Apparent temperature of fragments in the breakup of spectator residues

Type2diabetesmellitusandriskoforalcancerandprecancerouslesions:Ameta-analysisofobservational

studies

YuhuaGong,BenjuanWei,LiYu,WeijuanPan

DepartmentofStomatology,RenJiHospital,SchoolofMedicine,ShanghaiJiaoTongUniversity,Shanghai,China

articleinfosummary

Objective:Associationsbetweentype2diabetesmellitus(type2DM)andriskoforalcancerandprecan-cerouslesionshavebeenreportedwithcontroversial ndings.Weperformedameta-analysistoexploretheseassociations.

Methods:Weidenti edstudiesbyaliteraturesearchofMEDLINEandEMBASEthroughMay31,2014,andbysearchingthereferencelistsofpertinentarticles.Summaryrelativerisk(SRR)with95%con denceinterval(CI)wascalculatedwitharandom-effectsmodel.Between-studyheterogeneitywasassessedusingtheCochran’sQandI2statistics.

Results:Atotalof13studies(4case-controland9cohortstudies)ontheassociationbetweentype2DMandoralcancerwereincluded.Overallanalysisfoundthatcomparedwithnon-diabeticindividuals,indi-vidualswithtype2DMhadasigni cantlyelevatedincidenceoforalcancer(SRR=1.15,95%CI:1.02–1.29;Pheterogeneity=0.277,I2=15.4%;10studies).Subgroupanalysesfoundthatdurationoffollow-up(P11years)signi cantlyalteredthispositiveassociation.Type2DMwasassociatedwithincreasedoralcancermortality(SRR=1.41,95%CI:1.16–1.72;4studies).Meta-analysisofthefourcase-controlstudiesshowedapositiveassociationbetweentype2DMandriskoforalprecancerouslesions(SRR=1.85,95%CI:1.23–2.80;Pheterogeneity=0.038,I2=57.5%).Nosigni cantpublicbiaswasfoundacrossthesestud-ies.

Conclusions:These ndingsofthismeta-analysisindicatethatcomparedwithnon-diabeticindividuals,individualswithtype2DMhaveanelevatedriskoforalcancerandprecancerouslesionsdevelopment.

Ó2015ElsevierLtd.Allrightsreserved.

Articlehistory:

Received8October2014

Receivedinrevisedform22December2014Accepted2January2015

Availableonline31January2015Keywords:

Type2diabetesmellitusOralcancerMeta-analysis

Precancerouslesions

Introduction

Oralcancer(OC)representstheeighthmostfrequentcancerworldwide,whichincludescancersofthelip,gums,tongue,softhardpalate,etc.[1].Thegeographicareawiththehighestinci-denceandmortalityfromthisdeadlydiseaseisMelanesia,fol-lowedbysouthcentralAsia.InChina,oralcancerwasreported3.29per100,000asincidencerateand1.49per100,000asmortal-ityratein2008[2].Despitetheadvancesindiagnosisandtreat-ment,the5-yearsurvivalrateforpatientswithOCisstilllowinmanypartsoftheworld[3].Oralprecancerouslesionshavebeenwellrecognizedastheprecursorsoforalcancer[4],whichincludeoralleukoplakia,erythroplakia,andsubmucous brosis,etc.Recently,progresseshavebeenmadethroughepidemiologicalstudiesinvestigatingenvironmentalriskfactorsfororalcancerCorrespondingauthorat:DepartmentofStomatology,RenJiHospital,Schoolofmedicine,ShanghaiJiaoTongUniversity,160PujianRoad,PudongDistrict,200127,Shanghai,China.Tel./fax:+8602168383204.

E-mailaddress:weijpan@http://www.77cn.com.cn(W.Pan).

http://www.77cn.com.cn/10.1016/j.oraloncology.2015.01.0031368-8375/Ó2015ElsevierLtd.Allrightsreserved.

andprecancerouslesions,andthewelldocumentedfactorsincludecigarettesmoking,alcoholconsumption,betel-quidchewingandsometypesofviralinfections[5–8].

Ithasbeenshownthattype2diabetes(type2DM)areriskfac-torsforseveralmalignancies,includingcancersofthebreast[9],endometrium[10],pancreas[11,12],andliver[13].Thehypothe-sizedbiologicalmechanismsisrelatedtotheeffectofinsulinandinsulin-likegrowthfactors(IGFs)axis,whichwouldtriggerintra-cellularsignalingcascadeswithmitogenicandantiapoptoticeffects[14].Additionally,thein ammation-mediatedcarcinogene-sisisalsoawell-knownempiricalfact[14].

Isthereanycorrelationbetweentype2DMandcarcinogenesisoftheoralcavity?Inconsistentresultshavebeenreportedfortheseassociations[15–31].Campbellandhiscoauthorsprospectivelyenrolledacohortof1,053,831U.S.adults,andobservedatotalof1182deathsfromoralcancersafter28yearsoffollow-up[29].Dia-beticmenhadasigni cantriskofOCmortalitythandidnon-dia-beticmen(relativerisk[RR]=1.44,95%con denceinterval[CI]:1.07–1.94),whilediabeticwomenhadanon-signi cantlyincreasedriskthannon-diabeticwomen(RR=1.43,95%CI:

Y.Gongetal./OralOncology51(2015)332–340333

0.94–2.20).SimilarresultswerealsoobservedinthestudybyWideroffetal.[17].However,anon-signi cantlyincreasedriskassociationbetweendiabetesandOCwasobservedinmostoftheincludedstudies,andeven,asigni cantlyinverseassociationwasshowninthestudybyHjalgrimetal.[16].

