adults body mass index and waist-to-hip ratio, inKorean marker of type 2 diabetes, independentof Thoracic-to-hip circumference ratio as anovel Clinical Practice Diabetes Researchand

Thoracic-to-hip circumference ratio as a

novel marker of type 2 diabetes, independent of body mass index and waist-to-hip ratio, in Korean adults

Duong Duc Pham

, BonCho Ku

, Chol Shin

, Nam H. Cho

, Seongwon Cha

, Jong Yeol Kim

*

aKoreaInstituteofOrientalMedicine,461-24Jeonmin-dong,Yuseong-gu,Daejeon,RepublicofKorea

bNationalHospitalofTraditionalMedicine,29NguyenBinhKhiem,Hanoi,VietNam

cDepartmentofInternalMedicine,KoreaUniversityAnsanHospital,516,Gojan-1-dong,Danwon-gu,Ansan425-707, RepublicofKorea

dDepartmentofPreventiveMedicine,AjouUniversitySchoolofMedicine,Suwon,RepublicofKorea

article info

Articlehistory:

Received9January2013 Receivedinrevisedform 23September2013

Accepted20December2013 Availableonline7February2014

Keywords:

Anthropometricindex Upper-to-lowerratio Type2diabetes

abstract

Aims:Wecompareduppertrunkanthropometricindiceswithoverallandcentralobesity indicatorstopredictthepresenceoftype2diabetesinmiddle-agedandelderlyKorean individuals.

Methods:Thiscross-sectionalinvestigationincluded4079ruralandurbanparticipantsaged 40–80years.Neck,thoracic,waist(WC),andhipcircumferencesweremeasuredbyareliable andstandardizedmethod.Theneck-to-hipratio,thethoracic-to-hipratio(THR),andthe waist-to-hipratio(WHR)werecalculated.A75-goralglucosetolerancetestwasperformed.

Type2diabeteswasdefinedbasedontheguidelines oftheWorldHealthOrganization (1999).

Results: ThereceiveroperatorcharacteristiccurveanalysisindicatedthatTHRandWHR werebetterthanbodymassindex(BMI)andotheranthropometricindicesatpredictingthe presenceoftype2diabetes.Theadjustedoddsratios(OR)acrossquartilesofTHRwere slightlyhigherthantheORsforWHR,particularlyinthehighestquartile(oddsratiosand 95%CI:2.11(1.47–3.04)versus1.95(1.37–2.77)inmen;3.40(2.18–5.31)versus2.31(1.48–3.60) inwomen).TheassociationsofTHRandWHRwithtype2diabetesremainedsignificant, despiteaslightattenuationafteramultivariateadjustmentforBMI.ThejointeffectofBMI andTHRontheriskoftype2diabeteswaslargerthanthatofBMIandWHR.

Conclusions: THRmaybeanovelmarkeroftype2diabetes,particularlyinwomen,andits associationwithdiabeteswasindependentofBMIandWHR.

*Correspondingauthorat:DepartmentofMedicalResearch,KoreaInstituteofOrientalMedicine(KIOM),461-24Jeonmin-dong,Yuseong- gu,Daejeon305-811,RepublicofKorea.Tel.:+82428689489;fax:+82428689480.

E-mailaddress:ssmed@kiom.re.kr(J.Y.Kim).

ContentsavailableatScienceDirect

Diabetes Research and Clinical Practice

j o u r n a lh o m e p a g e :w w w . e l s e v i e r . c o m / l o c a t e / d i a b r e s

0168-8227

http://dx.doi.org/10.1016/j.diabres.2013.12.022

#2014TheAuthors.PublishedbyElsevierIrelandLtd.Open access under CC BY-NC-SA license.

#2014TheAuthors.PublishedbyElsevierIrelandLtd.Open access under CC BY-NC-SA license.

1. Introduction

Type2diabeteshasbecomeaglobalepidemichealthproblem worldwide,anditiscloselyrelatedtonumerouscardiometabolic complicationsofobesity[1].Bodymassindex(BMI)hasbeenthe mostcommonlyusedmeasureofoverallobesitytoreflecttotal bodyfat,althoughtheindexcannotdistinguishbetweenmuscle andfatmass[2].Manystudieshavedemonstratedthatexcessive accumulation of adipose tissue in particular body regions contributestometaboliccomplications[3,4].

Vague [5] first proposed that those who have fat predominantlyaccumulatedintheupperbodyratherthan the lower body are more susceptible to metabolic dis- turbances.Forseveraldecades,numerousstudiesexamined this hazardous body shape phenotype; however, most emphasizedtheimpactofbodyfatcentralization,measured bywaistcircumference(WC)orwaist-to-hipratio(WHR),on the risk of metabolic complications [6,7]. Freedman and Rimm demonstrated that fat accumulation in various regionsoftheuppertrunkhaddifferentdiabetescorrelates independentofabdominalfat[8].In recentyears,interest has increased in the cardiometabolic correlates of upper body subcutaneous (sc) [9,10], intra-thoracic [11,12], and pericardialfat[13].Kangetal.foundthatandroidfat,afat depot located just above the central fat depot, is more accurate than abdominal visceral adipose tissue (VAT) at predictingmetabolicsyndrome[14].

