Quantification of the visceral and subcutaneous fat by computed tomography: Interobserver correlation of a single slice technique

ORIGINAL

ARTICLE

/

Gastrointestinal

imaging

Quantification

of

the

visceral

and

subcutaneous

fat

by

computed

tomography:

Interobserver

correlation

of

a

single

slice

technique

D.

Sottier

,

J.-M.

Petit

,

S.

Guiu

,

S.

Hamza

,

H.

Benhamiche

,

P.

Hillon

,

J.-P.

Cercueil

,

D.

Krausé

,

B.

Guiu

a_{DépartementdeRadiodiagnosticetd’ImagerieMédicalediagnostiqueetthérapeutique,CHU} deDijon,2,boulevardMaréchal-de-Lattre-de-Tassigny,BP77908,21079DijonCedex,France b_{Départementd’Endocrinologie,DiabétologieetMaladiesMétaboliques,CHUdeDijon,2,} boulevardMaréchal-de-Lattre-de-Tassigny,BP77908,21079DijonCedex,France

c_{INSERMU866,FacultédeMédecine,Dijon,France}

d_{Départementd’OncologieMédicale,CentreGeorges-Franc¸ois-Leclerc,Dijon,France} e_{Départementd’Hépatologie,CHUdeDijon,2,boulevardMaréchal-de-Lattre-de-Tassigny,BP} 77908,21079DijonCedex,France

KEYWORDS Computed tomography; Subcutaneousfat; Visceralfat; Area; Cancer Abstract

Purpose:Toassesstheinterobserverreproducibilityofthequantificationofthevisceraland

subcutaneousfatbycomputedtomographyfromanumbilicalsliceandstudytheeffectofthe leveloftheslice(slicegoingthroughthenavelversusaslicegoingthroughdiscL3—L4).

Materialsandmethods:Forty-fourbreastcancerpatientswhohadaCT-scanwereincludedin

thisstudy.Thisisadoubleblind(juniorversussenior)retrospectivestudy todeterminethe interobserver reproducibility.A juniorobserverstudiedthe variationbetween two levelsof slicebyselectinganimagegoingthroughL3—L4andthenavel.

Results:Themeasurementofthefatobtainedfromanumbilicalsliceseemedtobewell

cor-related andconsistent with thatobtained from aslicewith a discreference (L3—L4). The interobserverreproducibilityisgoodforthequantificationoftheumbilicalfat(Spearmanand Linat0.9921and0.985[P<0.001]forthevisceralfat).

Conclusion:The interobserverreproducibilityofthesinglesliceCT-scanmeasurementgoing

throughthenavel(easilydetected)isexcellentandmay thereforebeusedinoncologyasa predictivetooltomeasureacharacteristicofthehostandnotthetumor.

©2013Éditionsfrançaisesderadiologie.PublishedbyElsevierMassonSAS.Allrightsreserved.

∗_{Correspondingauthor.DépartementdeRadiodiagnosticetd’ImagerieMédicalediagnostiqueetthérapeutique,CHUdeDijon,2,boulevard} Maréchal-de-Lattre-de-Tassigny,BP77908,21079DijonCedex,France.

E-mailaddress:[email protected](B.Guiu).

2211-5684/$—seefrontmatter©2013Éditionsfrançaisesderadiologie.PublishedbyElsevierMassonSAS.Allrightsreserved.

http://dx.doi.org/10.1016/j.diii.2013.04.006

CORE Metadata, citation and similar papers at core.ac.uk

Obesity, andmore specificallytheaccumulation of vis-ceralfat, is afactor of riskassociated witha greatmany cancers [1,2]. It is also associated with a higher risk of recurrence after treatment [3] and death [4,5]. Visceral fat is even thought to be a factor of risk independent of the development of cancer of the colon and pancreas [6].

More recently, visceral fat has been shown to be the first predictive biomarker of the efficacy of antian-giogenics in cancer of the colon and kidney [7,8]. Its evaluationis therefore of majorimportancein the treat-ment.

Severalstudieshavebeencarriedoutonthemethodology toquantifyandmeasureabdominalfatbycomputed tomo-graphy[9—15].Inparticular,acalculationbasedonasingle sliceareahasbeenshowntobesufficient[16].Insome stud-ies,thenavalwasusedasapointofreference[9,11,13—15] while in others, a bone or disc wasused aspoint of ref-erence[17,18].As farasweareaware,theinterobserver reproducibilityofthecalculationoftheareaofvisceralfat bycomputedtomographyhasneverbeenstudiedinthe lit-erature.However,itisabasicelementinthereliabilityof apredictivemarker.

