the mediating effects of dietary habits on the relationship between television viewing and body mass...

The mediating effects of dietary habits on therelationship between television viewing and bodymass index among youthValerie Carson1 and Ian Janssen1,2

1School of Kinesiology and Health Studies, Queen’s University, Kingston, ON, Canada2Department of Community Health andEpidemiology, Queen’s University, Kingston, ON, Canada

Received 7 October 2011; revised 11 January 2012; accepted 16 January 2012

Summary

Objective: There is evidence to suggest that excessive television viewing is an independent determinantof obesity in young people. However, the pathways between television viewing and obesity are not fullyunderstood. Therefore, the purpose of this study was to examine whether the relationship between televisionand body mass index (BMI) is mediated by television snacking and junk food consumption.

Methods: Results are based on 15 973 youth in grades 6–10 who participated in the Canadian 2009/2010 Health Behaviour in School-Aged Children Survey (HBSC). Participants self-reported their weight andheight and BMI z-scores were calculated based on World Health Organization growth standards. Partici-pants reported the frequency of snacking while watching television and the frequency of eating junk food(sweets, soft drinks, baked goods, French fries, potato chips). Total hours per week of television werecalculated. A contemporary multiple mediation analysis was used to examine associations.

Results: A modest positive relationship was observed between television viewing and BMI. The mean BMIz-score was 0.15 units higher in youth in the highest television viewing quartile by comparison with the youthin the lowest quartile. However, contrary to our hypothesis, television snacking and junk food consumptionwere not significant positive mediators of the television and BMI relationship.

Conclusion: The pathways between television viewing and obesity are complicated and remain poorlyunderstood. Future research using longitudinal or experimental designs, more precise measurement toolsand formal mediation analyses is needed. This research should consider mediators related to both energyintake and expenditure.

Keywords: Adolescent, food habits, obesity, television.

Introduction

To develop effective interventions and public healthstrategies to combat the global pediatric obesity

crisis, it is necessary to understand the determi-nants of obesity and the pathways by which thesedeterminants exude their influence. Three recent lit-erature reviews found a consistent association

Address for correspondence: Dr I Janssen, School of Kinesiology and Health Studies, Queen's University, 28 Division Street, Kingston,ON, K7L 3N6, Canada. E-mail: [email protected]© 2012 The AuthorsPediatric Obesity © 2012 International Association for the Study of Obesity. Pediatric Obesity ••, ••–••

What is already known about this subject• Excessive television viewing is associated with obesityamong the pediatric population.• Although there is limited supporting evidence, it isthought that poor dietary habits explain the relationshipbetween television viewing and obesity.

What this study adds• This was one of the first studies to formally test themediating effects of dietary habits on the television and BMIrelationship.• Dietary habits did not explain the positive relationshipbetween television viewing and BMI in this study.

PEDIATRICOBESITY ORIGINALARTICLE doi:10.1111/j.2047-6310.2012.00049.x

OR

IGIN

AL

AR

TIC

LE

between excessive television viewing and obesityamong children and youth; however, similar findingswere not observed for other screen time behaviourssuch as computer and video game use (1–3). Simi-larly, a recent study of a representative sample ofAmerican children and adolescents reported thattelevision use but not computer use or the overallvolume of sedentary behaviour was associated withobesity (4).

It has been suggested that the relationship be-tween television viewing and obesity is more likelycaused by an increased energy intake thandecreased energy expenditure. Although televisionviewing may be at the lowest end of the energyexpenditure continuum (5), increased televisionviewing was associated with a minimal reduction in24-h energy expenditure in one study (6). Conversely,excessive television viewing among children andyouth has been linked to a higher intake of energy-dense foods, drinks and fast foods (7). This may beattributable in part to the high exposure to junk foodadvertisements (8,9), which may influence youngpeople's food preferences and purchase requests(8). Furthermore, television may act as a prompt toeat and a distraction when eating; thereby, providinga context for snacking and overeating (7,10,11).Amongst the screen time behaviours, this context issomewhat unique to television viewing becauseyoung people's hands are constantly available to usefor eating, unlike video games and computers (12).

