[PDF]

Fruit and vegetable consumption and mortality from all causes, cardiovascular disease, and cancer: systematic review and dose-response meta-analysis of prospective cohort studies PDF Free Download

1 / 22
0 views•22 pages

Fruit and vegetable consumption and mortality from all causes, cardiovascular disease, and cancer: systematic review and dose-response meta-analysis of prospective cohort studies PDF Free Download

Fruit and vegetable consumption and mortality from all causes, cardiovascular disease, and cancer: systematic review and dose-response meta-analysis of prospective cohort studies PDF free Download. Think more deeply and widely.

4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 1/22
BMJ.2014;349:g4490.
Publishedonline2014Jul29.doi:10.1136/bmj.g4490
PMCID:PMC4115152
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovascular
disease,andcancer:systematicreviewanddoseresponsemetaanalysisof
prospectivecohortstudies
XiaWang,instructorinnutrition, YingyingOuyang,researchfellow, JunLiu,researchfellow, MinminZhu,instructor
inbiostatistics, GangZhao,instructorinmedicine, WeiBao,postdoctoralfellow, andFrankBHu,professor
DepartmentofMaternalandChildHealthCare,SchoolofPublicHealth,ShandongUniversity,Jinan,China
DepartmentofNutritionandFoodHygiene,SchoolofPublicHealth,TongjiMedicalCollege,HuazhongUniversityofScienceandTechnology,
Wuhan,China
DepartmentofEpidemiologyandBiostatistics,SchoolofPublicHealth,TongjiMedicalCollege,HuazhongUniversityofScienceandTechnology,
Wuhan,China
DepartmentofCardiovascularSciences,ShandongProvincialHospitalaffiliatedtoShandongUniversity,Jinan,China
EpidemiologyBranch,DivisionofIntramuralPopulationHealthResearch,EuniceKennedyShriverNationalInstituteofChildHealthandHuman
Development,NationalInstitutesofHealth,Rockville,MD20852,USA
DepartmentsofNutritionandEpidemiology,HarvardSchoolofPublicHealth,Boston,MA02115,USA
Correspondenceto:FBHu;Email:nhbfh@channing.harvard.eduandWBao;Email:wei.bao@nih.gov
Accepted2014Jun23.
Copyright©Wangetal2014
ThisisanOpenAccessarticledistributedinaccordancewiththeCreativeCommonsAttributionNonCommercial(CCBYNC3.0)license,which
permitsotherstodistribute,remix,adapt,builduponthisworknoncommercially,andlicensetheirderivativeworksondifferentterms,provided
theoriginalworkisproperlycitedandtheuseisnoncommercial.See:http://creativecommons.org/licenses/bync/3.0/.
Abstract
ObjectiveToexamineandquantifythepotentialdoseresponserelationbetweenfruitandvegetable
consumptionandriskofallcause,cardiovascular,andcancermortality.
DatasourcesMedline,Embase,andtheCochranelibrarysearchedupto30August2013withoutlanguage
restrictions.Referencelistsofretrievedarticles.
StudyselectionProspectivecohortstudiesthatreportedriskestimatesforallcause,cardiovascular,and
cancermortalitybylevelsoffruitandvegetableconsumption.
DatasynthesisRandomeffectsmodelswereusedtocalculatepooledhazardratiosand95%confidence
intervalsandtoincorporatevariationbetweenstudies.Thelinearandnonlineardoseresponserelationswere
evaluatedwithdatafromcategoriesoffruitandvegetableconsumptionineachstudy.
ResultsSixteenprospectivecohortstudieswereeligibleinthismetaanalysis.Duringfollowupperiods
rangingfrom4.6to26yearstherewere56 423deaths(11 512fromcardiovasculardiseaseand16 817from
cancer)among833 234participants.Higherconsumptionoffruitandvegetableswassignificantlyassociated
withalowerriskofallcausemortality.Pooledhazardratiosofallcausemortalitywere0.95(95%
confidenceinterval0.92to0.98)foranincrementofoneservingadayoffruitandvegetables(P=0.001),
0.94(0.90to0.98)forfruit(P=0.002),and0.95(0.92to0.99)forvegetables(P=0.006).Therewasa
thresholdaroundfiveservingsoffruitandvegetablesaday,afterwhichtheriskofallcausemortalitydidnot
reducefurther.Asignificantinverseassociationwasobservedforcardiovascularmortality(hazardratiofor
eachadditionalservingadayoffruitandvegetables0.96,95%confidenceinterval0.92to0.99),while
1,2 2 2
3 4 5 6
1
2
3
4
5
6
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 2/22
higherconsumptionoffruitandvegetableswasnotappreciablyassociatedwithriskofcancermortality.
ConclusionsThismetaanalysisprovidesfurtherevidencethatahigherconsumptionoffruitandvegetables
isassociatedwithalowerriskofallcausemortality,particularlycardiovascularmortality.
Introduction
Increasedconsumptionoffruitandvegetableshasbeenrecommendedasakeycomponentofahealthydiet
forthepreventionofchronicdiseases. Cardiovasculardiseaseandcancerarethetwoleadingcausesof
deathworldwide. Factorsthatcanreducetheoccurrenceoftheseimportantdiseasescouldcontributeto
importantimprovementsinhealthandlongevity.
Inrecentyears,therehasbeengrowingevidencethatfruitandvegetableconsumptionisrelatedtomortality,
includingmortalityfromcardiovasculardiseaseandcancer. Theresults,however,arenotentirely
consistent.Whileseveralstudiesfoundthatconsumptionwasassociatedwithalowerriskofmortality,
nosignificantdifferencesinriskofmortalitywereobservedbetweenvegetariansandnonvegetariansina
Britishpopulation. Inmoststudies,theassociationhasbeenexaminedbycategorisingthemainvariable
intofourthsorfifthsofdailyconsumption. Thereexistsmuchuncertaintyaboutthedoseresponse
relationbetweenconsumptionandtheriskofmortality,especiallyforcancer,asrecentlargeprospective
studieshavefoundnoorminimaleffectsofconsumptiononoverallcancerincidenceormortality.
Understandingtherelationbetweenfruitandvegetableconsumptionandmortalityisimportantforguiding
consumerchoicesandprioritisingdietaryguidelinestoreducerisk.Weperformedametaanalysisof
prospectivecohortstudiestoquantifythedoseresponserelationbetweenfruitandvegetableconsumption
andriskofallcause,cardiovascular,andcancermortality.
Methods
Searchstrategy
WefollowedtheMetaanalysisofObservationalStudiesinEpidemiology(MOOSE) forperformingand
reportingthepresentmetaanalysis.Wecarriedoutametaanalysisofprospectivecohortstudiesthat
examinedtheassociationsoffruitandvegetableconsumptionwithriskofallcause,cardiovascular,and
cancermortality.Wesystematicallysearcheddatabases,includingMedline(from1950),Embase(from
1980),andtheCochraneLibrary(from1960),betweenMay2013and30August2013(lastdatesearched).
WeusedasearchstrategythatincludedtruncatedfreetextandexplodedMeSHtermsrelevantto“fruits”,
“vegetables”,“cardiovasculardisease,”“coronarydisease,”“myocardialischemia,”“stroke,”“neoplasms,”
“causeofdeath,”“mortality,”“humans,”“epidemiology,”“followupstudies,”“prospectivestudies,”and
theirvariants.Norestrictionswereimposedonlanguageofpublications.Weindentifiedadditionalarticlesby
manuallysearchingthereferencelistsfromrecentreviewsandtheextractedpapers.
Studyselection
Weexcludedletters,comments,reviews,metaanalyses,ecologicalstudies,andanimalstudies.Studieswere
includediftheywerecohortstudies,studiedtheeffectsoflevelsoffruitandvegetableconsumption,and
reportedmortalityfromallcauses,cardiovasculardisease,orcancerastheoutcomesofinterest.Toidentify
eligiblestudies,weusedatwostepselectionprocess.Twoindependentinvestigators(GZ,JL)conductedan
initialscreeningofalltitlesorabstractsandthenevaluatedallpotentiallyrelevantarticlesbasedonfulltext
reviews.Studieswereexcludediftheydidnotmeetallcriteria.
Dataextractionandqualityassessment
Dataextractionwasconductedwithastandardiseddatacollectionform.Twoauthors(GZ,JL)
independentlyperformedtheextractionofdata.Weusedhazardratiosasameasureoftheassociation.The
1 2
3
456
789
10
11 12 13
14 15 16
17
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 3/22
primaryexposurevariablewastotalfruitandvegetableconsumption,butwealsoexaminedtotalfruitand
totalvegetablesseparately.Outcomesofinterestinthisstudywereallcause,cardiovascular,andcancer
mortality.AlloutcomeswereclassifiedbasedontheWorldHealthOrganization’sinternationalclassification
ofdiseasecriteria.
