2.52www.00271

AcceptedManuscriptNonalcoholicFattyLiverDiseaseandRiskofIncidentCardiovascularDisease:AMeta-AnalysisofObservationalStudiesGiovanniTargher,ChristopherD.
Byrne,AmedeoLonardo,GiacomoZoppini,CorradoBarbuiPII:S0168-8278(16)30199-4DOI:http://dx.
doi.
org/10.
1016/j.
jhep.
2016.
05.
013Reference:JHEPAT6109Toappearin:JournalofHepatologyReceivedDate:21March2016RevisedDate:30April2016AcceptedDate:9May2016Pleasecitethisarticleas:Targher,G.
,Byrne,C.
D.
,Lonardo,A.
,Zoppini,G.
,Barbui,C.
,NonalcoholicFattyLiverDiseaseandRiskofIncidentCardiovascularDisease:AMeta-AnalysisofObservationalStudies,JournalofHepatology(2016),doi:http://dx.
doi.
org/10.
1016/j.
jhep.
2016.
05.
013ThisisaPDFfileofanuneditedmanuscriptthathasbeenacceptedforpublication.
Asaservicetoourcustomersweareprovidingthisearlyversionofthemanuscript.
Themanuscriptwillundergocopyediting,typesetting,andreviewoftheresultingproofbeforeitispublishedinitsfinalform.
Pleasenotethatduringtheproductionprocesserrorsmaybediscoveredwhichcouldaffectthecontent,andalllegaldisclaimersthatapplytothejournalpertain.
1NonalcoholicFattyLiverDiseaseandRiskofIncidentCardiovascularDisease:AMeta-AnalysisofObservationalStudies.
GiovanniTargher,MD1,ChristopherD.
Byrne,MBBCh2,3,AmedeoLonardo,MD4,GiacomoZoppini,MD1,andCorradoBarbui,MD51SectionofEndocrinology,DiabetesandMetabolism,DepartmentofMedicine,UniversityandAziendaOspedalieraUniversitariaIntegrataofVerona,Verona,Italy2NutritionandMetabolism,FacultyofMedicine,UniversityofSouthampton,Southampton,UK3SouthamptonNationalInstituteforHealthResearchBiomedicalResearchCentre,UniversityHospitalSouthampton,UK4OutpatientLiverClinicandDivisionofInternalMedicine,DepartmentofBiomedical,MetabolicandNeuralSciences,NOCSAE,AziendaUSLandUniversityofModenaandReggioEmilia,Baggiovara,Modena,Italy5WHOCollaboratingCentreforResearchandTraininginMentalHealthandServiceEvaluation,SectionofPsychiatry,UniversityofVerona,Verona,ItalyWordcount:abstract250;text6243(includingabstract,references,figurelegends,andlaysummary);n.
2Tables+n.
3Figures+online-onlyMaterial(n.
2supplementalTables+n.
2supplementalFigures).
Runningtitle:NAFLDandriskofcardiovasculareventsDisclosures:nothingtodisclose.
AuthorContributions:studyconceptanddesign:GT,CDB,AL;acquisitionofdata:GT,GZ;statisticalanalysisofdata:CB;analysisandinterpretationofdata:GT,CDB,AL;draftingofthemanuscript:GT,CB;criticalrevisionofthemanuscriptforimportantintellectualcontent:AL,CDB,GZ.
Addressforcorrespondence:Prof.
GiovanniTargher,MDSectionofEndocrinology,DiabetesandMetabolismDepartmentofMedicineUniversityandAziendaOspedalieraUniversitariaIntegrataPiazzaleStefani,137126Verona,ItalyE-mail:giovanni.
targher@univr.
it2AbstractBackground&Aims:Therehavebeenmanystudiesoftheeffectsofnonalcoholicfattyliverdisease(NAFLD)andtheriskofcardiovasculardisease(CVD),butthesehaveproducedconflictingresults.
Weperformedameta-analysisofthesestudiestoquantifythemagnitudeoftheassociationbetweenNAFLD(andNAFLDseverity)andriskofCVDevents.
Methods:WesearchedPubMed,Googlescholar,andWebofSciencedatabasesusingterms"NAFLD","cardiovascularevents","cardiovascularmortality","prognosis"andtheircombinationstoidentifyobservationalstudiespublishedthroughJanuary2016.
Weincludedonlyobservationalstudiesconductedinadults>18yearsandinwhichNAFLDwasdiagnosedonimagingorhistology.
Datafromselectedstudieswereextractedandmeta-analysiswasthenperformedusingrandomeffectsmodelling.
Results:Atotalof16unique,observationalprospectiveandretrospectivestudieswith34,043adultindividuals(36.
3%withNAFLD)andapproximately2,600CVDoutcomes(>70%CVDdeaths)overamedianperiodof6.
9yearswereincludedinthefinalanalysis.
PatientswithNAFLDhadahigherriskoffataland/ornon-fatalCVDeventsthanthosewithoutNAFLD(randomeffectoddsratio[OR]1.
64,95%CI1.
26-2.
13).
Patientswithmore'severe'NAFLDwerealsomorelikelytodevelopfatalandnon-fatalCVDevents(OR2.
58;1.
78-3.
75).
Sensitivityanalysesdidnotalterthesefindings.
FunnelplotandEgger'stestdidnotrevealsignificantpublicationbias.
Conclusions:NAFLDisassociatedwithanincreasedriskoffatalandnon-fatalCVDevents.
However,theobservationaldesignofthestudiesincludeddoesnotallowtodrawdefinitivecausalinferences.
Keywords:NAFLD;cardiovasculardisease;CVDevents;mortality;meta-analysis3IntroductionNon-alcoholicfattyliverdisease(NAFLD)isaclinico-pathologicalsyndromethatrangesfromsimplesteatosistononalcoholicsteatohepatitis(NASH)withvaryingamountsoffibrosis,andcirrhosis.
1NAFLDisbecomingthemostcommoncauseofchronicliverdiseaseworldwide,affectingupto30%oftheadultpopulationintheUnitedStatesandEurope.
1-3Overthepastdecade,ithasbecomeincreasinglyclearthatNAFLDisnotonlyassociatedwithanincreasedriskofliver-relatedmorbidityormortality,butalsoitisamultisystemdiseasethataffectsavarietyofextra-hepaticorgansystems,includingthecardiovascularsystem.
3-7Arecentcomprehensivemeta-analysisinvolving27cross-sectionalstudieshasshownthatNAFLDwasassociatedwithvariousmarkersofsubclinicalatherosclerosis,suchasincreasedcarotidarteryintimal-medialthickness,impairedflow-mediatedvasodilation,increasedarterialstiffnessorincreasedcoronaryarterycalcification.