Thepurposeofthepresentstudywastosummarizeallavailableevidencefromobservationalstudiestoestimatetheriskoforalcancerandprecancerouslesionsinpatientswithtype2DMfollow-ingthemeta-analysisofobservationalstudiesinepidemiology(MOOSE)guidelines[32].MaterialsandmethodsDatasourcesandsearches

Toidentifyrelevantstudies,twoinvestigators(G.Y.H.andW.B.J.)independentlyconductedasystematicliteraturesearchofMED-LINE(fromJanuary1,1966)andEMBASE(fromJanuary1,1974),throughMay31,2014.Inaddition,amanualreviewofreferencesfromprimaryorreviewarticleswasperformedtoidentifyanyaddi-tionalstudies.Therelevantstudiesweresearchedwiththefollow-ingtextwordand/orMedicalSubjectHeading(MeSH)terms:(1)‘‘diabetes’’;(2)‘‘oralcancer’’OR‘‘oralcarcinoma’’OR‘‘mouthneo-plasm’’OR‘‘oralleukoplakia’’OR‘‘oralerythroplakia’’OR‘‘oralsub-mucous brosis’’;and(3)‘‘risk’’OR‘‘incidence’’OR‘‘prevalence’’OR‘‘mortality’’.Nolanguagerestrictionswereimposed.Studyselection

Studieswereincludedinthismeta-analysisif:(1)theyhadori-ginaldatafromcase-controlorcohortstudies;(2)theexposureofinterestwastype2DM(ormainlytype2DM);(3)theprimaryout-comewasclearlyde nedasoralcancersorprecancerouslesions;and(4)studiesshouldreporteitheradjustedoddsratios,rateratio,hazardratio(HR),orstandardizedincidence/mortalityratios(SIR/SMR)withtheir95%CIs(ordatatocalculatethem).Twoauthors(G.Y.H.andW.B.J.)independentlyevaluatedallofthestudiesretrievedfromthedatabases;incaseofdisagreementoruncer-tainty,athirdreviewer(P.W.J.)wasconsulted.Weexcluded2arti-clesthatreportedtype1DMandOCrisk[33,34].Ifastudyappearedinmorethanonearticle,datafromthemostrecentpub-licationwereusedforthestatisticalanalysis[35,36].Dataextraction

Thefollowingdatawereextractedindependentlybytwoinves-tigatorsusingastandardizeddatacollectionformforeachstudy:thedesigntype(case-controlorcohortstudy),the rstauthor’slastname,yearofpublication,countryoforigin,samplesizeandnum-berofcases,ageandgenderofthesubjects,durationoffollow-upincohortstudies,assessmentofexposureandoutcome,covariatesadjustedorbymatching,andtheeffectestimateswith95%CIs.Fromeachstudy,weextractedtheriskestimatesthatre ectedthegreatestdegreeofadjustmentsforpotentialconfounders.Sex-speci criskestimateswereextractedwheneveravailable.Ifstudiesreportedbothincidenceandmortalityrate,weextractedtheboth[23].Onestudyreportedriskestimationsforbothyoung-(age<30years)andold-onset(ageP30years)DM,andweextractedonlytheriskestimationfortheold-onsetDM,becausemostindividualswithyoung-onsetDMaretype1DM[16].Qualityassessmentforindividualstudies

Toassessthestudyquality,twoofus(G.Y.H.andW.B.J.)adoptedtheNewcastle-OttawaqualityassessmentScale(NOS)

[37].TheNOSuses3parametersofqualityforcase-controlorcohortstudies:selection,comparability,andexposure/outcomeassessment.TheNOSscaleassignsamaximumof4starsforselec-tion,2starsforcomparability,and3starsfortheexposure/out-come.Thetotalscorewas9stars,andastudywith7ormorestarswasde nedasahigh-qualitystudy.StudieswereconsideredaslowqualityiftheycouldnotbeevaluatedbytheNOSduetoinsuf cientinformation.Statisticalanalysis

Wedividedepidemiologicstudiesintothreegeneraltypesaccordingtothemeasurementofriskestimations:case–controlstudies(oddsratio),cohortstudiesusingnon-diabeticpopulationcomparisons(rateratioandHR)andusingexternalgeneralpopu-lationcomparisons(SIR/SMR).Becausetheabsoluteriskoforalcancerislow,alltheabovemeasuresyieldsimilarestimatesofRR[38].AllstatisticalanalyseswereperformedusingSTATA,ver-sion11.0(STATA,CollegeStation,TX,USA).SummaryRRs(SRRs)withtheircorresponding95%CIswerederivedwiththemethodofDerSimonianandLairdusingtheassumptionsofarandomeffectsmodel,whichincorporatesbetween-studyvariability[39].Atwo-tailedP<0.05wasconsideredstatisticallysigni cant.

Inassessingheterogeneityamongstudies,weusedtheCochranQandI2statistics.TheI2statisticistheproportionoftotalvariationcontributedbybetween-studyvariation,whichhasbeensuggestedthatI2valuesof25%,50%,and75%areassignedtolow,moderate,andhighheterogeneity,respectively[40].Toexplorethesourcesofheterogeneity,subgroupandmeta-regressionanalyseswereper-formedaccordingtosex,geographiclocations,publicationyear,methodsofDMascertainment,studyqualityscore,durationoffol-low-up,thenumberofcases,de nitionofoutcome(incidencevs.mortality)andadjustmentsforconfoundingfactorsincludingsmoking,bodymassindex(BMI),andalcoholuse.Sensitivityanal-yseswereperformedbyexcludingonestudyinthemeta-analysisandcalculatingapooledestimatefortheremainderofthestudiestoevaluatewhethertheresultsweresigni cantlyaffectedbyasin-glestudy.PublicationbiaswasassessedbyusingfunnelplotsandthefurtherBegg’sadjustedrankcorrelationandEgger’regressionasymmetrytests[41,42].Results

Searchresults,characteristicsandqualityassessment

Thesearchstrategygenerated831citationsofwhich29wereconsideredofpotentialvalueandthefulltextwasretrievedfordetailedevaluation(Fig.1).Eighteenofthese29articlesweresub-sequentlyexcluded:11studiesdidnotevaluatethisassociation,3studiesreportedthesamepopulation,3studiesdidnotreportRRand/or95%CI,and2studiesreportedyoung-onsetDM.Additional6articleswereincludedfromreferencereview.Thus,atotalof17articlesprovideddatatoinvestigatetheassociationbetweentype2DMandoralcancer(n=13)orprecancerouslesions(n=4;Tables1and2).

Fourstudiesreportedtheassociationbetweentype2DMandriskorprecancerouslesions,allofwhichhadacase-control/cross-sectionaldesignandwerepublishedbetween2004and2010[19–21,24].Thefourstudiesreportedatotalof1407caseswithoralprecancerouslesions(1137caseswithoralleukoplakia,100caseswithoralerythroplakiaand170caseswithoralsubmu-cous brosis).DMstatuswasascertainedbyself-report[19,21]andmedicalexamination[20,24]intwostudies,respectively.Diagno-sisofprecancerouslesionswasbasedonhistologicalormedicalexamination.Adjustmentsweremadeforpotentialconfoundersof1ormorefactorsinallstudies.

