Salt Content Impacts Food Preferences and Intake among Children

含盐量影响儿童的食物偏好和摄入量

SaltContentImpactsFoodPreferencesandIntakeamongChildren

SofiaBouhlal1,2,3,ClaireChabanet1,2,3,SylvieIssanchou1,2,3,SophieNicklaus1,2,3*

tetdel’Alimentation,Dijon,France,2INRA,UMR1324,CentredesSciencesduGou tetdel’Alimentation,Dijon,France,1CNRS,UMR6265,CentredesSciencesduGou

´ tetdel’Alimentation,Dijon,France3UniversitedeBourgogne,CentredesSciencesduGou

Introduction

Therelationshipbetweendietarysodium(Na)intakeandtheprevalenceofsomediet-relateddiseaseshaslongbeenrecognized[1,2].Thus,reducingdietarysodiumintakehasbecomeapublichealthpriorityinseveralcountries.Consequently,somegovern-mentshaveenactedpoliciesaimedatreducingthesalt(NaCl)contentoffoods[3],assodiumlevelsareoftenincreasedbysaltinfood(1gofNaCl=400mgofNa).Giventheimportanceofchildhoodinestablishingfoodpreferencesandhealthyeatinghabits[4,5],analyzingtheroleofsaltinchildren’sfoodpreferencesandintakeisimportant.Therecommendedsaltintakeforchildrenaged8–11yearsis5gperday[6].InFrance,saltintakeinthispopulationis6gperdayandincreaseswithage[7].Saltcontributestothetasteoffoodsandmakesthemmoreenjoyable[8].Forchildren,theroleoffoodsensoryproperties(i.e.,taste,texture,smell)oneatingbehavioriskey[9];thus,understandingtheimpactofsaltreductiononfoodpreferencesandfoodintakeisrelevant[10,11].

Inadults,apositiveassociationwasfoundbetweenperceivedsaltinessandpreferencesforsnackfoodswithvaryingsaltcontents,andapositivelinkwasalsofoundbetweenlikingsaltysnacksanddietarysodiumintake[12].Moreover,saltinessimpactedprefer-encesdifferentlydependingonthetypeoffoodtested,supportingthehypothesisthattherearefood-specificpreferencesforhighsaltcontent[8,13].Inaddition,anotherstudyofadultsfoundadifferentimpactofsaltcontentonpreferencesandintake:researcherscomparedtheintakeoffoodswithvaryingsaltcontenttotheirhedonicratingsafterasip-and-spittasting.Theoptimalsalinityrevealedbythehedonicratingwashigherthantheoptimallevelrevealedbytheintakedata[14].Astudyof10-to12-year-oldchildrenshowedthatthosewholikedsaltyfoods(asreportedbytheirmother)ingestedmoresalt(asestimatedbyurinarysodiumexcretion)[15].Anotherstudyshowedapositiveeffectofsaltcontentonfoodintakein2-to3-year-olds[16].However,itcouldnotaddresstheroleofhedonicresponsesinchildrenthisyoungformethodologicalreasons[17].Onemightwonderifthefoodintakeofolderchildrenwouldreflectthesameresponsetosaltcontentvariationsandtowhatextenttheirintakewouldrelatetotheirlikingofthefood.Theobjectiveofthepresentstudyisthustoevaluatetheimpactofsaltcontentonschool-agechildren’sfoodintakeandfoodhedonicratings.Basedontheresultsobservedwiththetoddlerpopulation[16],onewouldhypothesizethatsaltcontentwouldalsohaveapositiveeffectonintakeinanoldergroupofchildren(e.g.,childrenaged8–11years)andthatthiseffectwouldbefoodspecific.Tothebestofourknowledge,

such

含盐量影响儿童的食物偏好和摄入量

aninvestigationoftheinfluenceofsaltcontentonfoodhedonicratingsandintakehasneverbeenconductedinthesamegroupofchildren.

Thus,thefirstobjectiveofthepresentstudywastoevaluatetheeffectofsaltcontentonschool-agechildren’smeasuredintakeandhedonicresponsesfortwofoods(greenbeansrepresentingvegetablesandpastarepresentinggrains).Thesecondobjectivewastodeterminetherelativeimportanceofseveralfactorspotentiallyinfluencingchildren’sfoodintake(saltcontent,hedonicresponses,hungerandweightstatus).