8Alltheseassociationswereindependentofmultiplecardio-metabolicriskfactorsacrossawiderangeofpatientpopulations.
8Severalstudieshavealsodemonstratedthattheprevalenceofclinicallymanifestcardiovasculardisease(CVD)wasalsosignificantlyincreasedamongpatientswithNAFLD(asreviewedelsewhere).
5,6Worryingly,NAFLDwasalsoassociatedwithahigherprevalenceofhigh-riskandvulnerablecoronaryarteryplaques,independentlyoftraditionalCVDriskfactorsandtheextentandseverityofcoronaryatherosclerosis.
9Althoughthecross-sectionalassociationbetweenNAFLDandincreasedCVDprevalenceisstrongandconsistent,itremainsuncertainwhetherthepresenceofNAFLDpredictsincidentCVDeventsorwhetherthemoresevereformsofNAFLDareassociatedwithanevenhigherriskoffutureCVDevents.
Moreover,themechanismslinkingNAFLDtoCVDarecontroversialandseveralputativemechanismshavebeenproposedwhich,however,aretobetracedbacktoliverhistologicchanges,insulinresistanceandoxidativestress.
10Inthiscontext,wehavecarriedoutacomprehensivesystematicreviewandmeta-analysisofpublishedobservationalstudiestogaugepreciselythenatureandmagnitudeoftheassociationbetweenNAFLDandtheriskofincidentCVDevents.
WehavealsoinvestigatedwhethertheseverityofNAFLDisassociatedwithahigherriskofCVDevents.
Clarificationof4theseissuesmayhaveimportantclinicalimplicationsformanagementofpatientswithNAFLD.
MethodsRegistrationofreviewprotocolTheprotocolforthisreviewwasregisteredinadvancewithPROSPERO(InternationalProspectiveRegisterofSystematicReviews,#CRD42016033481).
Typeofstudies,inclusionandexclusioncriteriaanddefinitionofsevereNAFLDStudieswereincludediftheywereobservational,prospectiveorretrospectivestudiesthatreportedfataland/ornon-fatalCVDeventsinadultpatients(>18yearsold)withNAFLDascomparedwithadultindividualswithoutNAFLD.
Studyparticipantswereofeithersexwithnorestrictionsintermsofcomorbidconditions.
WeincludedonlystudiesinwhichthediagnosisofNAFLDwasbasedoneitherradiologicalimagingorhistologyintheabsenceofcompetingcausesofhepaticsteatosis.
Exclusioncriteriawereasfollows:1)studiesthatusedonlyserumliverenzymelevelstodiagnoseNAFLD;2)studiesconductedinpaediatricpopulation(70%CVDdeaths).
StudieswerecarriedoutinUnitedStates,Europe(Denmark,Finland,Germany,Italy,andSweden),Asia(China,SouthKorea,Turkey,andJapan)andAfrica(Egypt).
Mostofthesestudiesincludedmiddle-agedsubjectspredominantlyofmalesex.
Thelengthofthefollow-upperiodrangedfrom3to26.
4years(exceptforthestudybyEmreetal.
34whoconsideredonlyin-hospitaleventsasoutcome);medianfollow-upwasof6.
9years(inter-quartilerange:4.
5-10.
6years).
Ofthe17comparisons,7employedfatalCVDeventsasoutcomemeasure,5fatalandnon-fatalCVDevents(combinedendpoint),and5non-fatalCVDeventsasoutcomemeasure(Table1).
CVDeventswerevalidatedbymedicalrecords,deathcertificatesorendpointcommitteesusingtheInternationalClassificationofDiseasesdiagnosiscodes.
8Ofthe16includedstudies,10receivedeightorninestarsattheNOS,indicatinglowriskofbias(supplementaryTable1).
Comparabilityofcohortswasjudgedathighriskofbiasintwostudies(supplementaryFigure1andsupplementaryTable2).
NAFLDandriskofincidentCVDeventsThedistributionofstudiesbyestimateoftheassociationbetweenNAFLDandriskofCVDeventsisplottedinFigure2.
Sixteenstudies(17comparisons)provideddatasuitableforthepooledprimaryanalysis.
NAFLDwassignificantlyassociatedwithanincreasedriskoffataland/ornon-fatalCVDevents(randomeffectOR1.
64,95%CI1.
26-2.
13,I2=86%).
Whenthiscomparisonwasstratifiedbyoutcome(i.
e.
,analysingseparatelythepublishedstudiesthathadeithernonfatalCVDevents,orfatalCVDevents,orbothasprimaryoutcomes),thepresenceofNAFLDwassignificantlyassociatedwithbothanincreasedriskoffatalandnon-fatalCVDeventsconsideredtogether(randomeffectOR1.
63,95%CI1.
06-2.
48,I2=83%)andanincreasedriskofnon-fatalCVDevents(randomeffectOR2.
52,95%CI1.
52-4.
18,I2=61%);however,theassociationbetweenNAFLDandfatalCVDevents(whentheanalysiswasrestrictedonlytostudieswithCVDmortalityastheprimaryoutcome)wasnotstatisticallysignificant(randomeffectOR1.
31,95%CI0.
87-1.
97,I2=90%)(Figure2).
TheEgger'sregressiontestdidnotshowstatisticallysignificantasymmetryofthefunnelplot,thussuggestingthatpublicationbiaswasunlikely(supplementaryFigure2).
SevereNAFLDandriskofincidentCVDeventsSixstudies(7comparisons)reporteddataonpatientswith'severe'NAFLD,definedeitherbypresenceofhepaticsteatosisonimagingpluseitherelevatedGGTlevelsorhighNAFLDfibrosisscoreorhighhepaticFDGuptakeonpositronemissiontomographyorbyincreasingfibrosisstageonliverhistology.
ThedistributionofstudiesbyestimateoftheassociationbetweensevereNAFLDandriskofincidentCVDeventsisplottedinFigure3.
Comparedwiththenon-NAFLDgroup,thepresenceofmore'severe'NAFLDwassignificantlyassociatedwithanincreasedriskofCVDmortality(randomeffectOR3.
28,95%CI2.
26-4.
77,I2=0)aswellaswithanincreasedriskoffatalandnon-fatalCVDeventsconsideredtogether(randomeffectOR1.
94,95%CI1.
17-3.
21,I2=23%).