Bodygirthmeasurementsatmultipleregionsofthetrunk canprovideinformationonparticularlocalfataccumulation, whereasgirthratiosmayreflectthefatdistributionpattern and body shape. Numerous epidemiological studies have usedWHRasadeterminantofthefatdistributionpattern;

however,WHR,whichcontainsinformationonabdominalfat and hip fat, does not reflect the concrete image of fat distribution. In addition, whether body fat distribution provides more accurate informationon the risk oftype 2 diabetes than total body fat remains controversial [15].

Becauseoftheincreasingevidenceofmetaboliccorrelates ofuppertrunkfat,weassesstheassociationofuppertrunk circumferences andtheir ratiosto hip circumference(HC) withtype2diabetesandexaminewhethertheseanthropo- metricindicescanpredicttype2diabetesmoreaccurately thanoverallobesity(assessed by BMI)andcentral obesity (assessedbyWCandWHR)indicators.

2. Materials and methods

Thecurrentstudywasdesignedasacross-sectionalinves- tigationaspartoftheKoreanHealthandGenomeEpidemiol- ogyStudy(KHGES),anongoingperspectivecohortstudy[16].

A total of 7629 participants aged 10 years or more were recruited using a two-stage cluster sampling method by telephoneormailfromJune2009toDecember2011atthe KoreaUniversityAnsanHospitalandtheCenterforClinical EpidemiologyofAjouUniversityHospital.TheAnsancohort datarepresentanurbancommunity,andtheAnsungcohort

data represent a rural community. In this analysis, we included only participantsaged 40–80 yearsandexcluded participantswhohadbeendiagnosedwithdiabetesand/or werereceivingmedicationforhypertension,diabetes,dysli- pidemia, andwho hadmissingdata onbody composition indices,bodymeasurements,bloodpressure,fastingplasma glucose(FPG),2-hpost-loadplasmaglucose,orsmokingand drinkinginformation.Theanalysiswasperformedon4079 participants (1967men and2112women). This study was approvedbytheNationalInstituteofHealthEthicsandthe InstitutionalReviewBoardoftheKoreanHealthandGenomic Study, Ajou University. All of the participants gave their writteninformedconsent.

2.2. Anthropometricindicesandbodycomposition measurements

Body measurements were measured horizontally with a tapelinetothenearest1mmbytrainedoperatorsfollowing thestandardizedoperatingproceduredevelopedbyJangetal.

[17].Ithasbeenrevealedthat thesemeasurementshavea highreliabilitywithrelativetotaltechnicalerrorofmeasure- mentwere0.68%–2.18%[17].Neckcircumference(NC)was measuredatthelowermarginofthethyroidcartilagewhile participants sat on a chair with their heads positioned horizontally. Thoracic circumference (ThC), WC, and HC weremeasuredwhileparticipantsstooderectandbreathed naturally.Thetapelinewaspositionedattheleftandright prominencesofthe7th–8thcostochondraljunctions,andthe ThCvaluewasrecordedbetweeninspirationandexpiration.

WCandHCweremeasuredattheumbilicusscarandupper marginofthepubis,respectively.Theneck-to-hipcircum- ference (NHR), thoracic-to-hip circumference (THR), and waist-to-hip circumference (WHR) ratios were defined as NC, ThC, and WC divided by HC, respectively. Thus, the upper-trunk-relatedanthropometricindiceswereNC,ThC, NHR,andTHR.

Bodyheightandweightweremeasuredusingadigitalscale (GL-150;GTechInternationalCo.,Ltd.,Uijeongbu).Participants worecasualclothing.Themandibularplanewasparalleltothe floor. BMI was calculated as weight (kg) divided by height squared (m²). Body fat mass (BFM) was assessed using a bioelectricalimpedanceanalyzer(BIA)(INBODY720,Biospace Korea).Inaddition,bodyfatpercent(BFP)wascalculatedby dividingBFMbybodyweight.

2.2.1. Definitionoftype2diabetes

Themeasurementsofbloodpressure,fastingand2-hpost- load glucose concentration, and lipid profiles have been describedindetailinapreviousstudy[18].Type2diabetes was definedasaFPG126mg/dl [7.0mmol/l]and/ora 2-h post-loadplasmaglucose200mg/dl[11.1mmol/l],according tothe1999WorldHealthOrganizationdiagnosticcriteria[19].

2.3. Confoundingfactors

Participantswereclassifiedascurrentsmokersiftheysmoked currently,ex-smokersiftheyhadsmokedpreviouslybuthad quit,andnonsmokersiftheyhadneversmoked.Thesame classificationwasappliedforalcoholconsumption.