Table1

Descriptivecharacteristicofcase-controlstudiesfordiabetesmellitusandoralcancerandprecancerouslesions.Author/year/countryPrecancerouslesionsDietrich/2004/USA(19)

NumberofcasesOL,65

Age(year),sex

DiabetesassessmentSR

Outcome

ascertainmentMedical

examinationPathologicalMedical

examination

Adjustedrisk

estimations(95%CI)OR:3.03(1.28–7.21)

Adjustments/matchments

Age:57.5Male:69.2%Age:55.2Male:NAAge:NAMale:NAAge:55.2Male:55.3%

Age,sex,smoking,alcohol,education,BMI

Ujpal/2004/Hungary(20)Dikshit/2006/India(21)

OL,22OL,927OE,100OS,170OL,123

MRSR

OR:2.85(1.18–6.88)OR:1.4(1.0–1.8)2.7(1.4–5.1)0.4(0.1–1.3)

OR:1.93(1.09–3.41)

Age,sex

Age,education,smoking,durationofchewing,alcohol,BMI

Meisel/2010/PolandandGerman(24)Oralcancer

LaVecchia/1994/Italy(15)Kuriki/2007/Japan(22)

MR

Medical

examination

Age,sex,smoking,alcohol,education

OPC,181OC,77Age:53

Male:61.3%Age:>57Male:30%

SRSR

PathologicalCancerregistry

OR:0.5(0.2–1.1)M,FOR:

2.03(0.87–4.78)M1.33(0.17–10.5)FOR:

0.9(0.4–2.1)M1.0(0.3–3.0)F

OR:1.58(1.15–2.18)

Age,sex

Age,BMI,smoking,physicalexercise,bowelmovement,FHC,dietaryfactors

Li/2011/USA(25)OC,80Age,46.8Male:38.1%

SRSelf-report

Age,race/ethnicity,healthinsurance,smoking,drinking,BMI,physicalactivity

Bosetti/2012/ItalyandSwitzerland(28)

OPC,1468Age,58Male:81%

SRSelf-report

Age,sex,studycenter,yearofinterview,BMI,education,alcohol,smoking

Abbreviations:BMI,bodymassindex;RR,relativerisk;NA,notapplicable;M,male;F,female;OC:oralcancer.OPC:oral-pharyngealcancer;OL:oralleukoplakia;OS:oralsubmucous brosis;OE:oralerythroplakia;MR,medicalrecord;SR,self-report;FHC,familyhistoryofcancer;OR,oddsratio.

Thirteenstudies(4case-controland9cohortstudies)reportedtheassociationbetweentype2DMandoralcancer,withatotalofmorethan4.8millionparticipantsand6,465caseswithoralcancer.ThemethodsofDMascertainmentwerebasedonbloodglucoselevelstestormedicalrecordsin7stud-iesandonself-reportintheother6studies.Theascertainmentofoutcomewasbasedondiseaseordeathregistryinallstudies.

ThequalityscoresofeachstudyaresummarizedinSupplemen-taryTable1.Thequalityscoresrangedfrom5to9,withthemed-ianscore7.Themajorityofincludedstudies(13/17)wereofhighquality(NOSscoreP

7).

Table 2 Descriptive characteristics of cohort studies of diabetes and oral cancer. Author/year/country Wideroff/1997/Denmark (17) Participants, age, sex N= 109,581 Age:64(m); 69(f) Male:49% N= 1499 Age:>30 Male: NA N= 7148 Age:67 Male:50% N= 549,944 Age: 44.8 Male: 50% N= 715,061 Age: 55 Male: 52% N= 2.5 million

Age:>30 Male: 54.3% N= 154,975 Age: 63.1 Male:47.3% Tseng/2013/Taiwan(31) N= 998,540 Age: NA Male: 50% N= 494,867 Age, 50–71 MR 28 Diabetes assessment MR Follow up, ys 17 Outcome ascertainment ICD-7 Number of cases OPC, 172 Adjusted risk estimations (95% CI) SIR: 1.2(1.0–1.4) M 1.2(0.9–1.6) F Adjustments/matchments Age, sex, calendar year

Hjalgrim/1997/Denmark (16)

MR MR

8.6 10

Verlato/2003/Italy (18)

Cancer registry Mortality Records

OPC, 1 OPC, 14

SIR:0.5(0.01–0.8) SMR: 1.12 (0.61–1.88) M,F 1.22 (0.67–2.05)M RR: 1.89 (0.70–5.10) M 1.89 (0.28–13.0) F HR: 1.38 (0.90–2.12)

Age, sex Age, smoking, BMI

Y. Gong et al./ Oral Oncology 51 (2015) 332–340

Stocks/2009/Three European countries(23)

MR

10.4

Cancer/death registry

OPC,581

Age, sex, BMI, and smoking status

Seshasai/2011/19 countries (26)

SR/MR

13.6

Death registry

OC, 475

Age, sex Smoking status, and BMI Age, sex, time period in single calendar years and district of residence

Wotton/2011/England(27)

MR

1963–1998 1999–2008

Cancer registry

OPC, 571

RR: ORLS1: 1.04 (0.63–1.63) ORLS2: 0.95 (0.43–1.84) RR: 1.44 (1.07–1.94) M 1.43 (0.94–2.20) F

Campbell/2012/USA(29)

SR

12.1

Medical Record

OPC, 743(M); 439(F)

Age, education, BMI, physical activity, NSAIDs use Alcohol use, FHC, endoscopy history Age, obesity, hypertension, smoking, alcohol, stroke, ischemic heart disease

ICD-9

OC, 198

RR: 1.195 (0.892–1.601)M 1.223 (0.789–1.895)F HR: 0.90 (0.73–1.10) All

Lai/2013/USA(30)

SR

11

Linkage

OC, 1465

Age; sex; BMI; race; education; marital status; FHC; health status Intake of red meat, white meat, fruits, vegetables, alcohol, coffee; physical activity; cigarette smoking; multivitamin use

Male: 59.7%

0.84 (0.66–1.06) M 1.15 (0.76–1.73) F

Abbreviations: BMI, body mass index; NA, not applicable; M, male; F, female; OC: oral cancer. OPC: oral-pharyngeal cancer; MR, medical record; SR, self report; FHC, family history of cancer; NASIDs, non-steroidal antiin ammatory drugs; ICD, International Classi cation of Diseases; RR, rate ratio; HR, hazard ratio; SIR/SMR, standardized incidence/mortality ratios.

