Overweight and fat distribution Associationswithaspectsof morbidity /<=,/!/ '--/ ^o c\$i Promotor: dr. J. G. A.J. Hautvast hoogleraar indeleervandevoedingende voedselbereiding Co-promotor: dr. P.Deurenberg universitair hoofddocent J. C. Seidell Overweight and fat distributionAssociations with aspects of morbidity Proefschrift Terverkrijging vandegraadvan doctor inde landbouwwetenschappen, opgezagvanderector magnificus, dr. C. C. Oosterlee, inhetopenbaar te verdedigen opvrijdag 3oktober 1986 desnamiddagstevieruur indeaula vande Landbouwuniversiteit te Wageningen \W.HutQH Aanmijn ouders en Ellen Financial support bythe Dutch Preventiefonds throughout the study and theNetherlands Heart Foundation for thepublication thisthesisisgratefully acknowledged. I'. V...IOTHEEK LA1TOB0TJWHOGKSCHOOL WAGENINGKN Ur/072^ STELLINGEN l.Bijherzieningvanhetadviesinzakeadipositasvande gezondheidsraad (1984)verdienendevolgendepuntenextra aandachtindeaanbevelingen: -Hetbelangvanpreventievemaatregelenvoorovergewicht. -Hetbelangvandevetverdelinginderelatieovergewichten morbiditeit. - Leeftijdsspecifieke grenswaardenvoorovergewicht. 2.Deweegschaal iseenbelangrijk instrumentvoordehuisarts, eenmeetlintismogelijknogbelangrijker. 3.De verschillen inpsychosociale problematiek rond overgewicht tussenmannenenvrouwengevenaandat'gewichtsbeleving'een onderwerpiswatdringend emancipatiebehoeft. 4.De vetzuursamenstellingvanvetweefseluithetomentumzal eerderveranderingen indevetzuursamenstellingvande voeding reflecterendandievanvetweefseluithetfemorale engluteale gebied. 5.Hetfeitdatbijhetouderworden relatiefmeervetinde buikholtewordt opgeslagenendeopvattingdatjuistdit intra-abdominalevetweefsel ongunstige eigenschappen heeft zijnintegenspraakmetdebevindingdathet gezondheidsrisicobijeenbepaaldematevanovergewicht afneemtbijtoenemendeleeftijd. 6.Correlationsontheaggregate levelofnationsdonotprove anything;theironlypurposeinresearchistostimulatethe generationofhypotheses F.deWaard.NutritionandCancer1986;8:5-8. 7.Hetzouiniedersbelang zijnwanneerdeadviezenvande Gezondheidsraad,hoedeskundigzeookopgesteldmogenzijn, voorpublicatieaanmultidisciplinairepanels zoudenworden voorgelegd. 8.Hetalgemeengebruikvanmultipele regressie vergelijkingen bij studiesnaardedeterminantenvanhart-envaatziekten waarbij,naasteenmaatvoor obesitas,ookdeandere bekende risicofactorenvoordeeindpuntenwordenbetrokken,geveneen sterkeonderschattingvanhetgezondheidsrisico behorendbij eenbepaaldematevanovergewicht. 9.Hetverouderenvaneenpopulatie gaatgepaardmeteen toenemende heterogeniteitvanfysiologische variabelen.Het opstellenvanvoedingsrichtlijnen voorouderenzalgebaseerd moetenzijnopdeze heterogeniteit. BI!-.I.IOTHEEK IANDBOUWKOG* •'HOOl WAGENINGEN /o^V w 10.Andthenthere issignificance -aword that isprobably the single greatest intellectual pathogen inbothbiological and statistical domainstoday. FeinsteinA.R.'Clinicalepidemiology'.Philadelphia,W.B. Saunders1981. 11.Theissuesofpreventivemedicine have little todowith science, relative risks,and risk factors.Theycouldbemore profitablydebatedwithin the framework towhich theybelong - ethics,politics,andvested interests. Skrabanek P.Lancet1986;i:143-4. 12.Een 'population strategy'eneen 'high risk strategy' voor hetgunstigbeinvloedenvan risicofactoren zijngeen elkaar uitsluitende alternatievenmaardienenbeiden tegelijkertijd teworden toegepast. zieook:LewisB,MannJ.I.,Mancini M.Lancet1986;i:956-9. 13.Hetschrijvenvaneenonderzoeksvoorstel zouopgenomenmoeten zijnalsonderdeel vaneenverplicht onderwijselementinhet studiecurriculum 'Voedingvandemens'. 14.Diegenendievanvermagereneenafval-racemakenlopeneen grote kansalseerstenaantekomen. Stellingen behorend bijhetproefschriftvanJaapC.Seidell. Overweight and fatdistribution -associationswithaspectsof morbidity. Wageningen, 3oktober 1986. CONTENTS Abstract Voorwoord Chapter 1.Obesityand fatdistribution inrelationtohealthcurrent insightsand recommendations (WrldRevNutrDiet1986 (inpress)) 1 Chapter 2.Associationsofmoderateandsevereoverweightwith self reported illnessandmedical careinDutchadults (AmJPublicHealth1986;76:264-9) 38 Chapter 3.Utilizationofprimaryhealthcareofoverweightand non-overweight subjects-a sixyear follow-upstudy (submitted) 56 Chapter 4.Overweightandchronic illness-a restrospective cohortstudywitha follow-upof6-17 years,inmen andwomenof initially20-50yearsofage (JChronDis1986 (inpress)) 72 Chapter 5.The relationbetweenoverweightandsubjectivehealth - withattentiontotheeffectsofage,socialclass, slimmingbehaviorand smokinghabitsonthis relation (AmJPublicHealth1986;76: (inpress)) 88 Chapter 6. Fatdistributionofoverweightpersons inrelationto morbidityand subjectivehealth (IntJObesity 1985;9:363-74) 104 Chapter 7.Assessmentof intra-abdominal fat- relationbetween anthropometryandcomputed tomography (AmJClinNutr (inpress)) GeneralDiscussion 123 135 Samenvatting 139 Summary 143 CurriculumVitae 146 VOORWOORD Inditproefschriftwordenenkeleonderzoekenbeschrevennaardeassofciaties tussenovergewichtenparametersvandegezondheidstoestand.HoeweldeVakgroep HumaneVoedingvandeLandbouwHogeschool teWageningendethuisbasiswasvoor deonderzoeken,werdhetgrootstedeelvanhetwerkuitgevoerd insamenwerking methetCentraalBureauvoordeStatistiek,hetNijmeegsUniversitairHuisartsen InstituutenhetInstituutvoorRadiodiagnostiekvandeKatholiekeUniversiteit teNijmegen. Indegenoemdeinstituten ishetleeuwendeelvande onderzoeksgegevens verzameld.Hetisnatuurlijk onmogelijkallebetrokkenmedewerkersvandeze institutenmetnametenoemen.Eenaantalwaar ik rechtstreeksmeeheb samengewerktheb,wilikhieronderbedankenvoordeplezierigewijzewaaropde samenwerkingsteedsisverlopen.Nietalleenbijhetverzamelenofhetter beschikking stellenvandegegevensbenikdoorvelengeholpenmaarookbijhet verwerkenvandiegegevens,de interpretatievande resultatenenhet schrijven vandeartikelen zijnvelenbetrokkengeweest. Ineerste instantiegaatmijndankuitnaarmijnpromotorProf.dr.J.G.A.J. Hautvastenmijnco-referentdr.P.Deurenbergdiebeideneenvoortduijende stimulanswarenvoorhetonderzoek enmijsteedsopuitstekendewijzehebben begeleid.Hetvertrouwendatzijinmijsteldenhebiksteedszeergewaardeerd. DepraktischeentheoretischemedischekennisvanKarelBakxenArieOosterlee wasnietalleenonontbeerlijkvoorhetonderzoekmaarvooraldeenthousiaste wijzewaaropzijdezekenniswistenovertebrengenmaaktede samenwerkingmet hen steedsextraprettigenleerzaam. JanBuremadie zowelinwetenschappelijk alsmuzikaalopzichtdebassocontinuo was inhetONNO-projectdank ikvoordenauwgezetheidwaarophij statistische adviezengafenmanuscriptencorrigeerde. Voordevele inspirerendediscussiesenmuzikaleavondurenwil ikMattiRookus enWijavanStaverendanken.OokdevelegesprekkenmetJanWeststrateenhet kritischcommentaarvanMartijnKatanwarenvanbelang. Hetmeestetypewerkvoorditproefschrift isopvoortreffelijkewijzeverzorgd doorWilKnuiman-HijlenAnnemarie Zijlmans.DeEngelsetekstenwerden gecorrigeerddoorSeamusWardenJ.Burrough.Demedewerkersvandetekenkamer vanhetBiotechnion stondengarantvoorgrafischwerkvanhogekwaliteit. DeonnavolgbaremanierwaaropBenScholteenGebcaVelemacomputerproblemenin hetalgemeenendievanmijinhetbijzonderwistenoptelossendientteworden vermeldevenalsdewijzewaaropLisettedeGrootdeC.B.S.bestandenwistuitte pluizen. Dedoctoraalstudentendieikhebbegeleid:EllydeBoer,Gerdavan'tBos,Hans Groenendijk,GerdaDoornbos,LeoHenskensenRitaSchoutenwilikbedankenvoor nunbelangrijkebijdragenaanallerleionderdelenvanverschillendeonderzoeken. Ookwilikgraagallenognietgenoemdevasteentijdelijkemedewerkersvande VakgroepHumaneVoedingbedankenvoordefijnesfeerwaarinikhebkunnen werken.Specialedanknogvoordemedewerkersvanhetsecretariaatendeheer Middelburgdie,wanneererpractischeproblemenwaren,altijdtijdvrijmaakten ommedaarmeetehelpen. VandevelemedewerkersvanhetNijmeegsUniversitairHuisartsenInstituutwil ikvooralProf.dr.F.J.A.Huygen,HenkvandenHoogenenPierreHoppener bedanken.Zijzijnhetgeweestdiehetmogelijkhebbengemaaktomgegevensvan devoorepidemiologischonderzoekzeerwaardevolleContinueMorbiditeits Registratietegebruiken.Debijdezeregistratieaangeslotenartsen,ennun assistentes,hebbenveelwerkvoormijverzet.DorienKaaijkassisteerdebijde anthropometricineenvandepraktijken. VandemedewerkersvandeafdelingGezondheidsstatistiekenvanhetCentraal BureauvoordeStatistiekwilikiniedergevalJanvanSonsbeekenJaapvande BergbedankenvoordehulpbijhetverwerkenvandeGezondheidsenquetes. ProfessorJ.H.Ruijsendrs.MartinThijssenbegeleiddenmijbijde werkzaamhedenmetdeCT-scanner.Deartsenenlaborantenvandeafdeling Rontgen-diagnostiekwarenbijdepraktischeuitvoeringeenenormesteun. Prof.dr.W.P.T.James(Edinburgh)benikerkentelijkvoordeinspiratiebijhet opzettenvanhetonderzoek.Dediverseledenvanbegeleidingscommisieswilik dankenvoornuninteresseinmijnwerkzaamheden.Naasthetwetenschappelijkwerk werdiksteedsgemotiveerddoormijnouders,mijnbroerRoel(dieookdeomslag vanhetproefschriftverzorgde),enzusLydia. TenslottewilikEllenbedankenvoordesteundoordikendun. OVERWEIGHTAND FATDISTRIBUTION -ASSOCIATIONSWITHASPECTSOF MORBIDITY THESIS,DEPARTMENTOFHUNANNUTRITION,AGRICULTURAL UNIVERSITY,WAGENINGEN, THE NETHERLANDS, 3OKTOBER 1986. J.C. Seidell ABSTRACT This thesis reportsontheassociation betweenestimatesof theamount;aftd distribution of fatmasswithaspects ofmorbidity inDutchadults.A literature reviewonthe current insights intotheseassociations isincluded.The results of several cross-sectional and retrospective cohort studiesarepresented. The prevalence of severeoverweight,defined asaBodyMass Index (BMI=kg/m )higher than 30kg/m ,isabout 6% inwomenand 4%inmen,while about_34%of themen and 24%of thewomenaremoderately overweight (BMI25-30kg/m ).Among overweight persons,especiallywhen severelyoverweight,theprevalence and incidence ofvarious chronic disorders anduseofaspects ofmedical cairewas higher than innon-overweight persons.Forgoutandarteriosclerotic disease, overweight appeared tobea risk factoratlower levelsofBMI inmen than in women.The.incidencewasparticularly high inmenwithan initial BMI between 27-30kg/m .Theassociation betweenBMIand subjective healthwasmore pronounced inwomen than inmen.Thiswastrue for certain somatic aswell as psychosomatic complaints. Fatdistribution,asmeasuredwith thewaist-hip circumference ratio or waist-thigh circumference ratio,was showntobe related totheprevalence of certain chronicdisorders inmenandwomen.Theseassociationswere independent ofageandBMI.A study inwhichComputed Tomography scans,takenatthe level of theL4vertebra,were related toanthropometric measurements, revealed that correlations of circumference ratioswith theamountof intra-abdominal fatwere higher thanwith subcutaneous abdominal fat. Itisconcluded thatoverweight isrelated toaspects ofmorbidity butthat it maybeparticularly relevant toinclude fatdistribution measurements ihthe evaluation ofhealth risks inoverweight subjects. -1- CHAPTER1 CBESITYAND FATDISTRIBUTION INRELATIONTOHEALTH CURRENT INSIGHTSAND RECOMMENDATIONS J.C.Seidell,P.Deurenberg,J.G.A.J.Hautvast WrldRevNutrDiet (inpress) GENERAL INTRODUCTION Obesity researchhas receivedmuchattentionduring thepastdecades,during whichnewevidence foraetiologic andprognostic aspectsofobesityand underlying mechanismshasbeenaccumulating. The recommendations for prevention and therapy,however,havenot changeddramatically. InFebruary 1985,a consensuspanelattheNational Institute ofHealth inBethesda,Maryland, discussed theavailable knowledge onthehealth consequences ofobesity and formulated recommendations for thetreatment ofobesity.Someaspectsof the evidence that formed thebasis forthese recommendations arediscussed inthe firstpartof this review. The consensuspanel concluded that recentevidence strongly suggests that the regionaldistribution ofbody fatisan importantdeterminant for the occurrence ofclinical correlates ofobesity suchascardiovascular disease,strokeand diabetes.The current knowledge oftheassessment and consequences of typesof fatdistribution isdiscussed inthe second partofthis review. Itcanbe expected that,inthenear future,measuresof fatdistributionwill be included in recommendations forthediagnosisand treatment ofhuman obesity. 