28, 48 of the measure used. In total, seven studies used a 24‐h food recall to assess dietary intake 33, 34, 35, 41, 46, 47, 50, with five studies reporting the recall data from one occasion, and one study each reporting the recall data from either two 41 or three 35 occasions. Other commonly used measures were self‐administered questionnaires (n = 10) or food frequency questionnaires (n = 8), with four studies reporting both the reliability and validity of these measures 31, 40, 44, 49 and one study reporting only the validity 52. Out of the 21 studies that measured adiposity, nearly all studies used objective measures with only four studies using self‐reported height and weight to calculate BMI 36, 38, 43, 49. Four studies measured blood biochemistry profiles 40, 46, 47, 48, with one of those studies reporting the risk of the metabolic syndrome 46. Regarding the analyses, only 10 studies both adequately described and used appropriate statistical methods (including taking into account clustering effects due to sampling design) and addressed missing data where relevant 28, 29, 30, 31, 36, 41, 42, 47, 48, 49.

Main findings

The associations between sedentary behaviour and cardiometabolic markers adjusting for dietary intake are reported in Tables 2 and 3. Because of the majority of studies reporting on adiposity outcomes, results for adiposity and other cardiometabolic outcomes were presented separately. Table 2 summarizes the associations between various screen time behaviours and total sedentary behaviour with measures of adiposity (e.g. BMI, waist circumference), adjusting for dietary intake. Table 3 summarizes the studies reporting on associations between screen time behaviours with the other cardiometabolic outcomes (e.g. blood pressure, blood glucose, lipids), adjusting for dietary intake.

Table 2

Associations between various sedentary behaviours and adiposity, adjusting for dietary intake

Author [quality %]	Dietary factors included in analysisb	TV	Video use or VG	PC	Total ST	Total SB
Arango 30 [80]	Soft drink, high‐fat foods, fast foods	0 TVg WC +TVb WC			0 ST g WC + STb BMI + STb WC
Berkey 49, a [80]	EI		+ Vidb SR‐BMI		+ STg SR‐BMI
Hsu 42 [80]	FV, sweets, snacks, fast food				+ STBMI
Peart 41 [80]	EI, fat intake	+ TV BMI		0 PC BMI
Tang 31 [80]	Soft drink, FV	+ TV BMI		0 PC BMI
Must 28, a [80]	% CHO, protein and fat, FV, soft drink				0 ST BMI 0 ST %BF	0 SB BMI
Nasreddine 50 [74]	EI, cereals, dairies, meat, legumes/nuts, FV, added fats/oils, fast food, sugar/sweets, SSBs					+ SB BMI + SB WC + SB WtH
Al‐Haifi 32 [74]	FV, SSBs (including energy drinks), milk, fast food, cakes/doughnuts, sweets	0 TV BMI 0 TV WC		0 PC BMI 0 PC WC
Kunesova 52 [74]	Fish intake/meat fat intake			+ PC BMI
Forshee 33 [74]	All beverages (including SSBs, coffee/tea, beer)	0 TVBMI
Crespo 34 [74]	EI	+ TVb/g BMI
Carson 29 [70]	Junk food (including sweets, soft drinks, cakes/pastries, potato chips, French fries)	+ TVBMI
Vik 43 [70]	FV, unhealthy snacks (including soft drinks, candy and potato chips)				+ ST SR‐BMI
Zhang 35 [67]	EI, % fat intake	0 TV BMI
Fulton 51, a [64]	EI					0 SB FMI + SB BMI + SB FFM
Hume 44 [64]	SSBs, unhealthy snacks (including soda, candy, potato chips)				0 STb/g BMI 0 STb/g SF 0 STb WC + STg WC
Janssen 36 [64]	FV, sweets, soft drink, cake, potato chips	+ TV b/g SR‐BMI		0 PC b/g SR‐BMI
Giammettei 37 64]	Soft drink	+ TV BMI	0 VG BMI	0 PC BMI
Shan 45 [60]	Fast food, snacks, SSBs, alcohol				+ STBMI
Delva 38 [60]	FV	+TVb/g SR‐BMI
Elgar 39, a [60]	Snacks	+ TV BMI
Summary of results
0 No significant association		5	1	6	6	2
+ Significant association, independent of diet		9	1	1	7	5

Longitudinal study.

Dietary variables adjusted for in the analyses involving the sedentary behaviour and health outcome variables.

