BMJ2013;346:e8539doi:10.1136/bmj.e8539(Published7January2013) Page1of13 Research RESEARCH Egg consumption and risk of coronary heart disease and stroke: dose-response meta-analysis of prospective cohort studies OPENACCESS Ying Rong doctoral student12, Li Chen research fellow12, Tingting Zhu research fellow12, Yadong Song research fellow12, Miao Yu research fellow12, Zhilei Shan research fellow12, Amanda Sands doctoral student3, Frank B Hu professor3, Liegang Liu professor12 1DepartmentofNutritionandFoodHygiene,HubeiKeyLaboratoryofFoodNutritionandSafety,TongjiMedicalCollege,HuazhongUniversityof ScienceandTechnology,430030Wuhan,People’sRepublicofChina;2MinistryofEducationKeyLabofEnvironmentandHealth,SchoolofPublic Health,TongjiMedicalCollege,People’sRepublicofChina;3DepartmentsofNutritionandEpidemiology,HarvardSchoolofPublicHealth,Boston, MA,USA Abstract ConclusionsHigherconsumptionofeggs(uptooneeggperday)is ObjectiveToinvestigateandquantifythepotentialdose-response notassociatedwithincreasedriskofcoronaryheartdiseaseorstroke. associationbetweeneggconsumptionandriskofcoronaryheartdisease Theincreasedriskofcoronaryheartdiseaseamongdiabeticpatients andstroke. andreducedriskofhemorrhagicstrokeassociatedwithhigheregg consumptioninsubgroupanalyseswarrantfurtherstudies. DesignDose-responsemeta-analysisofprospectivecohortstudies. DatasourcesPubMedandEmbasepriortoJune2012andreferences Introduction ofrelevantoriginalpapersandreviewarticles. Cardiovasculardiseaseisnowapublichealthcrisis,affecting EligibilitycriteriaforselectingstudiesProspectivecohortstudies millionsofpeopleinbothdevelopedanddevelopingcountries. withrelativerisksand95%confidenceintervalsofcoronaryheartdisease Althoughtherateofdeathattributabletothediseasehasdeclined orstrokeforthreeormorecategoriesofeggconsumption. indevelopedcountriesinthepastseveraldecades,itisstillthe ResultsEightarticleswith17reports(nineforcoronaryheartdisease, leadingcauseofdeathandextortsaheavysocialandeconomic eightforstroke)wereeligibleforinclusioninthemeta-analysis(3081 tollglobally.1-3Inlowandmiddleincomecountries,the 269personyearsand5847incidentcasesforcoronaryheartdisease, prevalenceofcardiovasculardiseasehasincreaseddramatically. and4148095personyearsand7579incidentcasesforstroke).No By2020,thediseaseisforecastedtobethemajorcauseof evidenceofacurvelinearassociationwasseenbetweenegg morbidityandmortalityinmostdevelopingnations.4 consumptionandriskofcoronaryheartdiseaseorstroke(P=0.67and Inrecentdecades,concernhasmountedregardingthehigh P=0.27fornon-linearity,respectively).Thesummaryrelativeriskof prevalenceandcostsassociatedwithcardiovasculardisease, coronaryheartdiseaseforanincreaseofoneeggconsumedperday withgrowinginterestinalteringriskfactorsandreversingthis was0.99(95%confidenceinterval0.85to1.15;P=0.88forlineartrend) globalepidemic.Amongtheknownriskfactorsfor withoutheterogeneityamongstudies(P=0.97,I2=0%).Forstroke,the cardiovasculardisease,levelsoflowdensitylipoprotein(LDL) combinedrelativeriskforanincreaseofoneeggconsumedperday cholesterolhavearousedparticularattention.IntheWomen’s was0.91(0.81to1.02;P=0.10forlineartrend)withoutheterogeneity HealthStudy,afterameanfollow-upofeightyears,participants amongstudies(P=0.46,I2=0%).Inasubgroupanalysisofdiabetic withthehighestlevelsofLDLcholesterolshowedanotably populations,therelativeriskofcoronaryheartdiseasecomparingthe higherriskofcardiovasculareventsthanthosewiththelowest highestwiththelowesteggconsumptionwas1.54(1.14to2.09;P=0.01). levels.5Inaddition,severalmeta-analysesofobservational Inaddition,peoplewithhighereggconsumptionhada25%(0.57to studiesandrandomizedcontrolledtrialshavefoundthata 0.99;P=0.04)lowerriskofdevelopinghemorrhagicstroke. reductioninconcentrationsofLDLcholesterolcould significantlyreducetheriskofcoronaryheartdiseaseandstroke Correspondence to: L Liu [email protected] Extra material supplied by the author (see http://www.bmj.com/content/346/bmj.e8539?tab=related#webextra) Web appendix:Web tables No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe:http://www.bmj.com/subscribe BMJ2013;346:e8539doi:10.1136/bmj.e8539(Published7January2013) Page2of13 RESEARCH incidenceandmortality.6-9Dietisanimportantdeterminantof Data extraction serumcholesterol,butdietarycholesterolhasonlyamodest Dataextractionwascarriedoutindependentlybytwoauthors contributiontoplasmaconcentrationsofLDLcholesterol.10On (YRandLC)usingastandardextractionform.Weextracted theotherhand,dietarycholesterolmayprompttheoxidationof thefollowinginformationfromeachstudy:authors,yearof LDLandincreasepostprandiallipemia,whichcouldraisethe publication,studyname,studylocation,yearsoffollow-up, riskofvasculardisease.11Tominimizetheelevationofblood samplesize(numberofparticipantsandincidentcases), cholesterolandreducetheriskofcardiovasculardisease,the participants’characteristics(ageandsex),endpoints(coronary AmericanHeartAssociation(AHA)hasrecommendedthe heartdisease,stroke,orboth),outcomesascertainment,egg publictoconsumelessthan300mg/dayofcholesterol.1213 consumptioncategories,covariatesadjustedinthemultivariable Sinceeggsareamajorsourceofdietarycholesterol,withone analysis,andrelativerisks(95%confidenceintervals)forall largeeggcontainingalmost210mgofcholesterol,thepublic categoriesofeggconsumption. hasbeenrecommendedtolimiteggconsumptionunlessthe Qualityassessmentwasperformedaccordingtothe intakeofotherfoodshighincholesterolisrestricted.14However, Newcastle-Ottawaqualityassessmentscale,22whichisa eggsarealsoaninexpensiveandlowcaloriesourceofmany validatedscalefornon-randomisedstudiesinmeta-analyses. othernutrients,includingminerals,proteins,andunsaturated Thisscaleawardsamaximumofninepointstoeachstudy:four fattyacids,whichcouldlowertheriskofcardiovascular forselectionofparticipantsandmeasurementofexposure,two disease.15Additionally,inpopulationsfollowingacarbohydrate forcomparabilityofcohortsonthebasisofthedesignor restricteddiet,dietarycholesterolfromeggscouldincrease analysis,andthreeforassessmentofoutcomesandadequacy plasmaconcentrationsofhighdensitylipoprotein(HDL) offollow-up.Weassignedscoresof0-3,3.5-6,and6.5-9for cholesterol,16whichhasbeensuggestedtoprotectagainst low,moderate,andhighqualityofstudies,respectively.When