335

336Y.Gongetal./OralOncology51(2015)332–340

Type2DMandriskofprecancerouslesions

Fig.2showedtheresultsofmeta-analysisofthefourstudiesthatreportedtheriskoforalprecancerouslesionsinpatientswithtype2DM[19–21,24].Onestudypresentedriskestimationsforthreetypesoforalprecancerouslesions(oralleukoplakia,oralerythropla-kia,andoralsubmucous brosis),respectively,weextractedthethreeRRstocalculatethepooledriskestimations[21].Theran-dom-effectsmodelfoundapositiveassociationbetweentype2DMandriskoforalprecancerouslesionswithevidenceofheteroge-neityamongstudies(SRR=1.85,95%CI:1.23–2.80;Pheterogeneity=0.038,I2=57.5%).Whenrestrictingthemeta-analysisforonlyoralleukoplakia,asimilarSRRwasobservedwithno/lowheterogeneity(SRR=1.88,95%CI:1.29–2.72;Pheterogeneity=0.182,I2=38.3%).Type2DMandoralcancerincidence

Inanalysisofthe10studiesontheassociationbetweentype2DMandincidenceoforalcancer,weobtainedtheSRRof1.15(95%CI:1.02–1.29)inarandom-effectsmodelwhencomparingdiabeticindividualstonon-diabeticindividuals,withnoevidentheterogene-ityamongstudies(Q=17.73,Pheterogeneity=0.277,I2=15.4%;Fig.3).Type2DMandoralcancermortality

Inanalysisofthe4studiesontheassociationbetweentype2DMandoralcancermortality,weobtainedtheSRRof1.41(95%CI:1.16–1.72)inarandom-effectsmodelwhencomparingdiabeticindividualstonon-diabeticindividuals,withnoevidentheteroge-neityamongstudies(Q=3.52,Pheterogeneity=0.475,I2=0;Fig.4).Publicationbias

Theshapeofthefunnelplotsfortheseincludedstudiesontheassociationoftype2DMandincidenceoforalcancerseemedsym-metrical,andtheBegg’sadjustedrankcorrelationtest(P=0.392)andtheEgger’sregressiontest(P=0.176)indicatedtherewerenoevidentpublicationbias.Wedidnotevaluatethisbiasfor

precancerouslesionsandOCmortalityduetothesmallnumberofstudiesincluded.Subgroupanalyses

Table3showedthesubgroupmeta-analysesaccordingtostudydesign,sex,geographicalarea,studyqualityscore,DMassessment,durationoffollow-upandconfoundersadjustedbyBMI,smokingandalcoholuse.Whenstratifyingaccordingtostudydesign,wefoundasigni cantlyincreasedriskofOCindiabeticindividualscomparedwithnon-diabeticindividualsinthecohortstudies(SRRs=1.11,95%CIs:1.00–1.23,Pheterogeneity=0.561,I2=0),butnotinthecase-control(SRRs=1.18,95%CIs:0.77–1.80,Pheterogeneity=0.145,I2=39.1%).Femaleswithtype2DMhadasigni cantlyelevatedriskofOCdevelopment(SRRs=1.21,95%CIs:1.00–1.46),andmaleswithtype2DMhadanincreased,butnotsigni cantly,riskofOCdevelopment(SRRs=1.07,95%CIs:0.83–1.37).Whensubgroupanalysiswasstrati edbydurationoffollow-up,adiffer-entassociationbetweenDMandtheriskofOCwasfound(SRRs=0.93,95%CIs:0.77–1.13forduration<11years,andSRRs=1.19,95%CIs:1.06–1.34fordurationP11years).Asigni -cantriskassociationbetweentype2DMandOCdevelopmentwasobservedinstudieswithhighquality(SRRs=1.22,95%CIs:1.08–1.38),butnotinstudieswithlowquality(SRRs=1.03,95%CIs:0.85–1.25).Whetherornotcontrolledforsmoking,BMI,oralcoholusedidnotalteredtheassociationbetweenahistoryoftype2DMandriskofOC.

Sensitivityandmeta-regressionanalysis

Wethenconductedameta-regressionanalysistoinvestigatetheimpactoftheabovestudycharacteristicsontheRRs.Durationoffollow-up(Pdifference=0.068)signi cantlyalteredthesummaryriskassociationbetweentype2DMandriskofOC(Table3).WealsoconductedasensitivityanalysisbyomittingonestudyatatimeandcalculatingtheSRRsfortheremainderofstudies,andfoundthattherewerenochangesinthedirectionofeffectwhenanyonestudywasexcluded(datanotshown).

Y.Gongetal./OralOncology51(2015)332–340337

Discussion

Thecurrentmeta-analysis,whichisthe rstandmostcompre-hensivetodate,indicatethatcomparedwithnon-diabeticindivid-uals,individualswithtype2DMhaveanapproximately15%increasedriskoforalcancerdevelopment,andtheincreasedriskswerestatisticallysigni cantafteradjustingfortobaccoandalcoholuse,twomajorriskfactorsforthiscancer.Type2DMmayalso

338Y.Gongetal./OralOncology51(2015)332–340

Table3

Subgroupanalysisfortheassociationoftype2diabetesandoralcancerincidence.Characteristic

Subgroup

Refs.(n)

SRR(95%CI)

TestsforheterogeneityQ

Design

PublicationyearLocationsSex

Durationoffollow-up,yStudyqualityscoreDiabetesascertainmentsNo.ofcasesAdjustmentsAlcoholuseBMI/smoking

Case-controlCohort

Before2000Since2000Asia

Non-AsiaMenWomen<11P11

High(NOSscore>6)Low(NOSscore66)MRSR<300P300YesNoYesNo

[4][6][3][7][2][8][7][6][4][6][7][3][4][6][6][4][4][6][6][4]