Two previous studies have tested whether theinfluence of television on body mass index (BMI)and obesity is mediated by dietary habits (10,13).Francis and colleagues examined whether the rela-tionship between television viewing at age 7 and thechange in BMI between 5 and 9 years of age wasmediated by snacking and fat from snacks at age 7among 173 girls (10). Modest positive mediatingeffects (path coefficients ~.28) of snacking andfat from snacking were observed among girls fromoverweight families but not among girls from non-overweight families (10). Ekelund and colleaguesconsidered whether eating meals while watchingtelevision mediated the relationship between televi-sion viewing and adiposity among ~2000 childrenand youth (13). The relationship between televisionviewing and adiposity was attenuated after the eatingmeals while watching television variable was addedto the model (13). However, the results of these twostudies should be interrupted with caution. Specifi-cally, the generalizability of the findings is limited inthe first study because of the small and homoge-neous sample (10). As the second study did notconduct a formal mediation analysis, it is unknown

whether eating behaviours were mediating or con-founding the relationship (13).

The purpose of this study was to use a contem-porary mediation analysis to examine whether therelationship between television and BMI is mediatedby television snacking and junk food consumption.We examined these relations in a large and repre-sentative sample of Canadian youth. The findingsof this study may help us further understand themechanisms by which excessive television viewinginfluences childhood obesity.

MethodsStudy overview

The study was based on a sample of Canadian youthin grades 6–10 who participated in the HealthBehaviour in School-Aged Children Survey (HBSC).Responses from the HBSC questionnaire wereused to estimate BMI, television viewing, frequencyof snacking while watching television, frequency ofeating junk food and potential confounders. Linearregression models and a contemporary multiplemediation analysis approach (14,15) were used toassess the total, direct and indirect effect of televisionviewing on BMI. The ‘total effect’ represents the rela-tionship between television viewing and BMI, withouttaking into account mediator variables. The ‘directeffect’ represents the relationship between televisionviewing and BMI after accounting for the televisionsnacking and junk food mediator variables (seeFig. 1). The ‘indirect effect’ represents the relation-ship between television viewing and BMI that occursthrough the television snacking and junk food media-tor variables (see Fig. 1).

Study sample

The study is based on the Canadian 2009/2010HBSC. The HBSC is a World Health Organization(WHO)-sponsored cross-sectional survey conductedin over 40 countries. It consists of a classroom basedquestionnaire about health behaviours, lifestylefactors and demographics. The Canadian samplewas designed according to the international HBSCprotocol in that a cluster design was used withschool class being the basic cluster, the distri-bution reflected the distribution of Canadians ingrades 6–10. The sample was weighted to accountfor oversampling of certain provinces and territories.The total sample consisted of 26 078 students whoattended 436 different schools across Canada. Weexcluded participants who did not respond to themeasures of interest, leaving 15 973. In comparisonwith the total sample, the final sample was slightly

2 | V. Carson & I. JanssenO

RIG

INA

LA

RT

ICL

E

© 2012 The AuthorsPediatric Obesity © 2012 International Association for the Study of Obesity. Pediatric Obesity ••, ••–••

older (0.5 years) and included slightly more female(2.8%) and Caucasian participants (9.1%). Ethicsapproval was obtained from the Queen's UniversityGeneral Research Ethics Board. Consent wasobtained from participating school boards, individualschools, parents and students.

Study measuresBMI

Participants self-reported their weight in pounds orkilograms and their height in feet/inches or centime-tres. BMI was calculated (kg m-2), age- and sex-specific BMI z-scores were calculated based onWHO growth standards, and participants were clas-sified as non-overweight, overweight or obese basedon these same standards (16).