vasculardisease.1718Therefore,someorganizationshave studieshadseveraladjustmentmodels,weextractedthosethat recommendedthatreducingeggintakemightnotbeimportant reflectedthemaximumextentofadjustmentforpotentially forhealthypeoplewithnormallevelsofcholesterolinthe confoundingvariables. blood.19Foodbaseddietaryguidelinesfromcountriesincluding Forstudiesthatreportedeggintakeasservingsperweekorday, Nepal,Thailand,andSouthAfricarecommendconsumingeggs everydayorregularlyaspartofahealthydiet.20 weassumedthateachservingwasequivalenttooneegg.For studiesthatlackedtheunitofconsumption,thecategorieswere Severalprospectivecohortstudieshaveexaminedtheassociation estimatedbymultiplyingthefrequencyofconsumptionfrom betweeneggconsumptionandriskofcoronaryheartdisease thefoodfrequencyquestionnaireswithanaverageportionsize andstroke.However,therelationbetweeneggconsumption accordingtothemeanintakederivedfromthe24hdiaries.We andriskofcardiovasculardiseaseremainscontroversial. contactedtheauthorsifthedataofinterestwerenotdirectly Therefore,weconductedadose-responsemeta-analysisof showninthepublications.Toresolvediscrepancies,weused prospectivecohortstudiestoquantifytheassociationbetween groupconsensusandconsultedathirdreviewer. eggconsumptionandriskofcoronaryheartdiseaseandstroke. Statistical analysis Methods Inthismeta-analysis,therelativerisksand95%confidence Search strategy intervalswereconsideredastheeffectsizeforallstudies,and thehazardratiosweredeemedequivalenttorelativerisks.Any WeconductedaliteraturesearchofPubMed(Medline)and resultsstratifiedbysexweretreatedastwoseparatereports. EmbasefromJanuary1966throughJune2012forprospective Thosearticlesreportingbothcoronaryheartdiseaseandstroke cohortstudiesexaminingtheassociationbetweenegg werealsotreatedastwoseparatereports.Owingtothedistinct consumptionandriskofCHDandstroke.PubMedsearchterms cut-offpointsforcategoriesindifferentarticles,wecomputed were“CardiovascularDiseases”[MeSH]or“Stroke”[MeSH]or arelativeriskwith95%confidenceintervalforanincreased “CoronaryDisease”[MeSH]or“myocardialinfarction”[MeSH] intakeofoneeggperdayforeachreport.Themethoddescribed orCHD)andegg.SimilarsearchtermswereusedforEmbase. byGreenlandandLongnecker23andOrsiniandcolleagues24was Inaddition,wescrutinizedreferencesfromrelevantoriginal usedtocalculatethetrendfromthecorrelatedestimatesforlog papersandreviewarticlestoidentifyfurtherpertinentstudies. relativeriskacrosscategoriesofeggconsumption.Theamount Nolanguagerestrictionswereimposed.Wefollowedthe ofeggconsumption,distributionsofcasesandpersonyears, standardcriteriaforconductingmeta-analysesofobservational studiesandreportingtheresults.21 andrelativerisksand95%confidenceintervalswereextracted accordingtothismethod. Study selection Themedianormeaneggconsumptionineachcategorywas usedasthecorrespondingdoseofconsumption.Themidpoint Studieswereincludedinthismeta-analysisiftheysatisfiedthe oftheupperandlowerboundarieswasconsideredthedoseof followingcriteria:thestudydesignwasprospective,the eachcategoryifthemedianormeanintakepercategorywas exposureofinterestwaseggconsumption,theoutcomewas notavailable.Ifthehighestcategorywasopenended,the coronaryheartdiseaseorstroke,andtheinvestigatorsreported midpointofthecategorywassetat1.5timesthelower relativeriskswith95%confidenceintervalsforatleastthree boundary.Ifthenumberofcasesandpersonyearswerenot quantitativecategoriesofeggintake.Additionally,weexcluded available,weusedtherelativeriskscomparingthehighestversus reviews,editorials,non-humanstudies,andletterswithout lowestcategoriesofeggintaketoobtainasummaryestimate. sufficientdata.Studiesofotherexposuresanddiseaseswere alsoexcluded.Ifstudypopulationswerereportedmorethan Inaddition,weevaluatedapotentialcurvelinearassociation once,weusedtheresultwiththelongestfollow-uptime. betweeneggconsumptionandriskofcoronaryheartdisease andstroke,usingrestrictedcubicsplineswiththreeknotsat percentiles10%,50%,and90%ofthedistribution.25APvalue forcurvelinearityornon-linearitywascalculatedbytestingthe No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe:http://www.bmj.com/subscribe BMJ2013;346:e8539doi:10.1136/bmj.e8539(Published7January2013) Page3of13 RESEARCH nullhypothesisthatthecoefficientofthesecondsplineisequal assessment(score0-9)yieldedascoreof6.5orabove(high tozero. quality)forallstudies,withanaveragescoreof7.6(web TheheterogeneityamongstudieswasestimatedbytheCochran appendix,tablesAandB). QtestandI2statistic.26Heterogeneitywasconfirmedwitha significancelevelofP≤0.10.TheI2statisticdescribesthe Association between egg consumption and risk of coronary heart disease percentageoftotalvariationinpointestimatesthatcanbe attributedtoheterogeneity.FortheI2metric,weconsidered Sixarticles34363740-42withninereportswereincludedinthe low,moderate,andhighI2valuestobe25%,50%,and75%, dose-responseanalysisofeggconsumptionandriskofcoronary respectively.2627Weusedafixedeffectmodel(Mantel-Haenszel heartdisease.Usingarestrictedcubicsplinesmodel,wefound method)whenheterogeneitywasnegligible,andarandomeffect noevidenceofacurvelinearassociationbetweenegg model(DerSimonianandLairdmethod)whenheterogeneity consumptionandriskofcoronaryheartdisease(P=0.67for wassignificant.28Forestplotsandfunnelplotswereusedto non-linearity;fig2⇓).Thesummaryrelativeriskofcoronary examinetheoveralleffectandassessthepublicationbias, heartdiseaseforanincreaseofoneeggperdaywas0.99(95% respectively. confidenceinterval0.85to1.15;P=0.88forlineartrend).We Wealsoconductedanalysesstratifiedbysex,studylocation, sawnoheterogeneityamongstudies(P=0.97,I2=0%;fig3⇓). numberofcasesandparticipants,durationoffollow-up,repeated Additionally,BeggandEggerregressiontestsprovidedno eggconsumptionmeasurements,studyquality,andwhether evidenceofsubstantialpublicationbias(P>0.05forbothtests). dietvariablesorcholesterollevelswerecontrolledforinmodels. Amongtheincludedstudies,twoarticles3437(fourtotalreports) AllstatisticalanalyseswereperformedwithStataversion11 examinedtherelationbetweeneggconsumptionandriskof (StataCorp),andalltestsweretwosidedwithasignificance coronaryheartdiseaseinpopulationswithdiabetes.Owingto levelof0.05. thelackofdataforpersonyearsindiabeticpopulations,we