1.OBESITY,CURRENT RECOMMENDATIONS 1.1.Definition andassessment ofoverweight andobesity;agreement onan international standard Bray [25]haspointed outthat itisimportant todistinguish between overweight andobesity.Overweight isdefined asan increase inbodyweight above some arbitrary standarddefined in relation tohealth.Obesity isdefined asan abnormally highproportion ofbody fat.Theterm 'abnormal'indicates that obesity corresponds totheupper end of rangeof thebody fat percentages observedinaparticularpopulation.Overweightandobesitycanbedistinguished onlyiftheamountofbodyfatcanbemeasured.Themethodsforassessingbody fatcanbedividedintodirectandindirectmethods.Thesemethodshaverecently beenwellreviewedbyGarrow [51].Table1givesasummaryofmethodscurrently inuse.Theyareofvaryingaccuracyanditisdifficulttopinpointoneofthe indirectmethodsasthemostvalidone. Table1.Quantitativemethodsforthedeterminationofbodyfatinhuman (adaptedfromBray1985)[25] 1)Directcarcassanalyis 2)Bodydensityandbodyvolume(byunderwaterweighing;plethysmography) 3)Isotopeorchemicaldilution -Amountofbodyfat,bymeasuringthedistributionofcyclopropaneor Krypton -Amountofleanbodymass,bymeasurementofPotassium(40K)inthebody -Amountofbodywater,bymeasuringthedistributionoftritiatedwater (3H20)orantipyrine 4)Anthropometricmeasurements -Heightandweighttablesorindices(sometimesincludingmeasuresofframe size) -Skinfoldthickness(usuallytriceps,biceps,supra-iliac,subscapular skinfolds) -Circumferencemeasurements(nowoftenusedfortheassessmentofbodyfat distribution,usuallythigh,hipsandwaistcircumferences) 5)Electricalconductivityofthebody -Totalbodyelectricalconductivity(TOBEC) -Bioelectricimpedance 6)Imagingtechniques -Ultra-soundwaves -Computerizedtomography (CT)ornuclearmagneticresonance(NMR) ThedifficultyofvalidatingthemethodshasrecentlybeendiscussedbyGarrow [51].Anewmethodofdeterminingbodyfatcontent,whichallowsspecification ofbodyfatdistributionoverdifferentbodyfatdepotsiscomputedtonography. Thismethodwillbediscussedinmoredetailinsection2.1.3ofthisreview. Thismethodcanprobablyberegardedasthebestindirectmethodavailable,but itinvolves radiationofsubjects.NuclearMagneticResonance imaging (NMR)may beagoodalternative tocomputed tomography [43]butthis requires further investigation. Currently,weight-height indicesarethemostwidelyusedmeasuresofadiposity andoverweight.Theyhavetheadvantageofeasycalculationandcanbeused in epidemiologic studiesonlargegroupsof individuals.Theysuffer,however,from misclassificationbias.Weight-height indicesareconstructed onthebasisof threecriteria. a)Zerocorrelationwithbodyheight. b)Highcorrelationwithbodyfatness. c)Correlationwithmortality. Thesecriteriawillnowbediscussed. a)Onthebasisofthecriterionofnocorrelationwithheight,Benn [11] proposedweight/height"asanobesity index,inwhichp isapopulationspecificvalue,obtainedbyminimizing thecorrelationofthe indexwith height [11].Garrowargued thatadisadvantage ofthisindex isthatitcan beapplied toindividualsonlyafter thecorrectvalue forphasbeen determined intheappropriatepopulation [51].Anotherdisadvantage isthat results fromdifferentpopulationsaredifficult tocompareand to interpret [49].Quetelet [81]observed in1869thatindividuals'weightswere roughly proportional toheightsquared.Since thentheindexweight/height hasbeen 2 showntosatisfy thecriterionoflowcorrelationwithheight.Useofa fixed pvalue (i.e.2)doesnothavethedisadvantagesoftheBennIndex. b)Although thecorrelationofQuetelet'sIndex,orthe 'BodyMass Index',with thepercentageofbodyfat (asdeterminedby,forinstance,under-water weighing)isgenerally foundnottobeveryhigh (around0.7 - 0.8),other indices rarely showhigher correlations [49].GarrowandWebster [51a] criticallyexamined the relationofQuetelet'sindextobodyfatand demonstrated that threestandardmethods (underwaterweighing,water dilutionorpotassium counting),yield considerablydifferent resultsand,in addition,showed thatQuetelet'sIndexshouldnotbe related topercentage body fat,because itisameasureoftheamountoffat (kg).Indiceswhich takemeasuresof framesize intoaccount,shownomeaningful improvement in estimatingbody fatness inyoungadults [10,86].Theimportanceof frame size inotheragegroups requiresfurther investigation[90]. c)On thebasisofanalysisofexistingmortalitydata,Garrow [50]proposed a classificationofobesityonthebasisofQuetelet's Index (Table2 ) . Table2.Garrow'sclassificationofobesityincategoriesofBodyMassIndexor Quetelet'sIndex(kg/m)[50] BMI(kg/m) InterpretationofBMIin category <20 Underweight Grade0obesity 20-24.9 Acceptableweight Grade1obesity 25-29.9 Moderateoverweight Grade2obesity 30-39.9 Severeoverweight,obese Grade3obesity >,40 Morbidlyobese Someinvestigatorshavepreferredtouse27kg/m asacut-offpointfor obesityinmen[65,98].AsQuetelet'sIndexhasbeenshowntobesomewhat inferiortoBenn'sIndexwith1.5asvalueforpinwomen,someinvestigators 15 haveusedweight/height" forwomen [97].Tomakecomparisonsbetweenmenand womenmoreeasytheytransformedthedifferentindicesintostandardscores.The advantagesofthisapproachremaintobeproven. Sofar,Quetelet'sIndexhasbeenacceptedbyexpertcommitteesasthemost usefulmeasureofobesitywhenonlyheightandweightdataareavailable[29, 53,87,91]. OtherindicesasBroca'sindexortheponderalindexarenow generallyconsideredtobelesssuitableindicesofbodyfatness. 1.2.Prevalenceofoverweightandobesityinindustrialcountries Table3.Prevalenceofmoderateandsevereoverweight(obesity)insome industrializedcountries(inpercentages) Quetelet'sIndex UnitedKingdom TheNetherlands UnitedStates Australia (kg/m2) men women men women men women menwomen moderateoverweight;25-29.9 34 24 34 24 31 24 12 12 severe overweight:>,30 34; 24 Inthe fourcountries shownintable 3,moderateoverweightappearstobemore common inmenthaninwomen.InEuropeancountries,obesity isslightlymore common inwomenthaninmen.Obesityappearstobemoreprevalent intheUnited States inbothmenandwomenthaninanyoftheother industrial countries in the table 1.3.Body fatnessandmortality-epidemiological evidence 1.3.1.Epidemiological evidence Oneof themost comprehensive reviewsonobesityandmortalitywaspublished by SimopoulosandVan Itallie in1984 [90].Their reviewisbasedonthe conclusionsofaworkshoponbodyweight,healthand longevityheld inBethesda, Maryland,in1982.They reviewedanumber of largepopulation studies inwhich a relationshipbetweenoverweightandmortalityhadbeendemonstrated.Some characteristics ofthese studies,towhichweadda recentlypublished population study fromNorway [107],are shownintable4. Itistypical forallthese studiesthatlargepopulationshad tobe followed up foratleast10yearsbeforeanyeffectofbodyweightonmortalitywas detectable.Bjorntorp [15]hasproposed thatthedelayedeffectmaybeexplained bythepossibility thatexcessmortality (especially from cardiovascular disease)ispresentmainly inasubgroupoftheobesepopulation.Lumpingall typesofobesityintoonecategorymaycausestatisticaldilutionofaneffect ofbody fatnessonmortality inasubpopulationoftheobese.Another featureof the studies summarized intable 4istheJ-shaped relationbetweenBodyMass Indexandmortality.ExcessmortalityatbothendsoftherangeofBodyMass Indexareevaluated insection1.3.2. Smoking appears tobean importantconfounding factor inthe relationship betweenbodyfatnessandmortality,andneedstobetakenintoaccountassessing the riskofoverweightandobesity. Theavailableepidemiological evidencehasledtotheconclusion that 'overweightpersons (i.e.personswhoseweightsareaboveaverage)tendtodie younger thanaverageweightpersons,particularly thosewhoareoverweightat younger ages' [90].Similar conclusionsweredrawnby theRoyalCollegeof Physicians in1983 [87). It remainstobedetermined,however,whetherpersonswhoaremoderately overweightbutnotobesemightbenefitfromweight reductionwhenthere isno indicationofabnormalities (e.g.hypertension,impairedglucosetolerance, -6- Table 4.Someaspects ofstudiesdemonstratingarelationshipbetweenanobesity indexandmortality (adapted fromSimopoulosand VanItallie [90] Typeofstudies Mathematical Weightwith Sexandage Average/ Remarkson Follow-up relation minimum difference max studypopulation after which mortality of relation follow- BMIand up(yrs) an increased risk became mortality apparent(yrs Life insurance statistics 4.2 -BuildStudy 1979 J-shaped weightbelow million average moreapparent 6.6/22 inmenage15- Policy-holders, mainlymale,middle-*- 39thanin class,ofapparently age40-69 - Providentmutual weightbelow lifestudy goodhealth onlymen /34 Malepolicy holders average Prospective studies 750000 -AmericanCancer J-shaped 10-20%belowof Society 5209 -FraminghamHeart J-shaped Study 3983 -Manitoba Study 6/14 Predominantly middl* averageweight classhealthyvolunteers (non-smokers) self-reporteddata + 20%below effect inmen averageweight decreasedwith (non-smokers) age belowaverage most apparent -/26 -/26 Residents ofFramin^haa Malepilots inmen20-40yrs 1.2 -Norway J-shaped million BMI's between flattercurve 23and 27kg/m forwomen 10/15 formalesand minimum increase General population females J-shaped (after aroundaBMI of 2 23kg/m' adjustmentof weightchange inadulthood BMI=BodyMass Index (weight/height inkg/m) -/10 Middleagedmen hyperuricaemiaorhyperlipoproteinemia)andnoevidenceofclinicalillnessora geneticpredispositionforsuchillness.ThispointhasbeenstressedbyBerger etal[12]andAndres[2]whohavesincebeenfrequentlymisquotedasbeing proponentsoftheviewthatoverweightsubjectsshouldnotbetreatedunless 2 theyarereallyobese(BMI>30kg/m). Themainproblemofoverweightsubjectsisthattheyareatriskofbecoming obese,andforpreventionpurposesthisgrouprequiresattention[50,59]. More importantisthefindingthatfatdistribution,eveninverymoderatedegreesof overweight,seemstohaveasubstantialimpactonhealthrisk.Are-evaluation ofexistingevidencemightleadtonewdiagnosticcriteriaforestimatingthe riskofoverweightandobeseindividuals(Seesection2formoredetails). Anotherfactorthatneedstobetakenintoaccountintheevaluationof overweightindividualsistheirage,asithasbeenshownthatweightatminimum mortalityincreaseswithage[2].InfacttheNIHconsensusdevelopment conferenceconcludedfromtheFraminghamandAmericanCancerSocietiesstudy thattheincreaseinmortalitywithrelativeweightissteeperforthoseunder 50yearsofage,suggestingincreaseddesirabilityoftreatmentinyoungerage groups. 