Werecordedthefollowingcharacteristicsintheidentifiedstudies:firstauthor,cohortname,country,
publicationyear,ageatentry,sex,samplesizeofthecohort,outcomes,durationoffollowup,methodof
assessmentoffruitandvegetableconsumption,ascertainmentofoutcomes,andvariablesthatenteredintothe
multivariablemodelaspotentialconfounders.Regardinginclusionofstudiesandinterpretationofdata,a
thirdinvestigator(YYO)wasconsultedtoresolvediscrepancies.Anydisagreementsweresettledthrough
consensuswithallthreeauthors.WeassessedstudyqualitywiththeNewcastleOttawaqualityassessment
scale. Thesystemallowedatotalscoreofupto9points(9representingthehighestquality).Wederiveda
scorethatsummarisedeightaspectsofeachstudy.
Statisticalmethods
WeusedSTATAversion12.0(StataCorpLP,CollegeStation,TX)toanalysethedata.Inthismetaanalysis,
weusedhazardratiosand95%confidenceintervalsasameasureoftheeffectsizeforallstudies.Thestudy
byStrandhagenandcolleaguesdidnotreporttherelativerisksorhazardratiosandconfidenceintervalsin
eachconsumptioncategory, sowecalculatedtherelativerisksbyusingthetotalnumberofpatientsandthe
numbersofevents. AnotherstudybyColditzandcolleaguesreportedonlyhazardratiosbutnot
corresponding95%confidenceintervals. Wecalculated95%confidenceintervalsbyusingPvaluesand
effectestimates. Allotherstudiesincludedinthemetaanalysisreportedhazardratios,estimatedfromCox
proportionalhazardsmodels,andcorresponding95%confidenceintervals.Weusedtheresultsofthe
originalstudiesfrommultivariablemodelswiththemostcompleteadjustmentforpotentialconfounders.We
usedtheinversevarianceweightedmethodtoobtainoverallhazardratiosand95%confidenceintervalsfor
anincreaseinconsumptionofoneservingadayoffruitandvegetable.Arandomeffectsmodelaccounted
forvariationbetweenstudiesasthiscanprovidemoreconservativeresultsthanafixedeffectsmodel.
Thedoseresponserelationwasestimatedbyusinggeneralisedleastsquarestrendestimation,accordingto
themethodsdevelopedbyGreenlandandLongnecker. Weusedthetwostagegeneralisedleast
squarestrendestimationmethod,whichfirstestimatedstudyspecificslopelinesandthencombinedwith
studiesinwhichtheslopesweredirectlyreportedtoobtainanoverallaverageslope. Dataontheamount
offruitandvegetableconsumption,distributionsofcasesandpersonyears,andhazardratiosand95%
confidenceintervalswereextractedtoapplythismethod.Weassignedthemedianconsumptionineach
categoryoffruitand/orvegetableconsumptiontothecorrespondinghazardratioforeachstudy.Ifmedians
forthatcategorywerenotreported,weestimatedapproximatemediansbyusingthemidpointofthelower
andupperbounds.Ifthehighestcategoryofthestudieswasopenended,weconsideredthedifferencefrom
thelowestrangetothemediantobeequivalenttothesamedifferenceintheclosestadjacentcategory.Iffruit
andvegetableconsumptionwasreportedbyservingsortimes,weconverteditintothestandardservingfor
thedoseresponseanalysis,whichwasdefinedas77gforvegetablesand80gforfruit.
Inaddition,weexaminednonlinearassociationsbetweenfruitandvegetableconsumptionandallcause
mortalityusingatwostagerandomeffectsdoseresponsemetaanalysis.Thiswasdonebymodelling
consumptionwiththeuseofrestrictedcubicsplineswiththreeknotsatfixedcentiles(10%,50%,and90%)
ofthedistribution. Wefirstestimatedarestrictedcubicsplinemodelwithageneralisedleastsquares
regression,consideringthecorrelationwithineachsetofpublishedhazardratios. Wethencombinedthe
studyspecificestimates,usingtherestrictedmaximumlikelihoodmethodinamultivariaterandomeffects
metaanalysis. Weestimatedthepooledhazardratiosforservingsoffruitandvegetablesadayusinga
proceduretotabulateandplotresults. Atestforanonlinearrelationwascalculatedbymakingthe
coefficientofthesecondsplineequaltozero.
18
9
19
20
21
22
23 24 25
25
26
27 28
25
22
29
30 2
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 4/22
WeevaluatedheterogeneitybetweenstudieswithCochran’sQtest(P<0.10) andusedI toquantifythe
proportionofthetotalvariationduetothatheterogeneity. Toexplorethesourcesofheterogeneityamong
studiesandtesttherobustnessoftheassociations,weconductedsubgroupanalysesandseveralsensitivity
analyses.Wealsoexaminedtheinfluenceofindividualstudiesontheoverallriskestimate,whichwas
investigatedbyrecalculatingthepooledestimatesfortheremainderofthestudiesbyomittingonestudyat
eachturn.
Potentialpublicationbiaswasassessedbytheapplicationofcontourenhancedfunnelplots, Egger’slinear
regressiontest, andBegg’srankcorrelationtestattheP<0.10levelofsignificance. Ifpublicationbias
wasindicated,wefurtherevaluatedthenumberofmissingstudiesinametaanalysisbytheapplicationofthe
trimandfillmethodandrecalculatedthepooledrisksestimatewiththeadditionofthosemissingstudies.
Exceptwhereotherwisespecified,aPvalue<0.05wasconsideredsignificant.
Results
Literaturesearch
Figure1showsstudyselectionprocessandresultsfromtheliteraturesearch.Weidentified2019articlesfrom
theMedlinedatabase,3037articlesfromtheEmbasedatabase,and2361articlesfromtheCochraneLibrary.
Afterexclusionofduplicatesandpapersthatdidnotmeettheinclusioncriteria,weobtained25fullarticles
ofpotentiallyrelevantstudies.Afterfulltextreviews,weexcludednine:onearticlewithinsufficientdatafor
specificlevelsoffruitandvegetableconsumption ;sixarticlesthatreportedresultsofmixeddiet,
specificcancermortality, oramarkeroffruitandvegetableconsumption orenrolledpatientswith
diabetes ;andtwofurtherarticles withdatafromthesamecohortusedinotherstudies.Sixteenarticles
with56 423deaths(11 512fromcardiovasculardiseaseand16 817fromcancer)reportedfrom833 234
participantswereincludedforthemetaanalysis.
Characteristicsoftheincludedstudies
Tables1and2showthecharacteristicsoftheincludedstudies,allofwhichhadaprospectivecohortdesign.
Thetotalnumberofparticipants(from501to451 151)anddeaths(from42to25 682)variedwidelyacross
cohorts.Thedurationoffollowuprangedfrom4.6yearsto26years.Fourstudiesincludedonlymen,and
12studiesincludedmenandwomen.SixstudieswereconductedintheUnitedStates, fourin
Asiancountries, andsixinEurope. Threestudiesmeasuredconsumptionoffruit
andvegetablesbydietrecords, andallotherstudiesusedfoodfrequencyquestionnaires.
Allstudiesadjustedforage,exceptforonestudyofmenbornin1913becausealltheparticipantswereatthe
sameage. Mostcohortscontrolledforsomeconventionalriskfactors,includingbodymassindex(n=12),
smoking(n=13),andalcoholconsumption(n=12).Somecohortsalsoadjustedforphysicalactivity(n=8)
andenergyintake(n=9),butfewstudiesadjustedforotherdietaryvariablesornutrients(n=6).
Assessmentofstudyqualityyieldedanaveragescoreof7.1,and10studieshadascoreof6.5orabove(see
tableA,appendix1).
Fruitandvegetableconsumptionandriskofallcausemortality
Table3showstheresultsofthepooledanalysisforalltheincludedstudies.Therelationbetweenfruitand
vegetableconsumptionandriskofallcausemortalitywasevaluatedinsevenstudies,
comprising553 698participantsand42 219deaths.Thepooledhazardratioofallcausemortalitywas0.95
(95%confidenceinterval0.92to0.98;P=0.001;fig2)foranincrementofoneservingoffruitand
vegetablesaday,withsignificantheterogeneity(P<0.001,I =82%).