Television viewing

The amount of time spent watching television wasassessed using the following question: ‘About howmany hours a day do you usually watch television(including videos and DVDs) in your free time?’ Therewere nine response options for both weekday andweekend use ranging from, ‘None at all’ to ‘7 or morehours a day’. Weighted means for weekday andweekend use were used to calculate total weeklyhours. Then television use was categorized into quar-tiles. A previous validation study reported that a briefquestionnaire used to measure television use, similarto that used in the HBSC, was significantly correlated(r = 0.47) with television use measured by a weeklylog among 11- to 15-year-olds (17).

Television snacking and junk food mediators

Mediating variables included the frequency of snack-ing while watching television and the frequency ofeating junk food. Television snacking was assessedusing the following question: ‘How often do you eat asnack while you watch television (including videos andDVDs)?’ There were seven response options rangingfrom, ‘Never’ to ‘Everyday’. Junk food was assessedby using the following question, ‘How many times aweek do you usually eat or drink . . .?’ (i) ‘sweets(candy or chocolate)’; (ii) ‘coke or other soft drinks thatcontain sugar’; (iii) ‘cake, pastries or donuts’; (iv)‘potato chips’ and (v) ‘French fries’. Each of the fiveitems had seven response options ranging from‘Never’ to Everyday, more than once”. The responsesfrom the five items were averaged to create an overalljunk food score; higher scores reflected a great con-sumption of junk food. The internal consistency for thejunk food items in this study was a = 0.77.

Potential confounders

Confounders considered in the analyses were basedon assumptions of confounding (18) and previousliterature describing relationships between screentime and obesity (2,3). They consisted of gender, agein months, age2, race (Caucasian, other), family struc-ture (both parents, single parent, parent and stepparent, and other) (19), socioeconomic status (SES),currently dieting (yes or no) and dieting in the last12 months (yes or no). SES was measured using thepreviously validated family affluence scale thatincludes the summation of four items regarding family

Exposure: Television

Direct Effect Outcomes:BMI z-score

Indirect Effect a2

Snacking

Confounders:Gender

ageage2

RaceFamily Structure

Socioeconomic Status Currently Dieting

Dieting in the Past Year

Indirect Effect b1

Indirect Effect a1

Indirect Effect b2

Junk Food

Figure 1 The direct effect and indirect effect of television viewing on BMI z-scores, when considering televisionsnacking and junk food consumption as mediator variables (14).

Television, dietary habits and obesity | 3

OR

IGIN

AL

AR

TIC

LE


wealth (car ownership, bedroom sharing, holidaytravel and computer ownership) (20). Participantswere divided into low, medium and high SES groupsusing the established cut-points for this scale (20).

Statistical analysis

Analyses were completing using SAS version 9.2(SAS Institute Inc., Cary, NC, USA). Error estimatesand 95% confidence intervals (CIs) were adjusted inall regression models using SURVEYREG proce-dures to account for clustering by school class andsample weights.

To address the main study objective, the total,direct and indirect effects of television on BMIz-scores were estimated. The total and direct effectswere estimated using multiple linear regressionmodels. The indirect effects were estimated using abootstrap sampling procedure with 1000 resamples,which took into account multiple mediators (14,15).This procedure calculated point estimates and boot-strap percentile 95% CIs based on the product ofregression coefficients estimated in multiple linearregression models (indirect effects a1* b1 and a2* b2 inFig. 1) for each resample (14). There was evidence ofmediation if zero was not included in the CI (14).Multiple mediator models were used to estimate theindirect effect of one mediator variable (e.g., televi-sion snacking) while controlling for the effects ofother mediators (e.g. junk food) (15). Findings fromsimulation research suggest that this bootstrapsample procedure is one of the most valid andpowerful methods for testing mediation (21). Gender,age, age2, race, family structure, SES, currentlydieting and dieting in the past year were the con-founders included in the regression analyses.