1.3.2.Causesofdeathinoverweightandunderweightindividuals Asdiscussedintheprevioussection,ithasbeenshownthatmortalityratesare increasedatbothendsofthedistributionoftheBodyMassIndex.Themain causesofdeathintheunderweightandtheoverweightpartofthepopulationare dissimilar.Waaler[107]identifiedlungdiseases(includinglungcancer, tuberculosisandobstructivelungdisease)andstomachcancerastypicalcauses ofdeathofunderweightindividuals.Itmustbenoted,however,thatno adjustmentshadbeenmadeforpossibleconfoundingeffectsofsmoking. Cardiovasculardisease,cerebrovasculardiseaseanddiabeteswerefoundtobe characteristiccausesofdeathinoverweightsubjects.Similarconclusionshave beendrawnfromotherinvestigations,whereincreasedcancerincidencewas observedinunderweightsubjectsandincreasedcardiovascularriskwasapparent inoverweightsubjects[9,46].Suchillnessmightalsobethecauseof underweight. FromaNorwegianstudy[107]itwascalculatedthatobesityisresponsiblefor 4%ofalldeathsinwomenand5%inmen.Thisisofthesameorderasforlung cancerinmenandbreastcancerinwomen. LewandGarfinkel[74]intheirmuchcitedanalysisoftheAmericanCancer -8- Society Study found thatmenwithweightsmore than40%above theaveragewere at increased riskofcancer of the colon,rectumandprostate.Inobesewomen, increased risk for cancer ofthegallbladder,breast,cervix (especially post-menopausal),endometrium,uterusandovariumwas found (especially inlower socio-economicgroups).More evidence for increased risk ofcancer at particular sites inbothunderweight andoverweight subjects showsthat itisimpossible to give anoptimalweight forcancer risk,especially sincenutritional factorsmay be important intheassociation ofbodyweight andcancer risk [46].Garn proposed thatlowserumand tissue levelsofvitamins (particularly vitamin A) and lowcholesterol levelsmightexplaintheincreasedmortality risk inthe underweight partofthepopulation. Cardiovascular mortality Apart fromthepossible statistical dilutioneffectdue tothe failure to take theheterogeneity ofhumanobesity intoaccount,sophisticated analysesmay blurr thepicture. Ithasbeenproposed [65]thatobesity isnotamajor risk factor forcardiovascular disease since itprovednot tobea risk factor independently ofother risk factors,suchaselevated blood pressure,ECG abnormalities,cholesterol andglucose tolerance.Themechanism bywhich obesity could lead topremature cardiovascular deathprobably involves these risk factorsand itisnot fullyclearhowastatistically independent effectof Body Mass Indexoncardiovascular disease shouldbeinterpreted[87]. Brunzell [27]hasargued that,sincecardiovascular disease risk is particularly elevated inyoung adults,genetic factorsmayplayamajor role inthe association.Thishypothesis has tobe investigated further. Hubert etal [60]ina26yearof follow-up studyof 5209menandwomen, concluded thatMetropolitan relativeweight atbaseline examinationwasan independent predictor ofanginapectorisandother coronarydisease inmenand women. Inwomen itwasalsoapredictor of stroke,congestive failure and coronary and cardiovascular death.Inthis relationadjustmentsweremade for age,cholesterol,systolic blood pressure,smoking,leftventricular hypertrophy and glucose tolerance.Asdiscussed above,there isnoobvious underlying mechanism for the independent relationship isnotdirectly athand.Thejdataof theFraminghamstudy [60]also revealed thatweight gain inyoungadultyears conveyed an increased risk ofcardiovascular disease inmenandwomen,that could notbeattributed toinitialweight or to risk factors thatmay result fromweightgain. From this important study itwas concluded thatintervention inobesity isan -9advisable goal intheprimaryprevention ofcardiovascular disease. 1.4.Obesity andmorbidity Fromanextensive reviewof theliterature,Berger etal [13]concluded that the associations ofobesitywith several chronicdisorders arewellestablished, but that theevidence forother claimed associations isnotalwaysconclusive.A listing ofthedisorders found,orclaimed, tobeassociatedwithoverweight is shown intable 5. Table 5.Diseases andmetabolic abnormalities provenorclaimed tobe related toobesity (adapted fromBerger etal 1985 [13] Diseases related toobesity Diseases related toobesity (welldocumented evidence) (notwelldocumented evidence) hypertension gall stones hyperlipoproteinaemia kidney stones gout musculo-skeletaldisorders diabetesmellitis hernia glucose intolerance pulmonary insufficiency cardiovascular disease alveolar hypoventilation venous stasis haemmorrhoids severe impairment of self-image susceptibility topsychoneuroses impairment of sexual and reproductive functions riskofendometrial andbreast cancers An important problemwith thediseases thatarenot firmly related toobesity, shown intable 5,isthat theyaredifficult tostudy inprospective studies. Severalvariables,associatedwithenvironment andlife style,may confound the relation.Available evidence isbasedmainly on self-reported dataor case-control studies inselected populations.Aswill bediscussed inthe next section,fatdistributionmaycomplicate thedetectionofexistingassociations. Furthermore,slimming effortsmightcontribute totheonsetofdiseases such as gallstonedisease andmaythus lead tospuriousassociationswithoverweight. -10Clearly,therelationofbodyweighttomorbidityisaverycomplexfieldto study.TheNIHconsensuspanel,discussingthissubject,placedconsiderable weightonevidencefromthetwoNationalHealthandNutritionSurveys(NHANES) conductedbytheNationalCenterforHealthStatistics(NCHS)intheUSA.The NHANESII(1976-1980)studyrevealedstrongassociationsofhypertension, hypercholesterolemiaanddiabeteswithBMIvaluesatorabove27.8formenand 27.3forwomen(the85thpercentilelevelintheUSA),especiallyinyoung adults(ages20-44). Recommendationsforweightreduction,basedonincreasedriskofmorbidityand mortalitywithoverweight,weremadeforthefollowingcircumstances(adapted from[29]). (i)Excessbodyweightof20percentormore.ThiscorrespondstoaBMI above26.4formenand25.8forwomen(1959Metropolitantables),or above27.2formenand26.9forwomen(1983Metropolitantables).For extremeobesity,risksaremuchgreater,includingariskofdangerous cardiopulmonaryconditions, (ii)Familyhistoryorriskfactorsformaturityonset(Type-Il)diabetes, (iii)Highbloodpressure, (iv)Hypertriglyceridemiaorhypercholesterolemia. (v)Coronarydisease(oratherosclerosis), (vi)Gout, (vii)Functionalimpairmentduetoheartdisease,chronicobstructive pulmonarydisease,orosteoarthritis(spine,hipsandknees,whichbear weight), (viii)Historyofchildhoodobesity. Diagnosticandcardiopulmonaryfitnessassessmentsshouldinvolvethephysician, andprogramsofweightreductionshouldbringinotherhealthprofessionalsas well. Oneoftheresearchareasstressedbythepanelasbeinganimportantareafor futurestudywasresearchondeterminantsofbodyfatdistribution,andonthe mechanismsofitsadverseeffects. Wemayconcludethat,althoughsomeassociationsremaintobeclarified,obese oroverweightindividualswhohavemusculoskeletalorrespiratorycomplaints, elevatedbloodpressure,hyperlipoproteinemia,diabetestypeII,orglucose tolerancecanbeexpectedtobenefitfromweightreduction[16, 59]. -11- 2. FATDISTRIBUTION;ANEWDIMENSION INOBESITY RESEARCH 2.1. Introduction Intheprevious section some recommendations madebythe consensuspanel of the NIH inBethesda,Maryland in1985were cited.Fromthe reportsof thepanel it was clear that throughout thediscussions thetopic of fatdistributionwas a recurrent theme.Similarly,atmany important international congresses on obesity research (e.g.the International CongressonObesity inNewYork in1984 and the International SymposiumontheMetabolic Consequences ofObesity in Marseille in1985)increasing attention hasbeenpaid totheclassification of typesofbody fatdistribution and its relations todisease.Although the importance ofadistinctionbetween central andperipheral typesof fat distributionwas stressedbyVague intheearly fourties,the subjectof human fatdistribution hasbeenasomewhat neglected subject inobesity research until recentyears. Investigators frommany countries (especially Sweden,Franceand theUnited States)arenowclarifying themetabolicbasis for theobserved relations betweendifferent typesof fatdistribution andvarious clinical associates of obesity. Table 6showsasummaryof someofthe relationships established between a fat distribution type, inwhich fat ispredominantly stored intheabdominal region, andvariousdiseases. Table 6.Diseasesormetabolic abnormalities shown tobe related to abdominal fat distribution Men Women Prospective Only incross- Prospective Only incross- studies sectional studies studies sectional studies ischemic heart arthrosis myocardial hypertension disease hypertension infarction glucose intolerance stroke glucose intolerance angina hyperinsulemia diabetes hyperinsulemia pectoris menstrual stroke abnormalities unclear:gout,gallstones,kidney stones -12- Inthenextsectionofthisreviewwewillfirstdiscusssomeofthemethods usedfortheclassificationoftypesoffatdistribution,andthenconsiderthe functionalcharacteristicsoffatstoredindifferentregionsofthebody. Finally,therelations,summarizedintable6,willbediscussedinsomedetail. 2.1.Anoverviewofmethodsusedfortheclassificationoftypesoffat distribution 2.1.1.Subcutaneousfatpatterning JeanVaguecanberegardedasoneofthepioneersinresearchonbodyfat distribution.Heintroduced,intheearlyfourties,theconceptsofandroidand gynoidtypesoffatdistribution [101].Forclassificationpurposeshedeveloped anIndexofMasculineDifferentiation (IMD)basedontheaverageofthenapeto sacrumskinfoldratio(correctedforthetotalthicknessofthefatinthetwo regions)andthe'Brachio/FemoralAdipoMusculoRatio'(B/FAMR).THEB/FAMR wascalculatedfromskinfoldsattheproximalpartsofthearmandthighand circumferencesofthelimbsatthesamelevel [103].Relativelyhighvaluesof theIMDweretypicallyfoundinmen(androidfatdistribution)whilelower valueswerefoundtobecharacteristicinwomen(gynoidfatdistribution). Vague'sobservationsofassociationsofthesetypesoffatdistributionwith specificdisordersandmetabolicabnormalitiesstimulatedresearchinthisarea. Othermethodsusedforthedescriptionofsubcutaneousfatpatternsinclude Z-scorepatternprofiles [45],ratiosofskinfoldssuchastriceps/subgcapular ratios[42],andtheratioofsumoftrunkskinfoldstothesumofskinfoldson extremities[36,23,30].Principalcomponentsanalysisofsubcutaneousfat siteshasalsobeenusedinthestudyofhumanfatpatterning,andthe usefulnessofthisapproachhasrecentlybeenconfirmed[34,79].Joosetal [62]useddiscriminantanalysisofskinfoldthicknessessinordertodistinguish diabeticsfromnondiabetics.Itwasshownthatinmenacontrastofsubscapular withwaistandlegskinfoldsandinwomenofsubscapularandleg(calf) skinfoldsprovidedthebestbasisfortheclassificationofsubjectsinto diabeticsandnon-diabetics. Inobesityresearchthesemethodshavecertaindrawbacks,sinceskinfold thicknessmeasurementsinobesesubjectsareoftendifficulttomeasureandhave poorreproducibility[26]. Anotherdrawbackisthatmeasurementofthedistributionofsubcutaneousfat alonefailstotaketheintra-abdominalfatdepotintoaccount. -13- 2.1.2.Circumferenceratiosfordescribingfatdistributionpatterns Methodsofdescribingmorethanthesubcutaneousfatpatternaloneinvolve,for instance,somatotyping.Insuchmethods,thatwerefrequentlyusedinthe fiftiesandsixtiesofthiscentury,classificationoftypicalformsofbody buildismadebytrainedraters.CraigandBayer[31]usedphotographsand outlinesofobesewomenandclassifiedthemintocategoriesrangingfrom hyperfemininetohypofeminine.Damonetal[32]classifiedFraminghammales accordingtothetypologyofSheldonandrelateddifferenttypesofbodybuild tohealth. Ashwelletal[6]redevelopedtheconceptsofgynoidandandroidfat distributionandratedpicturesofobesewomenasandroidandgynoid. DiscriminantanalysiswasusedtoconstructaFatDistributionscore(FD-score) basedonbodydimensionmeasurements.Withthisscoreasimilarclassification ofobesesubjectsintotypesoffatdistributionwasobtainedcomparedtothe classificationbasedonthephotographsandtheclassificationsystemthatwas developedbyVague.ThescoreofAshwelletalwascalculatedfromtheratioof waistdiametertothighdiameter [6].Latertheyprovedthatthis'Fat Distributionscore'washighlycorrelatedwiththeratioofwaisttothigh circumferences [7]. InSweden,collaboratorsinresearchonfatcellmetabolismandmetabolic complicationsfoundthatthewaistcircumferencewashighlycorrelatedwith metaboliccomplications.Asthewaistcircumferencewascloselydependentonthe sizeoftheindividual,theratioofthewaisttohipscircumferencewasusedin theanalysesinordertocorrectforbodysize[68].Inanepidemiologicalstudy of15532obesewomeninaslimmingorganizationintheUSA,theratioof reportedwaistgirthandhipsgirthwasusedasanindexofbodyfat distributionas'itwassimilartotheindexreportedbyAswelletal,andthe IndexproposedbyVague'[57].Evansetal[38]laterconfirmedthatthe simplestmethodfordescribingfatdistribution(i.e.waist/hipscircumference ratio)wasatthesametimethemostusefulonewithrespecttometabolic complications[38].Later,thisratioofwaisttohipscircumferencewasfound tobeastrongpredictorofmortalityinmenandwomen[73,72].Waist/thigh ratioisprobablyanequallyusefulmeasureoffatdistribution[16,88] althoughinpracticesubjectsmaybemorereluctanttohavetheirthigh circumferencemeasuredthantheirhipscircumference. -14Thewaist/hipsandwaist/thighcircumference ratiosdoesmeasurethe predominance of fatstorage ineithertheabdominal regionor,respectively,the gluteal and femoral region.Theabdominal fatdepot indicatedbythemagnitude ofthewaistcircumference includesboththesubcutaneous fatdepotandthe intra-abdominal fatdepot.These fatdepotshavedifferentmorphological and metabolic characteristics,and itmaybe importantto distinguish thetwoin studiesofmetabolic complications,morbidityandmortality.Computed tomography hasbeenfound tobeauseful toolforestimating theamountofabdominalfat stored internallyandsubcutaneously. 