30 2
31
32
33 34
35
36 37 38 39
40 5
41 4 42
7 8 9 11 12 13 20 43 44 45 46 47 48 49 50 51
7 8 20 45 46 51
11 12 13 47 9 43 44 48 49 50
45 48 51
9
11 12 13 43 45
48
7 8 43 44 45 46 48
2
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 5/22
Begg’srankcorrelationtestindicatednopublicationbias(P=0.76),butEgger’slinearregressiontest
indicatedpossiblepublicationbiasfortheassociation(P=0.006).Weusedthetrimandfillmethodto
recalculateourpooledriskestimate.Theanalysissuggestedthattheimputedriskestimatewas0.95(95%
confidenceinterval0.92to0.98),whichisidenticaltoouroriginalriskestimate.Nomissingstudieswere
imputedinthecontourenhancedfunnelplot(seefigA,appendix2).
Usingarestrictedcubicsplinesmodel,weobservedsomeevidenceofacurvilinearassociationbetweentotal
consumptionoffruitandvegetablesandriskofallcausemortality(P=0.01fornonlinearity;fig3).There
wasadoseresponserelationbetweenconsumptionoffruitandvegetablesanddecreasingriskofallcause
mortalityatconsumptionbelowfiveservingsaday,buttheriskdidnotdecreasefurtherwithfiveormore
servingsaday.Comparedwithpeoplewhohadnodailyconsumptionoffruitandvegetables,theestimated
hazardratiosofallcausemortalitywere0.92(95%confidenceinterval0.90to0.95)foroneserving/dayof
fruitandvegetables,0.85(0.81to0.90)fortwoservings/day,0.79(0.73to0.86)forthreeservings/day,0.76
(0.69to0.83)forfourservings/day,0.74(0.66to0.82)forfiveservings/day,and0.74(0.65to0.82)forsix
ormoreservings/day.
Sevenstudiesreporteddataforfruitorvegetableconsumptionandallcausemortality. The
summaryestimateswere0.94(95%confidenceinterval0.90to0.98;P=0.002;fig4)foreachadditional
servingadayoffruitand0.95(0.92to0.99;P=0.006;fig5)forvegetables.Therewassignificant
heterogeneityforbothanalyses(P<0.001,I =77%forfruitandP<0.001,I =86%forvegetables).
Forfruitconsumption,wefoundnosignificantevidenceofpublicationbiasusingBegg’srankcorrelation
test(P=0.54),butEgger’slinearregressiontestindicatedpossiblepublicationbias(P=0.02).Theapplication
ofthetrimandfillmethoddidnotchangetheriskestimate(pooledhazardratio0.94,95%confidence
interval0.90to0.98).Nomissingstudieswereimputedinthecontourenhancedfunnelplots(seefigB,
appendix2).Forvegetableconsumption,nosignificantpublicationbiaswasdetected(BeggtestP=0.13;
EggertestP=0.36).
Wealsofoundevidenceofanonlinearassociationforfruit(P=0.01fornonlinearity;seefigC,appendix2)
orvegetableconsumption(P=0.01fornonlinearity;seefigD,appendix2)withriskofallcausemortality.A
lowerriskofallcausemortalitywasobservedinassociationwithhigherfruitconsumptionatabouttwo
servingsaday(hazardratio0.83,0.78to0.88)andvegetableconsumptionataboutthreeservingsaday
(0.75,0.66to0.86).
Fruitandvegetableconsumptionandriskofcardiovascularmortality
Totalconsumptionoffruitandvegetableswasinvestigatedinfourstudies withatotalof469 551
participantsand6893cardiovasculardeaths.Fruitconsumptionwasexaminedinsixstudies
with677 674participantsand9744cardiovasculardeaths,andvegetableconsumptionwasexaminedinsix
studies withof677 674participantsand9744cardiovasculardeaths.
Theaveragereductionintheriskofcardiovasculardiseasemortalitywas4%(hazardratio0.96,95%
confidenceinterval0.92to0.99;P=0.02)foreachadditionalservingperdayoffruitandvegetables
combined(seefigE,appendix2),5%(0.95,0.91to1.00;P=0.03)forfruitconsumption(seefigF,appendix
2),and4%(0.96,0.93to0.99;P=0.01)forvegetableconsumption(seefigG,appendix2).
Fruitandvegetableconsumptionandriskofcancermortality
Twostudiesreporteddataontheassociationbetweencancermortalityandtotalconsumptionoffruitand
vegetables, sevenstudiesonfruit, andeightstudiesonvegetables.
Thehazardratioforcancermortalitywas0.97(95%confidenceinterval0.90to1.03;P=0.31;seefigH,
appendix2)foreachadditionalserving/dayoftotalconsumption,0.99(0.97to1.00;P=0.06;seefigI,
appendix2)forfruit,and0.99(0.97to1.01;P=0.19;seefigJ,appendix2)forvegetables.
9 11 13 43 45 49 51
2 2
7 8 43 48
9 11 12 13 43 51
9 11 12 13 43 51
8 43 9 13 43 47 49 50 51 9 13 20 43 47 49 50
51
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 6/22
Subgroupandsensitivityanalyses
Table4showsthedifferentsubgroupanalysesofstudiesonallcausemortality.Theassociationsbetween
fruitandvegetableconsumptionandriskofallcausemortalitydidnotdiffersubstantiallybystudylocation,
sex,numberofparticipants,durationoffollowup,orstudyquality.
Toexplorepotentialsourceofheterogeneityacrossstudies,wecarriedoutseveralsensitivityanalyses(see
tableB,appendix1).Heterogeneitybetweenstudieswasmainlycausedbyonelargestudy. Forallcause
mortality,afterweexcludedthisstudyfromtheanalysistherewasnolongeranyevidenceofsignificant
heterogeneityforthecombinedconsumptionoffruitandvegetables(I 82%0%),orforfruitconsumption
(I 77%0%).Exclusionofthisstudyfromthepooledestimatehadlittleimpactontheoveralleffectsize.
Toexaminetheimpactofmultivariableadjustment,weconductedadditionalsensitivityanalysesby
excludingstudiesthatdidnotadjustforphysicalactivityorenergyintake.Wealsoexaminedstudieswith
someformofadjustmentforsocioeconomicstatus(suchaseducationorincomelevel).Overall,the
sensitivityanalysesdidnotleadtoanychangeinthesignificanceordirectionofeffectfortheassociation
betweenthecombinedconsumptionoffruitandvegetableandriskofallcausemortality(seetableB,
appendix1).
Tofurtherconfirmtherobustnessoftheresults,weconductedaseriesofsensitivityanalysesbyexcluding
onestudythatdidnotreporthazardratios. Exclusionofthisstudydidnotappreciablyaltertheresultsfor
total,cardiovascular,orcancermortality(seetableC,appendix1).
Discussion
Thismetaanalysissupportsthetheorythathigherconsumptionoffruitandvegetablesisassociatedwitha
reducedriskofmortalityfromallcausesandcardiovasculardisease.Theriskofallcausemortalitywas
decreasedby5%foreachadditionalservingadayoffruitandvegetables,by6%forfruitconsumption,and
by5%forvegetableconsumption.Weobservedathresholdataroundfiveservingsaday,afterwhichthere
wasnofurtherreductioninrisk.Whilewefoundasignificantinverseassociationforcardiovascular
mortality,higherconsumptionwasnotappreciablyassociatedwithriskofcancermortality.
Explorationofheterogeneityandpublicationbias
Inthemetaanalysisofallcausemortality,onestudywithalargesamplesizecontributedtomostofthe
observedheterogeneity. Ourresultsshowedthatabout82%ofthevarianceinheterogeneitywasduetothe
variationbetweenstudies. Sensitivityanalysesshowedthatexclusionofthislargestudydidnot
appreciablyalterthepooledhazardratios.
Therewerenomissingstudiesimputedinregionsofthecontourenhancedfunnelplots.Egger’slinear
regressiontestindicatedthatthePvaluewassignificant.NopublicationbiaswasfoundforBegg’srank
correlationtest.Also,theapplicationofthetrimandfillmethoddidnotchangetheaverageeffectsize,
furthersuggestingthatresultswerenotaffectedbypublicationbias.