Thepurposeofthisstudyistoassessthisinterobserver reproducibilitybycomputedtomographyfromanumbilical sliceand studythe effectofthe levelof the sliceonthe quantificationof thefat by comparingthis umbilical area withthatofonegoingthroughdiscL3—L4.

Figure 1. Computed tomography image in axial slice passing throughthenavel showingthe differentstepsinthecalculation ofthetotalfatandsubcutaneousfataftersegmentationofthefat densitypixels(₋₁₉₀ HUto₋₃₀HU)withImage Jsoftware.The visceralfatisobtainedbysubtraction.

Materials

and

methods

Eligible

patients

In astudy ontheantiangiogenictreatment of breast

can-cer,forty-foursuccessivepatientswithhistologically-proven

breast cancer benefited from pre-therapeutic computed

tomographyimagingandweretherebyincludedinthisstudy.

These patients provided their written consent to use the

clinicaldataandimagingwhilerespectingtheiranonymity. This is a double blind retrospective study (junior ver-sus senior) oninterobserver reproducibility. The variation

betweentwoareaswasstudiedbyasinglejuniorobserver

whoselectedan imagegoingthroughL3—L4andan image

throughthenavel.

Measurement

of

the

visceral

and

subcutaneous

fat

The segmentationof thefatwasdeterminedbycomputer

tomography (CT)beforetreatmentontheentire abdomen

in patients placed in decubitus dorsal. The two levels of

slice were selected,at theumbilical level and at L3—L4,

enabling singleslice segmentation of the fat. The images

acquired werethen postprocessedwithImage Jsoftware

(http://rsb.info.nih.gov/ij/).Withthissoftware,itwas pos-sibletomeasurethepixelsindensitiesbetween−190and −30Hounsfieldunits(HU)inordertodefinethefat compart-ments(subcutaneous,visceral)anddefine anareainmm2

foreachofthem(Figs.1and2).

Statistical

analysis

ThemainpurposeofthisstudywastoshowthataCTslice passingthroughthenavelwasareproduciblemethodinthe determinationoftheareaofvisceralfat.

TheslicepassingthroughdiscL3—L4waschosenbecause itrepresentsthelimitoftheupperabdomen,afixedmarker

bydefinition(discmarker).TheupperlevelsfromT12—L2

werenotselectedbecausetheliverisatoobigpartofthe image,therebylimitingthestudyofthevisceralfat.Asto thelowerlevels,asofL5,thisistheareaofsubcutaneous fatwhichishighlyinfluencedbythefatfromthebuttocks.

The navel level has been validated in several studies

[9,11,13—15] andis very easily found during thescrolling oftheaxialslices(asopposedtotheinter-vertebraldiscs). For this reason,this level wasused tostudy the visceral fat.Forcertainauthors,thepositionofthenavelmayvary accordingtothepatient’smorphotype.

Themean,minimumandmaximumvaluesoftheareaof total,subcutaneousandvisceralfatwerecompared.Their coefficient ofcorrelation (Spearman’scoefficient) andthe concordance(Lin’scoefficient)werealsostudied.

Results

Comparative

study

of

two

levels

of

slice

with

different

markers

Thetotalfat(Fig.3),subcutaneousfat(Fig.4)andvisceral fat(Fig.5)aswellastheircorrelation (Spearman’s

coeffi-Figure2. Computedtomographyimagesinaxialslicepassingthroughthenavel(a)andL3—L4disc(b)withresultsaftersegmentationof thefatdensitypixelswithImageJsoftwareanddeterminationofthetotalandsubcutaneousfat.Thevisceralfatisobtainedbysubtraction.

Table1 Areasoftotal,subcutaneousandvisceralfatassessedattheslicepassingthroughthenavelandbyaL3—L4 discreference.

Mean(min—max) Coef.Correl.(Spearman) Coef.Concord.(Lin)

Totalfat Navel 40,059mm2_{(6456—81,498mm}2_{) 0.9215(P<0.0001) 0.903(P<0.001)}

DiscL3—L4 35,786mm2_{(6168—69,841mm}2₎

Subcutaneousfat Navel 27,431mm2_{(5425—64,003mm}2_{) 0.8778(P<0.0001) 0.832(P<0.001)}

DiscL3—L4 23,212mm2_{(5000—47,355mm}2₎

Visceralfat Navel 12,636mm2_{(1031—32,508mm}2_{) 0.9376(P<0.0001) 0.930(P<0.001)}

DiscL3—L4 12,574mm2_{(955—34,293mm}2₎

Coef.:coefficient;Correl.:correlation;min:minimum;max:maximum.

cient)andconcordance(Lin’scoefficient)arepresentedin

Table1.