2.1.3.Computed tomography infatdistribution research In1982,Borkanetal [21]published the resultsofan investigationwhich demonstrated thevalueofCTscanning inassessingabdominal fatcontent.Data from8malepatients revealed thatasingle scanattheleveloftheulnbilicus provided themostuseful informationonintra-abdominaland subcutaneous abdominal fat.Dixon [35]comparedCT-scansof25menand 25womenatthislevel (whichcorresponds inmostcasestotheleveloftheL4vertebra)andobserved thatmenhad significantlymore fatwithintheabdominal cavity.Thetotal amountofbody fatinthis regionwas foundtobe similar inmenandwomen,but inwomenagreaterproportionofthe fatwasstoredsubcutaneously.Grauer etal [55]studied 50adultmalesand62adultfemalesandassessed fatdistribution atseveralothervertebral levels (Ll,L3and L5). Theyconfirmed thatthe proportionofsubcutaneous fatisgreater inwomen thaninmenbut foundthat total fatvolume inwomenwashigherthan inmen.Similar resultswere obtained fromobese subjects [100].Borkanetal [22]showed thattotal fatmassinmen didnotdifferbetweenmiddleaged (meanage46years)andoldermen (meanage 69years),butthattheproportionofsubcutaneousabdominal fatwaslowerand, correspondingly, theproportionof intra-abdominal fatwashigher intheolder men.Aswell etal [8]studied in28womenthe relationshipsbetweenthe circumference ratiosdescribedaboveand the ratioof subcutaneousabdominal fat tosubcutaneous fatattheumbilicus (L4)level.Therewasa significant correlationbetween thewaist/hips circumference ratioand theproportionof intra-abdominal fat.These correlations remained significantafter adjustments forageanddegreeofobesity.Theauthorssuggest thattheobserved relationshipbetweenahighwaist/hips ratioandmetabolic complicationsmight, at least inwomen,reflecta relativelylargeamountof intra-abdominalfat. -15- Ifthisisproventobethecase,itmightprovideafurtherexplanationforthe higherrisksthatmenhaveformetaboliccomplications,comparedtowomen,at thesamedegreeofobesity [68].Ontheotherhand,thereisadiscrepancy betweentheobservedinternalizationoffatwithincreasingageandthedecrease inmorbidityriskofobesitywithadvancingage.Furtherresearchisneededto clarifythisapparentparadox.Figures1and2illustratetheuseofcomputed tomographyperformedontwomalepatients.Thepatientinfigure2isolderand fatterthanthepatientinfigure1.Thelargeramountofbodyfatinthe abdominalregioninpatient2isclearlypredominantlystoredasintraabdominalfat. Figures1and2: ComputedtomographicscansatL4levelintwomalepatients. Figure1.Age:33years;BMI:21.1kg/m;Surfaceofintra-abdominalbodyfat: 2 2 30cm;Surfaceofsubcutaneousbodyfat:124cm. -16- Figure2.Age:55years;BMI:26.5kg/m;Surfaceofintra-abdominalbodyfat: 2 2 157cm;Surfaceofsubcutaneousbodyfat:179cm. -17- 2.2.Functional characteristicsoffatstored indifferent regions 2.2.1.Morphologyofadipocytes indifferent regions Ithasbeenknown foralongtimethatwomenhavemorebodyfatthanmen [77]. Sjostrometal [92]showed,inastudyof11maleand12femalestudents,whose weightswerewithinthenormal range,thatwomenhadmore subcutaneousbody fat thanmenand thatthisdifferencecouldbeattributed toalarger fatcell number inwomenthaninmen.Thiswasdemonstrated fortheepigastric, hypogastricand femoral regionbutnot forthegluteal region.Inthegluteal regionthedifference insubcutaneous fatthicknessbetweenmenandwomenwas due tolarger fatcell size inwomen.Krotkiewski etal [67]foundthat non-obesemiddleagedwomenhadmorebody fatthannon-obeseyoungwomen,owing mainly tothelarger fatcellsintheolderwomen,particularly intheabdominal region.Thisfinding indicated thatfatcellsoftheabdominal regionaremore responsivetonutritionalorhormonal factorsthanfatcellsfromotherregions. WhenKrotkiewski etal [68]extended their studiestoalargepopulation (n= 930)ofobese individualstheyfoundthatfatcellweightinobesemenwas similar inall four studied regionsbut fatcellnumberwas found tobehighest intheabdominal region.Inobesewomen,asinwomenofnormalweight,fatcell sizeand fatcellweightwerehighest intheglutealregion. Thus, thetypicalcharacteristicsoffatcellmorphologyand fatdistribution in menandwomenofnormalweight remainspresent inobesity.Itwas shownthat in theobese,the increasingamountofbody fatleadstoanincreased fatcell size,uptoacriticalvalueof 0.7 -0.8 ugper cell.After thisincrease,a more rapidincreaseof fatcellnumberwasobserved inall regions.This finding challenged thehypothesis (whichwashotlydebated someyearsago)that fatcell number isacharacteristicdetermined early inlifeandpredisposing fat infants tolifelongobesity.Thishypothesishasnowbeengenerally rejected [56,63]. Thedifferencesbetweenmenandwomenwith regardtofatcellmorphologyand fat distributionmaybe,according toBjorntorp [14],explainedbythespecific controlof femoral-gluteal fatcellsby femalesexhormones. The roleofhormonal regulationofbody fatdistributionand fatcell sizewas furtherelucidatedbywork ofRebuffe-Scriveetal [83]who foundthat fatcells were larger inbothabdominaland femoral regionsinlactatingwomenthanin pregnantandnon-pregnantwomen.Lipo-proteinLipase (LPL)activitywashigher inthe femoral than intheabdominal region innon-pregnant andpregnantwomen. Thus, fatispreferentially stored inthe femoral regionundernormal conditions and inpregnancy inwomen.Duringlactation,however,LPLactivitywasmarkedly reduced inthefemoral regionwhile therewasnochange intheabdominal region. Basaland catecholamine stimulated lipolysiswere foundtobemoreactive inthe abdominal region (ashadalsobeenobserved inother studies [5,76,69]). Duringlactation,incontrasttootherconditions,lipolysis increased inthe femoral regions.These findingsledtothehypothesisthatthe characteristic female fatstorage intheglutealand femoral fatdepotshavea functionfor the storageof triglyceridesthatcanbeused toprovidetheextraenergyneeded for lactation.Thiscanprobablybe regardedasanadaptationthatwasadvantageous inpre-industrialcivilizationsbutisnolongerofmuchgreatphysiological importanceand is,ironically,thecauseofcosmetic concern inwomenliving in populationswhere food isabundant inallseasons. Ithasbeen reported that inthepost-menopausalstatewomenbecomemore similar tomenwith regard toregionaldifferences infatcellmorphology,andmaybe influencedbytheuseofestrogenmedication [14,15].Other findingswhich supportthehypothesisofhormonal regulationofbody fatdistributionwere publishedbyEvansetal [39].Theystudiedadipocytemorphologyandmetabolic profiles in80healthypre-menopausalwhitewomenwithbodyweights ranging from normal tomarkedlyobese.They reported thatincreasingandrogenicity (reflected indecreasedplasma- sexhormonebindingglobulinandan increased percentage of freetestosterone)wascorrelatedwithincreasingwaist-hipscircumference ratioandan increase infatcellsizeintheabdominalbutnotinthefemoral region. Althoughthedetailsofthemechanisms involved remaintobeelucidated, evidence sofar strongly suggeststhatspecificendocrine regulationof abdominaland femoraladipocytesexplainsatleastsomeofthedifferences betweenmenandwomenandbetweentypesofbodyfatdistributionwithregard to the regionalvariation inmorphologyanddistributionofadipocytes. 2.2.2.Additional factorsthatmay influence regionaldifferences infat celldistributionandmetabolism 2.2.2.1.Weight reductionand fasting Ashasbeendiscussed intheprevioussections,theabdominal fatdepot appearstobeamorelabiledepotoftriglycerides.Itmightthereforebe expected thatweight reductionwould leadtoa reduction incell size, particularly intheabdominal region.Thishypothesis issupportedby findings -19ofArner etal [4],whoobserved that,in25obesemiddleagedwomen,fatcell volumedecreased intheabdominalbutnotinthefemoral regionduringoneweek of fasting.LPLactivitydecreased,and lipolyticactivity increased,inthe femoral regionas resultof fasting,buttheseeffectswere foundtobemuch morepronounced intheabdominal region.Similarobservationsweremadeby Smithetal [94]who studied seriouslyobesepatientsbeforeandafter jejuno-ilealby-pass.Fatpatternswould thereforebe likelytobealtered after substantialweight reduction.Manystudiesinthepast,however,have beenunable toconfirm this. Forexample,Edwards [37]observed in1950that in womenwho lostmorethan12.7kgtherewasabasic fatpatternthat remained stableoverabroadweight range.G a m [45]studied 13menduring controlled weight lossandfoundthatthe relative fatpattern remainedconstant.Craig andBayer [31]re-examined 50women2-10yearsafter conspiciousweight lossor gain,andconcludedonthebasisofoutlinesthatwhetherweightwas lostor gained thedistributionoffatandtheandrogynicpattern remained thesame. Vague in1974 reported that inastudyof13obesemales,28obese femalesand 21leanfemalesduringweight lossorgaintheadipomuscular ratiosin individualsweremoreorlesscharacteristicandwere inclined toreappear whateverweightvariationswere imposedonthesubject [105].Ashwelletal [6] foundnocorrelationoftheir 'FatDistributionScore'withweight lossand proposed that fatdistributionwas relativelyconstantandpossiblygenetically determined.Thework ofBorjeson [20]in1976onmonozygoticanddizygotic twinssuggested that fatdistributionmight indeedhaveanimportantgenetic component.A recent investigationofweight lossin187severelyobesewomenby Lanskaetal [69]demonstrated thattheratioofwaisttohipsgirthwasnota usefulprognostic indicator ofweightchange forthesewomenwith,itshouldbe noted, refractory severeobesity.Theassumptionthatobesity ismore catching intheabdominal regionandthatfemoral regionsaremore resistanttoslimming [3]maybeunwarranted [52].On theotherhand,Albrink andMeigsshowed,in 1964, that in419male factoryworkers,skinfold thicknessofthetrunkwas correlatedwiththeamountofweightgained sinceagetwenty five,but forearm skinfold thicknessshowednosuchcorrelation [1].Theyhypothesized that forearmskinfold thickness reflected innatelifelongobesity,while skinfold thicknessof thetrunk reflected,atleast inpart,obesityacquiredduring adultlife.Geneticallydetermined fatstorage capacity indifferent regions maybecomeapparentonly inlaterlifeundertheinfluencesofenvironmental factors.Thishypothesis,thatfatintheextremities isincreased ininherited butnot inacquiredobesity,isstrengthenedbythe reportofStrendberg -20- (quotedbyWells[108])ofafullthicknessskingraftfromtheabdominalwall tothehandofaslimyounggirl.Asanadultthisgirlacquiredexcessweight aboutthegirthandtheskingraftofthehandalsobecameobese,althoughthe neighboringforearmadiposetissuedidnotincreaseinsize.Theresult,a grotesqueboxinggloveappearancetothehand,suggeststhatthereisagreater potentialforabdominalfattoexpandthanforfatoftheextremities.Thus, environmentalinfluencesmaybenecessaryfortheexpressionofgeneticfactors thatdeterminefatdistribution.Ifthisisindeedthecase,weightgaininthe 'innate'obeseshouldnotinfluencefatdistribution,butinthosewithadult onsetobesityitshoulddoso.Theconflictingresultsdescribedabovecouldbe duetodifferencesintheaetiologyoftheobesesubjectsthatwerestudied. Anotherfeatureofthestudiesdescribedaboveisthattheyconsideredonly subcutaneousfatpatterningbeforeandafterweightloss,andmayhavefailed totakealterationsintheintra-abdominalfatdepotintoaccount.Aswas discussedintheprevioussection,thisdepot,ofallthemajorfatdepots,is themostlabilepooloftriglycerides.Studiesinwhichsubjectswithaclear abdominalfatdistributionarecomparedtothosewithfatstoragepredominantly inglutealandfemoraldepots,beforeandaftercontrolledweightreduction, areneededtoclarifythisissue.Althoughitwillbedifficulttoidentify themwithoutmisclassification,individualswhoaregeneticallypredisposed towardsobesityshouldbeseparatedfromthosewithnofamilyhistoryof obesityatall.Insuchstudies,notonlyshouldsubcutaneousfatdistribution bestudied;variationintheamountoffatwithintheabdominalcavityshould alsobeincluded.StudiesofBouchardetal[24]suggestthatabout60%ofthe subcutaneousfatpatternmightbeexplainedbygeneticandenvironmental factors,whicharebothequallyimportant.Theimportantimplicationsforthe treatmentofabdominalobesitymeanthatpriorityshouldbegiventostudiesof theeffectsofweightreductionbyvariousmethodsonfatdistributionand associatedriskfactors. 