1.18(0.77–1.80)1.11(1.00–1.23)1.12(0.88–1.42)1.15(0.99–1.35)1.26(1.00–1.59)1.11(0.96–1.30)1.07(0.83–1.37)1.21(1.00–1.46)0.93(0.77–1.13)1.19(1.06–1.34)1.22(1.08–1.38)1.03(0.85–1.25)1.19(1.04–1.36)1.13(0.90–1.42)1.18(1.05–1.33)1.16(0.87–1.55)1.15(0.92–1.43)1.18(1.03–1.34)1.20(1.00–1.46)1.15(1.00–1.31)

8.27.734.5113.021.3315.5914.550.382.8910.6110.613.652.3514.716.6710.167.828.3512.714.99

Pheterogeneity0.1450.5610.2110.2920.7210.1570.0250.9600.5760.9870.4760.3020.8850.0650.7560.0710.0940.5530.1760.417

I2(%)39.1033.615.5029.458.8000017.8045.6050.846.9029.60

0.3530.7440.4250.5640.0680.1060.7820.831Pdifference

0.9080.498

Abbreviations:SRR,summaryrelativerisk;NOS,Newcastle-Ottawaqualityassessmentscale;BMI,bodymassindex;MR,medicalrecord;SR,self-report.

increaseOCmortality.Inaddition,onlylimitedevidencefromfourcase-controlorcrosssectionalstudiesindicatesasigni cantlyincreasedriskassociationbetweendiabetesandoralprecancerouslesions.

Comparedwiththeriskassociationbetweentype2DMandoralcancer,wefoundasigni cantlystrongerriskfororalprecancerouslesions(SRR:1.85vs.1.15).Reasonsforthedisparityintheriskassociationbetweentype2DMandoralcancerandprecancerouslesionsarenotclear.However,itmaybeinferredthattype2DMwouldhavethemaineffectonearlystageofcarcinogenesisoftheoralcavity.Also,theincludedstudiesforprecancerouslesionsaresmallsamplesize(whichcouldbesubjecttoatypeIerror)andallbasedonaretrospectivedesign(whichwouldin atethesum-maryRRs).Furthermore,bothpremalignantlesionsanddiabetesarechronic,andwecouldnotclearlyestablishthatthepremalig-nantlesionsdevelopedafterdiagnosisofdiabetesduetotheretro-spectivedesignnature.Indeed,Albrechtandothersfoundthatprecancerouslesionsoccurredmostofteninthesecondyearofestablisheddiabetes[43].Therefore,thisresultsneedtobeinter-pretedwithextremecaution,andmorestudieswithaprospectivedesignarewarranted,especiallyfortheassociationbetweentype2DMandoralprecancerouslesions.

Inthesubgroupanalysisaccordingtosex,wefoundthatdia-beticwomenhadanincreasedriskoforalcancer,whereasthisassociationwasattenuatedandnotstatisticallysigni cantindia-beticmen.Theseresultscouldnotbeexplainedbecausemenaremoreaffectedbyoralcancerthanwomen,withamaletofemaleratio1.45inJapan[44]tohighestof10.5inTaiwan[45].Itispos-siblethatthedifferentrateofreportedhistoryandthedifferentmanagementofdiabetesamongmalesandfemalesmaycontributetothegenderdifference[21]:menarelikelytoundergobettermetaboliccontrol,leadingtolesslevelsofinsulinandreducedoxi-dativedamagetoDNA[46,47].Similarly,thegenderdifferenceobservedfortheeffectoftype2DMwasalsofoundoncoloncan-cer,kidneycancer,andnon-melanomaskincancer[48].

Ouranalysisfoundthatdurationoffollow-uphadasigni cantin uenceonthemagnitudeanddirectionoftheobservedassocia-tionbetweentype2DMandOCrisk;cohortstudieswithlongdurationoffollow-up(P11years)indicatedtype2DMasariskfactorforOCdevelopment,whichwasnotmadeforstudieswithshortdurationoffollow-up(<11years).Theseresultsindicatethereisthetimecumulativeeffectoftype2DMonoralcancerriskandsupportacausalrelationshipbetweentype2DMandOC.Ourresultsofanindependenteffectoftype2DMonriskoforalcancerandprecancerouslesionsarebiologicallyplausible.Type2DMisusuallyassociatedwithhyperglycemiaandinsulinresis-tance,compensatoryhyperinsulinemia,andoxidativedamagetoDNA[46,47].Theelevatedcirculatinginsulinlevelshavebeenshowntoincreasethebioavailableinsulin-likegrowthfactor-1(IGF-1)concentrations,bothofwhichcouldstimulategrowththroughcellularproliferation,inhibitionofapoptosis[49,50],andincreasemitogenesisincelllinesfromvariousepithelialtumorsincludingoralcancer[51].Inaddition,insulinresistanceleadstoincreasedreleaseofmultiplepro-in ammatorycytokines,includ-ingtumornecrosisfactor-alphaandinterleukin-6[52,53].Thesepro-in ammatorycytokinesfavorthedevelopmentofin amma-tionandsubsequentmalignanttransformationintheoralcavity[53].

Ourmeta-analysishasseveralstrengths.First,thenumberofsamplesizeincludedwaslarge(morethan4.8millionsubjects,6465oralcancercasesand1407caseswithprecancerouslesions),suggestingthesolidevidenceinevaluatingtheepidemiologicasso-ciationbetweenDMandriskoforalcancerandprecancerouslesions.Second,therewasnosigni cantevidenceofheterogeneityandpublicationbias,andour ndingswerestableandrobustinsensitivityanalyses.Third,basedontheNOS,moststudies(10/13)wereofhighquality.

Thereareseveralpotentiallimitationstotheresultsofthismeta-analysis.First,asameta-analysisofobservationaldata,thepossibilityofrecallandselectionbiasescannotberuledout,espe-ciallyforcase-controlstudies.Thismeta-analysisofcase-controlstudiesfoundanullassociationbetweentype2DMandOCrisk,whichmayalsobeduetothelowstatisticalpower(includingonlyfourcase-controlstudieswith1806OCcases).Cohortstudiesarelesssusceptiblebothbiasesduetotheirnaturethancase-controlstudies.However,cohortstudiesmightbepronetobein uencedbydetectionbiasbecausepatientswithdiabetesareunderincreasedmedicalsurveillanceandthustheiroralcancermightbemorelikelytobediagnosedatanearlierstageinpatientswithdiabetesthaninthosewithoutdiabetes.