Experimentallunches:measurementsoffoodintake.Theintakemeasurementsweretakeneveryotherweek

MaterialsandMethods

Thisstudyusedawithin-subjectcrossoverdesign.ItwasconductedfromOctobertoDecember2009inDijon(France).

1.Participants

Thestudywasconductedaccordingtothe

guidelineslaiddownintheDeclarationofHelsinkiandwasapprovedbyalocalethicscommittee(Comite´deProtectiondesPersonnesEstIBourgogne,2009/39;AFSSAPSIDRCB:2009-A00966-51).ThestudytookplaceinoneschoolinDijon(France).Parentsreceivedwritteninformationaboutthestudyandallowedtheirchildtoparticipatewithasignedconsentform.Twosetsofparentsrefusedtoallowtheirchildtoparticipate.

Parentsprovidedinformationabouttheirchild’sbodyweight(kg)andheight(m)basedontheirchild’shealthrecords;z-scoresforBMI(z-BMI)werethencalculated[18].Thesampleincludedatotalof106childrenbetweentheagesof8and11years.Theirparents’socioeconomicstatusrangedfrommid-leveltohigh.

Ethicsstatement.

2.ExperimentalDesign

Onthebasisofapreviousstudyconductedwithtoddlers,inwhichsaltcontentmodificationhadasignificantimpactongreenbeanandpastaintake[16],apoweranalysis[19]showedthat33childrenwerenecessarytodetectadifferenceof8gofgreenbeanintakebetweenanunsaltedvariantandavariantwith0.6%addedsalt(settingthesignificancelevelat0.05fortypeIerrorandthepowerat0.80).

Thechildren’sintakewasmonitoredfivetimesinaschoolcafeteriaatlunchtime.Afterthefivelunches,thechildrencametothelaboratorytoevaluatethetwotargetfoods.Thechildrenwereblindedtotheaimofthestudy.Theparentswereaskednottosharethestudyaimswiththeirchildren.Theschoolstaffmemberswereawarethatsaltcontentwasmodifiedbutwereblindedtotheexperimentaldesign.

Testfoods.Thetargetfoodsinwhichthesaltcontentwasmanipulatedweregreenbeansandpasta.Thesefoodsareusuallyservedintheschoolcafeteriawithaddedsaltandbutter.Thethreesaltconditionswerethesameasthoseusedinapreviousstudywith2-to3-year-oldtoddlers[16]:noaddedsalt,0%;0.6gofsaltaddedto100goffood;ortwicethisquantity,1.2gofsaltaddedto100goffood.Theselevelscorrespondto0,240and480mgofsodiumper100gofgreenbeansandto5,245and485mgofsodiumper100gofpasta,respectively.Theintermediatesaltcontent(0.6%)isthesamelevelthatistypicallyusedforfoodsinschoolcafeteriasandwasthusconsideredthereferencesaltlevel.Unseasonedfrozengreenbeans(CMICarrefour,LesUlis,France)anddriedpasta(Fusilli,Panzani,Lyon,France)wereboiledwithtapwater,withoutaddingsalt,oiloranyflavoring.Unsaltedbutter(2.5gper100goffood)wasaddedtothefoodforallconditions.Saltwasaddedbytheexperimentersaccordingtotheexperimen-taldesign.Thefoodandingredientswerestirredthoroughlytoensureahomogenousmixture.

atlunchtime(between11:45amand12:45pm).Saltshakersandsaltedseasonings(ketchup,mustard)werenotavailableduringtheexperiment.Intakeofeachfoodonthemenu(servingminusleftovers,ing)wasmeasuredforeachchildtothenearest1g(Soehnle,LeifheitGroup,Germany).Thefirstobservationinvolvedthecontrolcondition,inwhichthegreenbeansandpastawerepreparedwiththeintermediatesaltcontent.Foreachofthefollowingobservations,thesaltcontentvariedonlyinthegreenbeansorinthepasta.Forinstance,ifthesaltcontentwasdifferentfrom0.6%inthegreenbeans,thepastawasservedwith0.6%saltcontent.