Themeasurementofthefatdeterminedfroman umbil-icalsliceseemstobewellcorrelatedandwellconcordant withthatdeterminedfromaslicewithaL3—L4discmarker (fixbydefinition).

Figure3. CorrelationbetweentheumbilicalsliceandtheL3—L4 disc(totalfatvs.totalfat).

Study

of

the

interobserver

reproducibility

The total fat (Fig. 6), subcutaneous fat (Fig. 7) and vis-ceralfat(Fig.8)obtainedbyeachobserveraswellasthe studyoftheircorrelation(Spearman’scoefficient)andtheir concordance(Lin’scoefficient)arepresentedinTable2.

Figure4. CorrelationbetweentheumbilicalsliceandtheL3—L4 disc(subcutaneousfatvs.subcutaneousfat).

Figure5. CorrelationbetweentheumbilicalsliceandtheL3—L4 disc(visceralfatvs.visceralfat).

Figure6. Interobservercorrelationfortheumbilicalslice(total fatvs.totalfat).

The interobserverreproducibility is goodfor the quan-tificationofthefatatthenavellevel.

Onlyonepointonthesedifferentcurvesdoesnotappear

to be correlated and concordant. This involves the

un-blinding of one patient for which the observers analyzed

twoCT-scansobtainedatdifferentdates.

Figure7. Interobservercorrelationfortheumbilicalslice (sub-cutaneousfatvs.subcutaneousfat).

Figure8. Interobservercorrelationfor theumbilicalslice (vis-ceralfatvs.visceralfat).

Discussion

Obesity is now considered to be a factor of risk for a

greatmany cancers. The increasedincidencemay in part

accountfortheirincrease[2].Abdominalfatand,in partic-ular,thevisceralcontingentcontain‘‘metabolicallyactive’’ adipocytes that secrete pre-angiogenic and proliferative adipokins(proteinssecretedby theadiposetissue).These endocrineandparacrinesecretionsinpartaccountforthis obesity-relatedincreasedriskofcancer[6].

Table2 Interobserverreproducibilityofthetotal,subcutaneousandvisceralfatassessedinaslicepassingthroughthe navel.

Mean(min—max) Coef.Correl.(Spearman) Coef.Concord.(Lin)

Totalfat Observer1 40,059mm2_{(6456—81,498mm}2_{) 0.9726(P<0.0001) 0.981(P<0.001)}

Observer2 39,436mm2_{(5828—82,004mm}2₎

Subcutaneousfat Observer1 27,431mm2_{(5425—64,003mm}2_{) 0.9535(P<0.0001) 0.971(P<0.001)}

Observer2 27,067mm2_{(4786—65,358mm}2₎

Visceralfat Observer1 12,636mm2_{(1031—32,508mm}2_{) 0.9921(P<0.0001) 0.985(P<0.001)}

Observer2 12,369mm2_{(1042—30,083mm}2₎

Abdominalfatmaybequantifiedinseveralways: mea-surementoftheBMI,anthropometry(littleused)orimaging. TheuseoftheBMIisnotappropriatebecauseahighfigure isnotnecessarilyassociatedwithanincreaseinvisceralfat

[10]. Infact, the different anthropometricmeasurements (measurement of the circumference of the hip, waist or abdominalsagittaldiameter)arenotreliable[10,15,19]and often confuse visceral fat with subcutaneous fat. In real-ity,onlyimagingallowsboth compartmentstobestudied: subcutaneousandvisceralfat.

Sonography,aneasilyaccessibletechnique,canbeused tostudythevisceralfatquicklyineverydayclinicalpractice [10]. Nevertheless, it is not reproducible since it is very operator dependent. These measurementsare difficult to obtainandnotveryreliable[15]sinceitisdifficulttoobtain thesameplaneofsliceinallpatients,inparticularinobese patientswherethesonographyexaminationisgenerallyvery limited.

Computed tomography is the imaging technique most oftenusedsince,althoughirradiating,itisveryaccessible andreliable.MRimagingmayalsobeused[20]andhasthe advantagethatitisnotirradiating.Nevertheless,MRImay presentseveraldisadvantageswhencomparedwiththe CT-scan: thecost(moreexpensive thantheCT),accessibility andtechnicalobligations(specific sequencestolimit non-homogeneityoffieldforthesegmentationofthefat).The lastdisadvantageismajorsinceitisnotpossibletoquantity tovisceralfatretrospectivelyinMRI.Infact,quantification isonlypossiblewithsequencesacquiredinaspecificmanner. Severalstudies, especiallythoseabout therelationship between visceralfat and the metabolicsyndrome, or vis-ceral fat and factors of cardiovascular risk [9,11,13—15], havestudiedthequantificationandmeasurementof abdom-inalfatbyCT-scaninordertodeterminethemostreliable method(levelofslice,volumeacquisition...).