%BF, % body fat; +, significant positive association, independent of diet; 0, no significant association; b, boys only; BMI, objectively measured BMI; EI, energy intake; FFM, fat‐free mass; FMI, fat mass index; FV, fruits and vegetables; g, girls only; SF, skin‐folds; SR‐BMI, self‐reported BMI; SSB, sugar‐sweetened beverage; WC, waist circumference; WtH, waist to hip ratio.

Table 3

Associations between screen time behaviours with cardiometabolic risk markers and the metabolic syndrome, adjusting for dietary intake

Author	Dietary factors included in analysisa	Cardiometabolic risk markers								Met syndrome
		SBP	DBP	Insulin	Glucose	Total‐C	LDL‐C	HDL‐C	C‐Ratio
Kang 46 [84]	EI									0 1ST + 2ST
Sugiyama 47 [80]	EI, CHO, protein and fat intake	+ ST	0 ST
Goldfield 40 [77]	EI, % fat intake	0 TV + VG 0 PC 0 ST	0 TV 0 VG 0 PC 0 ST			0 TV 0 VG 0 PC 0 ST	0 TV 0 VG 0 PC 0 ST	0 TV 0 PC 0 ST	0 TV + VG 0 PC 0 ST
Hardy 48 [77]	EDNP foods (including confectionary, hot chips, salty snacks)	0 STb/g		0 ST g + ST b	0 ST b		0 ST b/g	0 ST b/g
Summary of results
0 No significant association		4	5	1	1	4	5	4	3	1
+ Significant association, independent of diet		2	0	1	0	0	0	0	1	1

Dietary variables adjusted for in the analyses involving the sedentary behaviour and health outcome variables.

+, significant positive association, independent of diet; 0, no significant association; 1, weekday only; 2, weekend day only; b, boys only; CHO, carbohydrate intake; C‐Ratio, cholesterol ratio (total cholesterol/HDL‐cholesterol); DBP, diastolic blood pressure; EDNP, energy‐dense nutritionally poor (foods); EI, energy intake; g, girls only; HDL‐C, high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol; SBP, systolic blood pressure; Total‐C, total cholesterol.

As shown in Table 2, strong evidence was observed between computer use and adiposity. Six of the seven studies (86%) examining computer use found no significant association with adiposity, with the regression analyses adjusting mainly for soft drink, energy‐dense snacks, and/or fruit and vegetable intake 31, 32, 36, 37, 41. Moderate evidence was found between both television viewing and total screen time with adiposity, with 64% 29, 30, 31, 34, 36, 37, 38, 39, 41 and 54% of the analyses 30, 42, 43, 44, 45, 49 respectively finding a significant and positive relationship, independent of dietary intake. However, no consistent pattern was observed according to the types of dietary variables adjusted for in the analyses. Additionally, there was moderate evidence to support a significant and positive relationship between total self‐reported sedentary behaviour and adiposity 50, 51, independent of total energy intake and other various food items. There was insufficient evidence of the relationship between video viewing and video game use 37, 49.

When examining the findings from the four longitudinal studies, no consistent relationship was observed 28, 39, 49, 51. Berkey et al. 49 and Elgar et al. 39 both reported positive longitudinal associations between screen time behaviours with adiposity, independent of total energy intake and intake of snacks, respectively; whereas Fulton et al. 51 and Must et al. 28 found no associations between overall sedentary behaviour and screen time with adiposity measures, independent of energy intake and percentage of macronutrient intake, respectively.

As shown in Table 3, three studies examined screen time behaviours with other cardiometabolic risk markers, such as blood pressure, insulin resistance, glucose tolerance and lipid levels 40, 47, 48, and one study examined total screen time and the risk of developing the metabolic syndrome 46. Overall, there was moderate (67%) to strong evidence (100%) to show no significant association between various screen time behaviours (e.g. television viewing, computer use) with systolic blood pressure 40, 48, diastolic blood pressure 40, 47, low‐density lipoprotein (LDL)‐cholesterol and high‐density lipoprotein (HDL)‐cholesterol 40, 48. Because of the limited number of studies performed, there was insufficient evidence to draw conclusions on the remaining cardiometabolic risk markers as well as the association between screen time and the risk of metabolic syndrome 46.

Additionally, as there were a limited number of studies reporting the unadjusted and adjusted findings, it was not possible to draw conclusions on whether dietary intake attenuated the sedentary behaviour and health relationship. However, of the three studies that did report the unadjusted findings, screen time was no longer associated with diastolic blood pressure 48 or the metabolic syndrome 46, and video game use was no longer negatively associated with HDL‐cholesterol 40 after adjusting for either energy‐dense, nutrient‐poor foods 48 or total energy and macronutrient intake 40, 46.