2.2.2.2.Age,degreeofobesity,ageofonsetofobesity,parity Studiesofsubcutaneousfatpatternshaverevealedthat,withincreasingage, fatshiftsfromtheextremitiestowardsthetrunk[47,48,93].Aswelletal [6]showedthatolder,fatterwomenaremorelikelytohaveacentraltypeof fatdistribution.ThiswasconfirmedinastudybySeidelletal[88]inwhich BodyMassIndexexplained15.9%ofthevariationinthewaist-thighratioof overweightwomen,andageexplained9.0%.Inoverweightmen,agewas -21- significantlycorrelatedwiththewaist-thighratio(r2=16.5%),andtoa smallbutsignificantextent,socialclasswasnegativelycorrelatedwiththe 2 waist-thighratio(r =3.3%). DatafromalargecrosssectionalsurveyintheUnitedStates(52953womenwho participatedinaTOPS,'TakeOffPoundsSensibly',programme)revealedsimilar results[70].Inthisstudy,inwhichself-reportedwaistandhipgirthswere used,thewaisttohipcircumferenceratioincreasedwithageandbodyweight. Theseeffectscouldnotbeaccountedforonthebasisofparity,menopausal statusorobesityhistory.Obesityhistorywasdefinedasthegreatestpercent over 'IdealBodyWeight'attainedduringtheteenageyears.Thestudyhasthe advantageofalargenumberofsubjects,buttheinformationwasobtainedfrom aratherselectivepopulationandmightbebiasedbyself-reportingofthe data.InthestudyofSeidellelal[88],self-reportedweighthistoryrevealed thatwomenbelongingtotheuppertertileofwaistthighratiohadalater onsetofobesitythanthoseinthelowertertileoftheratio.Inmen,nosuch aneffectofonsetofobesitywasfound.Suchfindingshaveledinvestigators tohypothesizethatobesity,acquiredearlyinlifewouldleadtogeneralized obesitywithexpansionofallfatdepots,whileinadultonsetobesitythis expansionwouldbeconfinedtotheabdominalregion.Asfatnessinchildrenis associatedwithearlymaturation,onemightexpectlong-termdifferencesinthe fatdistributionofearly-andlate-maturingindividuals[44].Theeffectof maturationonbodyfatdistributionhasbeenstudiedbyFrisanchoandFlegel [44]whoexaminedthedatafromtheNationalHealthandNutritionExamination SurveyIintheUnitedStates(NHANESI).Percentagetrunkfat(definedas subscapularskinfolddevidedbythesumofthetricepsandthesubscapular skinfold)wasfoundtoberelatedtoboneage(inchildren)andageofmenarche (inadultwomen). Theirresultssuggestedthatearlymaturationfrom7years onwards(exceptformalesaged13to16years)isassociatedwithlarger tricepsandsubscapularskinfoldthickness.Thiswasmoreevidentinthe subscapularskinfold.Theauthorsconcludedthatadvancedmaturationis associatedwithanaccentuationofthecentripetal (abdominal)distributionof fatduringadolescenceandadulthood.Theysuggestedthatenvironmental influencesmaybringabouttheexpressionofgeneticfactorsthatdeterminefat distribution.TheirpaperwascritisizedbyDeutchandMueller [34]whoargued thatinfemalesthedifferenceascribedtomaturitywaslikelytobea secondaryeffectofthecorrelationbetweenpercentagetrunkfatandbody fatness.IncontrasttoFrisanchoandFlegel,theyfoundthat,usingprincipal componentanalysisofsubcutaneousfatdistributioninadolesenceandyoung T-i- adults,obesity inadolescence andyoungadulthoodwasassociatedwith fat concentrated intheupper partof the trunk.Thiseffectwas independentof maturity,whichwas found tobeasignificant correlate of the trunk/extremity patterning componentbutnotof theupper/lower trunk-fatcomponent (inmales). Only long termprospective studies,inwhichbothbodyweight development and fatdistribution are followed, canclearlygive theanswer tothe question whether fatdistribution is related totheonsetofobesity,menopausal status, and/or parity. 2.2.2.3.Regionaldifferences infattyacidand glucosemetabolism in adipocytes. Itisdifficult tostudycells fromthe intra-abdominal fatdepot.Those who have studied these cells,obtained them fromtheomentummajor of patients undergoing abdominal surgery.Omental fatcellswere found todiffer from subcutaneous fatcells intheir lipolytic response.Ithasbeen demonstrated thatomental fatcellsaremore responsive tothe lipolytic effect of epinephrine andnor-epinephrine [54].Bolinder etal [18]showed that innonobeseadultmenandwomen,omental fatwasless responsive tothe antiliipolytic effect of insulin thansubcutaneous abdominal fat.Theydemonstrated that the difference could beexplained bydifferences ininsulin receptor affinity,and byadifference ininsulinactionat thepost-receptor level.Itcanbe expected, therefore,that invivo relatively largeamountsof free fatty acids will be released fromabdominal adipocytes intothebloodstream. From the intra-abdominal fatdepot the free fattyacidswilldraindirectly intothe portalvein,sotheliverwillbeexposed toan increased concentrationof free fattyacids. Fromthe subcutaneous fatdepot the released free fattyacidswill go intothe systemic circulation. Figure 3showsatentative overviewof the possible consequences of the releaseof largeamountsof free fattyacids (FFA) by theabdominal depots (adapted fromBjorntorp14,15). -23- Subcutaneous abdominal fat depot FFA insysfltemiccirculation high exposure of peripheral tissues toFFA deranged f glucose metabolism inperiphery iglyceridemia (VLDL increase) Intra-abdominal fatdepot Figure 3,Schematic representation of the (hypothetical) mechanismsby which increased fatmassmay lead tometabolic complications. (Adapted fromBjSrntorp(15). The relationbetweenpredominance of trunk fat (asopposed to faton the extremities) andplasma triglycerideswas shownbyAlbrink andMeigs in 1964 [1], and has sincebeen confirmed several times [40,64].Asdiscussed inthe previous section,Albrink andMeigs founda correlation of trunk fat (butnot forearm fat)withweight gain sinceage twenty-five.Inmenwith slender forearms,triglyceride concentration increased significantlywith abdominal skinfold thickness. Inmenwith fat forearmsnosuchcorrelationwas found. Albrink andMeigs suggested thatthis finding,and the fact thatmean triglyceride levels in innately thinmen (withslender forearms)who became obese later inlifewerehigher thaninthe innately fatmen (with fat forearms),mayhavebeenthe resultofoverloading ofexisting adipose cells. Despres etal [33]inastudyon234womenand 238menof18- 50yearsofage, confirmed thatabdominal and subscapular fatdepotsaremore closely associated with serum lipids thanother fatdepots,particularly inmen.Theabdominaland subscapular skinfoldswere foundtobepositively correlated with serum triglicerides and total cholesterol,negativelywithHDL-cholesterol,and most stronglywith the ratioofHDL-cholesterol tototal cholesterol.This regional trend remained significant after correction forconcomitant variables suchas age,cigarette smoking,habitual energy intakeand energy expenditure,maximal aerobic power andalcohol consumption. Suchassociationswere less clear in womenwho,although theywere fatter than themen inthestudy,had lower triglyceride and total cholesterol concentrations andhighervaluesof theHDL cholesterol/total cholesterol ratiothan themen.More recentlyotherparts of themechanismdescribed infigure 3havebeenproven.Thiebaud etal [99] demonstrated that increased plasma FFAlevels in25healthyyoungvolunteers, established bymeansof intralipid infusion,impaired theabilityof insulin to stimulate glucoseoxidation and storage.Thisinhibitoryeffectofelevated FFA levelsonglucose storage contributed more tothe impairment of glucose tolerance than thedefect inglucoseoxidation.They foundnoeffectofFFAon baseline insulin levels.Smithhas recently cited evidence [95]that perfusion with ahigh free fattyacid concentration inthe rat,reduced insulin clearance.Decreased hepatic insulin removal isan important causeof hyperinsulinemia inobesity,especially indiabetics [19].High FFAinthe portal veinmight thus,as suggested in figure 3,be responsible for hyperinsulinemia.Althoughmore evidence isneeded tounderstand the mechanisms described infigure 3,this fieldof research isdeveloping rapidly,and it can be expected thatwithina fewyears thetruecascadeofeventswill be established.Onequestionofparticular importance iswhichofthe two abdominal fatdepots (intra-abdominal or subcutaneous)causes the metabolic aberrations associated withacentral fatdistribution. -252.3.Clinical correlatesofhumanfatdistribution Thehypotheticalmechanisms showninfigure 3provideapossibleexplanation forthefindingsofepidemiological andclinical research revealingan associationbetweenabdominal obesityanddisorderssuchas ischemicheart disease,diabetesmellitus (type II), hypertension,and stroke.Theevidence for theseassociationswillnowbediscussed inmoredetail. 2.3.1. Ischemicheartdisease JeanVaguewasprobablythefirsttoreporta relationshipbetween fatlocated intheupperpartofthebodyandischemicheartdisease (characteristic inhis android typeofobesity).Hecalled fatmasslocated intheupperpartofthe body, 'diabetogenic'and 'atherogenic'fatmass [101,102,103,104,105].In 1959heandhisco-workerspublisheda studyon82obesemenand 158obese women [106].Theycompared subjectswithclinical symptomsofatherosclerotic disease tothosewithoutevidenceofsuchdisease.Itwas foundthatbetween thegroupstherewerenodifferences inageandtotal fatmass,butthatthe B/FAMRwassignificantlyhigher inthemenandwomenwithsymptomsof atheroscleroticdisease.Theyproposed,onthebasisofpreliminaryevidence, thatminorhypercorticism,thatwouldaccompanyandroidobesity,could,in part,be responsible fortheprogressionofobesity towardsatherosclerosis. His retrospective evidence ina rather selectivepopulationhaslong remained unrecognized.Damonetal [32]found thatinFraminghammalescasesofcoronary heartdiseasewere 'fatter,moremuscularandstockier'thanothers .The relationshipswerenotveryclearand itisdifficult totranslatebodybuild intotermsof fatdistribution. Only recentlyhave clear relationshipsbetweenabdominal fatdistributionand several cardiovascular endpointsbeenestablished inlargeprospectivestudies. InGoteborg, inSweden,investigators includedwaistandhips circumference measurements intheir setofanthropometricbaselinedata,when theystarted their long-term followupstudiesofmalesand females.A thirteen-year followupof 54-yearoldmenbornin1913revealed thata significant association between thewaist tohipscircumference ratioandtheoccurrenceof ischemic heartdisease.Infact,itproved tobethemostpowerfulpredictor for these endpoints,ofalltheanthropometricvariablesstudied (includingBodyMass Index)[73]. Thewaist tohips ratiopredicted ischemicheartdisease independentlyof -26measuresofobesity,butnotwhen risk factorslikesmoking,systolic blood pressure and serum cholesterolwere taken intoaccount.What isremarkable is that the risk for ischemicheartdiseasewashighest inmen inthe highest tertile ofwaist tohips ratioand inthelowest tertileofBodyMass Index. These findingswere confirmed intheParisprospective heart study [36J.In this studycircumferences hadnotbeenmeasured atbaseline examination,but an index of fatdistributionwasconstructed on thebasisofskinfoldsat several sites.The ratioof the sumof skinfoldsonthe trunk tothe sumof skinfolds on theextremitiesproved tobeanindependent predictor of cardiovascular endpoints ina10-year followupofmiddle-aged men.Lapidus etal [721 showed that ina12-year followupofwomenaged 38-60years,similar associations were found tobeevenmore pronounced than inmen.Thewaist/hips rati©was associated with the incidence ofmyocardial infarction andanginapectoris even whenage,bodymass index,smokinghabits,serumcholesterol and triglycerides, and systolic blood pressurewere taken intoaccount.Onedrawback of these studies isthat subsequent eventsarelinked only tomeasurements at baseline examination: changes inbodyweight and fatdistribution during follow-up cannot betaken intoaccount.Suchanapproachwould involve frequent re-examinations,whichare impossible inpractice.The studyofLarsonetal raise thequestionwhy relatively leanmenwithhighwaist/hips ratios,have thehighest risk fordeveloping coronaryheartdisease. Studies oneightnon-obesewomenwith familial partial lipodystrophy[85], which ischaracterized by fatstoragepredominantly as subcutaneous abdominal fat,andapatientwithWerner's syndrome [96],showed thateven innon-obese individuals an increased ratioofcentral toperipheral fat isassociated with metabolic complications. 