Resultsinrelationtootherstudies
AdherencetoaMediterraneandiet,witharelativelylargeamountoffruitandvegetables,hasbeenshownto
significantlydecreasetheriskoftotalmortalityandmortalityfromcardiovasculardiseases. Inolderpeople
fromseveralEuropeancountries,lowermortalitywasrelatedwithgreateradherencetoaMediterranean
diet. Overall,about1030%lowerriskofallcausemortalitywasfoundinmostprospectivestudies
comparinghighestwithlowestfruitandvegetableconsumption. Asmallstudy(n=713)thatused
totalserumconcentrationsofcarotenoidsasbiomarkersoffruitandvegetableconsumptionfurthersupported
ariskreductioninmortality,withahazardratioof0.50forwomeninthehighestthirdofserum
concentrationcomparedwiththelowestthird. Thefindingsfromourmetaanalysiswereconsistentwitha
43
2
2
9
43
52
53
38
4 7 8 9 11 49
5
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 7/22
recentstudyshowingthatalowerconsumptioncomparedwithfiveservingsoffruitandvegetablesaday
wasassociatedwithhighermortalityinadoseresponsemanner.
Forthereductionintotalmortality,wefoundathresholdofaroundfiveservingsadayoffruitand
vegetables,afterwhichtheriskofdeathdidnotreducefurther.Possiblemechanismsmightinvolvethe
availabilityofnutrientsandthedigestibilityoffruitandvegetables, butfurtherstudiesareneededto
confirmourresults.
Ourstudyalsoshowedthathigherconsumptionoffruitandvegetableswasassociatedwithlowerriskof
mortalityfromcardiovasculardiseasebutnotcancer.Theriskofcardiovascularmortalitywasdecreasedby
4%foreachservingadayofthecombinedconsumptionoffruitandvegetables,by5%forfruit,andby4%
forvegetables.Inapreviousmetaanalysisofeightcohortstudies,individualswhoconsumedmorethanfive
servingsadayhada26%lowerriskofstrokethanthosewhoconsumedlessthanthreeservingsaday. In
anothermetaanalysisofninecohortstudiesoncoronaryheartdisease, a4%lowerriskwasreportedfor
eachadditionalservingadayoffruitandvegetables,withastandardservingcalculatedas106g.Asforthe
mechanismsfortheinverseassociationbetweenconsumptionoffruitandvegetablesandcardiovascular
mortality,antioxidantcompoundsandpolyphenolsinfruitandvegetables—suchasvitaminC,carotenoids,
andflavonoids—havebeenshowntopreventtheoxidationofcholesterolandotherlipidsinthearteries
andtoincreasetheformationofendothelialprostacyclinthatinhibitsplateletaggregationandreduces
vasculartone. Resultsfromlargerandomisedcontrolledtrialshaveshownthatincreasedconsumptionof
fruitandvegetablescancontributetoasmalldecreaseinbloodpressure. Fruitandvegetablesaregood
sourcesofmagnesiumandpotassium,whichhavebeeninverselyassociatedwithmortalityinprevious
studies. Plasmaconcentrationsofantioxidants,suchasalphacaroteneandbetacarotene,increasein
parallelwithincreasedconsumptionoffruitandvegetables, andthiscouldreducetheriskofcancerand
cardiovasculardisease.SomeothercomponentsinfruitandvegetablessuchasvitaminC,carotenoids,and
otherphytochemicalsalsoprobablycontributetoareducedriskofmortality.
Theassociationbetweenhigherconsumptionoffruitandvegetablesandriskofcancerhasnotbeenfirmly
established.Resultsfromepidemiologicalstudiesareinconsistent,particularlyforhormonedependent
cancerssuchasbreastandprostatecancer. Severalstudiesshowednosignificantassociationbetweenthe
consumptionofeitherfruitorvegetablesandtotalcancerrisk, whereastheGreekEPICcohortstudy
reportedasignificantreductionintotalcancerriskassociatedwithhighconsumptionofbothfruitand
vegetables.IntheEPICstudyfrom10Europeancountries, therewasaweakinverseassociationbetween
highconsumptionandtotalcancerrisk.Inourstudy,thepooledresultsindicatethathigherconsumptionof
fruitandvegetableswasnotappreciablyassociatedwithriskofcancermortality,whichsuggeststhat
increasingtheamountoffruitandvegetablesaloneinanindividual’sdietmightnotprovideanappreciable
benefitonreducingcancermortality. Inadditiontotherecommendationofconsumingadequateamounts
offruitandvegetables,theadverseeffectsofobesity,physicalinactivity,smoking,andhighalcoholintakes
oncancermortalityshouldbefurtheremphasised.Itispossible,however,thatfruitandvegetable
consumptionmighthavestrongereffectsonspecificcancersites.Inaddition,differenttypesoffruitand
vegetablesmighthavedifferenteffectsoncancerrisk.Futurestudiesareneededtobemorespecificabout
typesofcancerandtheroleofdifferentgroupsoffruitandvegetables.
Strengthsandlimitationsofthereview
Thismetaanalysiswasbasedonseveralprospectivecohortstudiesfromvariouspopulations.Thecombined
samplesizewaslargeandthefollowupperiodwaslongenough.Theestimatesfromthefullyadjusted
modelsforeachstudywereusedinouranalysestoreducethepotentialofconfounding.Thedoseresponse
analysiswasconductedtoevaluatethelinearandnonlinearrelations.Thiscanhelptoquantifythe
associationsandtesttheshapeofthesepossibleassociations.Toexaminethepotentialsourcesof
heterogeneityandevaluaterobustnessinthesubgroups,weperformedseveralsensitivityanalyses.
44
54 55
26
56
57
58
59 60
61 62 63
60 64
4
65
16 66 67
15
65
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 8/22
Therewere,however,severallimitationsofthismetaanalysis.Becauseinmoststudiesfruitandvegetable
consumptionwasassessedbyfoodfrequencyquestionnaires,errorsinmeasurementwereinevitable.The
imprecisemeasurementofconsumptionmighthaveattenuatedthetrueassociations.Inaddition,fewstudies
adjustedforotherdietaryfactors,suchassaturatedfatintakeandconsumptionofprocessedmeat,etc.The
inverseassociationbetweenfruitandvegetableconsumptionandmortalitycouldberelatedtoagenerally
morehealthydietandlifestyle.Becauseallincludedstudieswereobservationalinnature,theresultscouldbe
subjecttoresidualorunmeasuredconfounding.Furthermore,thereweredifferencesinclassificationsoffruit
andvegetablesacrossstudies.Thetypesconsumeddifferedaccordingtogeographicallocations.These
factorscouldaffectourresults.Oursubgroupanalyses,however,showedthattheassociationsbetween
consumptionandriskofallcausemortalitydidnotdiffersignificantlybystudylocation.
Conclusions
Thismetaanalysisprovidesfurtherevidencethathigherconsumptionoffruitandvegetablesisassociated
withalowerriskofmortalityfromallcauses,particularlyfromcardiovasculardiseases.Theresultssupport
currentrecommendationstoincreaseconsumptiontopromotehealthandoveralllongevity.
Whatisalreadyknownonthistopic
Theassociationbetweenfruitandvegetableconsumptionandriskofmortalityhasbeenexamined
inmanystudies,butthedosedependencyofthisassociationhasnotbeendeterminedinameta
analysis
Whatthisstudyadds
Higherconsumptionoffruitandvegetablesisassociatedwithareducedriskofallcausemortality,
withanaveragereductioninriskof5%foreachadditionalservingaday(6%forfruitand5%for
vegetables)
Therewasathresholdaroundfiveservingsaday,afterwhichtheriskofdeathdidnotreduce
further
Therewasasignificantinverseassociationforcardiovascularmortality,buthigherconsumption
wasnotappreciablyassociatedwithcancermortality
WebExtra.Extramaterialsuppliedbytheauthor
Appendix1:SupplementarytablesAC
Appendix2:SupplementaryfiguresAJ
Notes
WethankFuzhongXue(directorandprofessorofstatistics,SchoolofPublicHealth,ShandongUniversity)
forprovidingstatisticaladvice.
Contributors:XW,GZ,YYO,andFBHcontributedtoconceptionanddesign.Allauthorswereinvolvedin
analysisandinterpretationofthedata.MMZ,XW,andWBdesignedandconductedthestatisticalanalysis.
XW,JL,WB,andGZdraftedthemanuscript,whichwascriticallyrevisedforimportantintellectualcontent
byXW,WB,andFBH.Allauthorsapprovedthefinalversion.FBHisguarantor.
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 9/22
Funding:ThisworkwasfundedbyNationalNaturalScienceFoundation(NSFC81071081)ofChina.WB
wassupportedbytheIntramuralResearchProgramoftheEuniceKennedyShriverNationalInstituteof
ChildHealthandHumanDevelopment,NationalInstitutesofHealth.Thefundershadnoroleinstudy
design,datacollectionandanalysis,preparationofthemanuscript,ordecisiontopublish.