2.4. Dataanalysis

ThedatawereanalyzedusingRsoftware,version2.14.1,ona Windows7platform.Thesignificancelevelwassetto0.05.A receiver operating characteristic (ROC) curve was used to estimatethe predictive performance ofBMI and anthropo- metricindicesandtocalculate theoptimalcutoffpointsin predicting type2diabetes. Morearea underthe ROCcurve (AUC) implies more predictive performance. The optimal cutoffforeachdeterminantoftype2diabeteswasdetermined bytheYoudenindex[20].Aunivariatelogisticregressionwas used to assess the association between a one standard deviation(SD)increaseinBMIandallanthropometricindices and type2 diabetes.A multiplelogistic regressionanalysis adjustedforage,habitualsmoking,alcoholconsumption,and study center was used to evaluate the risk increase by quartiles of BMI and all anthropometric indices. We used theAUCvalueandtheoddsratio(OR)todeterminetheupper trunkanthropometricindexthatwasmostcloselyassociated withtype2diabetes.Furtheradjustmentsforoverallobesity (assessedbyBMI)and centralobesity(assessedbyWCand WHR)indicatorswereincludedtoassesstheindependentand joint effects of the upper trunk anthropometric index in predictingtype2diabetes.

3. Results

Thecharacteristicsofparticipantsbygenderarepresentedin Table1.Comparedtothewomen,themenwereheavierand tallerandhad alowerBMI,bodyfatmass,andfatpercent.

Additionally, men had larger anthropometric indices and higherriskoftype2diabetesthanwomen.

3.1. Associationofanthropometricindiceswithtype2 diabetesandtheircutoffs

BMIandallanthropometricindiceswereassociatedwithtype 2diabetes.TheROCanalysisindicatedthatTHChadastronger relationshipwithtype2diabetesthanBMI,NC,orWC(men:

AUC=0.551 versus 0.533, 0.535, and 0.541; women:

AUC=0.578 versus 0.559, 0.545, and 0.559). THR and WHR (men:AUC=0.577and0.564;women:AUC=0.606and0.582) werestrongermarkersoftype2diabetesthanBMIandother anthropometricindices.Theoptimalcutoffpointsestimated bytheYoudenindexforpredictingdiabetesforTHRandWHR were0.93and0.93inmenand0.88and0.90inwomen(Table 2).EachincreaseintheSDoftheTHRwasassociatedwitha 1.29-fold(1.15–1.45)anda1.48-fold(1.29–1.70)increaseinthe oddsratiosoftype2diabetesinmenandwomen,respectively (p<0.0001),whereasthoseoftheWHRwere1.23-fold(1.09–

1.39) and 1.38 (1.20–1.58) in men (p=0.001) and women (p<0.0001), respectively (Table 2). After adjustments for potential confounders, including age, habitual smoking, alcoholconsumption,andstudycenter,THRand WHRstill hadastrongerassociationwithtype2diabetesthanBMIand other anthropometric indices. The adjusted ORs for the second,third, andfourth quartilesofTHR were1.48 (1.02–

2.14),1.91(1.33–2.75),and2.11(1.47–3.04)inmenandwere1.65 (1.05–2.59), 1.61 (1.02–2.56), and 3.40 (2.18–5.31) in women,

respectively,whereasthosevaluesforWHRwere1.53(1.07–

2.20), 1.47 (1.02–2.12), and 1.95 (1.37–2.77) inmen and 1.60 (1.03–2.49), 1.67 (1.07–2.61), and 2.31 (1.48–3.60) in women, respectively(Table3).

3.2. IndependentandjointeffectsoftheTHRandWHRin predictingtype2diabetes

Thesecondobjectiveofthisstudywastodeterminewhether THR,theupper-trunk-relatedanthropometricindexthatwas most related to type 2 diabetes, was an independent determinantoftype2diabetestooverallandcentralobesity indicators. Because the data showed that WHR was more closelyassociatedwithtype2diabetesthanWC,acomparison ofthemagnitudeoftheassociationwithtype2diabeteswas conductedbetweenBMI,WHR,andTHR.

Themultiplelogisticregressionanalysisshowedthatthe adjustment for potential confounders and BMI slightly attenuatedtheassociationbetweenTHRandWHRwithtype

Table1–Characteristicofstudyparticipants.

Men (n=1967)

Women (n=2112)

p

Studycenter

Ansann(%) 885(45.0) 1067(50.5) Ansungn(%) 1082(55.0) 1045(49.5)

Age(yrs) 57.9(8.3) 57.6(8.3) 0.27

Bodyheight(cm) 166.9(5.8) 154.4(5.7) <0.0001 Bodyweight(kg) 66.4(9.5) 57.5(8.1) <0.0001

BMI(kg/m²) 23.8(2.9) 24.1(3.1) 0.001

NC(cm) 37.0(2.2) 32.8(1.9) <0.0001

ThC(cm) 86.7(6.1) 79.1(7.2) <0.0001

WC(cm) 85.6(7.5) 83.7(8.3) <0.0001

HC(cm) 91.9(5.2) 92.3(5.3) 0.006

NHR 0.40(0.02) 0.36(0.02) <0.0001

THR 0.94(0.04) 0.86(0.06) <0.0001

WHR 0.93(0.05) 0.91(0.07) <0.0001

SBP(mmHg) 118.6(15.1) 114.8(16.5) <0.0001 DBP(mmHg) 79.2(9.8) 75.1(10.1) <0.0001 FPG(mg/dl) 98.2(16.4) 92.8(11.4) <0.0001 2hpost-loadglucose(mg/dl) 127.0(51.2) 128.7(40.8) <0.0001 Totalcholesterol(mg/dl) 191.8(32.9) 202.2(33.4) <0.0001 Triglycerides(mg/dl) 138.4(70.2) 120.1(57.5) <0.0001 HDLcholesterol(mg/dl) 45.3(12.0) 49.0(12.6) <0.0001 LDLcholesterol(mg/dl) 118.8(30.1) 129.2(29.9) <0.0001 Smoking,n(%)