obtainedthesummaryrelativeriskcomparingthehighestwith Results thelowesteggconsumptionforcoronaryheartdiseasein diabeticpatients(relativerisk1.54(1.14to2.09);P=0.01;table Literature search 3⇓;webappendix,tableC). Figure1⇓showstheresultsofliteratureresearchandselection. Association between egg consumption and Weidentified616articlesfromPubMedand824articlesfrom risk of stroke Embasepriorto20June2012.Afterexclusionofduplicates andstudiesthatdidnotfulfilltheinclusioncriteria,16remaining Thisdose-responseanalysisinvolvedsixarticles3437-394142with articlesseemedtoberelevantforthismeta-analysis.After eightreportsoneggconsumptionandstrokerisk.Wedidnot evaluatingthefulltextsofthese16publications,weexcluded findacurvelinearassociationbetweeneggconsumptionand eightarticlesasfollows.Twoarticles2930wereexcludedowing riskofstroke(P=0.27fornon-linearity;fig4⇓).Thecombined tolackofsufficientdataforestimationofrelativerisks.Another relativeriskofstrokeforanincrementofoneeggconsumed twoarticles3132wereexcludedbecausetheydidnotseparately perdaywas0.91(95%confidenceinterval0.81to1.02;P=0.10 reporttherelativerisksand95%confidenceintervalsfor forlineartrend;fig5⇓).Noheterogeneityofeffectestimates coronaryheartdiseaseorstroke.Wealsoexcludedonereport33 onrelativeriskswasobserved(P=0.46,I2=0%).Neitherthe becauseitwasthemeetingabstractofthestudybyScrafford BeggtestnortheEggertestforpublicationbiasreached andcolleagues.34AstudybyHeandcolleagues35wasexcluded significance(P>0.05forbothtests).Inaddition,three becauseitreportedtheintermediatefollow-upresultsofHealth articles343739withfourreportsprovidedinformationonfatal ProfessionalsFollow-upStudy.Thefinalmeta-analysisincluded stroke(pooledrelativerisk0.94(0.80to1.10);P=0.46;table eightarticles,ofwhichfour3436-38examinedmenandwomen 3).Moreover,fourarticles38394142reportedresultsfordifferent separately.ForthestudybyScraffordandcolleagues,34the typesofstroke,andthreearticles344142providedresultsforstroke estimateoftheassociationbetweeneggconsumptionandstroke inthosewithdiabetes.Forthesestudies,thecombinedrelative mortalityamongmenwasimprecisebecauseofsparsedata,and riskscomparingthehighestversuslowesteggintakewere0.75 thusforthisreportweincludedonlydataforwomen.Intotal, (0.57to0.99)forhemorrhagicstroke,0.91(0.82to1.01)for ourmeta-analysisincludedeightarticleswith17independent ischemicstroke,and0.80(0.29to2.15)fortotalstrokeamong reports. peoplewithdiabetes(table3;webappendix,tableD). Study characteristics Subgroup analyses Tables1⇓and2⇓showtheinformationextractedfromthe Subgroupanalyseswereconductedtoexaminethestabilityof includedstudies,allofwhichhadprospectivecohortdesigns theprimaryresults(table3).Theassociationsbetweenegg andparticipantswithnopriordiagnosesofcardiovascular consumptionandriskofcoronaryheartdiseaseandstrokewere diseaseatbaseline.Themeta-analysisconsistedof263938 similarinsubgroupanalyses,whichweredefinedbysex,study participantswith3081269personyearsoffollow-upfor location,numberofcasesorparticipants,durationoffollow-up, coronaryheartdisease,and210404patientswith4148095 repeatedeggconsumptionmeasurements,studyquality,and personyearsoffollow-upforstroke.Amongtheparticipants, whetherdietvariablesorcholesterollevelswerecontrolledfor wedocumented5847casesofcoronaryheartdiseaseduring inmodels.Anincrementofoneeggconsumedperdaydidnot follow-upperiodsrangingfromeightto20years,and7579 significantlyincreaseriskofcoronaryheartdiseaseorstroke casesofstrokeduringafollow-uprangingfrom8.8to22years. inanyofthecategories. Threecohorts373940wereamongAsians(Japan),andthe others3436384142wereconductedintheUnitedStates.Egg Discussion consumptionwasmeasuredbyfoodfrequencyquestionnaires inallstudies.Fourstudies36384042usedrepeatedmeasurements Thismeta-analysisidentifiednosignificantassociationbetween toupdatedietaryinformationtomoreaccuratelyreflectthe eggconsumptionandriskofcoronaryheartdiseaseorstroke. dietaryintakesoverfollow-up.Resultsofstudyquality Higherintakeofeggs(uptooneeggperday)wasnotassociated No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe:http://www.bmj.com/subscribe BMJ2013;346:e8539doi:10.1136/bmj.e8539(Published7January2013) Page4of13 RESEARCH withriskofcoronaryheartdiseaseorstroke.Similarresults cardiovasculardiseaserisk.Onestudyfoundthatparticipants wereobtainedinsubgroupanalyses.However,amongdiabetic withhighlevelsofcholesterolinthebloodweremorelikelyto participants,highereggconsumptionwasassociatedwitha reducetheireggconsumptionthanothers.40However,our significantlyelevatedriskofcoronaryheartdisease.Onthe subgroupanalysisshowedthattheassociationbetweenegg otherhand,highereggintakewasassociatedwithalowerrisk consumptionandcoronaryheartdiseasewassimilarinthe ofhemorrhagicstroke.Thesesubgroupresultsshouldbe models,withorwithoutadjustmentforcholesterollevels. interpretedwithcaution,becauseonlyafewstudiesfocusedon Recently,acrosssectionalstudyassessedthetotalplaquearea diabeticparticipantsandparticularstrokesubtypes. inpatientsattendingCanadianvascularpreventionclinicsto determinewhethertheatherosclerosisburdenwasrelatedto Results in relation to other studies dietaryeggintake.54Thestudyfoundastrongpositive Todate,themajorityofprospectivestudieshavefoundno associationbetweenthenumberofeggyolksandthedegreeof significantassociationbetweeneggconsumptionandriskof atherosclerosismeasuredbyplaqueareas.However,thestudy coronaryheartdiseaseorstroke.However,Burkeand didnotassessoradjustforotherdietaryorlifestylefactorsand colleagues43analyzeddatafrom514WesternAustralian didnotexaminehardcardiovasculardiseaseendpoints.The aborigineswithalmost14yearsoffollow-upandfoundthat crosssectionalnatureofthestudyalsolimitedcausal riskofcoronaryheartdiseaseincreasedinparticipants interpretationofthedata.Therefore,theresultsfromthiscross consumingeggsmorethantwiceperweek.Butthisstudywas sectionalanalysisshouldbeinterpretedwithcaution.55The smallandanalyzedmultipledietaryandlifestyleexposures. findingsfromourmeta-analysesofprospectivecohortstudies donotsupportapositiveassociationbetweeneggconsumption Somestudieshavefoundaninverseassociationbetweenegg andcardiovasculardiseaseoutcomesinthegeneralpopulation. consumptionandstrokerisk.Forexample,ananalysisofthe