Anexperimentalfixedmenuwasset.Allthemenuitemswerechosenbecausetheywereregularlyconsumedattheschoolcafeteriaandathome.Theyincludedavegetablesalad(asingleservingof50g;0.87kcal/gor3.64kJ/g;3.7mgofNa/g),chicken(asingleservingof50g;2.26kcal/gor9.46kJ/g;4.2mgofNa/g),greenbeans(aninitialservingof100gandasecondservingof100g;0.37kcal/gor1.55kJ/g),pasta(aninitialservingof100gandasecondservingof100g;1.4kcal/gor5.86kJ/g),cheese(asingleservingof18g;2.39kcal/gor10kJ/g;7.8mgofNa/g;LavachequiritH,FromageriesBEL,Paris,France),fruitpuree(asingleservingof100g;0.57kcal/gor2.38kJ/g;0.1mgofNa/g;HeroH,Allex,France),bread(adlibitum;2.65kcal/gor11.1kJ/g;6.5mgofNa/g)andwater(adlibitum).TheportionsizeswereadaptedfromtheFrenchguidelinesforschoolcanteenusage[20].Thevegetablesalad(ormace´doine)wasamixtureofvegetables,includingpeasandsmallpiecesofgreenbeans,carrotsandturnips;itwaspreparedwithmayonnaise(7g/100g).Thechickenwaspoachedandgrilledintheoven.Bothfoodswerepreparedbythecanteencookasusual,followingthesamerecipeforallfivelunches.Thefoodswerepre-portionedbytheexperimentersbeforelunchtimeandkeptwarm(+63uC)untillunchtime.Thechildrenwerefreetoaskforoneadditionalservingofbothtargetfoods,aswascustomaryattheschool,andtheexperimenterdocumentedanyadditionalservings.Theleftoversfromeachchildweremeasuredinakitchenareaoutofthechildren’ssight,sothattheywerenotawarethattheirfoodintakewasmonitored.

Sensoryevaluation.Thechildrenparticipatedinasensoryevaluationsessiontoevaluatethethreevariantsofeachtargetfood.Theygaveahedonicrating,apreferencerankingandaperceivedsaltinessranking.Thissensoryevaluationsessionwasorganizedaspartofaschoolouting.Ittookplaceat10:00amattheCentredesSciencesduGou tetdel’Alimentationlaboratorybecausethisisthetimeofdaywhenchildrenaremostalerttoperformsensorytesting[21].Duringthissession,thegroupwasdividedintotwosub-groups:onegroupstartedwithpastasamples,andtheothergroupstartedwithgreenbeansamplestocounteractpotentialordereffects.First,eachsub-groupgavetheirhedonicratingofthefoodonafivepointfacialscale(from‘verybad’,22,to‘verygood’,+2)[17].Thesampleswerepresentedoneatatime,followingaWilliamsLatinSquaredesign.Thechildrenwereinstructedtorinsetheirmouthswithwaterbetweenthetwosamples(EvianHwater,France;6.5mgofsodium/L).Second,theywereservedthethreevariantsconsecutivelyandaskedtorankthemaccordingtotheirpreference(from‘mostpreferred’,+1,to‘leastpreferred’,21)andthenaccordingtotheirsaltiness(from‘mostsalty’,+1,to‘leastsalty’,21).Tieswerenotallowed.Inthisstudy,bothanoverallhedonicratingandapreferencerankingweremeasuredbecausethesetwovariablesprovideduswithdifferentinformationaboutthechild’shedonicresponses.Thehedonicratingscoremadeitpossibleforachildtoscorea

含盐量影响儿童的食物偏好和摄入量

foodpoorly,regardlessofitssaltcontent,ifs/hedidnotlikeitingeneral.Thepreferenceranking‘forced’thechildtodecidewhichvariantwaspreferred,regardlessofher/hisopinionofthegivenfood.Thefoodswerekeptwarmafterthesaltwasadded,and30gwasservedina6oz.(175ml)isotherm-coveredcup(45uC).