Discussion

This systematic review identified and summarized all studies published prior to March 2014 that have accounted for dietary intake when exploring the association between sedentary behaviour and cardiometabolic health markers in adolescents. According to the 15‐item quality criteria checklist 26, 27, all 25 studies identified were considered to be of high quality. From the 21 studies examining adiposity outcomes, there was moderate to strong evidence to show significant and positive associations between television viewing, screen time and overall sedentary behaviour with adiposity, independent of dietary intake. This is concerning since the majority of Western adolescents are already exceeding the screen time recommendations 5, 11, 12, thus placing them at an increased risk of becoming overweight or obese. However, findings from the four papers examining other cardiometabolic markers found no significant associations between screen time behaviours with blood pressure, LDL‐cholesterol and HDL‐cholesterol.

Most studies found television viewing was associated with obesity in adolescents, irrespective of dietary intake. These findings are similar to other reviews involving children and adolescents, where positive associations between television viewing and adiposity have consistently been reported 6, 53, 54, 55, 56. For example, in the review by Tremblay et al., among the 170 studies examining body composition, the majority of studies found positive associations between sedentary behaviour (mainly assessed by television viewing) and BMI 6. Similarly, findings from a meta‐analysis revealed a small but significant relationship of 0.066 (95% CI = 0.056–0.078) between television viewing and body fatness among nearly 45,000 youth 53. However, although the majority of papers included in these reviews adjusted for physical activity, it is unknown whether the findings are also independent of dietary intake 6, 53.

Although television viewing was shown to be independently associated with adiposity, computer use was not. This is consistent with previous literature involving children and adolescents, where they report a lack of association between computer use and obesity 53, 54, 57. There may be a couple of reasons for this difference. Firstly, television viewing is at the lowest end of the energy expenditure spectrum and requires less energy expenditure than computer use 58. Secondly, there is increased exposure to junk food advertising while watching television, as opposed to using a computer use. This exposure to junk food advertising has been consistently linked with between‐meal snacking 59. Further, ‘mindless eating’ where children pay less attention to how much they consume is more likely to occur while watching television as opposed to using the computer 60.

Of the four studies that examined those cardiometabolic markers other than adiposity, no associations were reported between screen time behaviours with blood pressure, LDL‐cholesterol and HDL‐cholesterol. This finding is in contrast to a previous review by Tremblay et al. who reported consistent evidence on the positive association between screen time and blood pressure and total cholesterol in school‐age children and youth 6. However, it was not clear whether the studies included in that review adjusted for dietary covariates. A review by Chinapaw et al. reported similar findings to the current review where there was little evidence to draw conclusions on associations between ‘sedentary time’ (mainly measured by television viewing) and blood pressure or blood lipids 61.

The lack of findings in the present review could be attributed to a number of factors. It could be simply due to the limited number of studies available that have adjusted for dietary intake when reporting on cardiometabolic markers in young people. Alternatively, it could be due to the lack of lifetime exposure this age group has had to establish the risk markers for chronic disease, or to the lack of sensitive measures used in the studies (such as a one‐off fasted blood test) to capture the small adverse effects in their cardiometabolic profile. Thus, for future studies examining cardiometabolic markers in children or adolescents, it is recommended to use more sensitive measures such as a continuous glucose monitoring system or flow mediation dilation that are likely to detect the smaller fluctuations seen in a younger population group 62.

This review found that the majority of evidence relating to adiposity was independent of dietary intake. A possible explanation for this could be that the dietary variables adjusted for may not have been sufficient to attenuate the relationship. However, in the present review, because of the number of dietary variables examined across the included studies, it made it difficult to determine if independent (or null) associations were more commonly identified for one group of dietary variables over another (e.g. total energy intakes vs. macronutrient content). Additionally, because of the limited number of studies reporting the unadjusted and adjusted results (for dietary intake), it was not possible to draw conclusion as to whether a particular dietary variable attenuated the relationship 40, 46, 48. Further, it is also unknown whether these dietary variables reported were examined while participants were engaged in the screen time behaviours. If the behaviours were not measured concurrently, this suggests that dietary intake may be an indicator of an unhealthy lifestyle, rather than due to a direct association. Currently, there is limited research examining what dietary variables should be considered as covariates and what dietary variables, if any, mediate the relationship between sedentary behaviour and health in youth.