2.3.2.DiabetesMellitus,glucose intolerance andhyperinsulinemia DiabetesMellitus type IIwas the firstdisorder found tobeassociated with abdominal,upper bodyand/or android typesof fatdistribution was.Vague observed that inobesepatients increasing severityofdiabeteswas accompanied byanincrease inB/FAMR [106]. Lister andTanner [75]contpared 'acutely'diabetic patientswith 'non-acute'diabetics and found that thenonacutediabeticsweremoreendomorphicthanthosewhosediabeteshad appeared during early life.Endomorphyischaracterized by roundness of the body contour,atendency toobesity,asmooth skinwith finehair,and short tapering limbswith smallhandsand feet.Itisclear thatthesepersons,who -27canbe regarded astypical type IIdiabetics,would alsohaveahighwaist to thigh circumference ratio.Otherworkershavecompared skinfoldmeasurements of diabetics andnon-diabetics andconcluded thatdiabeticshave relatively larger skinfolds on the trunk thanontheextremities [42,62].Kissebah etal [66] later confirmed the findingsofVaguebut,insubsequent research, preferred theuse ofwaist tohips ratio formeasuring fatdistribution. Ina seriesof experiments onabout90womenKissebahandhisco-workersexamined the mechanismsunderlying the relationbetween fatdistribution anddiabetes [38, 39, 40,41].They concluded fromtheir experiments thatdiminished skeletal muscle insulin sensitivitywith impaired glucose storage capacity contributes tothe insulin resistance and,inturn,totheassociated glucose intolerance andhyperinsulinemiaofupper body segmentobesity (orabdominal obesity). This effectmaybe the resultofa reduction ininsulin receptor numberwhich could, inturn,besecondary topersistently elevated fastingplasma insulin levels [38, 39,40, 41]. Themechanism described above isbased on correlations betweenmany variables involved inthecomplex glucose-insulin regulatory system,andhas still tobe experimentally validated. Recently, Smith [95]cited evidence suggesting thathigh Free FattyAcid (FFA) concentrations,whichwere found tobeassociatedwith abdominal fat distribution (see figure 3 ) ,causeadecrease in insulin clearancewhich could lead tohyperinsulinemia. Lipolysisactivitymay,therefore,be thekey factor inthe relationbetweenabdominal obesityanddiabetes. Thus,although theepidemiological evidence for sucha relation is fairly strong and consistent,themechanism behind thecausal relationshipbetween an abdominal typeof fatdistribution anddiabetes appears tobevery complex. Theepidemiological evidence comes fromcross-sectional surveys [58]inwhich theprevalence of self-reported diabetes increased with waist/hips circumference ratio.This relationwas independent of thedegree of obesity. Ohlsenetal [80]showed that,intheir prospective studyofmiddle-aged men, the incidence ofdiabeteswasassociated withwaist/hips circumference ratios atbaseline examination. Inobese non-diabetic andotherwise apparently healthy subjects,the relation between glucose intolerance andhyperinsulinemia and thewaist tohips circumference hasalsobeendemonstrated[64]. 2.3.3. Hypertension Vague noted in1956 [102]thatandroid obesepersons hadatendency todevelop hypertension. In recentclinical andepidemiological studies thishasbeen confirmed [58,39,17,88].Itshouldbenoted thatKrotkiewski [68]f((>undthat body fatinobesewomenand fatcell size inobesemenwere, in stepwise regression analysis,more important than thewaist/thigh circumference ratio. Hartzetal [58],intheir large studyonTOPSmembers,compared obesewomen in theupper quartile ofwaist/hips circumference ratiowithwomen inthe lowest quartile of the ratio.They founda relative risk of 1.74,adjusted for effects ofageand relativeweight.Relative risk forwomen intheupper quartile of relativeweight,comparedwith thatof those inthelowestquartilewas 4.01 adjusted forageandwaist tohips ratio.Theeffects ofbody fat distribution andexcess fatmassthusseemtobe independent. Itisnotyetclearhowthis association ofbody fatdistribution canbeexplained. Thehyperinsulinemia, that isfound tobeassociatedwithabdominal obesity (figure 1), mightbe involved inthemechanism. Itiswellknownthathyperinsulinemia causes sodium retention andactsonthe sympathicnervous system. Effectsofhormonal factors onboth fatdistribution andbloodpressure cannotbe excluded. Sincehypertension isa risk factor forcardiovascular disease and cerebrovascular accidents,theobserved relationbetween abdominal fat distribution andhypertension may inpartexplain theepidemiological findings inwhich central fatdistribution hasbeen found tobeassociated with cardiovascular disease and stroke. 2.3.4.Other clinical correlates ofbody fat distribution AlthoughVague andhisco-workers played an important role in stimulating research onbody fatdistribution, someoftheir evidencewas poorly documented. Inhismuch cited paper in1956,Vague [102]stated that android obesitywasassociatedwithgout,uricacid crystalsand gallstones.Gynoid obesity,on theotherhand,wasassociatedwithdirectmechanical complications (respiratory and circulation problems).These findingswerebased on 'observations on thousandsofpatients'butnodatawerepresented. Vague proposed that theseassociationswerepossibly related tohormonal differences that caused both fatpatterns and thecomplications. Hartzetal [58]argued that ifVague'shypothesiswas true theassociations of body fatdistributionwithmenstrual abnormalitieswouldbeevenmore -29pronounced thanwith otherdisorders.They studied the relationshipbetween the waist tohip ratioand reportedmenstrual abnormalities (irregular cycles, oligomenorrhea andhirsutism) in11791womenaged 20- 39years.They confirmed that suchanassociationexisted,butthat itwasno stronger than thosewith diabetes,hypertension orgallbladder disease[58]. Therefore Hartzetalproposed another mechanismbywhichbody fat distribution influences synthesisofestradione fromplasma androstenedione.Adipose tissue seems indeed to influence thisconversion [59,78].Whether adipose tissue in different regionsbehavedifferently inthis respectneeds tobe investigated. Asmentioned above,Hartzetalobserved a relationof fatdistribution with gallbladder disease.Itwas suggested thatcholesterol metabolism maybe involved inthis relation. InaDutch study ithasbeen found thatarthrosis inobesemen ismore prevalent inthosewithahighwaist/thigh ratiothan inthosewitha low waist/thigh ratio [88].Silberberg, ina reviewof the relation between arthrosis and obesity,concluded thatarthrosis ismore closely associated with abnormalities in fatmetabolism thanwithweight-induced wear and tear [89].In fact,after adjustment for fatdistribution,Seidell etal [88]founda negative association betweenBodyMass Indexandarthrosis.For the interpretation of this finding itshould benoted intoaccount that this study involved obesemenonly.The same study revealed thata lowwaist/hips ratioor waist/thigh ratiowasassociatedwithahighprevalence ofvaricoseveins[88]. This isinagreementwith findingsbyVaguewhoproposed that 'venous insufficiency' wasoneof thecharacteristics ofgynoid obesity [102]. Whether this relation indicates thatwhen fat isstoredpredominantly inthe thighorhips regionblood flow inperipheral tissuesof theseareas is impaired remains tobeproven. 2.4.General conclusions The consistency of the relationships described intheprevious section leads to the conclusion thatpredominance of fat stored intheabdominal region is hazardous tohealth.Thusprevious research inwhichexcess fathasbeen treated asasimple entitymustbe interpretedwith caution.The implications of these findings forthediagnosis and treatment ofobese individuals have been summarized byBjorntorp [15].Heproposes thatabdominal obesity (awaist hips ratiogreater than 1.0 inmenand greater than 0.8 inwomen) isdangerous and should bediagnosed early andexamined forassociated risk factorsonwide -30indications (plasma insulinandlipids,glucose tolerance,bloodpressure). Future research should focusonmanipulating fatdistribution,andon evaluationof theeffectsof thismanipulationonmorbidityandmortality. REFERENCES 1.AlbrinkMJ,MeigsJW.Interrelationshipbetweenskinfold thickness,serum lipidsandblood sugar innormalmen.AmJClinNutr1964;15:255-261. 2.AndresR. Effectofobesityontotalmortality. IntJObes1980;4t386-389. 3.Arner P. Sitedifferences inhuman subcutaneousadipose tissuemetabolism inobesity.AesthPlastSurg1984;8:13-17. 4.Arner P,EngfeldtP,LithellH.Sitedifferences inthebasal metabolism of subcutaneous fatinobesewomen.JClinEndocrinolMetab 1981;53:948-951. 5.Arner P,BolinderJ,EngfeldtP,6stmanJ.Theanti lipolyticeffectof insulin inhumanadipose tissue inobesity,diabetesmellitus, hyperinsulinemia and starvation.Metabolism1981;30:753-759. 6.AshwellM,ChinnS,StalleyS,GarrowJS.Female fatdistribution-a photographicandcellularitystudy.IntJObes1978;2:289-302. 7.AshwellH,ChinnS,StalleyS,GarrowJS.Female fatdistribution-a simpleclassificationbasedontwocircumferencemeasurements.IntJObes 1982;6:143-152. 8.AshwellM,ColeTJ,DixonAK.Obesity:newinsight intothe anthropometric classificationof fatdistribution shownbycomputed tomography.BritMed J1985;290:1692-1694. 9.AvonsP,DucimetiereP,RakotovaR.Weightandmortality.Lancet 1983;i:1104. 10.BaeckeJAH,BuremaJ,Deurenberg P.Bodyfatness,relativeweightand framesize inyoungadults.BritJNutr1982;48:1-6. 11. BennRT.Somemathematicalpropertiesofweight-for-height indicesused as measuresofadiposity.BritJPrevSocMed1971;25:42-50. 12. BergerM,Berchtold P,GriesFA,ZimmermanH. Indicationsforthe treatmentofobesity.In:BjorntorpP,Cairella,HowardAN (eds). Recent advances inobesity research III.JohnLibbey,London,1981. 13.BergerM,JorgensV,KemmerFW,FinkeC,Berchtold P.Healthhazarsof overweight. In:vandenBergEME (ed).Proceedingsofthe fourth European nutrition conference: 31-37.Voorlichtingsbureauvoordevoeding.The Hague,theNetherlands,1985. -3114.BjorntorpP.Adipose tissue inobesity (Willendorflecture). In:Recent advances inobesity research:IV:163-170.HirschJ,Van ItallieTB (eds). JohnLibbey,London,Paris,1985. 15.BjorntorpP.Obesityandthe riskofcardiovascular disease.AmClinRes 1985;17:3-9. 16.BlackburnGL,ReadJL.Benefitsof reducing revisited.PostgradMedJ 1984;60(suppl.3):13-18. 17. BlairD,HabichtJ, SimsEAH,SylvesterD,AbrahamS.Evidence foran increased risk forhypertensionwithcentrally locatedbodyfatand the effectof raceand sexonthis risk.AmJEpidemiol1984;119:526-540. 18.Bolinder J,KagerL,OstmanJ,Arner P.Differencesatthe receptorand postreceptor levelsbetweenhumanomentalandsubcutaneousadipose tissue intheactionof insulinonlipolysis.Diabetes1983;32:117-123. 19.BonoraE,Zavaroni I,Coscelli C,ButturuniU.Decreasedhepatic insulin extraction insubjectswithmild glucose intolerance.Metabolism 1983;32: 438-446. 20. BorjesonM.Theaetiologyofobesity inchildren.ActaPaediatrica Scand 1976;65:279-287. 