ThirdNationalHealthandNutritionExaminationSurvey Subgroupanalyseshavesuggestedapositiveassociation 1988-1994(NHANESIII)datasetfoundasignificantinverse betweeneggconsumptionandcoronaryheartdiseaseriskin associationbetweenhighereggconsumptionandstroke diabeticpatients.Amongdiabeticpopulations,decreasedplasma mortalityamongmen.34AcohortstudyfromJapanfoundthat levelsofapolipoproteinE,togetherwithincreasedlevelsof increasedconsumptionofanimalproducts(includingeggs)was apolipoproteinC-IIIcouldleadtoabnormalcholesterol associatedwithreducedriskoftotalandhemorrhagicstroke transport,whichmightincreasetheriskofcoronaryheart death.39 disease.5657Theadverseeffectofeggconsumptionon lipoproteinprofileandglycemiccontrolcouldcontributetothe Weconsideredseveralpotentialreasonsforthelackofanoverall elevatedriskofcoronaryheartdiseaseindiabeticpopulations. associationbetweeneggconsumptionandcoronaryheartdisease orstroke.Althoughdietarycholesterolinfluencesplasma Inaddition,insulinsensitivitycouldinfluenceHDLmetabolism concentrationsofserumcholesterol,theeffectsarerelatively andcholesteroltransport.5859Riemensandcolleagues60found small.10Inaddition,epidemiologicstudieshavefoundweakor thatpeoplewithlowerinsulinsensitivityhadincreasedlevels littleassociationbetweendietarycholesterolintakeand ofplasmacholesterol,verylowdensitylipoproteincholesterol, cardiovasculardiseaserisk.10Apartfromdietarycholesterol, andLDLcholesterol,comparedwiththosewithhigherinsulin saturatedfatanddietarypatternsmightalsoinfluenceblood sensitivity.Activitiesofplasmalecithin,cholesterolacyl cholesterollevels,44-46suggestingthatcompliancewithgeneral transferase,phospholipidtransferprotein,andhepaticlipase dietaryrecommendationsinsteadofsimplyreducingegg werenegativelycorrelatedwithinsulinsensitivity,whichcould consumptioncouldhaveagreatereffectontheriskof haveenhancedreversecholesteroltransport.60Thesefindings cardiovasculardisease.Additionally,individualdifferencesin suggestabiologicalmechanismforpossibleadverseeffectsof responsetodietarycholesterolvarygreatly,whichcouldaffect insulinresistanceonriskofcoronaryheartdiseaseindiabetic theassociationbetweeneggconsumptionandriskofcoronary populationsthroughcholesterolmetabolism.Nonetheless,this heartdiseaseandstroke.Moreover,severalstudieshaveshown subgroupfindingofapositiveassociationbetweenegg thateggconsumptionfavorstheformationoflargerLDLand consumptionandcoronaryheartdiseaseriskwasbasedona HDLparticles,whichmightenhanceprotectionagainst smallnumberofstudiesandthusneedstobereplicatedinfurther atherosclerosis.4748 studies. Otherthancholesterol,eggsareagoodsourceofothernutrients Severalprospectivecohortstudiesshowedthathemorrhagic suchashighqualityproteinandvitaminD.IntheDiet,Obesity, strokehadaninverseassociationwithserumlevelsof andGene(Diogenes)Project,increasedproteinconsumption cholesterol.61-64Inparticular,theresultofameta-analysis togetherwithamodestreductioninglycemicindexwas including13cohortsfromChinaandJapanshowedthat beneficialforweightcontrol.49Substitutingproteinfor decreasedcholesterolconcentrationsconferredanincreased carbohydratealsopartlyresultedinlowerbloodpressure, riskofhemorrhagicstroke.65Ithasbeensuggestedthatlow improvedlipidslevels,andconcomitantlyreduced cholesterollevelspromotenecrosisofmedialmusclecellsand cardiovascularrisk.50HighervitaminDintakemighthave reduceplateletaggregability,whichcouldleadtoplasmatic beneficialeffectsonthereductionofvisceraladiposetissue51 arterionecrosisandtheincidenceofhemorrhagicstroke.6667It andothercardiovascularriskfactors52. isunclearwhethertheinverseassociationbetweenegg consumptionandhemorrhagicstrokeismediatedthroughlow Anotherpossibilityisthatlifestylefactorsassociatedwithegg levelsofserumcholesterolorothermechanisms.Sincethis consumptionmighthaveobscuredapositiveassociationbetween subgroupfindingwasbasedonasmallnumberofstudies,the eggconsumptionandriskofcoronaryheartdiseaseandstroke. resultsshouldbeinterpretedwithcaution. However,regulareggconsumptiontendstobeassociatedwith unhealthylifestylefactorssuchassmokingandphysical Strengths and limitations inactivity.343653Higherconsumptionofeggsisalsolikelytobe associatedwithincreasedconsumptionofredandprocessed Ourstudyhasseveralstrengths.Ourmeta-analysisincluded meats.36Theseconfoundingfactorstendtoexaggeraterather prospectivecohortstudieswithlargesamplesizeandlong thanmasktheassociationbetweeneggconsumptionand durationoffollow-up,whichsignificantlyincreasedthe No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe:http://www.bmj.com/subscribe BMJ2013;346:e8539doi:10.1136/bmj.e8539(Published7January2013) Page5of13 RESEARCH statisticalpowertodetectpotentialassociations.Weinvestigated (NSFC81072291),andNationalBasicResearchProgram adose-responserelationbetweeneggconsumptionandriskof (2009CB118803)ofChina.Thefundershadnoroleinstudydesign, coronaryheartdiseaseandstroke,allowingustoexaminethe datacollectionandanalysis,decisiontopublish,orpreparationofthe shapeofthispossibleassociation.Linearandnon-linearrelations manuscript. werealsotestedtoquantifytheassociations.Inaddition,we Competinginterests:AllauthorshavecompletedtheICMJEuniform usedmodelsadjustingformostestablishedriskfactorsanddid disclosureformatwww.icmje.org/coi_disclosure.pdf(availableon stratifiedanalysestoexplorewhethersomefactorscouldexplain requestfromthecorrespondingauthor)anddeclare:nosupportfrom theresults. anyorganizationforthesubmittedwork;nofinancialrelationshipswith Severallimitationsofourstudyshouldalsobeacknowledged. anyorganizationsthatmighthaveaninterestinthesubmittedworkin Firstly,errorsinmeasurementofeggintakeandotherdietary thepreviousthreeyears;andnootherrelationshipsoractivitiesthat habitscouldhaveattenuatedindividualstudyresultsandledto couldappeartohaveinfluencedthesubmittedwork. thenullassociationbetweeneggconsumptionandriskof Ethicalapproval:Ethicalapprovalnotneeded. coronaryheartdiseaseandstroke.Allthestudiesinouranalysis Datasharing:Noadditionaldataavailable. assessedeggconsumptionusingfoodfrequencyquestionnaires, severalofwhichhavebeenvalidatedwithreasonable 1 FordES,CapewellS.CoronaryheartdiseasemortalityamongyoungadultsintheU.S. reproducibilityandvalidityofselfreportedeggintake.However, from1980through2002:concealedlevelingofmortalityrates.JAmCollCardiol misreportingofintakewasstillinevitable.68-70 2007;50:2128-32. 