Nonsmoker 506(25.7) 2064(97.7) <0.0001

Ex-smoker 810(41.2) 16(0.8)

Currentsmoker 651(33.1) 32(1.5) Drinking,n(%)

Nondrinker 482(24.5) 1526(72.2) <0.0001

Ex-drinker 122(6.2) 33(1.6)

Currentdrinker 1363(69.3) 553(26.2)

Diabetes,n(%) 331(16.8) 225(10.7) <0.0001 Valuesaremeans(SD)orn(%).pvaluesarecalculatedfromt-tests orchi-squaretestsforanalysisofvariance.BMI,bodymassindex;

NC, neck circumference; WC, waist circumference; HC, hip circumference; ThC, thoracic circumference; NHR, neck-to-hip circumference ratio; THR, thoracic-to-hip circumference ratio;

WHR,waist-to-hipcircumferenceratio;SBP,systolicbloodpres- sure;DBP,diastolic bloodpressure;FPG,fastingplasmaglucose;

HDLcholesterol,high-densitylipoproteincholesterol;LDLcholes- terol,low-densitylipoproteincholesterol.

2diabetes(Supplementary,TableS1).Thecorrelationcoeffi- cientsbetweenBMIandTHRwere0.28 inmenand 0.45in women,whereasthecorrelationcoefficientsbetweenBMIand WHRwere0.49inmenand0.44inwomen(Supplementary, Table S2).Theadjusted ORsforthe THR werehigherthan those for the WHR in all models, particularly for women (Supplementary, Table S1). Furthermore, there was an increasein the prevalenceoftype2 diabetesby tertilesof THRwithineachtertileofWHR,particularlyinthehighest WHRtertile(Fig.1).WithineachtertileofWHR,the higher tertilesofTHRhadastrongerriskoftype2diabetes,including plasmafasting and 2-h post-load glucoseand triglycerides

concentrations, thanthelowertertiles(Supplementary,Fig.

S1).TheseresultsindicatetheindependenteffectofTHRon diabetes.

Supplementary material related to this article can be found, in the online version, at http://dx.doi.org/10.1016/

j.diabres.2013.12.022.

To examine the joint effects of BMI, THR, and WHRin predicting type 2diabetes, we stratified participants using internallycalculatedcutoffs.ThecutoffforBMI(25.7kg/m²for menand25.0kg/m²forwomen),THR(0.93formenand0.88for women),andWHR(0.93formenand0.90forwomen)were calculatedfromthedistributionofthestudysample(Table2).

Table2–AssociationofBMIandanthropometricindiceswithtype2diabetes.

OR(95%CI) AUC(95%CI) Cutoff Se Sp

Men

BMI(kg/m²) 1.12(1.00–1.26) 0.533(0.497–0.569) 25.7 0.33 0.76

NC(cm) 1.15(1.02–1.29) 0.535(0.499–0.570) 38.6 0.30 0.78

ThC(cm) 1.20(1.06–1.35) 0.551(0.516–0.586) 89.1 0.42 0.69

WC(cm) 1.16(1.03–1.30) 0.541(0.506–0.576) 88.0 0.45 0.63

NHR 1.18(1.04–1.33) 0.539(0.505–0.574) 0.42 0.25 0.83

THR 1.29(1.15–1.45) 0.577(0.544–0.610) 0.93 0.74 0.38

WHR 1.23(1.09–1.39) 0.564(0.530–0.598) 0.93 0.65 0.47

Women

BMI(kg/m²) 1.23(1.08–1.40) 0.559(0.518–0.600) 25.0 0.47 0.66

NC(cm) 1.19(1.04–1.36) 0.545(0.540–0.587) 33.3 0.48 0.63

ThC(cm) 1.30(1.14–1.49) 0.578(0.539–0.618) 80.5 0.52 0.62

WC(cm) 1.24(1.08–1.42) 0.559(0.520–0.599) 83.9 0.57 0.54

NHR 1.22(1.07–1.40) 0.548(0.507–0.588) 0.37 0.34 0.75

THR 1.48(1.29–1.70) 0.606(0.565–0.646) 0.88 0.50 0.69

WHR 1.38(1.20–1.58) 0.582(0.543–0.622) 0.90 0.65 0.49

OR(95%CI),oddsratio(95%confidenceinterval)fortype2diabeteswitheachstandarddeviationincreaseintheindependentvariables, calculatedbyunivariateanalysis;AUC(95%CI),area underthecurve values(95%confidenceinterval)calculatedbyreceiveroperating characteristiccurve;cut-off,cut-offpointforindependentvariablesinpredictingtype2diabetesdeterminedbyYoudenindex;Se,sensitivity;

Sp,specificity.ThetwohighestORandAUCvaluesareinbold.Forotherabbreviations,seeTable1.

Table3–Adjustedoddsratiosoftype2diabetesbyquartilesofBMIandanthropometricindicesformenandwomen.