Second,mostofthestudiesdidnotdistinguishbetweentype1andtype2DM,althoughweexcludedtwostudieswhichconsisted

Y.Gongetal./OralOncology51(2015)332–340339

ofpatientswithyoung-onsetdiabetes.Itshouldbenotedthatabout5–10%ofadultdiabeticshavetype1DM[54],whichmaynotincreasetheriskoforalcancer[33,34].Itisnotlikelythatthemixtureofthesetwoconditionstohavesigni cantlyaffectedourresults,becausetype2DMgenerallyaccountsforthemajorityofprevalentdiabetesparticularlyinolderindividuals.Further-more,theriskassociationswerelikelytobein uencedbyimpre-ciseassessmentsofDM,whichcouldhaveledtoskewingtheestimatedeffecttowardthenull.However,researcheshavefoundthatwhencomparedwithbloodglucosetestormedicalrecords,self-reportedresponsesfordiabetesarereliable,withhighspeci c-ityandsensitivity[55].Actually,subgroupanalysesindicatedthatthemethodsofDMascertainment(medicalrecordsorbloodglu-cosetestvs.self-report)didnotsigni cantlychangetheassocia-tionbetweenDMandOCrisk.

Third,diabeticpatientsmaytakeavarietyofmedications,includingmetformin,thiazolidinedionesandinsulin,themostcommonanti-diabeticdrugs.Althoughnode nitivelinksuggeststhatanti-diabeticmedicationcouldhelppromoteacancerpheno-type,however,abodyofresearches,todate,havesuggestedthatmetforminandthiazolidinedionescouldexertaprotectiveroleagainstthedevelopmentandprogressionofsomecancers[56,57].Inaddition,controversiessurroundingusageofinsulinhaveraisedsomeseriousconcernsandrethinking[58,59].Forexample,inaretrospectivecohortstudyofpeopletreatedinUKgeneralpractices,Currieandhiscoauthorsfoundthat,comparedwithmetforminmonotherapy,anincreasedriskofanysolidtumorswasobservedinsulfonylureamonotherapy(HR=1.36,95%CI:1.19–1.54)andinsulin-basedregimens(HR=1.42,95%CI:1.27–1.60).Useofinsulinanalogueswasnotassociatedwithanincreasedcancerriskascomparedwithhumaninsulin,withaHRof1.24(95%CI0.90–1.70)[59].Whereas,wecouldnotevaluatetheeffectsofthesemedicationsontheriskassociationbetweenDMandOCriskduetothelackofdatacollection.

Fourth,asameta-analysisofobservationalstudies,thepositiveassociationbetweendiabetesandriskofcancermaybeaccountedforbyunmeasuredfactors.TheonlyslightlyalteredeffectsizesbyadjustmentforBMI/obesity,smokingandalcoholusemaysuggestthattheassociationbetweendiabetesandOCriskisindependentofthesecofactors.Betel-quidchewinghasbeenindicatedasoneoftheprimaryriskfactorsfororalcavitycancer[60]andtype2DM[61].Noneoftheincludedstudies,however,tooktheeffectofthisfactorintoaccountinthecurrentanalysis,andthus,nocon-clusionswerepossiblyyieldedregardingitspotentialroleasaneffectmodi er.

Lastly,giventhatsmallstudieswithnullresultstendnottobepublished,publicationbiasmaybepresentevenifnopublicationbiaswasindicatedfrombothvisualizationofthefunnelplotandformalstatisticaltests[62].

Insummary,comparedwithnon-diabeticpatients,diabeticindividualsmayhaveasigni cantlyincreasedriskofdevelopingoralcancerandprecancerouslesions.FutureprospectivestudiesarewarrantedtoexploretheimpactofdurationofDM,useofanti-diabeticmedicationandtheimportantconfounders,especiallybetel-quidchewing.

Con ictofintereststatement

Therearenocon ictsofinteresttodeclare.

AcknowledgementsNofunding.

AppendixA.Supplementarymaterial

Supplementarydataassociatedwiththisarticlecanbefound,intheonlineversion,athttp://www.77cn.com.cn/10.1016/j.oraloncology.2015.01.003.

References

[1]SiegelR,WardE,BrawleyO,JemalA.Cancerstatistics,2011:theimpactof

eliminatingsocioeconomicandracialdisparitiesonprematurecancerdeaths.CACancerJClin2011;61:212–36.

[2]ChineseNationalCancerCenter.Oralcavityandpharynxbutnasopharynx.

Chinesecancerregistryannualreport,2011.Beijiing:MMSPress;2012.p.38–40.

[3]RogersSN,BrownJS,WoolgarJA,LoweD,MagennisP,ShawRJ,etal.Survival

followingprimarysurgeryfororalcancer.OralOncol2009;45:201–11.

[4]SilvermanJrS,GorskyM,LozadaF.Oralleukoplakiaandmalignant

transformation.Afollow-upstudyof257patients.Cancer1984;53:563–8.[5]KrishnaRaoSV,MejiaG,Roberts-ThomsonK,LoganR.Epidemiologyoforal

cancerinAsiainthepastdecade–anupdate(2000–2012).AsianPacJCancerPrev2013;14:5567–77.

[6]LiL,PsoterWJ,BuxoCJ,EliasA,CuadradoL,MorseDE.Smokinganddrinkingin

relationtooralpotentiallymalignantdisordersinPuertoRico:acase-controlstudy.BMCCancer2011;11:324.

[7]LeeCH,KoAM,YenCF,ChuKS,GaoYJ,WarnakulasuriyaS,etal.Betel-quid

dependenceandoralpotentiallymalignantdisordersinsixAsiancountries.BrJPsychiatry2012;201:383–91.

[8]ConwayDI,PetticrewM,MarlboroughH,BerthillerJ,HashibeM,Macpherson

LM.Socioeconomicinequalitiesandoralcancerrisk:asystematicreviewandmeta-analysisofcase-controlstudies.IntJCancer2008;122:2811–9.

[9]HeidemannC,BoeingH,PischonT,NothlingsU,JoostHG,SchulzeMB.

Associationofadiabetesriskscorewithriskofmyocardialinfarction,stroke,speci ctypesofcancer,andmortality:aprospectivestudyintheEuropeanProspectiveInvestigationintoCancerandNutrition(EPIC)-Potsdamcohort.EurJEpidemiol2009;24:281–8.