3.StatisticalAnalysis

First,univariateanalyseswerecarriedouttoinvestigatetheeffectsofsaltcontentandz-BMIontargetfoodintake,hedonicscoreandpreferencerankings.Second,amultivariateanalysiswasperformedusingStructuralEquationModeling(SEM)toevaluatetherespectiveweightsofthedifferentvariablesininfluencingtheintakeofthetargetfoods.Thesignificancelevelwassetatp,0.05.Univariateanalyses.TheanalyseswerecarriedoutusingSASSoftware(9.1,SASInstituteInc.,Cary,NC,USA).Intake(g)andhedonicscoredataforbothtargetfoodswereanalyzedusingalinearmixedmodel(SASMixedprocedure),with‘child’consid-eredasarandomeffect.Theprimaryfactortestedinthemodelwassaltcontent,andz-BMIwasintroducedasacovariate.Themodelwasthefollowing:Intake(orhedonicrating)=food+saltcontent+z-BMI+food6saltcontent+food6z-BMI+saltcontent6z-BMI+error.The‘Empirical’optionwasspecifiedtoincludeasandwichestimatorforthevariance-covariancematrixofthefixed-effectparameters.Ifasignificantfood6saltcontentinteractionwasfound,thedatawereanalyzedforeachtargetfoodseparatelyusingthemixedlinearmodel:Intake(orhedonicrating)=saltcontent+z-BMI+saltcontent6z-BMI+error.Thepreferencerankingsandsaltinessrankingsforeachfoodwereanalyzedaccordingtothemodel:Ranking=saltcontent+error.Ifasignificanteffectwasrevealed,multiplemeancomparisonswereappliedusingtheDunnetttest:thereferencesaltcontent(0.6%addedsalt)wascomparedtotheothervariants.Thedataarereportedasmeans6SEM.

Structuralequationmodeling(SEM)analysis.StructuralEquationModelingandbootstrapanalyseswereconductedwithR2.10.1,usingtheSEMpackage[22]toassessacausalpathwayderivedfromtheory[23].

Datasetstructureandvariabledefinition.TheSEMmethodcanhandlemultileveldata[24].Inthisstudy,thefollowingvariablesweretakenintoaccount:intake(ofvegetablesalad,chicken,greenbeansandpasta),hedonicscoreandpreferenceranking(ofgreenbeansandpastawithdifferentsaltcontents),saltcontentandz-BMI.Thereweretwolevelsofdata,assomevariableswererecordedatthemeallevel(level1;intakeofeachfooditem),whereasotherswererecordedatthechildlevel(level2;hedonicratingorpreferencerankingforgreenbeansorpastawithagivensaltcontent,z-BMI).Forgreenbeansandpasta,thehedonicratingofthevariantwith0.6%addedsaltwasusedasameasureofthe‘overallhedonicrating’ofeachfood,andthehedonicratingdifferencerelativetothevariantwith0.6%addedsaltwasusedasameasureof‘relativerating’.Alatentconstructcalled‘preference’wasdefinedusingtheobservedvariablesofpreferencerankingandrelativeassessment,astheyactuallymeasurethesameconstruct.Onelatentconstruct(‘preference’)wasdefinedforeachtargetfood.Anotherlatentconstructcalled‘hunger’wasdefinedatthemeallevelandwasthoughttoprovideinformationaboutthevariabilityinintakefromachildtoanother.Thisconstructwasthoughttohaveanimpactonthechild’sintakeofbothtargetfoodsandontheintakeofallfoodsconsumedbeforethemoratthesametime(vegetablesaladandchicken).

Becausewetookintoaccountthevariabilityinchildren’sintakeusingthelatentconstruct‘hunger’,andgiventherecognizedrelationshipbetweenfoodintakeandBMI[25],wehypothesizedthatz-BMIwouldberelatedtothe‘hunger’construct.

Thus,themodelingapproach,basedonacorrelationmatrix,allowedustoevaluateandcomparethevariablesinfluencingfoodintake[23]:saltcontent,hedonicratingandpreferenceranking,thechild’s‘hunger’andz-BMI.Pearson(whenbothvariableswerecontinuous),polyserial(whenonevariablewasdiscreteandtheotherwascontinuous)orpolychoriccorrelations(whenbothvariableswerediscrete)werecalculated[26].

Modelconstructionandestimation.Themaximumlike-lihoodestimatorwasusedtoestimatethemodelparameters(loadings).Thefirstmodel,basedontheoreticalconsiderations,includedaneffectoftheoverallhedonicratingofgreenbeansandpasta(i.e.,aratingofthe0.6%addedsaltvariant)onintake;thiseffectoccurredatthechildlevel.Italsoincludedthelatentconstruct‘preference’.Theeffectof‘preference’onintakewasconsideredforbothtargetfoods.Aquadraticeffectofsaltcontenton‘preference’wasaddedforgreenbeansandpasta.Thus,thesaltcontentvariableanditssquarewereincludedafterorthogo-nalization[27].Thereby,weconsideredthatahighersaltcontentcouldberelatedtoahigherpreferenceranking,whichinturncouldberelatedtohigherintake.