Quartile1 Quartile2 Quartile3 Quartile4

Men

BMI 1 0.79(0.56–1.13) 0.86(0.60–1.22) 1.43(1.02–2.00)^*

NC 1 0.85(0.61–1.21) 1.05(0.74–1.49) 1.67(1.19–2.34)^**

ThC 1 1.03(0.72–1.47) 1.25(0.88–1.78) 1.71(1.22–2.40)^**

WC 1 1.00(0.71–1.42) 1.21(0.85–1.72) 1.54(1.10–2.16)^*

NHR 1 1.19(0.84–1.69) 1.17(0.82–1.66) 1.53(1.09–2.15)^*

THR 1 1.48(1.02–2.14)^* 1.91(1.33–2.75)^*** 2.11(1.47–3.04)^***

WHR 1 1.53(1.07–2.20)^* 1.47(1.02–2.12)^* 1.95(1.37–2.77)^***

Women

BMI 1 1.14(0.75–1.75) 1.32(0.87–2.00) 1.68(1.13–2.51)^*

NC 1 0.90(0.59–1.37) 1.07(0.72–1.58) 1.70(1.16–2.49)^**

ThC 1 1.11(0.72–1.71) 1.53(1.01–2.32)^* 2.15(1.43–3.25)^***

WC 1 1.00(0.65–1.53) 1.48(0.99–2.20) 1.51(1.00–2.29)^*

NHR 1 1.28(0.85–1.93) 0.99(0.65–1.53) 1.86(1.25–2.77)^**

THR 1 1.65(1.05–2.59)^* 1.61(1.02–2.56)^* 3.40(2.18–5.31)^***

WHR 1 1.60(1.03–2.49)^* 1.67(1.07–2.61)^* 2.31(1.48–3.60)^***

Dataarepresentedasoddsratio(95%confidentinterval)adjustedforage,habitualsmoking,alcoholconsumption,andstudycenter.Quartiles 1,2,3,and4representthelowest,lowmedium,highmedium,andhighestquartiles,respectively,ofBMIandanthropometricindices.The lowestquartilewasthereferencecategory.Forotherabbreviations,seeTable1.

* Significantlevel:p<0.05.

** Significantlevel:p<0.01.

*** Significantlevel:p<0.001.

Fig.1–Prevalenceoftype2diabetesoftertilesofTHRacrosstertilesofWHR.THR,thoracic-to-hipratio(THR1,THR2,and THR3referstotertilesofTHRwithineachtertileofWHR);WHR,waist-to-hipratio(WHR1,WHR2,andWHR3referto tertilesofWHR).

Fig.2–JointandindependenteffectsofBMI,WHR,andTHRinpredictingtype2diabetesformenandwomen.Thecutoff pointforBMIwas25.7kg/m²formenand25.0kg/m²forwomen(seeTable2);cutoffpointsforWHR(seeTable2)were0.93 formenand0.90forwomen;cutoffpointsforTHR(Table2)were0.93formenand0.88forwomen.(H)meanshighcategory ofindependentvariables(e.g.,BMI(H)meansBMII25.7kg/m²formenandI25.0kg/m²forwomen);(L)meanslow categoryofindependentvariable(e.g.,BMI(L)meansBMI<25.7kg/m²formenand<25.0kg/m²forwomen).Thesample wasstratifiedinto8categorizeswith7dummies.ThegroupofsubjectswithlowBMI,lowWHR,andlowTHRwassetupas thereferencegroup.^*p<0.05;^**p<0.01;^***p<0.001,assessedbymultivariablelogisticregressionanalysisadjustedforage, smoking,drinking,andstudycenter.Barsrepresenttheadjustedoddsratiooftype2diabetes;numbersareoddsratio(95%

confidentinterval).

Thus,participantswerecategorizedinto8groupscodedas7 dummyvariables(Fig.2).Thegroupthatdidnothavehigh BMI, high WHR, or high THR was used as the reference categoryinthemultiplelogisticregressionanalysisadjusted forpotentialconfounders.AsshowninFig.2,thosewhohad onlyahigh BMIor high WHRor high THRdid nothave a significantlyhigherriskoftype2diabetesthanthereference group for both men and women. The adjusted ORs for individuals who had high BMI and high THR were 2.35 (1.25–4.43)formen(p<0.01)and3.05(1.09–8.51)forwomen (p<0.05).TheadjustedORsforindividualswithhighBMIand high WHRwere1.30 (0.60–2.83)for men(p=0.50) and1.11 (0.59–2.06)forwomen(p=0.75).Additionally,theriskincrease forindividualswithhighBMI,highTHR,andhighWHR(ORs were2.27 inmenand 3.34 inwomen)wasnot remarkably higherthanthatofindividualswhohadonlyhighBMIand highTHR.

4. Discussion

Many previous studies stressed the link between overall (assessed by BMI) and central (assessed by WC and WHR) obesity and increased risk of metabolic complications, including type 2 diabetes [21–24]. Debate continues about thebestanthropometricindicesfordiabetes.Recentstudies havefocusedonthecardiometaboliccorrelatesofuppertrunk fatanduppertrunk-relatedanthropometricindices,suchas NC [9,10], and the ‘‘protective’’ effect of hip size [24,25].