Sensitivityanalyses9Limitingtheanalysistohigh-qualitystudiesandlimitingtostudieswithadjustmentformultiplecovariatesprovidedoverallestimatesconsistentwiththeprimaryanalysis(Table2).
Additionally,excludingstudieswiththegeneralpopulationasreferencegroup,17,29andexcludingstudiesthatenrolledonlyparticipantswithdiabetes,hypertensionormyocardialinfarction22,25,29,31,34hadnoeffectonthecomparison(Table2).
Finally,eliminatingeachoftheincludedstudiesfromtheanalysishadnoeffectontheoverallriskofincidentCVDevents(datanotshown).
DiscussionSeveralstudieshaveassessedtheassociationbetweenNAFLDandtheriskofCVD.
ThedataonwhetherNAFLDbyitselfisassociatedwithanincreasedriskofCVDeventsanddeathremainsanissueofdebate.
TheresultsfromthesestudieshavebeenconflictingpartlyduetovariabilityinNAFLDdefinitionandCVDascertainment.
Apriornarrativereviewpublishedin2010byGhourietal.
concludedthatadiagnosisofNAFLDwasinsufficienttoconsiderpatientsasbeingathighriskforCVD,andthattheevidencebaseforCVDriskscreeningbasedonthepresenceofNAFLDwasweak.
38Thepresentedsystematicreviewandresultsofthemeta-analysisinvestigatingtherelationshipbetweenNAFLDandincidentCVDeventsisthemostcomprehensiveassessmentofthisrelationshiptodate.
ThedataproviderobustevidenceoftheassociationbetweenNAFLD(andNAFLDseverity)andriskofincidentCVDevents.
Indeed,theanalysisinvolvesatotalof16uniqueobservationalstudieswithaggregatedataon34,043adultparticipants(36.
3%withNAFLD)andapproximately2,600fatalandnon-fatalCVDoutcomes(>70%CVDdeaths)followed-upoveramedianperiodof6.
9years.
WefoundthatthepresenceofNAFLDwassignificantlyassociatedwitha64%increasedriskofacompositeendpointofCVD(i.
e.
,acombinedoutcomeinclusiveofCVDdeathandnon-fatalCVDeventssuchasmyocardialinfarction,angina,strokeorcoronaryrevascularization).
Whenthiscomparisonwasstratifiedbyoutcome,NAFLDwassignificantlyassociatedwithincreasedrisksofnon-fatalCVDevents,andfatalandnon-fatalCVDeventsconsideredtogether.
Moreover,stratifiedanalysisbyNAFLDseverityamongthe6cohortstudiesthatreported10dataonpatientswithmore'severe'NAFLDshowedthatpatientswithmore'severe'NAFLDhadahigherriskofdevelopingbothCVDmortality(randomeffectOR3.
28,95%CI2.
26-4.
77,I2=0)andfatalandnon-fatalCVDeventsconsideredtogether(randomeffectOR1.
94,95%CI1.
17-3.
21,I2=23%)comparedwithsubjectswithoutNAFLD.
Althoughfurtherlargerobservationalstudiesareneeded,theimagingdataalsosuggeststhatliverfatperseis(inourmeta-analysis)associatedwithanincreasedriskofCVDevents.
Obviously,theresultsofthisstratifiedanalysisshouldbeinterpretedwithcaution,becausewecombineddatafromstudiesreportinghistologicdata(fullyvalidatedtobeasurrogateforclinicaloutcomes)andstudiesthatusedimagingtechniqueswithnon-invasivescoringsystemsorserumbiomarkersfordefiningNAFLDseverity.
Inaddition,thecombinationofstudiesthatuseddifferentdiagnosticmodalitiestoestablishNAFLDseveritymightintroducesomebiasandheterogeneity.
Unfortunately,asdiscussedbelow,mostofthepublishedstudiesthatusedliverbiopsytodiagnoseNAFLD(i.
e.
,the'goldstandard'methodtodefineNAFLDseverity)lackedanadequatecontrolgroupandcannot,therefore,beincludedinthismeta-analysis.
Finally,thekeyquestionofwhethertheprognosticvalueofNAFLDinCVDdevelopmentisrestrictedtoNASHorisalsoassociatedwithsimplesteatosisremainsunresolved.
Todate,thereareonlyveryfewpublishedstudieswithsmallsamplesizesthathavespecificallycomparedtheriskofincidentCVDeventsbetweenpatientswithmoresevereNAFLDandthosewithmildNAFLD,33andmoreresearchisdefinitelyneededtoaddressthisissue.
Apriormeta-analysisdidnotfindanyassociationbetweentheseverityofNAFLDhistologyandriskofCVDevents.
39Collectively,ourfindingsextendtheresultsfromtwoprevioussmallermeta-analyses.
39,40Inthefirststudy,Mussoetal.
39in2011reportedthatNAFLD(definedbyeitherultrasonographyorhistology;n=7prospectivestudiesincluded)wassignificantlyassociatedwithanincreasedriskoffataland/ornon-fatalCVDevents(fixedeffectOR2.
05,95%CI1.
81-2.
31).
Inthesecondmeta-analysis,involving6studies(n=4cross-sectionalandn=2prospectivestudies),Luetal.
40reportedthatNAFLDdiagnosedonultrasonographywasassociatedwithanincreasedriskofCVD(randomeffectOR1.
50,95%CI1.
21-1.
87).
However,themeta-analysisbyLuetal.
includedalsocross-sectionalstudies,andbothofthesemeta-analyses(giventheverysmallnumberofeligiblestudies)didnotreportanalyses11stratifiedeitherbyoutcome(fatalvs.
nonfatalCVDevents)orbyNAFLDseverity.
39,40Morerecently,Younossietal.
havepublishedameta-analysisoftheglobalprevalence,incidence,progressionandoutcomesofNAFLD.
41Althoughtheauthorsdidnotshowanyspecificforestplotofcomparisonbetweenpublishedstudies,theystatedthatNAFLDwassignificantlyassociatedonlywithliver-relatedmortality,butnotwithCVDmortality.
Inparticular,inthismeta-analysisofobservationalstudies(n=6prospectivestudiesincludedfortheanalysisofCVDmortality)examiningtheassociationbetweenNAFLDandCVDmortality,theauthorsstatedthattherewasnosignificantassociationbetweenNAFLDandriskofCVDmortalitywhentheyincludedstudiesthatdefinedNAFLDbybothultrasonographyandserumliverenzymes;incontrast,ifNAFLDwasdiagnosedbyultrasonography,theincidencerateratioforCVDmortalitywasincreasedat1.