Thesameschemewasusedforgreenbeansandforpasta.Thecorrelationbetweentheoverallandrelativehedonicratingswasincluded.Thecorrelationbetweentheoverallhedonicratingofgreenbeansandtheoverallhedonicratingsofpastawasaddedtothemodeltoaccountforinter-raterdifferencesinusingthehedonicratingsscale.Theeffectsofvegetablesaladintakeontheintakeofgreenbeans,pastaandchickenwereadded(settothesamevalue)toaccountforthefactthatachildwhohadeatenalargeamountofsaladmightnothavebeenashungryfortheremainderofthemeal.

Thenestedmodelobtainedbyapplyingequalityconstraintstothegreenbeanandpastaparameters(theeffectofsaltcontenton‘preference’,theeffectof‘preference’onintake,theeffectoftheoverallhedonicratingonintake,‘preference’loadingsonpreferencerankingandrelativehedonicrating)wascomparedtotheunconstrainedmodel.

Finally,theeffectofz-BMIon‘hunger’wasinvestigated,hypothesizingthathigherz-BMIwouldcorrelatewithhigheroverallintake.Next,modificationindiceswereconsideredtoimprovethemodel,andtheyledustoconsideraspecificz-BMIeffectonthehedonicratingandintakeofpasta.Whenthesetwoeffectswereintroducedintothemodel,thez-BMIeffecton‘hunger’wasnolongersignificant,suggestingaspecificeffectofz-BMIonthehedonicratingandintakeofpastaratherthanageneraleffecton‘hunger’.Becausethisresultwasconsistentwithapreviousobservationintoddlers[16],thismodificationwasimplementedinthemodel.Themodificationindicesalsosuggestedalinkbetweentheoverallhedonicratingofgreenbeansandvegetablesaladintake;thislinkwasaddedtothemodel.Thiseffectmakessensebecausethevegetablesaladincludedpiecesofgreenbeans.

Comparisonsbetweenthetheoreticalmodelandthemodifiedmodelsweretestedwithlikelihoodratiotests(p,0.05).

Becausetheobservationswereclusteredbychild,theassump-tionofindependentobservationswasnotmet.Therefore,thestandarderrorsand95%confidenceintervalsoftheparameterswereobtainedusingbootstrapresampling(B=100),withtheresamplingperformedatthechildlevel[24,28].

Modelvalidationandindicatorsofmodelfit.Thefitisconsideredgoodwhenthemodelisparsimoniousandwhenthefittedcovariancematrixiscloseenoughtotheobservedcovariancematrix.Asrecommended,severalfitindicesthataddressdifferentaspectsofmodelfitarereportedhere:thechi-squaretest,theTucker-LewisNon-NormedFitIndex(NNFI),theBentler

含盐量影响儿童的食物偏好和摄入量

SaltImpactsChildren’sFoodPreferenceandIntake

ComparativeFitIndex(CFI)andtheRootMeanSquareErrorofApproximation(RMSEA).TheRMSEAisexpectedtobeascloseaspossibleto0,withavalueofabout0.05consideredgood,whiletheNNFIandtheCFIareexpectedtobeascloseaspossibleto1,withvaluesofabout0.90consideredacceptable.Thechi-squarestatisticisalsoreportedasarelevantindex,butitwasnotusedtoperformaformaltestbecausemodelsareusuallyrejected[29].

Results

Datafor75children(35girls)weretakenintoaccountinthepresentanalysis.Threehundredforty-ninemealswererecorded.One,3,17and54childrentookpartin2,3,4and5meals,respectively.Thechildren’smeanagewas9.460.1years(range:8to11years),withameanBMIof17.160.3kg/m2andameanz-BMIof0.4960.15SD(range:2.28to4.23SD).Ofthesechildren,11%wereconsideredoverweightorobese(z-BMI.+2SDs)[30].Overthefivelunches,theaveragetotalintakewas36769g(range:137to520g),andtheaveragetotalenergyintakewas41269kcal(range:172to570kcal)or1723638kJ(range:720to2384kJ).Apairedttestshowedthatoverthefivelunches,greenbeanintake(7564g;range:4to160g)waslowerthanpastaintake(13665g;range:61to200g)(t74=10.48;p,0.0001).