However, no studies have compared upper trunk-related anthropometricindicesandindicatorsofoverallandcentral obesityforpredictingtype2diabetes.Thepresentstudyisthe firstattempt tocompare theeffects ofuppertrunk-related anthropometricindicesandindicatorsofoverallandcentral obesity inpredicting type2 diabetes inKorean adults.We foundthat(i)THRistheupper-trunk-relatedanthropometric indexthatismostcloselyrelatedtotype2diabetesamongthe upper-trunk-relatedanthropometricindices;(ii)THRappears tobeanindependentdeterminantoftype2diabetestoBMI andWHRanditseffectmaybebetterthanthatofWHRin women; and (iii)BMI and THRcan predict type2 diabetes betterthanBMIandWHR.

Ourfindingssuggestedthatthestrengthoftheassociation betweendiabetesandThCisslightlystrongerthanthatofNC and WC. Moreover, THR isthe strongest markerof type2 diabetes among all investigated indices, particularly in women.ThCreflectsthesizeofthoraciccavityandthebody segmentwhereandroidfatismeasuredbydual-energyX-ray absorptiometry [14]. Evidence suggests that android fat components,suchaspericardial,hepatic,andintra-thoracic fats,areassociatedwithmetabolicdisturbancesindependent of total and abdominal fatness [11–13], and the metabolic correlatesofandroidfatarestrongerthanthoseofVAT[14].It has been indicated that hepatic fatness plays a unique important role indetermination oftype 2diabetes [26,27].

AlthoughWCisawidelyrecommendedindexfordetermining centralobesityandcanbeastrongdeterminantofdiabetes, thereisnoconsensusonthestandardmeasurementsitefor thisindex.WCmeasuredattheupperabdomen(justbelow the lowest rib) hasa stronger relationwith VAT than WC

measuredatthelowerabdomen(umbilicusoriliaccrest)[28].

The VAT measured at the upper abdomen is additionally associatedmorecloselywithobesity-relatedhealthrisksthan VATmeasuredatthelowerabdomen[29].Furthermore,WCis morerelatedtoabdominalscfatthanabdominalVAT[30],and VAT more strongly correlateswith insulin resistance than abdominal sc fat [31]. In this study, ThC measured at the prominenceofthe7th–8thcostochondraljunctions(approxi- mately5cmabovethelowestrib)maybemorerelatedwith androidfat/VATthanabdominalscfat.AlthoughThCandWC are highlycorrelated, they mayreflectdifferent aspects of body fat (android/VAT versus abdominal sc fat). Further studiesshouldfocusonthedifferentrelationshipsbetween ThCandWCwithabdominalscfatandVAT.

WeadditionallyfoundthatTHRandWHR,whicharethe ratios that include information on upper/central trunk measurementsandhipsize,hadastrongerrelationshipwith diabetes than BMI and other anthropometric indices; the effectofTHRwasslightlystrongerthanWHR(Table3).WCand HChaveindependentandoppositeassociationswithdiabetes risk.AlargerHCisinverselyassociatedwithcardiovascular risk,includingdiabetes.Theprotectiveeffectismostlikelya resultoftheincreaseingluteofemoralfatandmusclemass [25].ThecombinationofWCandHCinaratioreflectsnotonly the sizeofbody segmentsand localbody fatsbut alsothe pattern of fatdistribution. Several studies have reporteda stronger correlation between diabetes and WHR than WC, whichmaybearesultoftheintegrativeeffectsofbothwaist andhipsize[24,32,33].THRcanbeconsideredastheratioof androidfat,aharmfulfactor,togluteofemoral(gynoid)fat,a protectivefactor.Evidenceshowedthattheratioofandroidto gynoidfatwasassociatedwithinsulinresistance[34],andthe trunk-to-hipratio wasmorerelatedtobaselineinsulinand QUICKIlevelsthanabdominal-to-hipratios[35].AlthoughTHR andWHRwererelativelycorrelated(thePearson’scorrelation coefficients were 0.62 in men and 0.73 in women) in the currentstudy,thecorrelationbetweendiabetesandTHRwas slightlystrongerthanthatofWHRinallmodels,evenafter adjusting for BMI. Interestingly, the correlation between diabetes and THR remained within each tertile of WHR (Fig.1andS1).Furthermore,thecombinationofahighBMI andahighTHRresultedinahigherriskoftype2diabetesthan acombinationofahighBMIandahighWHR,possiblybecause ofthestrongercorrelationbetweendiabetesandandroidfat versusabdominalscfat.Takentogether,bodymeasurements at the upper trunk, such as ThC and THR, provide more accurate information on therisk ofdiabetesthan BMIand other anthropometric measurements, and they should be consideredinclinicalandepidemiologicalstudies.

Thisstudyhascertainlimitations.Althoughtheremaybe advantagesofTHRoverBMIandWHR,THRismoredifficultto measurebecauseitisnoteasytoidentifytheprominencesof the7th–8thcostochondraljunctionsinsomecircumstances, such as serious obesity. The operators must be trained thoroughly.Wewereunabletomeasurebodyfatcomposition to examine the relationship between ThC and THR with abdominalfatandVAT.Inthisstudy,WCwasmeasuredatthe umbilicuslevel thatmaycausemeasurementbias,particu- larly inobeseindividuals. Thesefindings originatedfroma cross-sectionalstudyand areunabletoconfirmanycausal

relationship.Because those whoareforty or olderareat a higherriskoftype2diabeteswithahigherdemandofdiabetes screeningthan thoseareatyoungerage,the presentstudy thereforefocusedonmiddle-agedandelderlyKoreanindivi- duals.Thus,theresultsmaynotbegeneralizabletopopula- tionsofotheragesandraces.