37(95%CI1.
23-1.
54)inpatientswithultrasound-diagnosedNAFLDascomparedtothecontrols.
41However,itisimportanttonotethatinDr.
Younossiandcolleagues'meta-analysis,thesamepopulation-basedcohortofindividuals,i.
e.
,theNationalHealthNutritionExaminationSurvey(NHANES)-IIIcohort,wasincludedmanytimes,andallstudiesthatenrolledonlypatientswithdiabetesorobesitywereexcluded.
41Asregardstothis,wepaidgreatattentioninexcludingfromourmeta-analysisallstudiesthatlackedanadequatecontrolgroup,orhadinadequatedataonoutcomesofinterest,orstudiesexhibitinganysignificantoverlapofpopulation.
Forexample,weincludedonlytwopublishedstudiesthatusedthesameNHANES-IIIdatabase(i.
e.
,theLazo'sstudyinthepooledanalysisandtheKim'sstudythatpresenteddataontheNAFLDfibrosisscoreinthestratifiedanalysisforexaminingtheassociationbetweenNAFLDseverityandCVDmortality),26,37butexcludedotherstudies42,43thathaveusedthesameNHANES-IIIdatabaseoftwoabove-mentionedstudies.
Forthesamereason,weexcludedthestudypublishedbyWongetal.
in2011,44giventhatitevaluatedthesamepatientpopulationasthatincludedinamorerecentstudy,withalongerfollow-upperiod,thatweincludedinourmeta-analysis.
31Again,wedidnotconsiderthestudiesbyEkstedtetal.
45andSderbergetal.
46giventhatbothcohortsofpatientswithbiopsy-confirmedNAFLDwereincludedinthemorerecentstudybyEkstedtetal.
(includedinourmeta-analysis).
33Finally,weexcluded,forexample,theretrospectivestudiesbyMatteonietal.
,47Rafiqetal.
,48Adamsetal.
,49andAnguloet12al.
,50whofollowed-uprelativelysmallcohortsofpatientswithbiopsy-provenNAFLD(recruitedfromtertiarygastroenterologycenters),becauseofthelackofanadequatecontrolgroup.
Althoughthisupdatedmeta-analysisofobservationalstudiesprovidesfurthersupportforexistenceofasignificantassociationbetweenNAFLDandincreasedriskoffatalandnon-fatalCVDevents,however,itisimportanttounderlinethatthequalityofpublishedstudiesisnotalwayshigh(only10studiesreceivedeightorninestarsattheNewcastle–OttawaScale;seesupplementaryTable1andsupplementaryFigure1)andthatcausalityremainstobeproveninhigh-qualityinterventionstudies.
AclearunderstandingofthepathophysiologicalpathwayslinkingNAFLDtoCVDeventsremainselusive,becauseoftheintricateinteractionsamongNAFLD,abdominalobesity,oxidativestressandinsulinresistance.
3,51However,thereisnowagrowingbodyofevidencesuggestingthatNAFLD,especiallyitsmoresevereforms(i.
e.
,NASHwithvaryingamountsoffibrosis),exacerbateshepatic/peripheralinsulinresistance,predisposestoatherogenicdyslipidemiaandreleasesavarietyofpro-inflammatory,vasoactiveandthrombogenicfactorsthatmaypromotethedevelopmentofCVD.
3-6,10AccumulatingevidencealsosuggeststhatNAFLDpatientshaveearlychangesincardiacsubstratemetabolism,producingmyocardialfunctional,structuralandarrhythmicconsequences.
6,52-54AlthoughallthesemechanismsplausiblylinkNAFLDtoCVD,nostudiestodatehaveprovenacause-and-effectrelationshipandfurtherresearchiscertainlyneededtogainmechanisticinsightsintothepathophysiologylinkingNAFLDtoCVD.
Collectively,ourfindingsindicatethatadiagnosisofNAFLDmayidentifyasubsetofthegeneralpopulation,whichisexposedtoanincreasedriskofCVDevents.
Additionally,ourfindingssuggestthatthemore'severe'formsofNAFLDareassociatedwithanevengreaterriskofCVDevents.
FurthertoprovidingevidencefortheneedforselectedcasefindingforNAFLDincertainhigh-riskgroups,2ourdataalsoimplythatpatientswithNAFLDshouldundergocarefulcardiovascularsurveillance.
Moreover,thosewiththemoresevereformsofNAFLDneedparticularattentiontoamelioratetheirhighriskofCVDevents.
5,33,45,46,50,55However,beforethisclinicalpracticestrategycanbeadvocated,largeandwell-designedinterventiontrialsareneededtotestwhethertreatingliverdiseaseinNAFLDmaydecreasetheriskofincidentCVDevents.
13Themainstrengthsandlimitationsofthisstudydeservemention.
Ourmeta-analysisprovidesthemostcomprehensiveassessmentandrobustevidencetodateoftheassociationbetweenNAFLD(asdiagnosedeitherbyimagingorbyhistology)andincidenceofmajorCVDeventsanddeath.
Itincludesmultiplecohortstudiesthathadrecruitedparticipantsfromgeneralpopulations,thereforereducinganyeffectsofclinicallyevidentpre-existingdiseaseonNAFLD.
Inaddition,asreportedinTable2,excludingstudiesthatenrolledonlypatientswithdiabetes,hypertensionormyocardialinfarctionprovidedoverallestimatesconsistentwiththeprimaryanalysis.
Moreover,itisimportanttounderlinethatweemployedstandardizedriskestimatesfromalleligiblestudiestoallowaconsistentcombinationofestimatesacrossstudies.
ThelargenumberoftotalCVDeventsprovidedhighstatisticalpowertoquantitativelyassesstheassociationbetweenNAFLDandCVDrisk.
Finally,selectivereportingofstudieswasnotaconcerninouranalyses,asourcomprehensivesearchandcontactwithinvestigatorsmadeitunlikelythatanypublishedreportwasmissedandvisualinspectionofplotsandformaltestsdemonstratednostatisticalevidenceofpublicationbias.
Asregardstothelimitationsofthismeta-analysis,wewereunabletofullyexaminetheimpactofadjustmentforallknownandpotentialCVDriskfactorsandalsocombinemodelsinstudiesthatadjustedforthesamesetofconfoundingfactors,becauseofthevaryingdegreeofconfounderadjustmentacrosstheindividualstudies(indeed,inalargepartofthepublishedstudiestheadjustmentforestablishedCVDriskfactorsandpotentialconfoundingfactorsisoftenincomplete–seeTable1).