Competinginterests:AllauthorshavecompletedtheICMJEuniformdisclosureformat
www.icmje.org/coi_disclosure.pdf(availableonrequestfromthecorrespondingauthor)anddeclare:no
supportfromanyorganizationforthesubmittedwork;nofinancialrelationshipswithanyorganizationsthat
mighthaveaninterestinthesubmittedworkinthepreviousthreeyears;andnootherrelationshipsor
activitiesthatcouldappeartohaveinfluencedthesubmittedwork.
Ethicalapproval:Notrequired.
Datasharing:Noadditionaldataavailable.
Transparency:Theleadauthor(themanuscript’sguarantor)affirmsthatthismanuscriptisanhonest,
accurate,andtransparentaccountofthestudybeingreported;thatnoimportantaspectsofthestudyhave
beenomitted;andthatanydiscrepanciesfromthestudyasplannedandregisteredhavebeenexplained.
Notes
Citethisas:BMJ2014;349:g4490
References
1.JointWHO/FAOExpertConsultation.Diet,nutritionandthepreventionofchronicdiseases.World
HealthOrganTechRepSer2003;916:1149.
2.NationalAcademyofSciencesCoDaH,NationalResearchCouncil.Dietandhealth:implicationsfor
reducingchronicdiseaserisk.NationalAcademyPress,1989.
3.WorldHealthOrganization.Thetop10causesofdeath.WHO,2012.
www.who.int/mediacentre/factsheets/fs310/en.
4.AgudoA,CabreraL,AmianoP,ArdanazE,BarricarteA,BerenguerT,etal.Fruitandvegetableintakes,
dietaryantioxidantnutrients,andtotalmortalityinSpanishadults:findingsfromtheSpanishcohortofthe
EuropeanProspectiveInvestigationintoCancerandNutrition(EPICSpain).AmJClinNutr2007;85:1634
42.[PubMed:17556703]
5.NicklettEJ,SembaRD,XueQL,TianJ,SunK,CappolaAR,etal.Fruitandvegetableintake,physical
activity,andmortalityinoldercommunitydwellingwomen.JAmGeriatrSoc2012;60:8628.
[PMCID:PMC3639003][PubMed:22587851]
6.TrichopoulouA,CostacouT,BamiaC,TrichopoulosD.AdherencetoaMediterraneandietandsurvival
inaGreekpopulation.NEnglJMed2003;348:2599608.[PubMed:12826634]
7.BazzanoLA,HeJ,OgdenLG,LoriaCM,VupputuriS,MyersL,etal.Fruitandvegetableintakeand
riskofcardiovasculardiseaseinUSadults:thefirstNationalHealthandNutritionExaminationSurvey
EpidemiologicFollowupStudy.AmJClinNutr2002;76:939.[PubMed:12081821]
8.GenkingerJM,PlatzEA,HoffmanSC,ComstockGW,HelzlsouerKJ.Fruit,vegetable,andantioxidant
intakeandallcause,cancer,andcardiovasculardiseasemortalityinacommunitydwellingpopulationin
WashingtonCounty,Maryland.AmJEpidemiol2004;160:122333.[PubMed:15583375]
9.StrandhagenE,HanssonPO,BosaeusI,IsakssonB,ErikssonH.Highfruitintakemayreducemortality
amongmiddleagedandelderlymen.TheStudyofMenBornin1913.EurJClinNutr2000;54:33741.
[PubMed:10745285]
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 10/22
10.KeyTJ,ApplebyPN,DaveyGK,AllenNE,SpencerEA,TravisRC.MortalityinBritishvegetarians:
reviewandpreliminaryresultsfromEPICOxford.AmJClinNutr2003;78:533S8S.[PubMed:12936946]
11.NaguraJ,IsoH,WatanabeY,MaruyamaK,DateC,ToyoshimaH,etal.Fruit,vegetableandbean
intakeandmortalityfromcardiovasculardiseaseamongJapanesemenandwomen:theJACCStudy.BrJ
Nutr2009;102:28592.[PubMed:19138438]
12.NakamuraK,NagataC,ObaS,TakatsukaN,ShimizuH.Fruitandvegetableintakeandmortalityfrom
cardiovasculardiseaseareinverselyassociatedinJapanesewomenbutnotinmen.JNutr2008;138:112934.
[PubMed:18492845]
13.ZhangX,ShuXO,XiangYB,YangG,LiH,GaoJ,etal.Cruciferousvegetableconsumptionis
associatedwithareducedriskoftotalandcardiovasculardiseasemortality.AmJClinNutr2011;94:2406.
[PMCID:PMC3127519][PubMed:21593509]
14.WillettWC.Fruits,vegetables,andcancerprevention:turmoilintheproducesection.JNatlCancer
2010;102:5101.
15.BoffettaP,CoutoE,WichmannJ,FerrariP,TrichopoulosD,BuenodeMesquitaHB,etal.Fruitand
vegetableintakeandoverallcancerriskintheEuropeanProspectiveInvestigationintoCancerandNutrition
(EPIC).JNatlCancer2010;102:52937.
16.GeorgeSM,ParkY,LeitzmannMF,FreedmanND,DowlingEC,ReedyJ,etal.Fruitandvegetable
intakeandriskofcancer:aprospectivecohortstudy.AmJClinNutr2009;89:34753.
[PMCID:PMC2647712][PubMed:19056579]
17.StroupDF,BerlinJA,MortonSC,OlkinI,WilliamsonGD,RennieD,etal.Metaanalysisof
observationalstudiesinepidemiology.JAMA2000;283:200812.[PubMed:10789670]
18.WellsG,SheaB,O’connellD,PetersonJ,WelchV,LososM,etal.TheNewcastleOttawaScale(NOS)
forassessingthequalityofnonrandomisedstudiesinmetaanalyses.2000.
www.ohri.ca/programs/clinical_epidemiology/oxford.htm.
19.McNuttL,WuC,XueX,HafnerJP.Estimatingtherelativeriskincohortstudiesandclinicaltrialsof
commonoutcomes.AmJEpidemiol2003;157:9403.[PubMed:12746247]
20.ColditzGA,BranchLG,LipnickRJ,WillettW,RosnerB,PosnerB,etal.Increasedgreenandyellow
vegetableintakeandloweredcancerdeathsinanelderlypopulation.AmJClinNutr1985;41:326.
[PubMed:3966422]
21.GardnerMJ,AltmanDG.ConfidenceintervalsratherthanPvalues:estimationratherthanhypothesis
testing.BMJ1986;292:74650.[PMCID:PMC1339793][PubMed:3082422]
22.JacksonD,WhiteIR,ThompsonSG.ExtendingDerSimonianandLaird’smethodologytoperform
multivariaterandomeffectsmetaanalyses.StatMed2010;29:128297.[PubMed:19408255]
23.BerlinJA,LongneckerMP,GreenlandS.Metaanalysisofepidemiologicdoseresponsedata.
Epidemiology1993:21828.
24.GreenlandS,LongneckerMP.Methodsfortrendestimationfromsummarizeddoseresponsedata,with
applicationstometaanalysis.AmJEpidemiol1992;135:13019.[PubMed:1626547]
25.OrsiniN,BelloccoR,GreenlandS.Generalizedleastsquaresfortrendestimationofsummarizeddose
responsedata.StataJ2006;6:4057.
26.HeFJ,NowsonCA,MacGregorGA.Fruitandvegetableconsumptionandstroke:metaanalysisof
cohortstudies.Lancet2006;367:3206.[PubMed:16443039]
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 11/22
27.OrsiniN,LiR,WolkA,KhudyakovP,SpiegelmanD.Metaanalysisforlinearandnonlineardose
responserelations:examples,anevaluationofapproximations,andsoftware.AmJEpidemiol2012;175:66
73.[PMCID:PMC3244608][PubMed:22135359]
28.HarreFE,LeeKL,PollockBG.Regressionmodelsinclinicalstudies:determiningrelationshipsbetween
predictorsandresponse.JNatlCancer1988;80:1198202.
29.OrsiniN,GreenlandS.Aproceduretotabulateandplotresultsafterflexiblemodelingofaquantitative
covariate.StataJ2011;11:129.
30.HedgesLV,OlkinI,StatistikerM.Statisticalmethodsformetaanalysis.AcademicPress,1985.