[10]SaltzmanBS,DohertyJA,HillDA,BeresfordSA,VoigtLF,ChenC,etal.Diabetes

andendometrialcancer:anevaluationofthemodifyingeffectsofotherknownriskfactors.AmJEpidemiol2008;167:607–14.

[11]BenQ,CaiQ,LiZ,YuanY,NingX,DengS,etal.Therelationshipbetweennew-onsetdiabetesmellitusandpancreaticcancerrisk:acase-controlstudy.EurJCancer2011;47:248–54.

[12]BenQ,XuM,NingX,LiuJ,HongS,HuangW,etal.Diabetesmellitusandriskof

pancreaticcancer:ameta-analysisofcohortstudies.EurJCancer2011;47:1928–37.

[13]WangCS,YaoWJ,ChangTT,WangST,ChouP.Theimpactoftype2diabeteson

thedevelopmentofhepatocellularcarcinomaindifferentviralhepatitisstatuses.CancerEpidemiolBiomarkersPrev2009;18:2054–60.

[14]NotoH,GotoA,TsujimotoT,OsameK,http://www.77cn.com.cntestinsightsintotheriskof

cancerindiabetes.JDiabetesInvest2013;4:225–32.

[15]LaVecchiaC,NegriE,FranceschiS,D’AvanzoB,BoyleP.Acase-controlstudyof

diabetesmellitusandcancerrisk.BrJCancer1994;70:950–3.

[16]HjalgrimH,FrischM,EkbomA,KyvikKO,MelbyeM,GreenA.Cancerand

diabetes–afollow-upstudyoftwopopulation-basedcohortsofdiabeticpatients.JInternMed1997;241:471–5.

[17]WideroffL,GridleyG,MellemkjaerL,ChowWH,LinetM,KeehnS,etal.Cancer

incidenceinapopulation-basedcohortofpatientshospitalizedwithdiabetesmellitusinDenmark.JNatlCancerInst1997;89:1360–5.

[18]VerlatoG,ZoppiniG,BonoraE,MuggeoM.Mortalityfromsite-speci c

malignanciesintype2diabeticpatientsfromVerona.DiabetesCare2003;26:1047–51.

[19]DietrichT,ReichartPA,ScheifeleC.Clinicalriskfactorsoforalleukoplakiaina

representativesampleoftheUSpopulation.OralOncol2004;40:158–63.

[20]UjpalM,MatosO,BibokG,SomogyiA,SzaboG,SubaZ.Diabetesandoral

tumorsinHungary:epidemiologicalcorrelations.DiabetesCare2004;27:770–4.

[21]DikshitRP,RamadasK,HashibeM,ThomasG,SomanathanT,

SankaranarayananR.Associationbetweendiabetesmellitusandpre-malignantoraldiseases:acrosssectionalstudyinKerala,India.IntJCancer2006;118:453–7.

[22]KurikiK,HiroseK,TajimaK.Diabetesandcancerriskforallandspeci csites

amongJapanesemenandwomen.EurJCancerPrev2007;16:83–9.

[23]StocksT,RappK,BjorgeT,ManjerJ,UlmerH,SelmerR,etal.Bloodglucoseand

riskofincidentandfatalcancerinthemetabolicsyndromeandcancerproject(me-can):analysisofsixprospectivecohorts.PLoSMed2009;6:e1000201.[24]MeiselP,DauM,SumnigW,HoltfreterB,HoushmandM,NauckM,etal.

Associationbetweenglycemia,serumlipoproteins,andtheriskoforalleukoplakia:thepopulation-basedStudyofHealthinPomerania(SHIP).DiabetesCare2010;33:1230–2.

[25]LiC,BalluzLS,FordES,OkoroCA,TsaiJ,ZhaoG,etal.Behavioralriskfactor

surveillancesystem.DiabetesCare2009;2011(34):1365–8.

[26]SeshasaiSR,KaptogeS,ThompsonA,DiAngelantonioE,GaoP,SarwarN,etal.

Diabetesmellitus,fastingglucose,andriskofcause-speci cdeath.NEnglJMed2011;364:829–41.

340Y.Gongetal./OralOncology51(2015)332–340

[27]WottonCJ,YeatesDG,GoldacreMJ.Cancerinpatientsadmittedtohospital

withdiabetesmellitusaged30yearsandover:recordlinkagestudies.Diabetologia2011;54:527–34.

[28]BosettiC,RosatoV,PoleselJ,LeviF,TalaminiR,MontellaM,etal.Diabetes

mellitusandcancerriskinanetworkofcase-controlstudies.NutrCancer2012;64:643–51.

[29]CampbellPT,NewtonCC,PatelAV,JacobsEJ,GapsturSM.Diabetesandcause-speci cmortalityinaprospectivecohortofonemillionu.s.Adults.DiabetesCare2012;35:1835–44.

[30]LaiGY,ParkY,HartgeP,HollenbeckAR,FreedmanND.Theassociation

betweenself-reporteddiabetesandcancerincidenceintheNIH-AARPdietandhealthstudy.JClinEndocrinolMetab2013;98:E497–502.

[31]TsengCH.OralcancerinTaiwan:isdiabetesariskfactor?ClinOralInvest

2013;17:1357–64.

[32]StroupDF,BerlinJA,MortonSC,OlkinI,WilliamsonGD,RennieD,etal.Meta-analysisofobservationalstudiesinepidemiology:aproposalforreporting.Meta-analysisofobservationalstudiesinepidemiology(MOOSE)group.Jama2000;283:2008–12.

[33]ZendehdelK,NyrenO,OstensonCG,AdamiHO,EkbomA,YeW.Cancer

incidenceinpatientswithtype1diabetesmellitus:apopulation-basedcohortstudyinSweden.JNatlCancerInst2003;95:1797–800.

[34]ShuX,JiJ,LiX,SundquistJ,SundquistK,HemminkiK.Cancerriskamong

patientshospitalizedforType1diabetesmellitus:apopulation-basedcohortstudyinSweden.DiabetMed2010;27:791–7.

[35]KaoCH,SunLM,ChenPC,LinMC,LiangJA,MuoCH,etal.Apopulation-based

cohortstudyinTaiwan–useofinsulinsensitizerscandecreasecancerriskindiabeticpatients?AnnOncol2013;24:523–30.