21.BorkanGA,GerzofSG,RobbinsAH,HultsDE,SilbertCK,SilbertJE. Assessmentofabdominal fatcontentbycomputed tomography.AmJClinNutr 1982;36:172-177. 22. BorkanGA,HultsDE,GerzofSG,RobbinsAH,SilbertCK.Age changesin bodycomposition revealedbycomputed tomography.JGerontology 1983;38:673-677. 23.BouchardC,SavardR,Despres J, TremblayA,LeblancC.Body composition inadopted andbiological siblings.HumBiol1985;57:61-75. 24. BouchardC. Inheritanceof fatdistributionandadipose tissuemetabolism. In:VagueJ,BjorntorpP,Guy-GrandBetal (eds)Metabolic complications ofhumanobesities:87-96.ExerptaMedica,Amsterdam,1985. 25.BrayGA.Obesity:definition,diagnosisanddisadvantages.MedJAust 1985;142:S2-S7. 26.BrayGA.GreenwayFL,MolitchME,DakinsWT,AtkinsonRL,HamiltonK.Use ofanthropometricmeasures toassessweight loss.AmJClinNutr 1978;31:769. 27. BrunzellJD.Obesityandcoronaryheartdisease-atargetedapproach. Arteriosclerosis 1984;4:180-182. 28.BuildStudy 1979.Chicago.SocietyofAduariesandAssociationofLife InsuranceMedicalDirectors,1980. -32- 29.BurtonBT,FosterWR.HirschJ,vanItallieTB.Healthimplicationsof obesity:AnNIHconsensusdevelopmentconference.IntJObesity1985;9: 155-159. 30.CairellaM,SianiV,D'Arcangell E;PapaF.Metabolic complicationsin obesity:correlationswith thedistributionsofadiposetissue.Int CongressonObesity [Abstract]NewYork1983. 31.Craig LS,BayerLM.Androgenicphenotypes inobesewomen.AmJPhysiol Anthropol1967;26:23-33. 32.DamonA,DamonST,HarpendingHC,KannelWB.Predicting coronaryheart disease frombodymeasurementsofFraminghammales.JChronDis1969;21: 781-802. 33.DespresJP,Allard C,TremblayA,TalbotJ,BouchardC.Evidence fora regionalcomponentofbodyfatnessintheassociationwith serumlipids in menandwomen.Metabolism;34:967-973. 34. DeutschMI.MuellerWH.Androgyny infatpatterning isassociatedwith obesity inadolescentsandyoungadults.AnnHumBiol1985;12:275-266. 35.DixonAK.Abdominal fatassessedbycomputed tomography: sexdifference in distribution.ClinRadiol1983;14:189-191. 36.DucimetiereP,AvonsP,CombienF,RichardJL.Corpulencehistoryand fat distribution inCHDetiology-theParisProspectiveStudy.EuropHeartJ 1983;4:8. 37. EdwardsDAW.Observationsonthedistributionofsubcutaneous fat.Clin Science1950;9:259-270. 38.EvansDJ,HoffmanRG,WilsonCR,KalkhoffRK,KissebahAH.Relationshipof regionalbody fatdistributiontoglucose toleranceand response todiet inpremenopausalwomen.ProcNorthAmAssoc StudyObesity,Poughkeepsie, NY [Abstract],1982. 39. EvansDJ,HoffmannRG,KalkhoffRK,KissebahAH.Relationshipofadrogenic activity tobody fattopography,fatcellmorphologyandmetabolic abervations inpremenopausalwomen.JClinEndocrinalMetab 1983;57:304-310. 40.EvansDJ,HoffmannRG,KalkhoffRK,KissebahAH.Relationshipofbody fat topographytoinsulinsensitivityandmetabolicprofilesinpremenopausal women.MetabClinExp1983;33:68-75. 41.EvansDJ,MurrayR,KissebahAH.Relationshipbetweenskeletalmuscle insulin resistance,insulin-mediated glucosedisposal,and insulinbinding - Effectsofobesityandbodyfattopography.JClinInvest 1984;74:1515-1525. -33- 42. FeldmanR,SenderAJ,SiegelaubAB,OaklandMS.Difference indiabeticand nondiabetic fatdistributionpatternsbyskinfoldmeasurements.Diabetes 1969;18:478-486. 43. FosterMA,HutchinsonJMS.NMR-imaging-methodandapplications.J BiochemEng1985;7:171-182. 44. FrisanchoAR,FlegelPN.Advancedmaturationassociatedwith centripetal fatpattern.HumBiol1982;54:717-727. 45.GarnSM.Relative fatpatterning:an individual characteristic.HumBiol 1955;27:75-89. 46.GarnSM,HawthorneVMetal. Fatnessandmortality inthewestof Scotland.AmJ. ClinNutr1983;38:313-9. 47.GarnSM,HarperRV.Fataccumulationandweightgainintheadultmale. HumBiol1955;27:39-49. 48.GarnSM,YoungRW.Concurrent fatlossand fatgain.AmJPhysiol AnthropolNS1956;14:497-504. 49.GarnSM,Pesick SD.Comparisonof theBennIndexandotherbodymass indicesinnutritional assessment.AmJClinNutr1982;36:573-575. 50.GarrowJS.Treatobesityseriously.A clinicalmanual.Edinburgh: Churchill Livingstone,1981. 51.GarrowJS.Indicesofadiposity.NutrAbstrRev1983,53:697-708. 51a.GarrowJS,WebsterJ.Quetelet's Index (W/H2)asameasureof fatness.Int J Obes1985;9:147-153. 52.GarrowJS.Losing fat (letter).Lancet 1985;i:387. 53.Gezondheidsraad.Advies inzakeadipositas.StaatsuitgeverijDenHaag, 1984. 54.Goldrick RB,McLoughlinGM.Lipolysisandlipogenesis fromglucose in human fatcellsofdifferent sizes.JClinInvest1970;49:1213-1223. 55.GrauerWO,MossAA,CannCE,GoldbergHI.Quantificationofbodyfat distribution intheabdomenusingcomputed tomography.AmJClinNutr 1984;39:631-637. 56.Gurr MI, JungRT,RobinsonMP,JamesWPT.Adipose tissuecellularity in man: the relationshipbetweenfatcell sizeandnumber,themassand distribution ofbodyfatandthehistoryofweightgainand loss. IntJ Obes1982;6:419-436. 57.HartzAJ,RupleyDC,KalkhoffRD,RimmAA.Relationshipofobesity to diabetes: influenceofobesity levelandbody fatdistribution.PrevMed 1983;12:351-357. -3458.HartzAJ,RupleyDC,RimmAA.Theassociationofgirthmeasurementswith disease in32856women.AmJEpidemiol1984;119:71-80. 59.HartzA,BarboriakP,WongA et al.Theassociationofobesitywith infertility and relatedmenstrualabnormalities inwomen.IntJObes1979; 2:57-73. 60.HubertHB,FeinleibM,McNamaraPH,CastelliWP.Obesityasan independent risk factor forcardiovascular disease:a26year follow-upof participants intheFraminghamHeartStudy.Circulation1983;67:968-977. 61. IsaksonB.Thecase formoderate energy restriction. In:VanItallieThB, HirschJ(eds)Recentadvances inobesity IV.JohnLibbey,Londons,Paris, 1985. 62. JoosSK,MullerWH,HanisCL,SchullWJ.Diabetesalert study:weight historyandupper bodyobesity indiabeticandnondiabeticmexican americanadults.AnnHumBiol1984;11:167-171. 63.JungRT,Gurr MI,RobinsonMP,JamesWPT.Doesadipocyte hypercellularity inobesityexist?BritMedJ1978;2:319-321. 64. KalkhoffRK,HartzAH,RupleyD,KissebahAH,Kelber S.Relationshipof body fatdistribution tobloodpressure,carbohydrate tolerance,and plasmalipids inhealthyobesewomen.JLabClinMed1983,102:621-627. 65. KeysA, FidanzaF,KarvonenMJ,KimuraN,TaylorHL.Indicesof relative weightandobesity.JChronDis1972;25:329-343. 66. KissebahAH,VydelingumN,MurrayR,EvansDJ,HartzAH,KalkhoffRK, AdamsPW.Relationofbody fatdistribution tometabolic complications of obesity.JClinEndocrinMetab1982;54:254-260. 67. KrotkiewskiM,SjostromL,BjorntorpP,SmithU.Regionaladiposetissue cellularity in relationtometabolism inyoungandmiddle-agedwomen. Metabolism1975;24:703-710. 68. Krotkiewski M,BjorntorpP,SjostromL,SmithU. Impactofobesityon metabolism inmenandwomen.JClinInvest1983;72:1150-1162. 69.LaFontanM,Dang-TranL,BerlanM.Alpha-adrenergicantilipolyticeffect ofadrenaline inhuman fatcellsofthethigh:comparisonwith adrenaline responsiveness ofdifferent fatdeposits.EurJClin Invest 1978;9:261-266. 70. LanskaDJ,LanskaMJ,HartzAJ,RimmAA.Factorsinfluencing anatomic locationof fattissue in52953women.IntJObes1985;9:29-38. 71.LanskaDJ,LanskaMJ,HartzAJ,KalkhoffRK,RupleyD,RimmAA.A Prospective studyofbody fatdistributionandweight loss. IntJObesity 1985;9:241-246. 72.LapidusL,BengtssonC,LarssonB,PennertK,RyboE,SjostromL. Distributionofadiposetissueand riskofcardiovascular diseaseand death:a 12year follow-upofparticipants inthepopulation studyof women inGothenburg,Sweden.BritMedJ1984;289:1261-1263. 73.LarssonB,SvardsuddK,WelinL,WilhelmsenL,BjorntorpP,TibblinG. Abdominaladipose tissuedistribution,obesity,and riskof cardiovascular diseaseanddeath:a13year followupofparticipants inthestudyofmen born in1913.BritMedJ1984;288:1401-1404. 74. LewEA,Garfinkel L.Variations inmortalitybyweightamong750.000men andwomen.JChronDis1979;32:563-576. 75. ListerJ,Tanner JM.Thephysiqueofdiabetics.Lancet1955,ii:1002-1004. 76. LithellH,BobergJ.Thelipoprotein-lipase activityofadipose tissue fromdifferent sites inobesewomenand relationshiptocell-size.IntJ Obes1978;2:47-53. 77. LjunggrenH.Sexdifference inbodycomposition.In:Brozek JC (Ed):Human bodycomposition.Oxford,PergamonPress:129,1965. 78.MacDonald PC,DemanCO,HemsellDLetal. Effectofobesityonconversion ofplasmaandrostenedione toestradione inpostmenopausalwomen,withand withoutendometrial cancer.AmJObstetGynaecol1978;4:448-455. 79.NorganNG,Ferro-LuzziA. Principal componentsasindicatorsofbody fatnessand subcutaneous fatpatterning.HumNutr:ClinNutr1985;39C:4553. 80.OhlsonLO,LarssonB,Svardsudd Ketal.Theinfluenceofbody fat distributionon the incidenceofDiabetesMellitus- 13.5yearsof follow-upoftheparticipants inthestudyofmenbornin1913. Diabetes1985;34:1055-1058. 81.QueteletLAJ.PhysiqueSociale,Brussels.CMuquardt,vol 2:92,1969. 82. RabkinSW,Mathewson FAL,HsuPH.Relationofbodyweighttodevelopment of ischemicheartdisease inacohortofyoungNorthAmericanMenaftera 26yearobservationperiod: theManitobaStudy.AmJCardiol1977;39:452458. 83. Rebuffe-ScriveM,EnkL,CronaN,LonnrothP,AbrahamssonL,SmithU, BjorntorpP.Fatcellmetabolism indifferent regionsinwomen-effects ofmenstrual cycle,pregnancy,and lactation.JClinInvest1985;75:19731976. 84. RhoadsGG,KaganA.The relationofcoronarydisease,stroke,and mortality toweight inyouthandmiddleage.Lancet1983;i:492-495. -36- 85. RobbinsDC,HortonES,Tulp0,SimsEAH.Familialpartial lipodystrophy: complicationsofobesity inthenon-obese?Metabolism1982;31:445-452. 86. RookusMA,BuremaJ,DeurenbergP,vanderWiel-WetzelsWAM.The impactof adjustmentofaweight-height index (W/H2)for framesizeonthe predictionofbody fatness.BritJNutr1985;54:335-342. 87. RoyalCollegeofPhysicans.Obesity.JRoyCollPhys1983;17:1-65. 88. SeidellJC,BakxJC,deBoerE,DeurenbergP,HautvastJGAJ.Fat distributionofoverweightpersons inrelationtomorbidityand subjective health.IntJObes1985;9:363-374. 89. SilberbergR.Obesityandosteoarthosis:301-315.InMancini,Lewis, Contaldo, (eds). medical complicationsofobesity,Academic Press,London, 1979. 90. SimopoulosAP,VanItallieThB.Bodyweight,health,and longevity.Am InternMed1984;100:285-295. 91. SimopoulosAP.Thehealth implicationsofoverweightandobesity.NutrRev 1985;43:53-40. 92. SjostromL,SmithU,Krotkiewski M,BjorntorpP.Cellularity indifferent regionsofadipose tissue inyoungmenandwomen.Metabolism1972;21:11431153. 93. SkerljB.Age changes infatdistribution inthe femalebody.ActaAnat 1959;38:56-63. 94. SmithU,HammerstenJ,BjorntorpP,KraiJG.Regionaldifferencesand effectofweight reductiononhuman fatcellmetabolism. EurJClin Invest 1979;9:327-332. 95. SmithU.Regionaldifferences inadipocytemetabolismandpossible consequences invivo.In:HirschJ,van ItallieTB (eds). Recent advances inobesity research IV:33-36.JohnLibbey,London,Paris,1985. 96. SmithU,DigirolamoM,BlohmeG,KraiJC,Tisell L.Possible systemic metaboliceffectsof regionaladiposity inapatientwithWerner's syndrome.IntJObes1980;4:153-163. 97. StewartAL,BrookRH,KaneRL.Conceptualization andMeasurementofHealth Habits forAdults intheHealth InsuranceStudy:Vol.II,Overweight. SantaMonica,CA:TheRand Corporation,R-2374/2-HEW, July1980. 98. StewartAL.BrookRH.Effectsofbeingoverweight.AmJPublicHealth 1983;73:171-178. 99. Thiebaud D,DeFronzoRA,JacotE,GolayA,AchesonK,Maeder E,Jequier E, FelberJP.Effectof longchaintriglyceride infusiononglucose metabolism inman.Metabolism 1982;31:1128-1136. 100. TokunagaK,MatsuzawaY,IshikawaK,TaruiS.A novel technique for the determinationofbody fatbycomputed tomography. IntJObes1983;7:437445. 101. VagueJ. Ladifferenciation sexuelle,facteurdeterminantdes formesde l'obesite. Pressemed1947,-53:339. 102. VagueJ. Thedegreeofmasculinedifferentiation ofobesity-a factor determining predisposition todiabetes,atherosclerosis,goutanduric calculus.AmJClinNutr1956;4:20-34. 103.VagueJ,BoyerJ,JubelinJ,NicolinoC,PintoC.Adipomuscular ratioin human subjects.In:physiopathology ofadipose tissue:360-386.VagueJ (ed). ExerptaMedica,Amsterdam,1969. 104. VagueJ,VagueP,BoyerJ,CloixMD.Anthropometry ofobesity,diabetes, adrenal andbeta-cell functions.InDiabetes:517.RodriguezRR (ed)ICS 231. ExerptaMedica,Amsterdam,1971. 