Univariateanalysis:Theimpactofsaltcontentonintake.Asignificantfoodxsaltcontentinteractioneffecton

pasta,thesignificantdifferenceinbetweentheintakeofthe1.2%variantandtheintakeofthe0.6%variantresultsina280mgdifferenceinsodiumintakeora0.7gdifferenceinsaltintake.Greenbeanintakewasnotrelatedtoz-BMI,whereaspastaintakewashigherwhenz-BMIwashigher(F(1,137)=0.08;p=0.77andF(1,134)=18.21;p,0.0001,respectively).Fora1SDincreaseinz-BMI,pastaintakeincreasedby15g.Theeffectofthesaltcontentxz-BMIinteractiononintakewasneithersignificantforgreenbeans(F(2,137)=0.01;p=0.99),norforpasta(F(1,132)=1.07;p=0.35).

Univariateanalysis:Theimpactofsaltcontentonhedonicratingandperceivedsaltiness.Forthehedonic

foodintakewasfound(F(2,271)=15.24;p,0.0001).Furtheranalyseswererunseparatelyforeachtargetfood.Thesaltcontentimpactedgreenbeanandpastaintakesignificantly(F(2,137)=5.65;p=0.0045andF(2,134)=12.26;p,0.0001,respectively)anddifferently,asfewerunsaltedgreenbeanswereconsumedthangreenbeanswith0.6%addedsalt,whereasmoreofthepastawith1.2%addedsaltwasconsumedthanthepastawith0.6%addedsalt(Figure1).Inthecaseofgreenbeans,thesignificantdifferencebetweentheintakeofthe0.6%variantandtheintakeofthe0%variantresultsina200mgdifferenceinsodiumintakeora0.5gdifferenceinsaltintake.Inthecaseof

ratingdata,thefoodeffect(F(1,74)=2.79;p=0.10)andtheinteractionbetweenfoodandsaltcontent(F(2,296)=1.17;p=0.31)werenotsignificant.Thefoodfactorwasremovedfromthemodel,andonlythesaltcontentfactorwastakenintoaccount.Thevariantswith0.6%and1.2%addedsaltwerelikedmorethantheunsaltedvariant(F(2,298)=32.32;p,0.0001;Figure2A).Intheinitialmodel,theeffectoftheinteractionbetweenfoodandz-BMIonthehedonicratingofthefoodwassignificant(F(1,292)=4.52;p=0.035),butinseparatemodelsforeachfood,theeffectofz-BMIonthehedonicratingwasnotsignificant(F(1,136)=1.22;p=0.27forgreenbeansandF(1,156)=0.27;p=0.60forpasta).Theeffectofthesaltcontentxz-BMIinteractiononhedonicratingwasneithersignificantforgreenbeans(F(2,136)=0.87;p=0.42),norforpasta(F(2,156)=0.66;p=0.52).

Fortherankingdata,thesaltcontentfactorwassignificant.Forbothtargetfoods,thevariantswith0.6%and1.2%addedsaltwerepreferredovertheunsaltedvariant(F(2,298)=53.48;p,0.0001).Themeanrankings(withscoresvaryingfrom21to1)forthevariantswith0%,0.6%and1.2%addedsaltwere20.61,0.39and0.23forthegreenbeansand20.60,0.24and0.36forthepasta,respectively.Whenaskedtoranktheirperceptionsofsaltiness,thechildrenwereabletodistinguishthethreevariantsforbothfoods(F(2,298)=328.86;p,0.0001;Figure2B

).

Figure1.Theimpactofaddedsaltongreenbeanandpastaintakeby8-to11-year-olds.Thevariationstestedwere0,0.6or1.2gofaddedsaltper100goffood.ThedataarepresentedasleastsquaresmeanswithSEM.The0%and1.2%addedsaltvariantswerecomparedtothe0.6%levelwithaDunnetttest:**p,0.01and***p,0.001.doi:10.1371/journal.pone.0053971.g001

含盐量影响儿童的食物偏好和摄入量

21%and26%ofthevarianceingreenbeanintakeandpastaintake,respectively.Theresultswerethesamewithouttheequalityconstraintsplacedonthegreenbeanandpastaparameters.Bootstrapresamplingshowedthatalloftheparametersweresignificant(p,0.05)excepttheeffectofz-BMIontheoverallpastahedonicrating(p=0.10),whichwassuggestedbythemodificationindices.