2 RogerVL,GoAS,Lloyd-JonesDM,BenjaminEJ,BerryJD,BordenWB,etal.Heart Thecookingmethodsofeggsandtheamountofsaltaddedto diseaseandstrokestatistics—2012update:areportfromtheAmericanHeartAssociation. Circulation2012;125:e2-220. eggswerenotavailableinmostoftheincludedstudies.The 3 RogerVL,GoAS,Lloyd-JonesDM,BenjaminEJ,BerryJD,BordenWB,etal.Executive nutrientcontentsofeggscouldalterdependingondifferent summary:heartdiseaseandstrokestatistics—2012update:areportfromtheAmerican HeartAssociation.Circulation2012;125:188-97. cookingmethodsorfeedingmethodsofchicken.Inaddition, 4 CelermajerDS,ChowCK,MarijonE,AnsteyNM,WooKS.Cardiovasculardiseasein wecouldnotuniformlyquantifythesizeofeggsineachstudy. thedevelopingworld:prevalences,patterns,andthepotentialofearlydiseasedetection. JAmCollCardiol2012;60:1207-16. Moreover,participantswithhighereggintakeconsumedmore 5 RidkerPM,RifaiN,RoseL,BuringJE,CookNR.ComparisonofC-reactiveproteinand dietarycholesterolandproteinbutfewercarbohydratesand low-densitylipoproteincholesterollevelsinthepredictionoffirstcardiovascularevents. weremorelikelytohavelowerlevelsofeducationthanthose NEnglJMed2002;347:1557-65. 6 CholesterolTreatmentTrialists’(CTT)Collaborators.TheeffectsofloweringLDL withlowereggintake.3436Severalstudiesadjustedforthose cholesterolwithstatintherapyinpeopleatlowriskofvasculardisease:meta-analysisof confoundingfactors.Toreducethisbias,weconducteda individualdatafrom27randomisedtrials.Lancet2012;380:581-90. 7 BaigentC,BlackwellL,EmbersonJ,HollandLE,ReithC,BhalaN,etal.Efficacyand stratifiedanalysisandfoundtheresultstoberobustindifferent safetyofmoreintensiveloweringofLDLcholesterol:ameta-analysisofdatafrom170,000 strataofcovariates. participantsin26randomisedtrials.Lancet2010;376:1670-81. 8 BrielM,Ferreira-GonzalezI,YouJJ,KaranicolasPJ,AklEA,WuP,etal.Association Secondly,duringthelongfollow-up,participantsmayhave betweenchangeinhighdensitylipoproteincholesterolandcardiovasculardisease morbidityandmortality:systematicreviewandmeta-regressionanalysis.BMJ changedtheirdiets.However,inourmeta-analysis,nearlyhalf 2009;338:b92. theincludedstudiesupdatedthedietinformationfromfood 9 SnidermanAD,WilliamsK,ContoisJH,MonroeHM,McQueenMJ,deGraafJ,etal.A frequencyquestionnaires.Stratifiedanalysisindicatedthatthe meta-analysisoflow-densitylipoproteincholesterol,non-high-densitylipoprotein cholesterol,andapolipoproteinBasmarkersofcardiovascularrisk.CircCardiovascQual associationsbetweeneggconsumptionandriskofcoronary Outcomes2011;4:337-45. heartdiseaseandstrokeweresimilar,regardlessofwhether 10 KanterMM,Kris-EthertonPM,FernandezML,VickersKC,KatzDL.Exploringthefactors thataffectbloodcholesterolandheartdiseaserisk:isdietarycholesterolasbadforyou repeatedeggconsumptionmeasurementswereconsidered. ashistoryleadsustobelieve?AdvNutr2012;3:711-7. 11 SpenceJD,JenkinsDJ,DavignonJ.Dietarycholesterolandeggyolks:notforpatients Thirdly,somestudiesconsideredtheintakeoffoodsinwhich atriskofvasculardisease.CanJCardiol2010;26:e336-9. eggwasthemainingredient.However,theresultssuggested 12 LichtensteinAH,AppelLJ,BrandsM,CarnethonM,DanielsS,FranchHA,etal.Summary thattheamountofeggsestimatedinotherfoodswasrelatively ofAmericanHeartAssociationDietandLifestyleRecommendationsrevision2006. ArteriosclerThrombVascBiol2006;26:2186-91. smallandwasunlikelytoaffecttheaforementionedassociations. 13 LichtensteinAH,AppelLJ,BrandsM,CarnethonM,DanielsS,FranchHA,etal.Dietand Finally,thestatisticalpowerwaslimitedinsubgroupanalyses lifestylerecommendationsrevision2006:ascientificstatementfromtheAmericanHeart AssociationNutritionCommittee.Circulation2006;114:82-96. ofdiabeticpatientsorsubtypesofstroke. 14 KraussRM,EckelRH,HowardB,AppelLJ,DanielsSR,DeckelbaumRJ,etal.AHA DietaryGuidelines:revision2000:astatementforhealthcareprofessionalsfromthe Conclusions NutritionCommitteeoftheAmericanHeartAssociation.Stroke2000;31:2751-66. 15 SongWO,KerverJM.NutritionalcontributionofeggstoAmericandiets.JAmCollNutr 2000;19(5suppl):556-62S. Insummary,resultsfromourmeta-analysisdonotsupportthat 16 MutungiG,RatliffJ,PuglisiM,Torres-GonzalezM,VaishnavU,LeiteJO,etal.Dietary highereggconsumptionisassociatedwithelevatedriskof cholesterolfromeggsincreasesplasmaHDLcholesterolinoverweightmenconsuming acarbohydrate-restricteddiet.JNutr2008;138:272-6. coronaryheartdiseaseandstroke.Subgroupanalysessuggest 17 DiAngelantonioE,SarwarN,PerryP,KaptogeS,RayKK,ThompsonA,etal.Major apositiveassociationbetweenhighereggintakeandriskof lipids,apolipoproteins,andriskofvasculardisease.JAMA2009;302:1993-2000. coronaryheartdiseaseindiabeticpatients,andaninverse 18 HuxleyRR,BarziF,LamTH,CzernichowS,FangX,WelbornT,etal.Isolatedlowlevels ofhigh-densitylipoproteincholesterolareassociatedwithanincreasedriskofcoronary associationbetweenhighereggconsumptionandincidenceof heartdisease:anindividualparticipantdatameta-analysisof23studiesintheAsia-Pacific hemorrhagicstroke.Studieswithlargersamplesizesandlonger region.Circulation2011;124:2056-64. 19 BetterHealthChannel(Australia).Cholesterol.2009.www.betterhealth.vic.gov.au/Bhcv2/ follow-uptimesarewarrantedtoconfirmthesesubgroupresults. bhcarticles.nsf/pages/Cholesterol_explained?open. 20 FoodandAgriculturalOrganizationoftheUnitedNations.Foodbaseddietaryguidelines bycountry.2009.www.fao.org/ag/humannutrition/nutritioneducation/fbdg/en/. WethankCatherineSauvaget,EricJGrant,andAdamMBernsteinfor 21 StroupDF,BerlinJA,MortonSC,OlkinI,WilliamsonGD,RennieD,etal.Meta-analysis providingdataforthemeta-analysis. ofobservationalstudiesinepidemiology:aproposalforreporting.Meta-analysisOf ObservationalStudiesinEpidemiology(MOOSE)group.JAMA2000;283:2008-12. Contributors:YRandLLconceivedthestudy.YRandLCsearchedthe 22 WellsGA,SheaB,O’ConnellD,PetersonJ,WelchV,LososM,etal.The databasesandcheckedthemaccordingtotheeligiblecriteriaand Newcastle-OttawaScale(NOS)forassessingthequalityofnonrandomizedstudiesin meta-analyses.2011.www.ohri.ca/programs/clinical_epidemiology/oxford.asp. exclusioncriteria.LLhelpeddevelopsearchstrategies.TZgaveadvice 23 GreenlandS,LongneckerMP.Methodsfortrendestimationfromsummarized onmeta-analysismethodology.YShelpedextractquantitativedatafrom dose-responsedata,withapplicationstometa-analysis.AmJEpidemiol1992;135:1301-9. 