Inconclusion,thesedataindicatethatTHRisassociated withtype2diabetesbeyondtheeffectsofoverall(assessedby BMI)and centralobesity(assessed byWHR)indicatorsina middle-agedand elderlyKorean population.Thesefindings underline the importance of upper trunk measurements, particularly ThC and THR, in epidemiological and clinical studies.

Conflict of interest statement

Theauthorsdeclarethattheyhavenoconflictofinterest.

Acknowledgements

ThisworkwassupportedbytheNationalResearchFoundation ofKorea(NRF),andthegrantwasfundedbytheKoreaMinistry of Education, Science and Technology (MEST) [No.

20120009001(2006-2005173)].Thisworkwasadditionallysup- portedbytheKoreaCentersforDiseaseControlandPreven- tion[2009-E00454-00,2010-E71001-00,and2011-E71004-00].

C.S.,N.H.C.,andJ.Y.K.designedandsupervisedtheKorean HealthandGenomeEpidemiologyStudy(KHGES),definedthe researchtheme,andeditedthemanuscript.D.D.P.designed methods,analyzedthedata,interpretedtheresults,andwrote themanuscript.B.C.K.co-analyzedthedata,interpretedthe results,andeditedthemanuscript.S.C.reviewedandedited themanuscriptandcontributedtodiscussion.

references

[1] WHO.Theglobalburdenofdisease2004update.Geneva, Switzerland:WHO;2004.

[2] WHOExpertConsultation.Appropriatebody-massindex forAsianpopulationsanditsimplicationsforpolicyand interventionstrategies.Lancet2004;363(9403):157–63.

[3] JensenMD.Roleofbodyfatdistributionandthemetabolic complicationsofobesity.JClinEndocrinolMetab2008;93(11 Suppl.1):S57–63.

[4] GargA.Regionaladiposityandinsulinresistance.JClin EndocrinolMetab2004;89(9):4206–10.

[5] VagueJ.Thedegreeofmasculinedifferentiationof obesities:afactordeterminingpredispositiontodiabetes, atherosclerosis,gout,anduriccalculousdisease.AmJClin Nutr1956;4(1):20–34.

[6] ParkSH,ChoiSJ,LeeKS,ParkHY.Waistcircumferenceand waist-to-heightratioaspredictorsofcardiovascular diseaseriskinKoreanadults.CircJ2009;73(9):1643–50.

[7] deKoningL,MerchantAT,PogueJ,AnandSS.Waist circumferenceandwaist-to-hipratioaspredictorsof cardiovascularevents:meta-regressionanalysisof prospectivestudies.EurHeartJ2007;28(7):850–6.

[8] FreedmanDS,RimmAA.Therelationofbodyfat distribution,asassessedbysixgirthmeasurements,to

diabetesmellitusinwomen.AmJPublicHealth 1989;79(6):715–20.

[9] PreisSR,MassaroJM,HoffmannU,D’AgostinoSrRB,Levy D,RobinsSJ,etal.Neckcircumferenceasanovelmeasure ofcardiometabolicrisk:theFraminghamHeartstudy.JClin EndocrinolMetab2010;95(8):3701–10.

[10] Ben-NounLL,LaorA.Relationshipbetweenchangesin neckcircumferenceandcardiovascularriskfactors.Exp ClinCardiol2006;11(1):14–20.

[11] ThanassoulisG,MassaroJM,HoffmannU,MahabadiAA, VasanRS,O’DonnellCJ,etal.Prevalence,distribution,and riskfactorcorrelatesofhighpericardialandintrathoracic fatdepotsintheFraminghamheartstudy.CircCardiovasc Imaging2010;3(5):559–66.

[12] FabbriniE,MagkosF,MohammedBS,PietkaT,Abumrad NA,PattersonBW,etal.Intrahepaticfat,notvisceralfat,is linkedwithmetaboliccomplicationsofobesity.ProcNatl AcadSciUSA2009;106(36):15430–5.

[13]RabkinSW.Epicardialfat:properties,functionand relationshiptoobesity.ObesRev2007;8(May(3)):

253–61.

[14] KangSM,YoonJW,AhnHY,KimSY,LeeKH,ShinH,etal.

Androidfatdepotismorecloselyassociatedwithmetabolic syndromethanabdominalvisceralfatinelderlypeople.

PLoSONE2011;6(11):e27694.

[15] MeisingerC,Do¨ringA,ThorandB,HeierM,Lo¨welH.Body fatdistributionandriskoftype2diabetesinthegeneral population:aretheredifferencesbetweenmenand women?TheMONICA/KORAAugsburgcohortstudy. AmJ ClinNutr2006;84(3):483–9.