31.HigginsJP,ThompsonSG,DeeksJJ,AltmanDG.Measuringinconsistencyinmetaanalyses.BMJ
2003;327:55760.[PMCID:PMC192859][PubMed:12958120]
32.PetersJL,SuttonAJ,JonesDR,AbramsKR,RushtonL.Contourenhancedmetaanalysisfunnelplots
helpdistinguishpublicationbiasfromothercausesofasymmetry.JClinEpidemiol2008;61:9916.
[PubMed:18538991]
33.EggerM,SmithGD,SchneiderM,MinderC.Biasinmetaanalysisdetectedbyasimple,graphicaltest.
BMJ1997;315:62934.[PMCID:PMC2127453][PubMed:9310563]
34.BeggCB,MazumdarM.Operatingcharacteristicsofarankcorrelationtestforpublicationbias.
Biometrics1994;50:1088101.[PubMed:7786990]
35.DuvalS,TweedieR.Trimandfill:asimplefunnelplotbasedmethodoftestingandadjustingfor
publicationbiasinmetaanalysis.Biometrics2000;56:45563.[PubMed:10877304]
36.VerlangieriA,KapeghianJ,ElDeanS,BushM.Fruitandvegetableconsumptionandcardiovascular
mortality.MedHypotheses1985;16:715.[PubMed:4000000]
37.KeyTJTM,ApplebyPN,BurrML.Dietaryhabitsandmortalityin11000vegetariansandhealth
consciouspeople:resultsofa17yearfollowup.BMJ1996;313:759.[PMCID:PMC2352199]
[PubMed:8842068]
38.KnoopsKT,deGrootLC,KromhoutD,PerrinA,MoreirasVarelaO,MenottiA,etal.Mediterranean
diet,lifestylefactors,and10yearmortalityinelderlyEuropeanmenandwomen.JAMA2004;292:14339.
[PubMed:15383513]
39.KushiLH,LewRA,StareFJ,EllisonCR,elLozyM,BourkeG,etal.Dietand20yearmortalityfrom
coronaryheartdisease:theIrelandBostonDietHeartStudy.NEnglJMed1985;312:8118.
[PubMed:2983212]
40.JansenMC,BuenodeMesquitaHB,RäsänenL,FidanzaF,NissinenAM,MenottiA,etal.Cohort
analysisoffruitandvegetableconsumptionandlungcancermortalityinEuropeanmen.IntJCancer
2001;92:9138.[PubMed:11351316]
41.NöthlingsU,SchulzeMB,WeikertC,BoeingH,vanderSchouwYT,BamiaC,etal.Intakeof
vegetables,legumes,andfruit,andriskforallcause,cardiovascular,andcancermortalityinaEuropean
diabeticpopulation.JNutr2008;138:77581.[PubMed:18356334]
42.CroweFL,RoddamAW,KeyTJ,ApplebyPN,OvervadK,JakobsenMU,etal.Fruitandvegetable
intakeandmortalityfromischaemicheartdisease:resultsfromtheEuropeanProspectiveInvestigationinto
CancerandNutrition(EPIC)Heartstudy.EurHeartJ2011;32:123543.[PMCID:PMC3094548]
[PubMed:21245490]
43.LeendersM,SluijsI,RosMM,BoshuizenHC,SiersemaPD,FerrariP,etal.Fruitandvegetable
consumptionandmortalityeuropeanprospectiveinvestigationintocancerandnutrition.AmJEpidemiol
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 12/22
2013;178:590602.[PubMed:23599238]
44.BellaviaA,LarssonSC,BottaiM,WolkA,OrsiniN.Fruitandvegetableconsumptionandallcause
mortality:adoseresponseanalysis.AmJClinNutr2013;98:4549.[PubMed:23803880]
45.TuckerKL,HallfrischJ,QiaoN,MullerD,AndresR,FlegJL.Thecombinationofhighfruitand
vegetableandlowsaturatedfatintakesismoreprotectiveagainstmortalityinagingmenthaniseitheralone:
theBaltimoreLongitudinalStudyofAging.JNutr2005;135:55661.[PubMed:15735093]
46.SteffenLM,JacobsDR,StevensJ,ShaharE,CarithersT,FolsomAR.Associationsofwholegrain,
refinedgrain,andfruitandvegetableconsumptionwithrisksofallcausemortalityandincidentcoronary
arterydiseaseandischemicstroke:theAtherosclerosisRiskinCommunities(ARIC)Study.AmJClinNutr
2003;78:38390.[PubMed:12936919]
47.SauvagetC,NaganoJ,HayashiM,SpencerE,ShimizuY,AllenN.Vegetablesandfruitintakeand
cancermortalityintheHiroshima/NagasakiLifeSpanStudy.BrJCancer2003;88:68994.
[PMCID:PMC2376354][PubMed:12618875]
48.RissanenTH,VoutilainenS,VirtanenJK,VenhoB,VanharantaM,MursuJ,etal.Lowintakeoffruits,
berriesandvegetablesisassociatedwithexcessmortalityinmen:theKuopioIschaemicHeartDiseaseRisk
Factor(KIHD)Study.JNutr2003;133:199204.[PubMed:12514290]
49.WhitemanD,MuirJ,JonesL,MurphyM,KeyT.Dietaryquestionsasdeterminantsofmortality:the
OXCHECKexperience.PublicHealthNutr1999;2:47787.[PubMed:10656467]
50.HertogM,BuenodeMesquitaHB,FehilyAM,SweetnamPM,ElwoodPC,KromhoutD.Fruitand
vegetableconsumptionandcancermortalityintheCaerphillyStudy.CancerEpidemBiomark1996;5:6737.
51.SahyounNR,JacquesPF,RussellRM.Carotenoids,VitaminsCandE,andmortalityinanelderly
population.AmJEpidemiol1996;144:50111.[PubMed:8781466]
52.CooryMD.Commenton:heterogeneityinmetaanalysisshouldbeexpectedandappropriately
quantified.IntJEpidemiol2010;39:932.[PubMed:19349478]
53.SofiF,CesariF,AbbateR,GensiniGF,CasiniA.AdherencetoMediterraneandietandhealthstatus:
metaanalysis.BMJ2008;337:a1344.[PMCID:PMC2533524][PubMed:18786971]
54.HolstB,WilliamsonG.Nutrientsandphytochemicals:frombioavailabilitytobioefficacybeyond
antioxidants.CurrOpinBiotech2008;19:7382.[PubMed:18406129]
55.OttenJJ,HellwigJP,MeyersLD.DRI,dietaryreferenceintakes:theessentialguidetonutrient
requirements.NationalAcademiesPress,2006.
56.DauchetL,AmouyelP,HercbergS,DallongevilleJ.Fruitandvegetableconsumptionandriskof
coronaryheartdisease:ametaanalysisofcohortstudies.JNutr2006;136:258893.[PubMed:16988131]
57.AsplundK.Antioxidantvitaminsinthepreventionofcardiovasculardisease:asystematicreview.J
InternMed2002;251:37292.[PubMed:11982737]
58.LeferA.Prostacyclin,highdensitylipoproteins,andmyocardialischemia.Circulation1990;81:20135.
[PubMed:2111745]
59.AppelLJ,MooreTJ,ObarzanekE,VollmerWM,SvetkeyLP,SacksFM,etal.Aclinicaltrialofthe
effectsofdietarypatternsonbloodpressure.NEnglJMed1997;336:111724.[PubMed:9099655]
60.JohnJ,ZieblandS,YudkinP,RoeL,NeilH.Effectsoffruitandvegetableconsumptiononplasma
antioxidantconcentrationsandbloodpressure:arandomisedcontrolledtrial.Lancet2002;359:196974.
[PubMed:12076551]
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 13/22
61.KaluzaJ,OrsiniN,LevitanEB,BrzozowskaA,RoszkowskiW,WolkA.Dietarycalciumand
magnesiumintakeandmortality:aprospectivestudyofmen.AmJEpidemiol2010;171:8017.
[PubMed:20172919]
62.ZhangW,IsoH,OhiraT,DateC,TamakoshiA.Associationsofdietarymagnesiumintakewith
mortalityfromcardiovasculardisease:theJACCstudy.Atherosclerosis2012;221:58795.
[PubMed:22341866]
63.YangQ,LiuT,KuklinaEV,FlandersWD,HongY,GillespieC,etal.Sodiumandpotassiumintakeand
mortalityamongUSadults:prospectivedatafromtheThirdNationalHealthandNutritionExamination
Survey.ArchInternMed2011;171:118391.[PubMed:21747015]
64.ZinoS,SkeaffM,WilliamsS,MannJ.Randomisedcontrolledtrialofeffectoffruitandvegetable
consumptiononplasmaconcentrationsoflipidsandantioxidants.BMJ1997;314:178791.