DespiteusingacompositeendpointofCVDtomaximizecomparability,wewereunabletoachievethisacrossallstudies,assomestudiesreportedcause-specificCVDoutcomes,suchasmyocardialinfarction,angina,stroke,orCVDdeath.
However,weperformedstratifiedanalysesofspecificCVDoutcomeswheneverpossible.
AplausibleexplanationforthedifferentialrelationshipofNAFLDwithCVDmortalityinpooledvs.
stratifiedanalysismaybedue,atleastinpart,tothedurationofNAFLD,giventhatcomparedtoNASH,simplesteatosistakesalongertimetoprogresstoadvancedfibrosisorcirrhosis,56andtherelativelyshortdurationoffollow-upofthepopulation-basedcohortstudiesthathaveevaluatedmortalityriskasanoutcome.
Weconsiderthatfurtherfollow-upstudiesinlargercohortsofpatientswithbiopsy-confirmed14NAFLD(withanadequategroupofcontrolsubjects)areneededinordertoprovewhetherNAFLDseveritycandifferentiallyaffectriskofincidentCVDevents.
Additionally,inthismeta-analysis,itwasnotpossibletocorrecttheestimatesfordynamicchangesinNAFLDstatusovertime.
ThismayhaveledtounderestimatetheassociationbetweenNAFLDandincidentCVDevents,becausedatainvolvingrepeatmeasurementsoffattyliverwerenotperformedinanyoftheeligiblestudies.
Hence,itisplausibletohypothesizethattheobservedassociationsmaybeevenstrongerthanthoseobservedinthepresentstudy.
AlthoughwefoundsignificantheterogeneitybetweenstudieswheninvestigatingassociationsinallpatientswithNAFLDandCVDevents,itisnoteworthythattherewasverylowheterogeneitybetweenstudies,andstrongerassociationsbetweenNAFLDandCVDrisk,whenwerestrictedtheanalysestostudieswithonlythemore'severe'formsofNAFLD.
Thus,itislikelythatthehighheterogeneityintheoverallanalysisreflectsamixofdifferentpeoplewithNAFLDfromsubjectswithsimplesteatosiswhogenerallyhavenothadthediseaseforlong,tosubjectswithNASHandadvancedfibrosiswhohavehadthediseaseformanyyears.
Wesystematicallyexploredandidentifiedpossiblesourcesofheterogeneityusingstratifiedanalyses,meta-regressionandsensitivityanalyses;moredetailedanalysesofthecausesofheterogeneitywillrequirecollaborativepoolingofindividualparticipantdatafromprospectivestudiesasthesebecomeavailableovertime.
Inconclusion,thefindingsofthisupdatedandlargemeta-analysisofobservationalstudiesindicatethatNAFLDissignificantlyassociatedwithanincreasedriskoffatalandnon-fatalCVDevents,andthatthisriskisprobablyhigherinpresenceofmoresevereliverdisease.
Someuncertainty,however,remainsastowhetherNAFLDisassociatedwithanincreasedriskoffatalandnon-fatalCVDeventsbeyondtheknowncardiovascularriskfactors.
Furthermore,itremainsuncertainastowhetherNASHisassociatedwithgreaterCVDriskthansimplesteatosis.
Additionalwell-controlledprospectivestudiesofamoreextensivepanelofknownCVDriskfactorsareneededtodrawfirmconclusionsaboutanyindependenthepaticcontributiontotheincreasedCVDriskobservedamongpatientswithNAFLD.
FurtherstudiesarealsoneededtoestablishwhetheraddingNAFLDtothecurrentlyavailableriskscoringsystemswillimproveCVDriskprediction.
Finally,inordertoassesscausal15relationshipsbetweenNAFLDandCVDevents,randomizeddouble-blindplacebocontrolledtrialswithCVDoutcomesthatfocusontreatmentsforliverdiseaseinNAFLD,areneeded.
AcknowledgementsGTissupportedinpartbygrantsfromtheUniversitySchoolofMedicineofVerona,Verona,Italy.
CDBissupportedinpartbytheSouthamptonNationalInstituteforHealthResearch(NIHR)BiomedicalResearchCentre.
References1.
RinellaME.
Nonalcoholicfattyliverdisease:asystematicreview.
JAMA2015;313:2263-2273.
2.
Non-alcoholicfattyliverdisease(NAFLD)studygroup,dedicatedtothememoryofProf.
PaolaLoria;LonardoA,BellentaniS,ArgoCK,etal.
Epidemiologicalmodifiersofnon-alcoholicfattyliverdisease:focusonhigh-riskgroups.
DigLiverDis2015;47:997-1006.
3.
AnsteeQM,TargherG,DayCP.
ProgressionofNAFLDtodiabetesmellitus,cardiovasculardiseaseorcirrhosis.
NatRevGastroenterolHepatol2013;10:330-344.
4.
TargherG,DayCP,BonoraE.
Riskofcardiovasculardiseaseinpatientswithnonalcoholicfattyliverdisease.
NEnglJMed2010;363:1341-1350.
5.
ByrneCD,TargherG.
NAFLD:amultisystemdisease.
JHepatol2015;62:S47-64.
6.
BallestriS,LonardoA,BonapaceS,etal.
Riskofcardiovascular,cardiacandarrhythmiccomplicationsinpatientswithnon-alcoholicfattyliverdisease.
WorldJGastroenterol2014;20:1724-1245.
7.
BallestriS,ZonaS,TargherG,etal.
Nonalcoholicfattyliverdiseaseisassociatedwithanalmosttwo-foldincreasedriskofincidenttype2diabetesandmetabolicsyndrome.
Evidencefromasystematicreviewandmeta-analysis.
JGastroenterolHepatol2015Dec14.
doi:10.
1111/jgh.
13264[Epubaheadofprint].
8.
OniET,AgatstonAS,BlahaMJ,etal.
Asystematicreview:burdenandseverityofsubclinicalcardiovasculardiseaseamongthosewithnonalcoholicfattyliver;shouldwecareAtherosclerosis2013;230:258-267.
9.
PuchnerSB,LuMT,MayrhoferT,etal.
High-riskcoronaryplaqueatcoronaryCTangiographyisassociatedwithnonalcoholicfattyliverdisease,independentofcoronaryplaqueandstenosisburden:resultsfromtheROMICATIItrial.
Radiology2015;274:693-701.
10.
LonardoA,SookoianS,PirolaCJ,etal.
Non-alcoholicfattyliverdiseaseandriskofcardiovasculardisease.