[36]ChiouWK,HuangBY,ChouWY,WengHF,LinJD.Incidencesofcancersin

http://www.77cn.com.cnnPacJCancerPrev2011;12:1577–81.

[37]WellsGASB,O’ConnellD,PetersonJ,WelchV,LososM,TugwellP.The

Newcastle-OttawaScale(NOS)forassessingthequalityofnonrandomisedstudiesinmeta-analyses,<http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp>,[accessed15.06.12].

[38]GreenlandS.Quantitativemethodsinthereviewofepidemiologicliterature.

EpidemiolRev1987;9:1–30.

[39]DerSimonianR,LairdN.Meta-analysisinclinicaltrials.ControlClinTrials

1986;7:177–88.

[40]HigginsJP,ThompsonSG,DeeksJJ,AltmanDG.Measuringinconsistencyin

meta-analyses.Bmj2003;327:557–60.

[41]BeggCB,MazumdarM.Operatingcharacteristicsofarankcorrelationtestfor

publicationbias.Biometrics1994;50:1088–101.

[42]EggerM,DaveySmithG,SchneiderM,MinderC.Biasinmeta-analysis

detectedbyasimple,graphicaltest.Bmj1997;315:629–34.

[43]AlbrechtM,BanoczyJ,DinyaE,TamasJrG.Occurrenceoforalleukoplakiaand

lichenplanusindiabetesmellitus.JOralPatholMed1992;21:364–6.

[44]AriyoshiY,ShimaharaM,OmuraK,YamamotoE,MizukiH,ChibaH,etal.

Epidemiologicalstudyofmalignanttumorsintheoralandmaxillofacialregion:surveyofmemberinstitutionsoftheJapaneseSocietyofOralandMaxillofacialSurgeons,2002.IntJClinOncol2008;13:220–8.

[45]ChiangCT,HwangYH,SuCC,TsaiKY,LianIeB,YuanTH,etal.Elucidatingthe

underlyingcausesoforalcancerthroughspatialclusteringinhigh-riskareas

ofTaiwanwithadistinctgenderratioofincidence.GeospatHealth2010;4:230–42.

[46]

TsaiPH,LiuJJ,YehCL,ChiuWC,YehSL.Effectsofglutaminesupplementationonoxidativestress-relatedgeneexpressionandantioxidantpropertiesinratswithstreptozotocin-inducedtype2diabetes.BrJNutr2012;107:1112–8.

[47]

FengB,RuizMA,ChakrabartiS.Oxidative-stress-inducedepigeneticchangesinchronicdiabeticcomplications.CanJPhysiolPharmacol2013;91:213–20.

[48]

NilsenTI,VattenLJ.Prospectivestudyofcolorectalcancerriskandphysicalactivity,diabetes,bloodglucoseandBMI:exploringthehyperinsulinaemiahypothesis.BrJCancer2001;84:417–22.

[49]GiovannucciE.Insulin,insulin-likegrowthfactorsandcoloncancer:areviewoftheevidence.JNutr2001;131:3109S–31020S.

[50]

SaydahSH,LoriaCM,EberhardtMS,BrancatiFL.AbnormalglucosetoleranceandtheriskofcancerdeathintheUnitedStates.AmJEpidemiol2003;157:1092–100.

[51]

YenYC,ShiahSG,ChuHC,HsuYM,HsiaoJR,ChangJY,etal.ReciprocalregulationofmicroRNA-99aandinsulin-likegrowthfactorIreceptorsignalinginoralsquamouscellcarcinomacells.MolCancer2014;13:6.

[52]

ChenFC,ZhangF,ZhangZJ,MengSY,WangY,XiangXR,etal.Tumornecrosisfactor-alphagenepolymorphismsandriskoforalcancer:http://www.77cn.com.cnnPacJCancerPrev2013;14:7243–9.

[53]

PivaMR,DESLB,Martins-FilhoPR,NonakaCF,DESST,DESAES,etal.Roleofin ammationinoralcarcinogenesis(PartII):CD8,FOXP3,TNF-alpha,TGF-betaandNF-kappaBexpression.OncolLett2013;5:1909–14.

[54]

EngelgauMM,GeissLS,SaaddineJB,BoyleJP,BenjaminSM,GreggEW,etal.TheevolvingdiabetesburdenintheUnitedStates.AnnInternMed2004;140:945–50.

[55]

OkuraY,UrbanLH,MahoneyDW,JacobsenSJ,RodehefferRJ.Agreementbetweenself-reportquestionnairesandmedicalrecorddatawassubstantialfordiabetes,hypertension,myocardialinfarctionandstrokebutnotforheartfailure.JClinEpidemiol2004;57:1096–103.

[56]SmithU,GaleEA.Doesdiabetestherapyin uencetheriskofcancer?Diabetologia2009;52:1699–708.

[57]

LibbyG,DonnellyLA,DonnanPT,AlessiDR,MorrisAD,EvansJM.Newusersofmetforminareatlowriskofincidentcancer:acohortstudyamongpeoplewithtype2diabetes.DiabetesCare2009;32:1620–5.

[58]BadrickE,RenehanAG.Diabetesandcancer:5yearsintotherecentcontroversy.EurJCancer2014;50:2119–25.

[59]CurrieCJ,PooleCD,GaleEA.Thein uenceofglucose-loweringtherapiesoncancerriskintype2diabetes.Diabetologia2009;52:1766–77.

[60]

PettiS,MasoodM,ScullyC.Themagnitudeoftobaccosmoking-betelquidchewing–alcoholdrinkinginteractioneffectonoralcancerinSouth-EastAsia.Ameta-analysisofobservationalstudies.PLoSOne2013;8:e78999.

[61]

YamadaT,HaraK,KadowakiT.Chewingbetelquidandtheriskofmetabolicdisease,cardiovasculardisease,andall-causemortality:ameta-analysis.PLoSOne2013;8:e70679.

[62]

SterneJAC,EggerM,MoherD,editors.Chapter10:Addressingreportingbiases.In:CochraneHandbookforSystematicReviewsofInterventions,Version5.0.1[updatedSeptember2008],HigginsJPT,GreenSeditors.TheCochraneCollaboration;2008,http://www.77cn.com.cn.

本文来源：https://www.bwwdw.com/article/3yai.html