[PMCID:PMC2126938][PubMed:9224079]
65.KeyT.Fruitandvegetablesandcancerrisk.BrJCancer2010;104:611.[PMCID:PMC3039795]
[PubMed:21119663]
66.HungH,JoshipuraKJ,JiangR,HuFB,HunterD,SmithWarnerSA,etal.Fruitandvegetableintake
andriskofmajorchronicdisease.JNatlCancer2004;96:157784.
67.BenetouV,OrfanosP,LagiouP,TrichopoulosD,BoffettaP,TrichopoulouA.Vegetablesandfruitsin
relationtocancerrisk:evidencefromtheGreekEPICcohortstudy.CancerEpidemBiomark2008;17:387
92.
FiguresandTables
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 14/22
Fig1Selectionofstudiesinvestigatingeffectoffruitandvegetableconsumptiononmortality
Table1 Characteristicsofstudiesincludedinmetaanalysisofassociationsoffruitandvegetable
consumptionwithriskofmortality
Firstauthor
Publication
year Country Studyname
Noof
participants
Ageat
baseline
(years)
Yearsof
followup
(person
years)
Mortality
endpoints(No
ofcases)
44
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 15/22
Bellavia 2013 Sweden Swedish
MammographyCohort
andSwedishMen
Cohort
71706* 4583 13(932
178†)
Allcause(11
439)
Leenders 2013 10
European
countries
EPIC 451151* 2570 13(5864
963†)
Allcause(25
682),CVD
(5125),cancer
(10438)
Zhang
(men)
2011 China ShanghaiMen’sHealth
Study
61500 4074 4.6(282
900)
Allcause
(1951),CVD
(635),cancer
(853)
Zhang
(women)
2011 China ShanghaiWomen’s
HealthStudy
74942 4070 10.2(764
408)
Allcause
(3442),CVD
(1023),cancer
(1485)
Nagura 2009 Japan JapanCollaborative
Cohort
59485* 4079 13(756
054)
Allcause
(7606),CVD
(2243)
Nakamura 2008 Japan TakayamaStudy 29079* ≥35 7(201
156)
CVD(384)
Tucker 2005 US Baltimore
LongitudinalStudyof
Aging
501‡ 3480 18(9018) Allcause(306)
Genkinger 2004 US OdysseyCohort 6151* 3093 12(14
960)
Allcause(910),
CVD(378),
cancer(307)
Steffen 2003 US AtherosclerosisRiskin
Communities
11940* 4564 11(131
340†)
Allcause(867)
Sauvaget 2003 Japan LifeSpanStudy 38540* 34103 18(693
720†)
Cancer(3136)
Bazzano 2002 US NHANESI 9608* 2574 19(159
304)
Allcause
(2530),CVD
(1145)
Rissanen 2002 Finland KuopioIschaemic
HeartDiseaseRisk
Factor
2641‡ 4260 12.8(33
800)
Allcause(485),
CVD(245)
Strandhagen 2000 Sweden StudyofMenBornin
1913
792‡ 5480 26(20
592)
Allcause(390),
CVD(226),
cancer(121)
Whiteman 1999 UK Oxfordand
CollaboratorsHealth
Check
11090* 3564 9(93464) Allcause(598),
cancer(257)
Hertog 1996 UK CaerphillyStudy 2112‡ 4569 13.8
(6874)
Cancer(114)
44
43
13
13
11
12
45
46
47
48
9
49
50
51
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 16/22
Sahyoun 1996 US Communitydwelling
volunteers
725* 60101 12(8700†) Allcause(217),
CVD(108),
cancer(64)
Colditz 1985 US Cohortstudyof
Massachusettsresidents
aged≥66
1271* ≥66 5(6355†) Cancer(42)
CVD=cardiovasculardisease;EPIC=EuropeanProspectiveInvestigationintoCancerandNutrition;
NHANESI=firstNationalHealthandNutritionExaminationSurvey.
*Menandwomen.
†Persontimeestimatedbymultiplyingnumberofparticipantsbyaveragefollowuptime.
‡Men.
Table2 Characteristicsofprospectivestudiesoffruitandvegetableconsumptioninrelationtomortality:
exposureandoutcomeassessment,exposurelevels,andcovariates
Study
Exposure
assessment
method
Measureof
associations
Outcome
assessment
Categoriesof
consumption
Covariatesinfullyadjusted
model
Bellavia,
201344
FFQ.Self
administered
HR Ascertained
throughlinkage
toSwedish
Registerof
DeathCausesat
NationalBoard
ofHealthand
Welfare
Fruitand
vegetables
(servings/day):0,
0.5,1,2,3,4,5,8
Age,sex,BMI,physicalactivity,
smoking,alcohol,energyintake,
educationallevel
Leenders,
2013
Country
specificFFQ
and7day
record
HR Recordlinkage
withcancer
registries,
boardsof
health,and
deathindices,or
activefollow
up
Fruitand
vegetables(g/day):
178.8,316.8,
468.4,725.4.Fruit
(g/day):74.6,
159.9,250.1,
403.0.Vegetables
(g/day):91.1,
151.6,215.6,339.4
Age,sex,centre,BMI,smoking
status,smokingduration,time
sincestoppedsmoking,numberof
cigarettessmokedaday,alcohol
consumption,physicalactivity,
education,processedmeat
consumption,vegetablesandfruit
()
Zhang,
2011
(men)
FFQ.
Interviewer
administered
HR Biennialhome
visits,record
linkageto
Shanghai
CancerRegistry
andShanghai
VitalStatistics
Registry,and
death
certificates
Fruit(g/day):14,
71,129,196,308.
Vegetables
(g/day):144,232,
307,398,583
Age,BMI,education,occupation,
familyincome,smoking,alcohol,
physicalactivity,multivitamin
supplementuse,intakesoftotal
energyandsaturatedfat,historyof
coronaryheartdisease,stroke,
hypertension,ordiabetes
51
20
43
13
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 17/22
Zhang,
2011
(women)
FFQ.
Interviewer
administered
HR Biennialhome
visits,record
linkageto
Shanghai
CancerRegistry
andShanghai
VitalStatistics
Registry,and
death
certificates
Fruit(g/day):62,
155,238,330,489.
Vegetables
(g/day):124,196,
261,345,506
Asabove
Nagura,
2009
FFQ.Self
administered,
butconfirmed
bycomparing
two
questionnaires
administered1
yearapart
HR Reviewing
death
certificates
Fruit
(servings/day):0.9,
2.3,3.9,5.9.
Vegetables
(servings/day):1.2,
2.3,3.4,5.2
Age,BMI,sex,smokingstatus,
alcoholintake,physicalactivity,
hoursofsleep,educationyears,
perceivedmentalstress,
cholesterolintake,SFAintake,n3
fattyacidsintake,sodiumintake
andhistoriesofhypertensionand
diabetes,vegetableandbeanintake
(forfruit)/fruitandbeanintake
(forvegetable)
Nakamura,
2008
FFQ.
Interviewer
administered.
Validityand
reproducibility
ofFFQ
demonstrated
bycomparingit
withother
dietary
assessment
methods
HR Datafrom
officeof
NationalVital
Statistics
Men:fruit
(servings/day):0.3,
0.7,1.3,2.6;
vegetables
(servings/day):2.2,
3.4,4,7.1.Women:
fruit
(servings/day):0.4,
0.9,1.5,2.7;
vegetables
(servings/day):2.5,
3.6,4.8,7.4
Age,BMI,totalenergy,marital
status,yearsofeducation,
smoking,alcohol,physical
activity,historyofhypertensionor
diabetes,menopausalstatus,
dietaryconfounders(totalprotein,
saturatedfat,sodiumintake)
Tucker,
2005
7daydiet
record.Self
completed,but
ambiguousor
incomplete
recordsclarified
bytelephone
interview
HR Causeofdeath
determinedby
consensusof
threephysicians
usingdeath
certificates,
hospitaland
physician
records,and
autopsydata
Fruitand
vegetables
(servings/day):0
5,≥5
Age,totalenergyintake,saturate
fat,BMI,smoking,alcohol,
physicalactivity,dietary
supplementuse,saturatedfat
intake,andseculartrend(yearof
firstvisitbeforevafter1980)
Genkinger,
2004
FFQ.Self
administered
HR Death
certificates
Fruitand
vegetables
(servings/day):
0.87,1.61,2.31,
3.21,4.89
Age,BMI,smokingstatus,
cholesterolconcentration,energy
intake
13
11
12
45
8
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 18/22
Steffen,
2003
FFQ.