24 OrsiniN,BelloccoR,GreenlandS.Generalizedleastsquaresfortrendestimationof somepapers.YS,MY,andZSanalyzedthedata.YRwrotethedraft summarizeddose-responsedata.StataJournal2006;6:40-57. ofthepaper.LC,TZ,YS,MY,ZS,AS,FBH,andLLcontributedto 25 HarrellFEJr,LeeKL,PollockBG.Regressionmodelsinclinicalstudies:determining relationshipsbetweenpredictorsandresponse.JNatlCancerInst1988;80:1198-202. writing,reviewing,orrevisingthepaper.LListheguarantor. 26 HigginsJP,ThompsonSG,DeeksJJ,AltmanDG.Measuringinconsistencyin Funding:ThisworkwasfundedbytheNationalScienceandTechnology meta-analyses.BMJ2003;327:557-60. 27 HigginsJP.Commentary:heterogeneityinmeta-analysisshouldbeexpectedand SupportProgram(2012BAI02B02),NationalNaturalScienceFoundation appropriatelyquantified.IntJEpidemiol2008;37:1158-60. No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe:http://www.bmj.com/subscribe BMJ2013;346:e8539doi:10.1136/bmj.e8539(Published7January2013) Page6of13 RESEARCH Whatisalreadyknownonthistopic Cardiovasculardiseaseaffectsmillionsofpeopleinbothdevelopedanddevelopingcountries Asamajorsourceofdietarycholesterol,eggshavebeeninvestigatedbyseveralepidemiologicstudiesinrelationtoriskofcoronary heartdiseaseandstroke However,whethereggconsumptionincreasesthefutureriskofcoronaryheartdiseaseandstrokeremainsunclear Whatthisstudyadds Consumptionofuptooneeggperdaywasnotassociatedwithincreasedriskofcoronaryheartdiseaseorstroke Subgroupanalysissuggestedthatconsumptionofuptooneeggperdaywasassociatedwithasignificantlyelevatedriskofcoronary heartdiseaseindiabeticpopulations,andareducedriskofhemorrhagicstroke 28 LauJ,IoannidisJP,SchmidCH.Quantitativesynthesisinsystematicreviews.AnnIntern 52 BrandenburgVM,VervloetMG,MarxN.TheroleofvitaminDincardiovasculardisease: Med1997;127:820-6. frompresentevidencetofutureperspectives.Atherosclerosis2012;225:253-63. 29 DawberTR,NickersonRJ,BrandFN,PoolJ.Eggs,serumcholesterol,andcoronary 53 DjousseL,GazianoJM.EggconsumptionandriskofheartfailureinthePhysicians’ heartdisease.AmJClinNutr1982;36:617-25. HealthStudy.Circulation2008;117:512-6. 30 SmithMA,FinnR,GreenJR.Eggandmeatconsumptioninmyocardialinfarction. 54 SpenceJD,JenkinsDJ,DavignonJ.Eggyolkconsumptionandcarotidplaque. Practitioner1983;227:673-4. Atherosclerosis2012;224:469-73. 31 HoustonDK,DingJ,LeeJS,GarciaM,KanayaAM,TylavskyFA,etal.Dietaryfatand 55 ZampelasA.Stillquestioningtheassociationbetweeneggconsumptionandtheriskof cholesterolandriskofcardiovasculardiseaseinolderadults:theHealthABCStudy.Nutr cardiovasculardiseases.Atherosclerosis2012;224:318-9. MetabCardiovascDis2011;21:430-7. 56 FieldingCJ,CastroGR,DonnerC,FieldingPE,ReavenGM.Distributionofapolipoprotein 32 ZazpeI,BeunzaJJ,Bes-RastrolloM,WarnbergJ,delaFuente-ArrillagaC,BenitoS,et Eintheplasmaofinsulin-dependentandnoninsulin-dependentdiabeticsanditsrelation al.EggconsumptionandriskofcardiovasculardiseaseintheSUNProject.EurJClin tocholesterolnettransport.JLipidRes1986;27:1052-61. Nutr2011;65:676-82. 57 VenkatesanS,ImrieH,ReadS,HallidayD.ApoCsubclassesfromnon-insulin-dependent 33 ScraffordC,TranN,BarrajL.Theimpactofeggconsumptiononhearthealthusingthe diabeticpatients—aquantitativecomparisonwithcontrolsubjects.BiochemSocTrans NHANESIIIfollow-upsurvey.FASEBJ2009;23(S1). 1995;23:278S. 34 ScraffordCG,TranNL,BarrajLM,MinkPJ.EggconsumptionandCHDandstroke 58 HowardBV.Insulinresistanceandlipidmetabolism.AmJCardiol1999;84:28-32J. mortality:aprospectivestudyofUSadults.PublicHealthNutr2011;14:261-70. 59 BorggreveSE,DeVriesR,DullaartRP.Alterationsinhigh-densitylipoproteinmetabolism 35 HeK,MerchantA,RimmEB,RosnerBA,StampferMJ,WillettWC,etal.Dietaryfatintake andreversecholesteroltransportininsulinresistanceandtype2diabetesmellitus:role andriskofstrokeinmaleUShealthcareprofessionals:14yearprospectivecohortstudy. oflipolyticenzymes,lecithin:cholesterolacyltransferaseandlipidtransferproteins.EurJ BMJ2003;327:777-82. ClinInvest2003;33:1051-69. 36 HuFB,StampferMJ,RimmEB,MansonJE,AscherioA,ColditzGA,etal.Aprospective 60 RiemensSC,VanTolA,StulpBK,DullaartRP.Influenceofinsulinsensitivityandthe studyofeggconsumptionandriskofcardiovasculardiseaseinmenandwomen.JAMA TaqIBcholesterylestertransferproteingenepolymorphismonplasmalecithin:cholesterol 1999;281:1387-94. acyltransferaseandlipidtransferproteinactivitiesandtheirresponsetohyperinsulinemia 37 NakamuraY,OkamuraT,TamakiS,KadowakiT,HayakawaT,KitaY,etal.Egg innon-diabeticmen.JLipidRes1999;40:1467-74. consumption,serumcholesterol,andcause-specificandall-causemortality:theNational 61 TanakaH,UedaY,HayashiM,DateC,BabaT,YamashitaH,etal.Riskfactorsfor IntegratedProjectforProspectiveObservationofNon-communicableDiseaseandIts cerebralhemorrhageandcerebralinfarctioninaJapaneseruralcommunity.Stroke TrendsintheAged,1980(NIPPONDATA80).AmJClinNutr2004;80:58-63. 1982;13:62-73. 38 BernsteinAM,PanA,RexrodeKM,StampferM,HuFB,MozaffarianD,etal.Dietary 62 IsoH,JacobsDRJr,WentworthD,NeatonJD,CohenJD.Serumcholesterollevelsand proteinsourcesandtheriskofstrokeinmenandwomen.Stroke2012;43:637-44. six-yearmortalityfromstrokein350,977menscreenedforthemultipleriskfactor 39 SauvagetC,NaganoJ,AllenN,GrantEJ,BeralV.Intakeofanimalproductsandstroke interventiontrial.NEnglJMed1989;320:904-10. mortalityintheHiroshima/NagasakiLifeSpanStudy.IntJEpidemiol2003;32:536-43. 63 YanoK,ReedDM,MacLeanCJ.SerumcholesterolandhemorrhagicstrokeintheHonolulu 40 NakamuraY,IsoH,KitaY,UeshimaH,OkadaK,KonishiM,etal.Eggconsumption, HeartProgram.Stroke1989;20:1460-5. serumtotalcholesterolconcentrationsandcoronaryheartdiseaseincidence:JapanPublic 64 NeatonJD,BlackburnH,JacobsD,KullerL,LeeDJ,SherwinR,etal.Serumcholesterol HealthCenter-basedprospectivestudy.BrJNutr2006;96:921-8. levelandmortalityfindingsformenscreenedintheMultipleRiskFactorInterventionTrial. 41 QureshiAI,SuriFK,AhmedS,NasarA,DivaniAA,KirmaniJF.Regulareggconsumption MultipleRiskFactorInterventionTrialResearchGroup.ArchInternMed doesnotincreasetheriskofstrokeandcardiovasculardiseases.MedSciMonit 1992;152:1490-500. 2007;13:CR1-8. 