[16] YooKY,ShinHR,ChangSH,ChoiBY,HongYC,KimHD, etal.GenomicepidemiologycohortsinKorea:presentand thefuture.AsianPacJCancerPrev2005;6(3):238–43.

[17] JangE,KimJY,LeeH,KimH,BaekY,LeeS.Astudyonthe reliabilityofsasangconstitutionalbodytrunk

measurement.EvidBasedComplementAlternatMed2012.

http://dx.doi.org/10.1155/2012/604842.

[18] ShinC,AbbottRD,LeeH,KimJ,KimmK.Prevalenceand correlatesoforthostatichypotensioninmiddle-agedmen andwomeninKorea:theKoreanHealthandGenome Study.JHumHypertens2004;18(10):717–23.

[19] WorldHealthOrganization.Definition,diagnosisand classificationofdiabetesmellitusanditscomplications:

reportofaWHOConsultation.Part1:diagnosisand classificationofdiabetesmellitus.Geneva,Switzerland:

WorldHealthOrganization;1999,Availableat:http://

whqlibdoc.who.int/hq/1999/WHO_NCD_NCS_99.2.pdf [accessed05.12.12].

[20] FlussR,FaraggiD,ReiserB.EstimationoftheYoudenindex anditsassociatedcutoffpoint.BiomJ2005;47(4):458–72.

[21] WangY,RimmEB,StampferMJ,WillettWC,HuFB.

Comparisonofabdominaladiposityandoverallobesityin predictingriskoftype2diabetesamongmen.AmJClin Nutr2005;81(3):555–63.

[22] RexrodeKM,BuringJE,MansonJE.Abdominalandtotal adiposityandriskofcoronaryheartdiseaseinmen.IntJ ObesRelatMetabDisord2001;25(7):1047–56.

[23] TaylorAE,EbrahimS,Ben-ShlomoY,MartinRM,Whincup PH,YarnellJW,etal.Comparisonoftheassociationsof bodymassindexandmeasuresofcentraladiposityandfat masswithcoronaryheartdisease,diabetes,andall-cause mortality:astudyusingdatafrom4UKcohorts.AmJClin Nutr2010;91(3):547–56.

[24] SnijderMB,ZimmetPZ,VisserM,DekkerJM,SeidellJC, ShawJE.Independentandoppositeassociationsofwaist andhipcircumferenceswithdiabetes,hypertensionand dyslipidemia:theAusDiabStudy.IntJObesRelatMetab Disord2004;28(March(3)):402–9.

[25] HeitmannBL,LissnerL.HipHipHurrah!Hipsizeinversely relatedtoheartdiseaseandtotalmortality.ObesRev 2011;12(6):478–81.

[26] StefanN,KantartzisK,MachannJ,SchickF,ThamerC, RittigK,etal.Identificationandcharacterizationof metabolicallybenignobesityinhumans.ArchInternMed 2008;168(15):1609–16.

[27] KantartzisK,MachannJ,SchickF,FritscheA,Ha¨ringHU, StefanN.Theimpactofliverfatvsvisceralfatin determiningcategoriesofprediabetes.Diabetologia 2010;53(5):882–9.

[28] Bosy-WestphalA,BookeCA,Blo¨ckerT,KosselE,GoeleK, LaterW,etal.Measurementsiteforwaistcircumference affectsitsaccuracyasanindexofvisceralandabdominal subcutaneousfatinaCaucasianpopulation.JNutr 2010;140(5):954–61.

[29] ShenW,PunyanityaM,ChenJ,GallagherD,AlbuJ,Pi- SunyerX,etal.Visceraladiposetissue:relationships betweensinglesliceareasatdifferentlocationsand obesity-relatedhealthrisks.IntJObes(Lond) 2007;31(5):763–9.

[30] OkaR,MiuraK,SakuraiM,NakamuraK,YagiK,Miyamoto S,etal.Comparisonofwaistcircumferencewithbodymass

indexforpredictingabdominaladiposetissue.DiabetesRes ClinPract2009;83(1):100–5.

[31] PreisSR,MassaroJM,RobinsSJ,HoffmannU,VasanRS, IrlbeckT,etal.Abdominalsubcutaneousandvisceral adiposetissueandinsulinresistanceintheFramingham heartstudy.Obesity(SilverSpring)2010;18(11):2191–8.

[32] LissnerL,BjorkelundC,HeitmannBL,SeidellJC,Bengtsson C.Largerhipcircumferenceindependentlypredictshealth andlongevityinaSwedishfemalecohort.ObesRes 2001;9:644–6.

[33] SeidellJC,HanTS,FeskensEJ,LeanME.Narrowhipsand broadwaistcircumferencesindependentlycontributeto increasedriskofnoninsulin-dependentdiabetesmellitus.J InternMed1997;242:401–6.

[34]AucouturierJ,MeyerM,ThivelD,TaillardatM,Duche´ P.

Effectofandroidtogynoidfatratiooninsulinresistance inobeseyouth.ArchPediatrAdolescMed

2009;163(9):826–31.

[35] PenaforteFR,JapurCC,Diez-GarciaRW,ChiarelloPG.

Uppertrunkfatassessmentanditsrelationshipwith metabolicandbiochemicalvariablesandbodyfatin polycysticovarysyndrome.JHumNutrDiet2011;24(1):

39–46.