Along with the analyses performed to address themain objective, some additional analyses were con-ducted. Specifically, linear regression was used toexamine the relationships between television viewing,television snacking and junk food (path a1 and a2 inFig. 1) and television snacking, junk food and BMIz-scores (path b1 and b2 in Fig. 1). These analysesalso controlled for the confounders used in the mainanalyses. Furthermore, participants' BMI z-scoreswere dichotomized into non-overweight (<+1 stan-dard deviation [SD]) and overweight/obese (+1 SD)and the analyses were repeated using logistic regres-sion models.

ResultsDescriptive analyses

Participant characteristics are in Table 1. Approxi-mately 48% were male and the average age was 14.

Participants watched a median of 16 h of televisionper week. According to the WHO BMI standards,18% of participants were overweight and 8% wereobese.

Relation between television viewingand dietary habits

After adjusting for confounders, a modest positiveassociation was observed between televisionviewing and television snacking. By comparison withtelevision viewing quartile 1, the frequency of televi-sion snacking on the 7-point scale was 0.65 (95%CI: 0.57, 0.74) units higher in quartile 2, 0.92 (0.84,1.01) units higher in quartile 3, and 1.35 (1.26, 1.45)units higher in quartile 4. A similar pattern wasobserved for junk food consumption; in comparisonwith television viewing quartile 1, the frequency ofjunk food consumption on the 7-point scale washigher in quartile 2 (0.19: 0.13, 0.24), quartile 3 (0.37:0.31, 0.42) and quartile 4 (0.61: 0.55, 0.67).

Relation between dietary habitsand BMI z-scores

Surprisingly, after adjusting for confounders, a smallnegative association between television snackingand BMI z-scores was observed. For every 1 unitincrease in the 7 point television snacking scale, the

Table 1 Participant characteristics of the 2009/2010Canadian Health Behaviour in School-Aged ChildrenSurvey

Total (N = 15,973)

Gender (%)Male 48.4Female 51.6

Age 14.1 (12.8–15.3)Race (%)

Caucasian 77.2Other 22.8

BMI (WHO; %)†

Non-overweight 73.9Overweight (+1 SD) 17.8Obese (+2 SD) 8.3

BMI (IOTF; %)†

Non-overweight 77.4Overweight 16.8Obese 5.8

Television, hours week-1 16.0 (7–24)

Data presented as prevalence for categorical variables and as median(interquartile range) for the continuous variables.†BMI = body mass index according to World Health Organization (WHO)(16) and International Obesity Task Force (IOTF) (22) classification.


RIG

INA

LA

RT

ICL

E


BMI z-scores decreased by 0.03 (-0.05, -0.01).Conversely, junk food consumption was not associ-ated with BMI z-scores. Similar magnitude and direc-tion of associations were seen when predictingoverweight/obesity with logistic regression models(data not shown).

Relation between television viewingand BMI z-scoresTotal effect

The association between television viewing and BMIz-scores, without taking into account mediator vari-ables but after adjusted for confounders, are shownin Table 2. There was a modest positive relationshipsuch that the mean BMI z-score was 0.15 unitshigher in youth in the highest television viewingquartile by comparison with the youth in the lowestquartile.

Direct effect

The associations between television viewing andBMI z-scores, after accounting for the televisionsnacking and junk food mediator variables and theconfounders, are also shown in Table 2. Similar tothe total effect, a positive relationship was observedfor the direct effect; however, the relationship wasstronger. For example, the mean BMI z-score was0.22 units higher in youth in the highest televisionviewing quartile by comparison with the youth in thelowest quartile.