65 Bloodpressure,cholesterol,andstrokeineasternAsia.EasternStrokeandCoronary 42 DjousseL,GazianoJM.Eggconsumptioninrelationtocardiovasculardiseaseand HeartDiseaseCollaborativeResearchGroup.Lancet1998;352:1801-7. mortality:thePhysicians’HealthStudy.AmJClinNutr2008;87:964-9. 66 OonedaG,YoshidaY,SuzukiK,ShinkaiH,HoriS,KoboriK,etal.Smoothmusclecells 43 BurkeV,ZhaoY,LeeAH,HunterE,SpargoRM,GraceyM,etal.Health-related inthedevelopmentofplasmaticarterionecrosis,arteriosclerosis,andarterialcontraction. behavioursaspredictorsofmortalityandmorbidityinAustralianAborigines.PrevMed BloodVessels1978;15:148-56. 2007;44:135-42. 67 TandonN,HarmonJT,RodbardD,JamiesonGA.Thrombinreceptorsdefine 44 SpadyDK,WoollettLA,DietschyJM.RegulationofplasmaLDL-cholesterollevelsby responsivenessofcholesterol-modifiedplatelets.JBiolChem1983;258:11840-5. dietarycholesterolandfattyacids.AnnuRevNutr1993;13:355-81. 68 WesterterpKR,GorisAH.Validityoftheassessmentofdietaryintake:problemsof 45 ClarkeR,FrostC,CollinsR,ApplebyP,PetoR.Dietarylipidsandbloodcholesterol: misreporting.CurrOpinClinNutrMetabCare2002;5:489-93. quantitativemeta-analysisofmetabolicwardstudies.BMJ1997;314:112-7. 69 BinghamSA,LubenR,WelchA,WarehamN,KhawKT,DayN.Areimprecisemethods 46 HowellWH,McNamaraDJ,ToscaMA,SmithBT,GainesJA.Plasmalipidandlipoprotein obscuringarelationbetweenfatandbreastcancer?Lancet2003;362:212-4. responsestodietaryfatandcholesterol:ameta-analysis.AmJClinNutr1997;65:1747-64. 70 KipnisV,FreedmanLS.Impactofexposuremeasurementerrorinnutritionalepidemiology. 47 GreeneCM,WatersD,ClarkRM,ContoisJH,FernandezML.PlasmaLDLandHDL JNatlCancerInst2008;100:1658-9. characteristicsandcarotenoidcontentarepositivelyinfluencedbyeggconsumptionin Accepted:10December2012 anelderlypopulation.NutrMetab(Lond)2006;3:6. 48 MutungiG,WatersD,RatliffJ,PuglisiM,ClarkRM,VolekJS,etal.Eggsdistinctly modulateplasmacarotenoidandlipoproteinsubclassesinadultmenfollowinga Citethisas:BMJ2013;346:e8539 carbohydrate-restricteddiet.JNutrBiochem2010;21:261-7. 49 LarsenTM,DalskovSM,vanBaakM,JebbSA,PapadakiA,PfeifferAF,etal.Dietswith Thisisanopen-accessarticledistributedunderthetermsoftheCreativeCommons highorlowproteincontentandglycemicindexforweight-lossmaintenance.NEnglJ AttributionNon-commercialLicense,whichpermitsuse,distribution,andreproductionin Med2010;363:2102-13. 50 AppelLJ,SacksFM,CareyVJ,ObarzanekE,SwainJF,MillerER3rd,etal.Effectsof anymedium,providedtheoriginalworkisproperlycited,theuseisnoncommercialand protein,monounsaturatedfat,andcarbohydrateintakeonbloodpressureandserum isotherwiseincompliancewiththelicense.See:http://creativecommons.org/licenses/by- lipids:resultsoftheOmniHeartrandomizedtrial.JAMA2005;294:2455-64. nc/2.0/andhttp://creativecommons.org/licenses/by-nc/2.0/legalcode. 51 RosenblumJL,CastroVM,MooreCE,KaplanLM.CalciumandvitaminDsupplementation isassociatedwithdecreasedabdominalvisceraladiposetissueinoverweightandobese adults.AmJClinNutr2012;95:101-8. No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe:http://www.bmj.com/subscribe BMJ2013;346:e8539doi:10.1136/bmj.e8539(Published7January2013) Page7of13 RESEARCH Tables Table1|Characteristicsofparticipantsandfollow-upinincludedstudiesofeggconsumptioninrelationtoriskofcoronaryheartdisease andstroke Ageat Meanlength Publication Sexof baseline Noof offollow-up Author year Studyname Country population (years) participants (years) Endpoints(Noofcases) Huetal36 1999 HealthProfessionals USA Male 40-75 37851 8 Coronaryheartdisease(866) Follow-upStudy Huetal36 1999 Nurses’HealthStudy USA Female 34-59 80082 14 Coronaryheartdisease(939) Sauvagetetal39 2003 Hiroshima/Nagasaki Japan Maleand 34-103 34807 16-17 Strokemortality(1259) LifeSpanStudy female Nakamuraetal37 2004 NIPPONDATA80 Japan Male 30+ 4077 14 Ischemicheartdisease mortality(39),stroke mortality(112) Nakamuraetal37 2004 NIPPONDATA80 Japan Female 30+ 5186 14 Ischemicheartdisease mortality(41),stroke mortality(107) Nakamuraetal40 2006 JapanPublicHealth Japan Maleand 40-69 90735 10.2 Coronaryheartdisease(462) Center-based female prospectivestudy Qureshietal41 2007 NHANES-I USA Maleand 25-74 9734 20 Coronaryarterydisease female (1584),stroke(655) Djousseetal42 2008 Physicians’Health USA Male 40-85 21327 20 Myocardialinfarction(1550), StudyI stroke(1342) Scraffordetal34 2011 NHANESIII USA Male 17+ 6833 8.8 Coronaryheartdisease mortality(198),stroke mortality(63) Scraffordetal34 2011 NHANESIII USA Female 17+ 8113 8.9 Coronaryheartdisease mortality(168),stroke mortality(74) Bernsteinetal38 2012 HealthProfessionals USA Male 40-75 43150 22 Stroke(1397) Follow-UpStudy Bernsteinetal38 2012 Nurses’HealthStudy USA Female 30-55 84010 26 Stroke(2633) NHANES-I=FirstNationalHealthandNutritionExaminationSurvey;NHANESIII=ThirdNationalHealthandNutritionExaminationSurvey;NIPPONDATA80=National IntegratedProjectforProspectiveObservationofNon-communicableDiseaseanditsTrendsintheAged,1980. No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe:http://www.bmj.com/subscribe BMJ2013;346:e8539doi:10.1136/bmj.e8539(Published7January2013) Page8of13 RESEARCH Table2|Outcomesandcovariatesofincludedstudiesofeggconsumptioninrelationtoriskofcoronaryheartdiseaseandstroke Study Endpoints Caseascertainment Dietarycategoryandrelativerisk(95%CI) Covariatesinfullyadjustedmodel Huetal,199936 Coronaryheart Selfreporteddiagnosis <1egg/week,1.0(reference);1,1.06(0.88to1.27); Age,bodymassindex,twoyeartime (men) disease orconfirmedbymedical 2-4,1.12(0.95to1.33);5-6,0.90(0.63to1.27);≥7, periods,smoking,parentalhistoryof recordsorautopsy 1.08(0.79to1.48) myocardialinfarction,multivitamin Huetal,199936 Coronaryheart Sameasabove <1egg/week,1.0(reference);1,0.82(0.67to1.00); supplementuse,alcoholconsumption, menopausalstatusand (women) disease 2-4,0.99(0.82to1.18);5-6,0.95(0.70to1.29);≥7, postmenopausalhormoneuse 0.82(0.60to1.13) (women),historyofhypertension, physicalactivity,andtotalenergyintake Sauvagetetal, Strokemortality Confirmedby Never,1.0(reference);≤1egg/week*,0.74(0.55to Age,sex,birthcohort,city,radiation 200339 nationwidefamily 1.00);2-4eggs/week,0.78(0.59to1.04);almostdaily, dose,bodymassindex,smoking, registrationsystem 0.72(0.54to0.97) alcoholhabits,educationlevel,history ofdiabetes,orhypertension