Metabolism2015Sep25.
pii:S0026-0495(15)00271-1.
doi:10.
1016/j.
metabol.
2015.
09.
017[Epubaheadofprint].
1611.
TahanV,CanbakanB,BalciH,etal.
Serumgamma-glutamyltranspeptidasedistinguishesnon-alcoholicfattyliverdiseaseathighrisk.
Hepatogastroenterology2008;55:1433-1438.
12.
CasteraL,VilgrainV,AnguloP.
NoninvasiveevaluationofNAFLD.
NatRevGastroenterolHepatol2013;10:666-675.
13.
KeramidaG,PottsJ,BushJ,etal.
Accumulationof(18)F-FDGintheliverinhepaticsteatosis.
AJRAmJRoentgenol2014;203:643-648.
14.
HongSP,NohTS,MoonSH,etal.
Hepaticglucoseuptakeisincreasedinassociationwithelevatedserumγ-glutamyltranspeptidaseandtriglyceride.
DigDisSci2014;59:607-613.
15.
HigginsJPT,GreenS,eds.
CochraneHandbookforSystematicReviewsofInterventionsVersion5.
1.
0[updatedMarch2011].
TheCochraneCollaboration,2011.
Availablefromwww.
cochrane-handbook.
org.
16.
WellsGA,SheaB,HigginsJP,SterneJ,TugwellP,ReevesBC.
Checklistsofmethodologicalissuesforreviewauthorstoconsiderwhenincludingnon-randomizedstudiesinsystematicreviews.
ResSynthMethods2013;4:63-77.
17.
GreenlandS.
Varianceestimatorsforattributablefractionestimatesconsistentinbothlargestrataandsparsedata.
StatMed1987;6:701-708.
18.
HigginsJP,ThompsonSG.
Quantifyingheterogeneityinameta-analysis.
StatMed2002;21:1539-1558.
19.
FurukawaTA,GuyattGH,GriffithLE.
Canweindividualizethe'numberneededtotreat'Anempiricalstudyofsummaryeffectmeasuresinmeta-analyses.
IntJEpidemiol2002;31:72-76.
20.
EggerM,SmithGD,PhillipsAN.
Meta-analysis:principlesandprocedures.
BMJ1997;315:1533-1537.
21.
JepsenP,VilstrupH,MellemkjaerL,etal.
Prognosisofpatientswithadiagnosisoffattyliver—aregistry-basedcohortstudy.
Hepatogastroenterology2003;50:2101-2104.
22.
TargherG,BertoliniL,RodellaS,etal.
Nonalcoholicfattyliverdiseaseisindependentlyassociatedwithanincreasedincidenceofcardiovasculareventsintype2diabeticpatients.
DiabetesCare2007;30:2119-2121.
23.
HamaguchiM,KojimaT,TakedaN,etal.
Nonalcoholicfattyliverdiseaseisanovelpredictorofcardiovasculardisease.
WorldJGastroenterol2007;13:1579-1584.
24.
HaringR,WallaschofskiH,NauckM,etal.
Ultrasonographichepaticsteatosisincreasespredictionofmortalityriskfromelevatedserumgamma-glutamyltranspeptidaselevels.
Hepatology2009;50:1403-1411.
25.
AdamsLA,HarmsenS,StSauverJL,etal.
Nonalcoholicfattyliverdiseaseincreasesriskofdeathamongpatientswithdiabetes:acommunity-basedcohortstudy.
AmJGastroenterol2010;105:1567-1573.
26.
LazoM,HernaezR,BonekampS,etal.
Non-alcoholicfattyliverdiseaseandmortalityamongUSadults:prospectivecohortstudy.
BMJ2011;343:d6891.
27.
ZhouYJ,LiYY,NieYQ,etal.
NaturalcourseofnonalcoholicfattyliverdiseaseinsouthernChina:aprospectivecohortstudy.
JDigDis2012;13:153-160.
1728.
ElAzeemHA,Khalekel-SA,El-AkabawyH,etal.
Associationbetweennonalcoholicfattyliverdiseaseandtheincidenceofcardiovascularandrenalevents.
JSaudiHeartAssoc2013;25:239-246.
29.
PistoP,SantaniemiM,BloiguR,UkkolaO,KesniemiYA.
Fattyliverpredictstheriskforcardiovasculareventsinmiddle-agedpopulation:apopulation-basedcohortstudy.
BMJOpen2014;4:e004973.
30.
PickhardtPJ,HahnL,MuozdelRioA,etal.
Naturalhistoryofhepaticsteatosis:observedoutcomesforsubsequentliverandcardiovascularcomplications.
AJRAmJRoentgenol2014;202:752-758.
31.
WongVW,WongGL,YeungJC,etal.
Long-termclinicaloutcomesafterfattyliverscreeninginpatientsundergoingcoronaryangiogram:aprospectivecohortstudy.
Hepatology2016;63:754-763.
32.
MoonSH,NohTS,ChoYS,etal.
Associationbetweennonalcoholicfattyliverdiseaseandcarotidarteryinflammationevaluatedby18F-fluoro-deoxyglucosepositronemissiontomography.
Angiology2015;66:472-480.
33.
EkstedtM,HagstrmH,NasrP,etal.
Fibrosisstageisthestrongestpredictorfordisease-specificmortalityinNAFLDafterupto33yearsoffollow-up.
Hepatology2015;61:1547-1554.
34.
EmreA,TerziS,CelikerE,etal.
ImpactofnonalcoholicfattyliverdiseaseonmyocardialperfusioninnondiabeticpatientsundergoingprimarypercutaneouscoronaryinterventionforST-segmentelevationmyocardialinfarction.
AmJCardiol2015;116:1810-1814.
35.
ZebI,LiD,BudoffMJ,KatzR,etal.
Nonalcoholicfattyliverdiseaseandincidentcardiacevents:TheMulti-EthnicStudyofAtherosclerosis.
JAmCollCardiol2016;67:1965-1966.
36.
FracanzaniAL,TiraboschiS,PisanoG,etal.
Progressionofcarotidvasculardamageandcardiovasculareventsinnon-alcoholicfattyliverdiseasepatientscomparedtothegeneralpopulationduring10yearsoffollow-up.
Atherosclerosis2016;246:208-213.
37.
KimD,KimWR,KimHJ,etal.
Associationbetweennon-invasivefibrosismarkersandmortalityamongadultswithnonalcoholicfattyliverdiseaseintheUnitedStates.
Hepatology2013;57:1357-1365.
38.
GhouriN,PreissD,SattarN.