Inthesedifferentstudies,thelevelofslicewasdebated. Theonemostoftenusedwasintheumbilicalsituation, vary-ingfromoneindividualtoanother,butingenerallocatedat L4—L5.Agreatmanyotherlevelswerestudied,inparticular thosepassingthroughtheinter-vertebraldiscsfromL1—L5. TheL4—L5levelwasnotfoundtobemostrepresentativeof theriskofobesityorcardiovascularrisk[12].

Measurement ofthevisceralfat onatargetsingleslice wasfoundtobeequalthatcarriedoutonseveralslices dur-ingvolumetricCTacquisitionsasregardstheobesity-related risk [16]. Certain authors have shown that the measure-mentofvisceralandsubcutaneousfat,carriedoutatL4—L5, L4—L5+5cmorL3—L4differverylittle[18].Wewantedto specifically study the umbilical level as it is very easy to detect on a CT-scan and therefore very easy in everyday practice,bycomparingitwithareputedlylessvariablelevel withadiscreference.

Inaddition,theinterobserverreproducibilityisan impor-tant aspect in the reliability of a predictive marker. However,asfarasweareaware,theinterobserver repro-ducibilityoftheareaofvisceralfathasneverbeenstudied intheliterature.

Ourworkrevealedthatthemeasurementofthefatbased onanumbilicalslicewaswellcorrelatedandwell concord-antwiththatbasedonaslicewithaL3—L4reference(fix by definition) andthat thereis a verygood interobserver correlationforthevisceralfat.

The interobserver correlation, as regardsthe subcuta-neousfat(Spearmanat0.9535withP<0.0001)andtotalfat (Spearmanat 0.9726withP<0.001),does notseem tobe asgoodevenifthepowerofthestudydidnotallowfora statisticalcomparison.Thismaybeduetothepresenceof intramuscularfat observed in case ofsarcopenia. Infact, thesubcutaneous(andtotal)fatcompartmentmayinclude thesefatty muscularpixelsthat theoreticallyrequire spe-cificoutlines(exclusions). These specificoutlines,related to the subcutaneous compartment (and therefore total) areable toaccount for the differences between the two observers.

Animportantpointinthediscussionistheexistenceof avariationinfatovertimeandduringthediseaseandthe treatmentthatmayaffectthedistributionandquantityof fat.Infact,duringourstudy,onepatientpresented consid-erable variationsin fat (subcutaneous, visceral andtotal) betweenthetwoobservers.AfterrereadingtheCTimaging (doubleblind study),we notedthat thetwoexaminations carriedoutatdifferentdateswerepost-treatment.

Rapidvariationsinweightgenerallyaccountforachange inthe quantity of subcutaneous fat.The evolutionof the neoplasticdisease often induces a change in the general state,accountingforaweightlosswithchangesinthefat compartments.The sameis trueofthetoxicityofcertain treatments.Therefore,itisimportanttostudythevisceral fatonthebaselineCT-scan,beforetreatment,ifwewant touseitasapredictivemarker.

TheCT-scanseemstobeaverygoodtechniqueforthe segmentationof theabdominalfatwithanumbilical slice asthishas severaladvantages: savings in time(the navel is easily detected on an axial slice), reproducibility in a standardizedactivity,accessibility, andverylow exposure toradiation(onlyonesliceisnecessary)which,considering theconstantincreaseinthenumberofCT-scansincurrent practiceandthenumberofpatientssufferingfromatumoral disease,isacrucialparameter.

Thelimitsofourstudyconsistofarelativelylownumber ofpatientsincludedandthelimitedcomparisonatonlytwo levelsofslice,evenifthesetwolevelsaremostoftenused intheliterature.As faraswe areaware,thisstudyisthe firsttoassess theinterobserverreproducibility ofasingle slicetechnique.

Conclusion

Bywayofconclusion,thisworkshowsthatthesingleslice

CTmeasurementtechniquedeterminedfromaslicepassing

throughaneasilydetectedlevel(thenavel)hasanexcellent interobserverreproducibilityandmaythereforebeusedin oncologyasapredictivetooltomeasureacharacteristicof thehostandnotthetumor.

Disclosure

of

interest

Theauthorsdeclarethattheyhavenoconflictsofinterest concerningthisarticle.

References

[1]BergstromA,HsiehCC,LindbladP,LuCM,CookNR,WolkA. Obesity and renalcell cancer —aquantitativereview. Br J Cancer2001;85(7):984—90.