Theprimarydeterminantofintakewasthelatentconstruct‘hunger’(r=0.46),followedbytheoverallhedonicratingofthefood(r=0.18)andthelatentconstruct‘preference’(r=0.10),whichwasrelatedtosaltcontentforbothgreenbeansandpasta.z-BMIimpactedpastaintake:thehigherthez-BMI,thehigherthepastaintake.

Discussion

Thepresentstudyaimedtoevaluatetheimpactofsaltonchildren’sfoodintake,takingintoaccountstherolesofhedonicresponses,hungerandweightstatus.Itshowedasignificantpositiveeffectofsaltcontentontheintakeofgreenbeansandpasta,althoughtheimpactwasdifferentforthesetwofoods.Areductioninthesaltcontentofgreenbeansdecreasedtheirintakeby21%,whereasanincreaseinthesaltcontentinpastaincreaseditsintakeby24%.Eventhoughthechildrenwereabletodistinguishbetweenthreevariantsoffoodsaccordingtotheirsaltiness,theypreferredthosewithmoreaddedsaltovertheunsaltedversionsofbothfoods.Asexpected,theresultsoftheSEMshowedthatsaltmodulatedfoodintakebymodifyingpreferencesandallowedustoevaluatetheweightofthedifferentfactorsthataffectedtargetfoodintake:1)thechild’shunger,2)thechild’shedonicratingofthefood,regardlessofitssaltcontentand3)thechild’sratingabouttheparticularsaltcontentofthefood.Inthisstudy,theeffectofsaltcontentonsensoryperceptions(hedonicrating,preferencerankingandperceivedsaltiness)andonintakewasinvestigatedinrealworldconditions.Todate,http://wendang.chazidian.comingadultparticipants,BellisleandLucascomparedsensoryevaluationdataandintakeofyogurtthatvariedinsugarcontentandofmashedpotatoesthatvariedinsaltcontent[14,31].Theseexperimentsshowedthattheoptimalamountsoftheseingredients(sugarandsalt)inagivenfoodareoverestimatediftheresultsfromthesensoryevaluationaretheonlyonesconsidered.Wedrewthesameconclusionforchildrenandpasta.Theminimumsaltcontentneededtooptimizehedonicratingscoresandpreferencerankingswouldbe0.6%addedsaltinbothfoods.Basedonintake,itseemspossibletoeliminatesaltfrompastabecausethevariantwith0%addedsaltwasconsumedasmuchasthevariantwith0.6%addedsaltandinalargequantity(134g).Forgreenbeans,furtherresearchmaybeneededtoexploretheimpactofareductioninsalttoalevelbelow0.6%.Itmaybepossibletorevealanintermediatesaltcontentlevelbetween0%and0.6%thatwouldnotnegativelyimpactintake.

Previousstudiesshowedthattheperceptionandacceptabilityofsaltinessmightbeproduct-specific[8,12,13,32]andthatthepreferenceforasaltytastecannotbegeneralizedtoallfoods[13,33].Thepresentresultsfromintakedataareinaccordancewiththesepreviousfindings.Thedifferencesintheimpactofthesaltcontentofvariousfoodscouldbeduetoadifferentimpactsofsaltcontentonthesensorycharacteristicsassociatedwiththefood[34].Forthefoodsinthisstudy,measuresoftasteintensityobtainedwithatrainedpanelreportedelsewhere[16]revealedthatthesaltcontentaffectedtheintensityofsaltinesssimilarlyinbothfoodsbutaffectedothersensorycharacteristicsinthe

green

Figure2.Theimpactofaddedsaltongreenbeanandpastasensoryevaluationby8-to11-year-olds.Thevariationstestedwere0,0.6or1.2gofaddedsaltper100goffood.LeastsquaresmeanswithSEMarepresentedfor(A)hedonicratingscores(from‘verybad’,22,to‘verygood’,2)and(B)saltinessrankings(from‘leastsalted’,21,to‘moresalted’,1).The0and1.2%addedsaltvariantswerecomparedtothe0.6%levelwithaDunnetttest:**p,0.01and***p,0.001.

doi:10.1371/journal.pone.0053971.g002

Structuralequationmodeling:Therelationshipbetweensaltcontent,hedonicratingandintake.SEMwasusedto

testthetheorythatsaltcontentinfluencesintakethroughitsimpactonpreferences.ThefinalmodelisrepresentedinFigure3.Allofthefitindicessuggestedagoodfit.Themodelexplained