Liverenzymes,nonalcoholicfattyliverdisease,andincidentcardiovasculardisease:anarrativereviewandclinicalperspectiveofprospectivedata.
Hepatology2010;52:1156-1161.
39.
MussoG,GambinoR,CassaderM,etal.
Meta-analysis:naturalhistoryofnon-alcoholicfattyliverdisease(NAFLD)anddiagnosticaccuracyofnon-invasivetestsforliverdiseaseseverity.
AnnMed2011;43:617-649.
40.
LuH,LiuH,ZouL,LuoS,SunL.
Independentassociationbetweennonalcoholicfattyliverdiseaseandcardiovasculardisease:asystematicreviewandmeta-analysis.
IntJEndocrinol2013;2013:124958.
41.
YounossiZM,KoenigAB,AbdelatifD,etal.
Globalepidemiologyofnon-alcoholicfattyliverdisease-meta-analyticassessmentofprevalence,incidenceandoutcomes.
Hepatology2015Dec28.
doi:10.
1002/hep.
28431[Epubaheadofprint].
1842.
StepanovaM,YounossiZM.
IndependentassociationbetweennonalcoholicfattyliverdiseaseandcardiovasculardiseaseintheUSpopulation.
ClinGastroenterolHepatol2012;10:646-650.
43.
YounossiZM,OtgonsurenM,VenkatesanC,etal.
Inpatientswithnon-alcoholicfattyliverdisease,metabolicallyabnormalindividualsareatahigherriskformortalitywhilemetabolicallynormalindividualsarenot.
Metabolism2013;62:352-360.
44.
WongVW,WongGL,YipGW,etal.
Coronaryarterydiseaseandcardiovascularoutcomesinpatientswithnon-alcoholicfattyliverdisease.
Gut2011;60:1721-1727.
45.
EkstedtM,FranzénLE,MathiesenUL,etal.
Long-termfollow-upofpatientswithNAFLDandelevatedliverenzymes.
Hepatology2006;44:865-873.
46.
SoderbergC,StlP,AsklingJ,etal.
Decreasedsurvivalofsubjectswithelevatedliverfunctiontestsduringa28-yearfollow-up.
Hepatology2010;51:595-602.
47.
MatteoniCA,YounossiZM,GramlichT,etal.
Nonalcoholicfattyliverdisease:aspectrumofclinicalandpathologicalseverity.
Gastroenterology1999;116:1413-1419.
48.
RafiqN,BaiC,FangY,etal.
Long-termfollow-upofpatientswithnonalcoholicfattyliver.
ClinGastroenterolHepatol2009;7:234-738.
49.
AdamsLA,LympJF,StSauverJ,etal.
Thenaturalhistoryofnonalcoholicfattyliverdisease:apopulation-basedcohortstudy.
Gastroenterology2005;129:113-121.
50.
AnguloP,KleinerDE,Dam-LarsenS,etal.
Liverfibrosis,butnootherhistologicfeatures,isassociatedwithlong-termoutcomesofpatientswithnonalcoholicfattyliverdisease.
Gastroenterology2015;149:389-397.
51.
ByrneCD,TargherG.
Ectopicfat,insulinresistance,andnonalcoholicfattyliverdisease:implicationsforcardiovasculardisease.
ArteriosclerThrombVascBiol2014;34:1155-1161.
52.
VanWagnerLB,WilcoxJE,ColangeloLA,etal.
Associationofnonalcoholicfattyliverdiseasewithsubclinicalmyocardialremodelinganddysfunction:apopulation-basedstudy.
Hepatology2015;62:773-783.
53.
TargherG,ValbusaF,BonapaceS,etal.
Nonalcoholicfattyliverdiseaseisassociatedwithanincreasedincidenceofatrialfibrillationinpatientswithtype2diabetes.
PLoSOne2013;8:e57183.
54.
MantovaniA,PernigoM,BergaminiC,etal.
Heartvalvecalcificationinpatientswithtype2diabetesandnonalcoholicfattyliverdisease.
Metabolism2015;64:879-887.
55.
LonardoA,BallestriS,TargherG,etal.
Diagnosisandmanagementofcardiovascularriskinnonalcoholicfattyliverdisease.
ExpertRevGastroenterolHepatol2015;9:629-650.
56.
SinghS,AllenAM,WangZ,etal.
Fibrosisprogressioninnonalcoholicfattylivervsnonalcoholicsteatohepatitis:asystematicreviewandmeta-analysisofpaired-biopsystudies.
ClinGastroenterolHepatol2015;13:643-654.
19FIGURELEGENDSFigure1.
Includedandexcludedstudies:theMOOSEflowdiagram.
Figure2.
Random-effectsmeta-analysisontheriskofincidentCVDevents(fatal,non-fatalorboth)associatedwithNAFLD.
ForestplotofcomparisonofpatientswithNAFLDversusthosewithoutNAFLD.
Figure3.
Random-effectsmeta-analysisontheriskoffatalandnon-fatalCVDeventsassociatedwithmore'severe'NAFLD(definedeitherbypresenceoffattyliveronimagingpluseitherelevatedserumgamma-glutamyltransferaseconcentrationsorhighNAFLDfibrosisscoreorhighFDGuptakeonpositronemissiontomography,orbyincreasingfibrosisstageonliverbiopsy).
20LAYSUMMARYThedataonwhetherNAFLDbyitselfisassociatedwithincreasedcardiovasculareventsanddeathremainsanissueofdebate.
Thefindingsofthisupdatedandlargemeta-analysisofobservationalstudiesindicatethatNAFLDissignificantlyassociatedwithanincreasedriskoffatalandnon-fatalcardiovascularevents.
However,theobservationaldesignofthestudiesincludeddoesnotallowustoprovethatNAFLDcausescardiovasculardisease.
ClinicianswhomanagepatientswithNAFLDshouldnotfocusonlyonliverdiseasebutshouldalsoconsidertheincreasedriskofcardiovasculardiseaseandundertakeearly,aggressiveriskfactormodification.
Recordsscreened(n=4569)Recordsexcludedbyspecificcriteria(n=4536):editorials,commentaries,reviews,casereports,pediatricpopulation,inadequatedefinitionofcases,inadeguateoutcomemeasuresFull-textarticlesassessedforeligibility(n=33)Full-textarticlesexcluded(n=17):studieswithoverlapofpopulationwithincludedstudies,lackofcontrolgroup,inadequatedataonoutcomesofinterestStudiesincludedinsystematicreview(n=16)Studiesincludedinmeta-analysis(n=16)Studiesincludedinmeta-analysis(n=16)-Community-basedcohort,n=6-Outpatientcohort,n=5-Hospital-based-cohort,n=2-Selectedcohort,n=3Figure1StudyorSubgroupFatalCVDevents(only)Adams2010Ekstedt2015Haring2009menHaring2009womenJepsen2003Lazo2011Zhou2012Subtotal(95%CI)Heterogeneity:Tau=0.