Nakamuraetal, Ischemicheart ConfirmedbyNational Ischemicheartdiseasemortality:seldom,1.18(0.26 Age,serumcreatinine,totalcholesterol, 200437(men) diseasemortality, VitalStatistics to5.42);1-2eggs/week,1.71(0.78to3.76);0.5 bloodglucose,bodymassindex,blood strokemortality eggs/day,1.49(0.63to3.48);1egg/day,1.0 pressures,useofbloodpressure (reference) loweringdrugs,smoking,andalcohol Strokemortality:seldom,0.93(0.36to2.40);1-2 intake eggs/week,1.09(0.69to1.72);0.5eggs/day,1.10 (0.68to1.76);1egg/day,1.0(reference);≥2eggs/day, 0.25(0.03to1.81) Nakamuraetal, Ischemicheart Sameasabove Ischemicheartdiseasemortality:seldom,1.42(0.56 200437(women) diseasemortality, to3.62);1-2eggs/week,0.64(0.28to1.44);0.5 strokemortality eggs/day,0.78(0.35to1.82);1egg/day,1.0 (reference);≥2eggs/day,1.27(0.16to9.80) Strokemortality:seldom,0.78(0.35to1.73);1-2 eggs/week,0.79(0.47to1.33);0.5eggs/day,1.46 (0.89to2.4);1egg/day,1.0(reference);≥2eggs/day, 1.22(0.29to5.17) Nakamuraetal, Coronaryheart Confirmedbymedical <1egg/week*,1.19(0.86to1.64);1-2eggs/week, Age;sex;bodymassindex; 200640 disease records,letter, 1.00(0.77to1.30);3-4eggs/week,1.00(0.79to1.26); hypertension;diabetes;useof telephoneordeath almostdaily,1.0(reference) cholesterolloweringdrugs;smoking; certificate alcoholdrinking;whetherparticipants intendedtoavoidcholesterolrichdiets; consumptionfrequenciesofmeat,fish, vegetables,andfruits;andcohorteffects Qureshietal, Coronaryartery Confirmedbymedical Coronaryarterydisease:<1egg/week,1.0(reference); Age,sex,raceorethnicity,systolicblood 200741 disease,stroke recordsordeath 1-6,1.0(0.9to1.1);>6,1.1(0.9to1.3) pressure,diabetesmellitus,serum certificate Stroke:<1egg/week,1.0(reference);1-6,0.9(0.7to cholesterol,smoking,bodymassindex, 1.0);>6,0.9(0.7to1.1) andeducationalstatus Djousseetal, Myocardialinfarction, Confirmedby Myocardialinfarction:<1egg/week,1.0(reference); Age,bodymassindex,smoking,history 200842 stroke physiciansormedical 1,1.12(0.96to1.31);2-4,1.16(1.00to1.36);5-6, ofhypertension,vitaminintake,alcohol records 1.18(0.93to1.49);≥7,0.90(0.72to1.14) consumption,vegetableconsumption, Stroke:<1egg/week,1.0(reference);1,0.96(0.82to breakfastcereal,physicalactivity, 1.13);2-4,1.06(0.91to1.24);5-6,1.13(0.89to1.42); treatmentarm,atrialfibrillation,diabetes ≥7,0.99(0.80to1.23) mellitus,hypercholesterolemia,and parentalhistoryofprematuremyocardial infarction Scraffordetal, Coronaryheart Notapplicable Coronaryheartdiseasemortality:<1egg/week,1.0 Age,energy,maritalstatus,educational 201134(men) diseasemortality, (reference);1-6,1.26(0.79to2.00);>6,1.13(0.61to status,raceorethnicity,smoking,body strokemortality 2.11) massindex,waisttohipratio,diabetes, Strokemortality:<1egg/week,1.0(reference);1-6, hypertension,anddietaryvariables 1.00(0.49to2.02);>6,0.27(0.10to0.73) Scraffordetal, Coronaryheart Notapplicable Coronaryheartdiseasemortality:<1egg/week,1.0 201134(women) diseasemortality, (reference);1-6,1.12(0.66to1.89);>6,0.92(0.27to strokemortality 3.11) Strokemortality:<1egg/week,1.0(reference);1-6, 0.93(0.46to1.90);>6,1.03(0.25to4.22) Bernsteinetal, Stroke Confirmedbymedical 0.14eggs/week†,1.0(reference);0.49,0.80(0.66to Age,timeperiod,bodymassindex, 201238(men) recordsorautopsy 0.97);1.19,0.88(0.73to1.05);3.01,0.80(0.66to smoking,physicalexercise,parental report 0.96);5.53,0.84(0.68to1.04) historyofearlymyocardialinfarction, menopausalstatusinwomen, Bernsteinetal, Stroke Sameasabove 0.49eggs/week†,1.0(reference);1.26,0.90(0.80to 201238(women) 1.01);2.17,0.94(0.83to1.05);3.01,0.86(0.76to multivitaminuse,vitaminEsupplement use,aspirinuseatleastonceperweek, 0.99);4.69,0.91(0.80to1.04) totalenergy,cerealfiber,alcohol,trans fat,fruitandvegetables,andother proteinsources No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe:http://www.bmj.com/subscribe BMJ2013;346:e8539doi:10.1136/bmj.e8539(Published7January2013) Page9of13 RESEARCH Table2(continued) Study Endpoints Caseascertainment Dietarycategoryandrelativerisk(95%CI) Covariatesinfullyadjustedmodel *Unitofeggconsumptionwasassumed. †Servingsperdayconvertedtoeggsconsumedperweek. No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe:http://www.bmj.com/subscribe BMJ2013;346:e8539doi:10.1136/bmj.e8539(Published7January2013) Page10of13 RESEARCH Table3|Stratifiedanalysesofrelativeriskofcoronaryheartdiseaseandstroke Noofreports* Relativerisk(95%CI) Pforheterogeneity I2 Pfortest Coronaryheartdisease Totalcases 9 0.99(0.85to1.15) 0.97 0.0 0.88 Fatalcases 4 1.18(0.71to1.96) 0.91 0.0 0.53 Coronaryheartdiseasewith 5 1.54(1.14to2.09) 0.59 0.0 0.01 diabetes† Subgroupanalysesfortotalcoronaryheartdisease Sex Male 4 1.06(0.89to1.25) 0.83 0.0 0.53 Female 3 0.99(0.83to1.20) 0.99 0.0 0.95 Studylocation USA 6 1.01(0.90to1.13) 0.96 0.0 0.89 Asia 3 0.95(0.64to1.41) 0.76 0.0 0.79 Noofparticipants >10000 4 0.97(0.81to1.15) 0.87 0.0 0.70 ≤10000 5 1.14(0.83to1.56) 0.93 0.0 0.42 Noofcases >500 4 1.00(0.89to1.13) 0.79 0.0 0.94 ≤500 5 0.96(0.67to1.37) 0.93 0.0 0.83 Durationoffollow-up >15years 2 1.02(0.89to1.16) 0.40 0.0 0.83 ≤15years 7 0.98(0.81to1.19) 0.98 0.0 0.85 Repeatedeggconsumptionmeasurements Yes 4 0.97(0.81to1.15) 0.87 0.0 0.70 No 5 1.14(0.83to1.56) 0.93 0.0 0.42 Studyquality Score≥8 5 0.97(0.80to1.17) 0.93 0.0 0.74 Score<8 4 1.06(0.88to1.28) 0.81 0.0 0.55 Controllingforotherdietvariablesinmodels‡ Yes 6 0.97(0.82to1.15) 0.97 0.0 0.73 No 5 1.14(0.84to1.55) 0.92 0.0 0.40 Controllingforserumcholesterollevelsoruseofcholesterolloweringdrugsinmodels Yes 5 0.97(0.75to1.26) 0.65 0.0 0.83 No 4 1.01(0.86to1.18) 0.97 0.0 0.94 Stroke Totalstroke 8 0.91(0.81to1.02) 0.46 0.0 0.10 Fatalstroke 4 0.94(0.81to1.10) 0.47 0.0 0.46 Hemorrhagicstroke† 3 0.75(0.57to0.99) 0.21 36.8 0.04 Ischemicstroke† 4 0.91(0.82to1.01) 0.79 0.0 0.08 Strokewithdiabetes† 3 0.80(0.29to2.15) 0.09 58.9 0.65 Subgroupanalysesfortotalstroke Sex Male 4 0.89(0.77to1.03) <0.001 90.5 0.13 Female 4 0.91(0.77to1.07) 0.24 29.0 0.25 Studylocation USA 5 0.90(0.79to1.03) 0.38 4.9 0.13 Asia 3 0.94(0.83to1.07) 0.28 20.6 0.34 Noofparticipants >10000 4 0.90(0.80to1.02) 0.25 26.9 0.10 ≤10000 4 0.97(0.75to1.27) 0.50 0.0 0.85 No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe:http://www.bmj.com/subscribe