Indirect effect

The indirect effects of snacking and junk food on therelationship between television viewing and BMIz-scores are shown in Table 3. Junk food consump-

Table 2 The total and direct effects of television viewingon BMI z-scores in the 2009/2010 Canadian HealthBehaviour in School-Aged Children Survey

Television Total effect Direct effect

b 95% CI† b 95% CI†

Quartile 1 Referent ReferentQuartile 2 0.06 -0.01, 0.14 0.10 0.02, 0.17*Quartile 3 0.12 0.05, 0.20* 0.17 0.10, 0.25*Quartile 4 0.15 0.06, 0.24* 0.22 0.14, 0.31*

*P � 0.05.†b = beta coefficient and 95% CI = 95% confidence interval.The Total effect model was adjusted for gender, age, age2, race, familystructure, socioeconomic status, currently dieting and dieting in the pastyear.The Direct effect model was adjusted for television snacking and junk foodconsumption in addition to gender, age, age2, race, family structure, socio-economic status, current dieting and dieting in the past year.

Table

3Th

ein

dire

ctef

fect

sof

snac

king

and

junk

food

onth

ere

latio

nshi

pbe

twee

nte

levi

sion

and

BM

Iz-

scor

esin

the

2009

/201

0C

anad

ian

Hea

lthB

ehav

iour

inS

choo

l-Age

dC

hild

ren

Sur

vey

Tele

visi

onS

pec

ific

ind

irect

effe

ctfo

rte

levi

sion

snac

king

Sp

ecifi

cin

dire

ctef

fect

for

junk

food

cons

ump

tion

Tota

lind

irect

effe

ct(s

nack

ing

+ju

nkfo

od)

Poi

ntes

timat

eP

erce

ntile

95%

CI†

Poi

ntes

timat

eP

erce

ntile

95%

CI†

Poi

ntes

timat

eP

erce

ntile

95%

CI†

Qua

rtile

1R

efer

ent

Ref

eren

tR

efer

ent

Qua

rtile

2-0

.03

-0.0

4,-0

.02*

<-0.

01-0

.01,

<0.0

1-0

.03

-0.0

4,-0

.02*

Qua

rtile

3-0

.04

-0.0

5,-0

.03*

<-0.

01-0

.01,

0.01

-0.0

4-0

.06,

-0.0

3*Q

uart

ile4

-0.0

6-0

.08,

-0.0

4*<-

0.01

-0.0

2,0.

01-0

.06

-0.0

9,-0

.04*

*P<

0.05

.† P

erce

ntile

95%

CI=

boot

stra

ppe

rcen

tile

95%

conf

iden

cein

terv

al.

Gen

der,

age,

age2 ,

race

,fa

mily

stru

ctur

e,so

cioe

cono

mic

stat

us,

curr

ently

diet

ing

and

diet

ing

inth

epa

stye

arw

ere

adju

sted

for

inal

lana

lyse

s.


OR

IGIN

AL

AR

TIC

LE


tion did not mediate the relationship between televi-sion viewing and BMI z-scores. Conversely, televisionsnacking did mediate the relationship; however, itwas a negative mediator. This implies that snackingdid not explain the positive relationship between tele-vision viewing and BMI.

Similar magnitude and direction of associationswere seen for total, direct and indirect effects whenthe categorical obesity measure was used in place ofcontinuous BMI z-scores (data not shown).

DiscussionThis study examined whether television snackingand junk food consumption mediated the relation-ship between television viewing and BMI in a largecross-sectional sample of Canadian youth. Althougha modest positive relationship was observed be-tween television viewing and BMI, this positive rela-tionship was not attributable to increased snackingand junk food consumption in youth with higher BMIvalues. In fact, to our surprise, a negative relation-ship, albeit a small one, was observed between tele-vision snacking and BMI.

Our finding that television viewing is positively re-lated to snacking (7) and BMI z-scores (1–3,23) inyouth is consistent with several literature reviews ofcross-sectional studies, longitudinal studies and ran-domized controlled trials. Meta-analyses conductedwithin the literature reviews examining the relation-ship between television viewing and BMI havereported small (1) to modest (23) effects, which isalso consistent with the present study. These find-ings suggest that interventions and initiatives aimingto reduce obesity in young people may have agreater impact if they target multiple behaviours,such as moderate-to-vigorous intensity physicalactivity, television viewing and dietary habits (24).