25;Chi=61.
73,df=6(P200mg/dl;mean52years,94%men4.
2(mean)Ultrasonography&positronemissiontomographywithF-18fluoro-2-deoxyglucose(FDG)(N=394withNAFLD)Non-fatalCVDevents(myocardialinfarction,angina,coronaryrevascularization)N=9CVDeventsAge,sex,andserumtriglyceridesNAFLDwithhighhepaticFDGuptakewasindependentlyassociatedwithnon-fatalCVDevents(adjustedHR4.
23;95%CI1.
05-17.
04)+23*Degreeofadjustment:0unadjusted;+adjustedforageand/orsex;++,furtheradjustmentfortraditionalCVDriskfactors;+++,furtheradjustmentfornon-traditionalCVDriskfactorsand/ormetabolicsyndrome.
Ekstedtetal.
201533Retrospectiveoutpatientcohort,n=229SwedenpatientswithNAFLDandelevatedserumliverenzymelevels(49%NASH);mean49years,66%men26.
4(mean)Histology(N=229withNAFLD)All-causeandCVDmortalityN=96totaldeaths(41CVDdeaths)ThereferencepopulationcomprisedallthoseofthesameageandsexlivinginthesamecountyaseachpatientwithNAFLDatbaselineIncreasedratesofall-cause,liver-relatedandCVDmortality(adjustedHR1.
55,95%CI1.
11-2.
15)inpatientswithNAFLDcomparedwiththegeneralcontrolpopulation.
Fibrosisstageonhistologysignificantlypredictedtheriskofall-cause,liver-relatedandCVDmortality(adjustedHR4.
36,95%CI2.
29-8.
29)+Emreetal.
201534Retrospectivehospital-basedcohort,n=186consecutiveTurkishnon-diabeticpatientsundergoingprimarypercutaneouscoronaryinterventionsforST-segmentelevationmyocardialinfarctionafterexcludingthosewithknownliverdiseasesorestablisheddiabetes;mean58years,76%menIn-hospitalcardiaceventsUltrasonography(N=75withNAFLD)In-hospitalCVDevents(acutemyocardialinfarction,acuteheartfailureordeath)N=32CVDevents(8CVDdeaths)Age,bodymassindex,totalcholesterol,HDLcholesterol,triglycerides,presenceofanteriorwallinfarction,andmulti-vesselcoronarydiseaseModerate-severeNAFLDwasindependentlyassociatedwithincreasedin-hospitalCVDevents(adjustedOR2.
45,95%CI1.
07-4.
87).
Moderate-severeNAFLDwasnotindependentlyassociatedwithCVDdeath(adjustedOR2.
24,95%CI0.
97-5.
16)+++Zebetal.
201635Prospectivecohortstudy,n=4119UnitedStatesparticipantsaged45to84yearswhowerefreeofCVDandknownliverdiseasesatbaseline(TheMulti-EthnicStudyofAtherosclerosis);mean62years,45%men7.
6years(median)Non-enhancedcomputedtomography(N=728withNAFLD)All-causemortalityandnon-fatalCVDevents(myocardialinfarction,resuscitatedcardiacarrest,anginaorcoronaryrevascularizationprocedures)N=253deathsand209nonfatalCVDeventsAge,sex,ethnicity,diabetes,hypertension,bodymassindex,lipids,smoking,familyhistoryofCHD,statinuse,Creactiveprotein,andcoronaryarterycalciumscoreoncardiacCTscansNAFLDwasindependentlyassociatedwithacompositeendpointinclusiveofall-causedeathandnonfatalCVDevents(adjustedHR1.
42,95%CI1.
00-2.
03)+++Fracanzanietal.
201636Prospectivecohortstudy,n=125ItalianpatientswithNAFLDand250age-andsex-matchedcontrolindividualswithoutknownliverdiseases;mean52years,87%men10yearsUltrasonographyorhistology(N=125withNAFLD)Non-fatalCVDevents(acutecoronarysyndrome,coronaryrevascularizationprocedures,ischemicstrokeortransitoryischemicattacks)N=35CVDeventsSex,smokinghistory,diabetes,hypertension,andcarotidatheroscleroticplaquesonultrasoundNAFLDwasindependentlyassociatedwithnon-fatalCVDevents(adjustedHR1.
99,95%CI1.
01-3.
91)+++Kimetal.
201337Population-basedcohort,n=11154UnitedStatesadults(NHANES1988-94);mean43years,48%men14.
5(median)Ultrasonography(N=4083withNAFLDthosewithmildsteatosiswereconsideredashavingNAFLD)All-causeandCVDmortalityN=1795totaldeaths(673CVDdeaths)Age,sex,race,education,income,diabetesstatus,hypertension,pre-existingCVD,lipid-loweringmedications,smoking,waistcircumference,alcoholintake,caffeineintake,totalcholesterol,HDLcholesterol,transferrinsaturation,C-reactiveproteinNAFLDwasnotassociatedwithincreasedall-causeandCVDmortality(adjustedHR0.
7595%CI0.
56-1.
01)inthewholecohort.
However,NAFLDwithadvancedhepaticfibrosis(definedbytheNAFLDfibrosisscore)wasindependentlyassociatedwithincreasedall-causeandCVDmortality(adjustedHR3.
46,95%CI1.
91-6.
25)+++24Table2.
Riskoffataland/ornon-fatalCVDeventsassociatedwithNAFLD(asdiagnosedeitherbyimagingorbyhistology):sensitivityanalyses.
AnalysisNumberofcomparisonsOverallORsorHRs(with95%confidenceintervals)PvaluesI2valuesIncludingonlyhigh-qualitystudiesattheNewcastle-Ottawascale111.
54(1.
13-2.
11)<0.
00186%Includingonlystudieswithfulladjustmentforcovariates61.
69(1.
11-2.
58)<0.
00178%Excludingstudieswiththegeneralpopulationasthereferencegroup151.
63(1.
19-2.
22)<0.
00186%Excludingstudieswithcohortsofparticipantswithdiabetes,hypertensionoracutemyocardialinfarction131.
57(1.
15-2.
14)<0.
00189%25Graphicalabstract