Interviewer
administered.
Onerepeated
measurement
duringfollow
up
HR Events
investigatedand
validatedby
usinghospital
records,and
deaths
investigatedand
validatedby
usingphysician
recordsand
nextofkin
interview
Fruitand
vegetables
(servings/day):1.5,
2.5,3.5,5.0,7.5
Age,BMI,race,sex,andtime
dependentenergyintake,
education,smoking,physical
activity,alcohol,hormone
replacementinwomen,waist:hip
ratio,systolicbloodpressure,and
useofantihypertensivedrugs
Sauvaget,
2003
FFQ.Self
administered.
Questionnaire
comparedwith
recordsof24
hourdietary
survey
HR Nationwide
family
registration
system
Fruit
(servings/week):
01,24,57;
vegetables
(servings/week):
01,24,57
Age,BMI,sex,radiationdose,city,
smokingstatus,alcoholhabits,
educationlevel
Bazzano,
2002
FFQ.
Interviewer
administered
HR Death
certificatesand
hospital
discharge
diagnosis
Fruitand
vegetables
(times/day):<1,1,
2,≥3
Age,sex,race,totalenergyintake,
historyofdiabetes,physical
activity,education,alcohol
consumption,smoking,vitamin
supplementuse
Rissanen,
2002
4dayfood
record.
Interviewer
administered
HR Linkageto
nationaldeath
registryusing
Finnishsocial
securitynumber
Fruitand
vegetables(g/day):
<133,133214,
215293,294408,
>408
Age,BMI,examinationyears,
urinaryexcretionofnicotine
metabolitesandalcohol
consumption,systolicanddiastolic
bloodpressure,diabetes,serum
LDL,HDLandtriglycerides,
maximaloxygenuptake,dietary
factors(energyadjustedintakesof
vitaminCandE,βcarotene,
lycopene,folateandfibre)
Strandhagen,
20009
FFQ.
Interviewer
administered
Total
numberof
patientsand
numbersof
events
reported
Death
certificatesand
autopsyrecords
Fruit
(times/week):01,
23,45,67;
vegetables
(times/week):01,
23,45,67
Sameageforallparticipants
Whiteman,
1999
Selfcompleted
simplefood
frequency
questions
HR Confirmedby
theOfficefor
National
Statistics
Fruit
(times/week):01,
13,47;
vegetables
(times/week):01,
13,47
Age,sex,smoking
Hertog, FFQ.Self HR Death Fruit(g/day):0 Age,BMI,smoking,socialclass,
46
47
7
48
49
50
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 19/22
1996 administered,
validated
againstweighed
7dayfood
record
certificates 27,2870,71118,
>118;vegetables
(times/week):0
79,80112,113
149,>149
alcohol,energyintake,fatintake
Sahyoun,
1996
3dayfood
record.Self
completed
HR Annualindexof
deathsanddeath
certificates
Fruit(g/day):
<163.8,163.8301,
301437.6,>437.6;
vegetables(g/day):
<89.2,89.2187,
187274.8,>274.8
Age,sex,diseasestatus,disabilities
affectingshopping
Colditz,
198520
FFQ.
Interviewer
administered
HR* Confirmedby
Massachusetts
vitalstatistics
register
Vegetables
(servings/day):
<0.7,0.71.0,1.1
1.5,1.62.1,≥2.2
Age
BMI=bodymassindex;FFQ=foodfrequencyquestionnaire;HDL=highdensitylipoprotein;HR=hazard
ratio;LDL=lowdensitylipoprotein;SFA=saturatedfattyacid.
*Confidenceintervalsnotreported.
Fig2Riskofallcausemortalityassociatedwithservings/dayoffruitandvegetables.Weightsarefromrandomeffects
analysis
Table3
 Metaanalysisoffruitandvegetableconsumptionandriskofallcause,cardiovascular,andcancermortality
Comparison Noofstudies Cases/participants
PooledHR*(95%
CI),Pvalue
Heterogeneity
(I ),Pvalue
Begg’stest,
Egger’stest
Allcausemortality
Fruitandvegetables
combined
7 42219/553698 0.95(0.92to0.98),
0.001
82,<0.001 0.76,0.006
Fruit 7 40192/660186 0.94(0.90to0.98),
0.002
77,<0.001 0.54,0.02
50
51
2
78434648
911134345
4951
911134345
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 20/22
Vegetables 7 40192/660186 0.95(0.92to0.99),
0.006
86,<0.001 0.13,0.36
Cardiovascularmortality
Fruitandvegetables
combined
4 6893/469551 0.96(0.92to0.99),
0.02
42,0.16 0.73,0.08
Fruit 6 9744/677674 0.95(0.91to1.00),
0.03
71,0.004 0.90,0.28
Vegetables 6 9744/677674 0.96(0.93to0.99),
0.01
63,0.02 0.88,0.73
Cancermortality
Fruitandvegetables
combined
2 10745/457302 0.97(0.90to1.03),
0.31
68,0.08 —
Fruit 7 16468/640852 0.99(0.97to1.00),
0.05
14,0.33 0.37,0.19
Vegetables 8 16510/642123 0.99(0.97to1.01),
0.19
37,0.13 0.27,0.13
HR=hazardratio.
*Perserving/day.
Fig3Doseresponserelationbetweenfruitandvegetableconsumptionandriskofallcausemortality
911134345
4951
784348
911134351
911134351
843
9134347
4951
913204347
4951
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 21/22
Fig4Riskofallcausemortalityassociatedwithservings/dayoffruit.Weightsarefromrandomeffectsanalysis
Fig5Riskofallcausemortalityassociatedwithservings/dayofvegetables.Weightsarefromrandomeffectsanalysis
Table4
 Stratifiedanalysisonassociationsoffruitandvegetableconsumptionandallcausemortality
Combined Fruits Vegetables
No
HR*(95%
CI)
P
value†
I
(%) No
HR*(95%
CI)
P
value†
I
(%) No
HR*(95%
CI)
P
value†
I
(%)
Location:
 US 4 0.95(0.92to
0.97)
0.25 27.4 2 0.94(0.86to
1.02)
0.80 0.0 2 0.85(0.67
to1.07)
0.03 79.6
 Europe 3 0.96(0.91to
1.00)
0.01 88.6 3 0.93(0.84to
1.02)
0.02 73.2 3 0.91(0.82
to1.01)
0.01 83.3
 Asia — — — — 2 0.93(0.90to
0.96)
0.28 13.5 2 0.98(0.92
to1.05)
0.01 95.6
Sex:
 Menand
women
5 0.96(0.93to
0.99)
0.01 86.5 5 0.94(0.90to
0.99)
0.01 82.3 5 0.95(0.92
to0.99)
0.01 90.6
 Men 2 0.94(0.89to 0.53 0.0 2 0.91(0.85to 0.65 0.0 2 0.95(0.90 0.75 0.0
2 2 2
4/13/2016
Fruitandvegetableconsumptionandmortalityfromallcauses,cardiovasculardisease,andcancer:systematicreviewanddoseresponsemetaanalysis…
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4115152/?report=printable 22/22
0.98) 0.98) to1.01)
Followuptime(years):
 ≥13 4 0.97(0.94to
1.00)
0.01 82.0 4 0.94(0.88to
1.00)
0.01 80.0 4 0.98(0.95
to1.02)
0.01 79.1
 <13 3 0.93(0.91to
0.96)
0.36 1.3 3 0.94(0.91to
0.97)
0.39 0.0 3 0.82(0.67
to1.00)
0.01 88.0
Studyqualityscore:
 ≥8 5 0.96(0.93to
0.99)
0.01 82.4 2 0.95(0.88to
1.03)
0.01 90.5 2 0.98(0.95
to1.04)
0.01 92.7
 <8 2 0.93(0.89to
0.96)
0.30 8.9 5 0.93(0.91to
0.96)
0.65 0.0 5 0.89(0.83
to0.96)
0.01 76.3
Noofparticipants:
 ≥10000 3 0.96(0.91to
1.00)
0.01 89.2 4 0.94(0.89to
0.99)
0.01 86.6 4 0.96(0.93
to1.00)
0.01 91.3
 <10000 4 0.94(0.91to
0.97)
0.20 36.1 3 0.92(0.87to
0.98)
0.80 0.0 3 0.90(0.80
to1.01)
0.04 69.0
HR=hazardratio.
*Perserving/day.
†Pforheterogeneity.
ArticlesfromTheBMJareprovidedherecourtesyofBMJGroup