[2]BianchiniF,KaaksR,VainioH.Overweight,obesity,andcancer risk.LancetOncol2002;3(9):565—74.

[3]OhkiT,TateishiR,ShiinaS,GotoE,SatoT,NakagawaH,etal. Visceral fat accumulation is an independent risk factor for hepatocellularcarcinomarecurrenceaftercurativetreatment inpatientswithsuspectedNASH.Gut2009;58(6):839—44.

[4]Calle EE, Rodriguez C, Walker-Thurmond K, Thun MJ. Overweight,obesity,andmortalityfromcancerina prospec-tively studied cohort of U.S. adults. N Engl J Med 2003;348(17):1625—38.

[5]RenehanAG,TysonM,EggerM,HellerRF,ZwahlenM. Body-mass index and incidence of cancer: a systematic review andmeta-analysisofprospectiveobservationalstudies.Lancet 2008;371(9612):569—78.

[6]SchliengerJL,LucaF,VinzioS,PradignacA.Obesityandcancer. RevMedInterne2009;30(9):776—82.

[7]GuiuB,PetitJM,BonnetainF,LadoireS,GuiuS,CercueilJP, etal.Visceralfatareaisanindependentpredictivebiomarker of outcome afterfirst-linebevacizumab-based treatmentin metastaticcolorectalcancer.Gut2010;59(3):341—7.

[8]Ladoire S, Bonnetain F, Gauthier M, Zanetta S, Petit JM, GuiuS, et al. Visceral fat area asa new independent pre-dictive factor of survival in patients with metastatic renal cellcarcinomatreatedwithantiangiogenicagents.Oncologist 2011;16(1):71—81.

[9]BorkanGA,GerzofSG,RobbinsAH,HultsDE,SilbertCK, Sil-bertJE. Assessmentof abdominalfat content bycomputed tomography.AmJClinNutr1982;36(1):172—7.

[10]Hirooka M, Kumagi T, Kurose K, Nakanishi S, Michitaka K, Matsuura B,et al.Atechnique for themeasurementof vis-ceral fat by ultrasonography: comparison of measurements by ultrasonography and computed tomography. Intern Med 2005;44(8):794—9.

[11]KobayashiJ,TadokoroN,WatanabeM,ShinomiyaM.Anovel methodofmeasuringintra-abdominalfatvolumeusing heli-cal computed tomography. Int J Obes Relat Metab Disord 2002;26(3):398—402.

[12]ShenW,PunyanityaM,ChenJ,GallagherD,AlbuJ,Pi-Sunyer X,etal.Visceraladiposetissue:relationshipsbetweensingle sliceareasat differentlocations and obesity-relatedhealth risks.IntJObes(Lond)2007;31(5):763—9.

[13]TairaK,HikitaM,KobayashiJ,BujoH,TakahashiK,MuranoS, etal.Delayedpost-prandiallipidmetabolisminsubjectswith intra-abdominalvisceral fat accumulation.Eur JClinInvest 1999;29(4):301—8.

[14]Tokunaga K, Matsuzawa Y, Ishikawa K, Tarui S. A novel technique for the determination of body fat by computed tomography.IntJObes1983;7(5):437—45.

[15]Yoshizumi T, Nakamura T, Yamane M, Islam AH, Menju M, YamasakiK,etal.Abdominalfat:standardizedtechniquefor measurementatCT.Radiology1999;211(1):283—6.

[16]ShenW,PunyanityaM,WangZ,GallagherD,St-OngeMP,Albu J, et al. Visceral adipose tissue: relations between single-slice areas and total volume. Am J Clin Nutr 2004;80(2): 271—8.

[17]Kuk JL, Church TS, Blair SN, Ross R. Does measurement site for visceraland abdominalsubcutaneous adiposetissue alterassociationswiththemetabolicsyndrome?DiabetesCare 2006;29(3):679—84.

[18]LeeS,JanssenI,RossR.Interindividualvariationinabdominal subcutaneousand visceraladipose tissue:influence of mea-surementsite.JApplPhysiol2004;97(3):948—54.

[19]Kvist H, Chowdhury B, Grangard U, Tylen U, Sjostrom L. Totalandvisceraladipose-tissuevolumesderivedfrom mea-surements with computed tomography in adult men and women: predictive equations. Am J Clin Nutr 1988;48(6): 1351—61.

[20]CarlierRY,DeTruchisP,RonzeS,MompointD,ValleeC,Melchior JC.MRIofintra-abdominalfatandHIV-associated lipodystro-phy:acasereview.JRadiol2007;88(7—8Pt1):947—56.