Also consistent with the findings regarding televi-sion snacking of the present study, two previouscross-sectional studies observed that the frequencyof snacking (25,26), frequency of television food con-sumption (26) and the percentage of energy intakefrom snacks (25) was a modest negative predictor ofoverweight and obesity among youth. The counterintuitive observation could potentially be explained byinformation bias and/or study design issues. Morespecifically, recall and social desirability biases asso-ciated with the self-report measures of televisionsnacking, height and weight in the present study mayhave resulted in measurement error. For example,previous studies have shown that overweight youthare more likely to underestimate their dietary intakeas well as their weight in comparison with non-

overweight youth (27–29). Furthermore, the presentstudy measured the frequency of snacking and notportion sizes or the type of food consumed whilesnacking. Therefore, it is possible that youth withgreater BMI z-scores ate larger portions of food ormore energy-dense foods when snacking, leading tohigher caloric intake of snacks compared with youthwith healthier BMI z-scores (26). Finally, the tempo-rality of relations between variables observed in thiscross-sectional study needs to be considered (30).Specifically, overweight youth may have altered theirdietary patterns in response to their weight (25,30),and even though we adjusted for dieting, residualconfounding may have occurred (18).

Compared with the previous two studies that founda modest negative relationship between snackingand overweight and obesity among youth, the nega-tive relationship observed in this study was small.Because of the sample size, the present study hadthe power to detect small associations. It is impor-tant to note that although this association was sta-tistically significant, it is likely too small to be ofclinical significance. In fact, the difference in averagesnacking score (measured on a 7-point scale)between other, overweight and obese groupswas � 0.1.

To our knowledge, only two previous studies havetested the mediating effects of dietary habits on therelationship between television viewing and BMIamong young people (10,13). Although both of thesestudies observed mediating effects (10), only one ofthem reported the magnitude and direction of theeffects (10). In this longitudinal study, modest positivepath coefficients (~0.28) were observed betweentelevision viewing, snacking frequency, fat fromsnacks, and change in BMI among girls from over-weight families (10). Thus, in contrast to the presentstudy, the mediating variables explained the positiveassociation between television and BMI. The moreprecise measures of snacking and BMI, the stratifi-cation of overweight and non-overweight families, aswell as the longitudinal study design may explain thedifference in findings from those observed in ourstudy (30).

Although we focused on the mediating effects ofdietary aspects (and therefore of energy intake) onthe relation between television and BMI, we cannotdiscount the potential relevance of the energy expen-diture aspects. Although one study found a minimalrelation between 24-h energy expenditure and tele-vision watching (6), even a modest change in energyexpenditure may have a meaningful impact on BMIwhen added up over weeks, months or even years.This is especially the case given the amount of time


RIG

INA

LA

RT

ICL

E


today's youth spend watching television (2.5–4 h day-1 on average in Canadian and Americanyouth) (31,32).

Strengths of this study include the use of largepopulation-based samples and the multiple mediatoranalyses. A primary limitation of this study is thecross-sectional design, which limits the ability tomake causal inferences about the relationships.Additionally, the biases associated with the use ofself-report data for all of the study variables may haveresulted in an under or overestimation of the trueassociations. The fact our final sample was not rep-resentative of the population in terms of age, gender,and race is another limitation. Furthermore, the rela-tionship between television viewing and obesity maydiffer by the type of television programming (33);however, the HBSC survey did not access thecontent type of television viewing. Finally, although avariety of potential confounders were considered, it ispossible that other unmeasured confounders had aninfluence on the relationships, such as growth rateand total energy intake.

ConclusionTelevision viewing was positively associated withBMI, snacking, and junk food consumption. How-ever, contrary to our hypothesis, television snackingand junk food consumption were not significant posi-tive mediators of the television viewing and BMI rela-tionship. Because of limitations within the existingliterature and an overall dearth of information, thepathways between television viewing and obesityappear complicated and remain poorly understood.Future research using longitudinal or experimentaldesigns, more precise measurement tools, andformal mediation analyses is needed that con-siders mediators related to both energy intake andexpenditure.