Another reason for the independent associations observed could be that dietary intake is simply not a strong driver of the relationship between screen time and adiposity. However, this explanation is at odds with previous literature where it is consistently reported that there are links between television viewing and elements of an unhealthy diet, such as lower fruit and vegetable consumption, higher consumption of energy‐dense snacks, drinks, and fast foods, and higher total energy intake 20. However, there are mixed findings on whether the same unhealthy dietary behaviours are linked with other sedentary behaviours such as video game use 63, 64, 65 and objectively measured total sedentary time 66.

Limitations of the studies in the review

Despite the high quality of studies included in this review, it is important to acknowledge that there were several methodological limitations. Firstly, the majority of papers included in the review were cross‐sectional, with the four longitudinal studies identified reporting inconsistent findings 28, 39, 49, 51. Therefore, it is unknown whether sedentary behaviour and dietary intake are predictive of health outcomes longitudinally. Secondly, there were considerable variations in how sedentary behaviour and dietary intake were examined and assessed. All of the studies included in our review used subjective measures to assess sedentary behaviour. Although these measures are useful to assess specific sedentary behaviour domains that could be targeted for interventions (e.g. reducing television viewing, or total screen time), they do not accurately capture the total time spent being sedentary. Thus, future studies are needed that also include an objective measure of overall sedentary time, such as accelerometers or inclinometers, to assess whether overall sedentary time has the same health implications as screen time behaviours.

Regarding dietary intake, there were two key limitations of the studies; the majority of studies only reported adjusted results and the studies reported on a vast array of dietary variables. By examining only the adjusted results, it is unknown whether the dietary variables attenuated the relationship between sedentary behaviour and adiposity. For future studies, it would be useful to report both the unadjusted and adjusted results when examining similar outcomes. Conducting mediation analyses would also help better understand the role of dietary intake in the sedentary behaviour and health relationship. For instance, was the relationship between sedentary behaviour and health explained partially or fully by dietary intake? The vast array of dietary variables made it difficult to identify common dietary variables adjusted for in the analyses among studies reporting positive vs. null findings. Future studies are needed examining the different types of dietary variables that may influence the sedentary behaviour and cardiometabolic health relationship. A suggested starting point would be to examine the dietary factors that have previously been shown to be associated with sedentary behaviour (e.g. total energy intake, fast food intake and snacking) 20 and adverse cardiometabolic health (e.g. energy‐dense foods and sugar‐sweetened beverages) 67, 68, 69. By examining specific dietary factors, it will provide much needed information on what dietary confounders to adjust for and thus provide greater consistency in future analyses.

Finally, there were inconsistencies in the analyses. For example, not all studies accounted for missing data where relevant or for clustering in their sampling design, thus potentially overestimating the statistical power of their findings. Additionally, not all studies adjusted for the same covariates, with only half of the studies adjusting for physical activity. However, it is interesting to note that there were no differences in findings between the studies that adjusted for physical activity and those studies that did not.

Strengths and limitations of review

Strengths of the review include broad search criteria examining many electronic sources, and a large number of studies were screened for eligibility by two reviewers. Additionally, an adapted quality assessment 26, 27 was applied to distinguish low‐ and high‐quality studies and was performed independently by two reviewers. However, we acknowledge that the cut‐off value to distinguish low and high quality is arbitrary. Other quality assessment tools such as the Effective Public Health Practice Project 70 and Downs & Black 71 were considered; however, these quality tools were mostly developed to assess randomized controlled trials and not observational studies. It is also possible that studies with significant findings were more likely to be included in the review. This was due to the inclusion criteria we applied to the review where studies were only included if they examined all three measures of interest in the analyses (sedentary behaviour, dietary intake and a cardiometabolic health outcome). However, if sedentary behaviour or dietary intake was previously found to be not significant in the initial correlation analyses with the health outcome, then typically these variables were not included in the regression analyses.

Conclusion

This systematic review found moderate to strong evidence of the relationships between self‐reported television viewing, total screen time and overall sedentary behaviour with adiposity, independent of dietary intake. It is important to understand the nature of these independent relationships to help with informing the design of future interventions. For example, how much focus needs to be on reducing sedentary behaviours or whether targeting both sedentary behaviour and dietary intake would be more beneficial. However, further research is still needed to examine whether the independent associations remain when using objective measures to assess overall sedentary time. Additionally, exploratory and longitudinal studies would help with understanding whether these behaviours are occurring simultaneously, and whether one behaviour potentially influences the other over time.

Conflict of interest statement

The authors have no conflicts of interest to disclose.

Table S1. Description of the characteristics and main findings.

Table S2. Overall scores of the methodological quality assessment (ranked from highest to lowest).

Click here for additional data file.