Conflict of InterestStatementThe authors declare that there are no conflicts ofinterest.

AcknowledgementsThe authors would like to thank Matthew King for hisassistance with the management of the HBSC study.This study was funded by an operating grant from theCanadian Institutes of Health Research (MOP 97962),a second operating grant co-funded by the CanadianInstitutes of Health Research and the Heart andStroke Foundation of Canada (PCR 101415), and a

contract from the Public Health Agency of Canada(Contract: HT089-05205/001/SS), the latter of whichfunds the Canadian version of the WHO-HBSC.The WHO-HBSC is a WHO/Euro collaborative study.International Coordinator of the 2009/2010 study:Candace Currie, University of St. Andrew's, Scotland;Data Bank Manager: Oddrun Samdal, University ofBergen, Norway. This publication reports data solelyfrom Canada (Principal Investigators: John Freemanand William Pickett). VC was supported by a CIHR –Frederick Banting and Charles Best Doctoral Award.IJ was supported by a Canada Research Chair and anEarly Researcher Award from the Ontario Ministry ofResearch and Innovation.

References

1. Marshall SJ, Biddle SJ, Gorely T, Cameron N, Murdey I.Relationships between media use, body fatness and physi-cal activity in children and youth: a meta-analysis. Int JObes Relat Metab Disord 2004; 28: 1238–1246.2. Rey-Lopez JP, Vicente-Rodriguez G, Biosca M,Moreno LA. Sedentary behaviour and obesity developmentin children and adolescents. Nutr Metab Cardiovasc Dis2008; 18: 242–251.3. Must A, Tybor DJ. Physical activity and sedentary beha-vior: a review of longitudinal studies of weight and adiposityin youth. Int J Obes (Lond) 2005; 29 (Suppl. 2): S84–S96.4. Carson V, Janssen I. Volume, patterns, and types ofsedentary behavior and cardio-metabolic health in childrenand adolescents. BMC Public Health 2011; 11: 274–284.5. Klesges RC, Shelton ML, Klesges LM. Effects of televi-sion on metabolic rate: potential implications for childhoodobesity. Pediatrics 1993; 91: 281–286.6. Grund A, Krause H, Siewers M, Rieckert H, Muller MJ.Is TV viewing an index of physical activity and fitness inoverweight and normal weight children? Public Health Nutr2001; 4: 1245–1251.7. Pearson N, Biddle SJ. Sedentary behavior and dietaryintake in children, adolescents, and adults a systematicreview. Am J Prev Med 2011; 41: 178–188.8. Coon KA, Tucker KL. Television and children's con-sumption patterns. A review of the literature. MinervaPediatr 2002; 54: 423–436.9. Harrison K, Marske AL. Nutritional content of foodsadvertised during the television programs children watchmost. Am J Public Health 2005; 95: 1568–1574.10. Francis LA, Lee Y, Birch LL. Parental weight statusand girls' television viewing, snacking, and body massindexes. Obes Res 2003; 11: 143–151.11. Williams DM, Raynor HA, Ciccolo JT. A review of TVviewing and its association with health outcomes in adults.Am J Lifestyle Med 2008; 2: 250–259.12. Owen N, Healy GN, Matthews CE, Dunstan DW. Toomuch sitting: the population health science of sedentarybehavior. Exerc Sport Sci Rev 2010; 38: 105–113.


OR

IGIN

AL

AR

TIC

LE


13. Ekelund U, Brage S, Froberg K, et al. TV viewing andphysical activity are independently associated with meta-bolic risk in children: the European Youth Heart Study.PLoS Med 2006; 3: e488.14. Hayes AF. Beyond Baron and Kenny: statisticalmediation analysis in the new millennium. CommunMonogr 2009; 76: 408–420.15. Preacher KJ, Hayes AF. Asymptotic and resamplingstrategies for assessing and comparing indirect effects inmultiple mediator models. Behav Res Methods 2008; 40:879–891.16. de Onis M, Onyango AW, Borghi E, Siyam A,Nishida C, Siekmann J. Development of a WHO growthreference for school-aged children and adolescents. BullWorld Health Organ 2007; 85: 660–667.17. Schmitz KH, Harnack L, Fulton JE, et al. Reliability andvalidity of a brief questionnaire to assess television viewingand computer use by middle school children. J Sch Health2004; 74: 370–377.18. Rothman KJ, Greenland S, Lash TL. Modern Epide-miology, Third Edition. Lippincott Williams & Wilkins: Phila-delphia, PA, 2008.19. Griesbach D, Amos A, Currie C. Adolescent smokingand family structure in Europe. Soc Sci Med 2003; 56:42–52.20. Boyce W, Torsheim T, Currie C, Zambin A. The familyaffluence scale as a measure of national wealth: validationof an adolescent self-report measure. Soc Indic Res 2006;78: 473–487.21. MacKinnon DP, Lockwood CM, Hoffman JM,West SG, Sheets V. A comparison of methods to testmediation and other intervening variable effects. PsycholMethods 2002; 7: 83–104.22. Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishinga standard definition for child overweight and obesity world-wide: international survey. BMJ 2000; 320: 1240–1243.23. Tremblay MS, Leblanc AG, Kho ME, et al. Systematicreview of sedentary behaviour and health indicators inschool-aged children and youth. Int J Behav Nutr Phys Act2011; 8: 98–119.

24. Driskell MM, Dyment S, Mauriello L, Castle P,Sherman K. Relationships among multiple behaviors forchildhood and adolescent obesity prevention. Prev Med2008; 46: 209–215.25. Keast DR, Nicklas TA, O'Neil CE. Snacking is associ-ated with reduced risk of overweight and reduced abdomi-nal obesity in adolescents: national Health and NutritionExamination Survey (NHANES) 1999–2004. Am J Clin Nutr2010; 92: 428–435.26. Vader AM, Walters ST, Harris TR, Hoelscher DM. Tele-vision viewing and snacking behaviors of fourth- andeighth-grade schoolchildren in Texas. Prev Chronic Dis2009; 6: A89.27. Fisher JO, Johnson RK, Lindquist C, Birch LL,Goran MI. Influence of body composition on the accuracyof reported energy intake in children. Obes Res 2000; 8:597–603.28. Bandini LG, Schoeller DA, Cyr HN, Dietz WH. Validityof reported energy intake in obese and nonobese adoles-cents. Am J Clin Nutr 1990; 52: 421–425.29. Sherry B, Jefferds ME, Grummer-Strawn LM. Accu-racy of adolescent self-report of height and weight inassessing overweight status: a literature review. ArchPediatr Adolesc Med 2007; 161: 1154–1161.30. Rodriguez G, Moreno LA. Is dietary intake able toexplain differences in body fatness in children andadolescents? Nutr Metab Cardiovasc Dis 2006; 16: 294–301.31. Carson V, Pickett W, Janssen I. Screen time and riskbehaviors in 10- to 16-year-old Canadian youth. Prev Med2010; 52: 99–103.32. Rideout VJ, Foehr UG, Roberts DF. Generation M2

Media in the lives of 8- to 18-year-olds: a Kaiser FamilyFoundation study. 2010. Available at: http://www.kff.org/entmedia/mh012010pkg.cfm. Accessed September,2011.33. Zimmerman FJ, Bell JF. Associations of televisioncontent type and obesity in children. Am J Public Health2010; 100: 334–340.


RIG

INA

LA

RT

ICL

E


the mediating effects of dietary habits on the relationship between television viewing and body mass...

Documents