105. VagueJ,RubinP,JubelinJ, Lam-vanG,Aubert F,WassermanAM, Fondarai J.Regulationoftheadiposemass:histometricandanthropometricaspects. In: JVague,JBoyer (eds). The regulationoftheadipose tissuemass:296310. ExerptaMedicaAmerica. 106. VagueJ, CombesR,Tramoni,Angeletti S,RubinP.Thediabetogenic adipose mass. In:ManciniM,LewisB,ContaldoF (eds).Medical complicationsof obesity:45-58.LondonAcademic Press,1979. 107. Waaler HT.Height,WeightandMortality.TheNorwegianexperience.Acta Med Scand 1984,-Suppl679:1-56. 108. WellsHG.Adipose tissue,aneglected subject.JAmMedAssoc1940;114: 2177. -38- CHAPTER 2 ASSOCIATIONSOFMODERATEANDSEVEREOVERWEIGHTWITHSELF-REPORTED ILLNESSAND MEDICALCAREINDUTCHADULTS J.C.Seidell,C.P.G.M. deGroot,J.L.A.vanSonsbeek,P.Deurenberg, J.G.A.J.Hautvast AmJPublHealth 1986;76:264-269 ABSTRACT Dataonheight,weight,illness,medicalcareconsumption,anddemographic variables for19,126Dutchadultsaged 20yearsorolderwereobtained from threeannualHealth InterviewSurveys.Dataonseverely (BodyMass Index 30.02 2 40.0kg/m )andmoderately (BMI25.0- 29.9kg/m )overweight subjectswere 2 comparedwiththoseonnon-overweightpersons (BMI20.0-24.9kg/m ),taking intoaccounteffectofsex,age,andeducational level.Inmen,severe overweightwasassociatedwithhypertension,especially inmenunder 50yearsof age. Inwomen,severeoverweightwasassociatedwithhypertension,diabetes, varicoseveins,asthma/bronchitis,andhemorrhoids.Increasedutilizationof medical careandmedicationswerealsoassociatedwithsevereoverweight.For moderatelyoverweight subjects,theseassociationswerelessclearorabsent. INTRODUCTION Obesity isconsidered tobeoneofthemostprevalentmedicalandpublichealth 1-3 problems inaffluent societies although thisassumptionhasbeen 45 criticized. ' The importanceofobesityasapublichealthproblemhasbeen reviewed recently inthe reportsofexpertcommitteesonoverweightand obesity fi ? 7 inEngland, and theUSA, and inTheNetherlands .These reports recommend the 2 useof theBodyMass Index (BMIorQuetelet'sIndexinkg/m )asacrude indicator ofthebody'sfatcontent.Fromdataavailable fromseveral relatively small surveys,itcanbeestimated thatalargeproportionoftheDutchadult 2 7910 population isatleastmoderately overweight (BMI>25kg/m ).'' Itis important toknow ifoverweight indifferentdegreesofseverity isassociated withahigherprevalence ofchronicdisordersorotherhealthcomplaints, -39related inturntoanincreaseduseofmedical servicesandmedicines,than couldbeexpectedinsubjectswithaBMIwithinthedesirable range.Evaluation oftheriskscouldbehelpfulindecidingwhether todirectpublichealth programsatreductionofbodyweightthroughoutthepopulation,orwhetherthe effort shouldbeconcentratedontheseverelyoverweight. Inthisstudy,theassociationsofBMI(calculated fromself-reportedheightand weight)with reported chronic illnessandaspectsofmedical consumption were investigated,usingdataonarepresentativepopulationof19,126Dutch adults. METHODS Subjects Subjectswere intervieweesintheannualContinuousDutchHealth Interview Surveys.Eachyear since1981,theDutchCentralBureauofStatistics (CBS)has approacheda representative sampleofthetotalnon-institutionalized population inTheNetherlands.Thestudydescribedhereexcluded thosewithoutDutch nationalityandthoseunder20yearsofage.SubjectswhoseBMIexceeded40 2 kg/m wereexcluded (0.8%ofthetotal studypopulation).Close inspectionof theirdataaroused strongsuspicionsaboutthevalidityofthereportedweight (e.g.,weightexpressedinpoundsinsteadofkilograms).Theremaining populationofthesurveysof1981,1982,and1983comprised9,369menand9,757 women. IndividualswithaBMIbelowthedesirable range (i.e.<20kg/m)were 2 excluded fromthecomparisonofseverelyandmoderately overweightwith non-overweight sincetheyareknowntohave increased riskofmorbidity. ' ' Theaveragenon-response ratesforthesesurveyswereabout33percent.About 22percent refusedtoparticipate,8percentwerenotathomeand3percent wereunabletoanswerbecauseofmedical,linguistic,orother reasons.More 1415 detailsofnon-respondersaredescribed elsewhere. ' Sampling, Variables Samplingwasdone intwosteps.Inthefirststep,arepresentative sampleof allmunicipalitiesinTheNetherlandswastaken.Inthesecond step,asampleof householdsoutofevery sampledmunicipalitywastaken.Trained interviewers visited thesehouseholdsthroughouttheyear,askedthequestionstothemembers ofthehousehold,andnotedtheanswersinaprestructuredquestionnaire. 16 Furtherdetailshavebeendescribedelsewhere. 111213 -40- Inconstructing thevariablesused inanalysis,agewascategorized into intervalsof 20-34, 35-49,50-64,and65yearsorolder.Education servedasan indicator ofsocio-economicstatus:lowlevel=primaryschooland some occupational training;middlelevel= secondaryschool;highlevel=university andother tertiary institutions.BodyMassIndex (orQuetelet's Index)was calculated fromself-reportedheight (without shoes)andweight (without clothes),dividingweight inkilogramsbythesquareoftheheight inmeters. Q ThreecategoriesofBMIweredistinguished,using theclassificationofGarrow. 2 - BMI=20.0-24.9kg/m (Desirable rangeorgrade 0obesity). 2 - BMI«25.0-29.9kg/m (Moderateoverweightorgrade Iobesity). 2 - BMI- 30.0- 40.0kg/m (Severeoverweightorgrade IIobesity). Subjectswereaskedwhether theysuffered fromlong-standing illnesses, handicaps,or theconsequencesofaccidentslistedonacardwith 19items.Each personcouldmentionup tofourillnesses. Subjectswereasked the reasonsforuptosixconsultationswith thegeneral practitioner and,inaseparatequestion,withmedical specialists,inthethree months (twomonths inthe1983survey)preceding theinterview. Consultations bytelephone (exceptthose formakingappointments)wereincluded.Treatmentsby medical specialistsduringhospitalizationwereexcluded,butoutpatient clinicaltreatments (e.g.,firstaidandx-ray)were included. Subjectswereaskedaboutanyhospitalization (formorethanonedayandone night)intheyearpreceding theinterview.Subjectswereaskedaboutany medicinesprescribedbyaphysician inthe14dayspreceding the interview (excludingoralcontraceptives),and inaseparatequestionabouttheuseof medicineswithoutapresciption (adescriptionofchronicdisordersand categoriesofmedicalcare isgiven inAppendix IenII). Analysis Data fromthe1981surveywereused tomakeapreliminary selectionofthe medical consumptionvariablesapparently associatedwithoverweight inanalysis of contingency tables.The strengthsoftheassociationswereevaluatedwith the extendedversionoftheMantel-Haenszel chi-square statistic for contingency tables. Whenage-specific or stratum-specific estimatesof riskweredifferent fromone (p<0.10),thedependentvariable (diseaseormedical consumption)was selected for furtherstudy. Ina stepwise forwardmultivariate logistic regressionanalysis (usingthe 18 packageprogramBMDPLR), dummyvariableswereused forcategoriesofBMI,age, andeducational leveland interaction terms (BMIxageandBMIx education).The -41threecategorizedindependentvariablesandtheinteractiontermswereadded(p <0.10)orremoved(p>0.15)inpvalueorder.Thedependentdichotomous variablewaspresenceorabsenceofaparticulardiseaseorreasonformedical careoruseofatypeofmedicine. UsingthecoefficientsofregressiononBMIcategories,oddsratioswere calculatedadjustedfortheothervariablesinthemodel,comparingsevereand moderateoverweighttonon-overweight.Wecalculated95percentconfidence intervalsfortheoddsratios,usingthestandarderrorsof,andthe correlationsbetween,themultiplelogisticregressioncoefficients,according 19 totheproceduredescribedbyLemeshowandHosmer. Thesameprocedureoflogisticregressionanalysiswasfollowedfortheselected dependentvariables,usingthedatafromthecombinedsurveys.Thedata presentedbelowwerecalculatedfromthecombinedsurveysunlessstated otherwise. RESULTS Accordingtothecriteriausedinthisstudy,about4percentofadultmenand 6percentofadultwomeninTheNetherlandsareseverelyoverweight.The prevalenceofmoderateoverweightcanbeestimatedas34percentinmenand24 percentinwomen.Intables1and2,somecharacteristicsofthestudy populationareshownforthecombinedsurveys.Theprevalenceofsevereand moderateoverweightisinverselyrelatedtoeducationallevel. Table1.Meanweight,height,bodymassindex(BMI)ofthestudypopulation, bysex,combinedsurveys1981-83 men n =9369 women n= 9757 Mean Sd Mean Sd Weight(kg) 76,5 10,0 65,0 10,4 Height(cm) 177,3 24,3 7,7 2,9 166,1 BMI(kg/m2) 23,6 6,4 3,7 Age(years) 43,9 16,5 44,9 17,0 Characteristics T~P -42- (V 01 < •P c ai H a <M IH •H Q u 0 iu „ M S m t_ >» ~» c x TJ <u <D •o o (H b i/> 1 c en in 10 & 0 o in in SO m ii • U-l V) c_ > • « i/i t_ 0 Oi ~* ON i i O eg in m • II II >» II 1 ' » 1=^ *""S* o 01 <D •H a) 4-) ai u c •rH C i s0 T3 C a) C 111 s IH 0 01 > i a; > 3 01 T3 01 c VI o a> o> s 0 o e 0 u M-l « 4-> c 0) L) 1-1 a; a. i -* § D O M fc, 10 •p <0 •a ' — • 5^ 0 u U -43- Table2.Distributionofeducational levelswithindifferent categoriesof bodymassindex (BHI),according tosex BMI (kg/m) n Lowlevel Middlelevel High level Men 20.0-24.9 5214 46.3% 36.4% 17.3% 25.0-29.9 3275 62.5% 27.5% 10.0% 30.0-40.0 368 70.9% 22.0% 7.1% TOTAL 8857 Women 20.0-24.9 5349 62.1% 27.0% 10.9% 25.0-29.9 2379 81.1% 14.1% 4.8% 30.0-40.0 581 80.0% 17.1% 2.8% TOTAL 8309 Figure 1showstheagegroup-specificdistributionsofBMIcategories foreach sex.For subjectsyounger than65years,theprevalence ofmoderate and severe overweight increaseswithage. Resultsofthe stepwise logistic regressionanalysis,using thedataofthe combined surveys,arepresented inTables 3and 4.Table 3showsthevariables included inthelogisticmodel.Only themodels inwhichBMIappeared are presented. Interactionsareevaluated onamultiplicative scale.Table 4shows theodds ratios (with95percentconfidence intervals)formoderatelyand severelyoverweight compared tonon-overweightsubjects. -44- Table 3.Variables included inaforward stepwisemultiple logistic regressionmodel Variables Gender Dependent variable Number ofcases Variables included inthemodel H description BMIx BMI age education Chronical illnessess men hypertension diabetes women hypertension diabetes varicoseveins generalpractitioner* Reason for consulting medical specialist Hospitalization* * 116 (1.3% 474 (5.7% 152 (1.8% 197 (2.4% asthma 218 (2.6% haemorrhoids 123 (1.5% atleast 1illness Reason for consulting 281 (3,2% 2519 (30.3% men trunk/extremities women cardiovascular complaints 339 (4.1% 310 (3.4% routinehandlings 818 (9.8% men trunk/extremities 127 (1.4% women at least1consultation women at least 1hospitalization 736 (8.9%) ** back/joints/muscles Onlymodelsthat included BMIarepresented Duringthreemonthspreceding the interview *** Inoneyearpreceding theinterview 147 (1.7% 1900 (22.9% BMI> age educati -45- Table4.Odds ratios (OR)adjustedforageandeducational leveland95% 1 1 confidence intervals(CI)formoderately andseverely overweight subjectsvssubjectsnotoverweight Odds ratios (OR)and95%confidence intervals (CJiof: Variables Gender Dependentvariable Chronic men illness Moderately Severely overweight overweight OR CI OR CI hypertension 1.64 1.27-2.13 2.48 1.57-3.90 diabetes 0.61 0.71-0.92 1.39 0.70-2.76 1.84 1.55-2.19 2.63 1.96-3.53 women hypertension diabetes 1.39 0.95-2.03 3.34 2.13-5.22 varicoseveins 1.53 1.21-1.94 3.06 2.03-4.62 asthma/bronchitis 1.07 0.78-1.47 1.80 1.18-2.76 haemorrhoids 0.75 0.47-1.17 2.28 1.37-3.80 atleast1illness 1.15 1.03-1.28 1.72 1.43-2.06 Reasonfor consulting men trunk/extremities 1.24 0.97-1.59 2.17 1.38-3.43 general women cardiovascular 1.32 1.03-1.68 1.93 1.36-2.73 routinematters 1.18 1.01-1.40 1.68 1.31-2.14 practitionar * Reasonfor men consulting trunk/extremities 1.50 1.04-2.17 2.27 1.14-4.50 back/joints/muscles 1.38 0.97-1.96 2.17 1.14-4.09 atleast 0.91 0.82-1.02 1.51 1.28-1.78 1.14 0.98-1.33 1.72 1.38-2.16 medical .,. * specialist women Hospitali- women atleast zation** 1consultation 1consultation *Consultationsinthreemonthsprecedingtheinterview (detaileddescription ofdependentvariablesavailableinAppendixII) **Hospitalizationinoneyearprecedingtheinterview (detailsinAppendixII) 1 ? Moderateoverweight:BMI 25.0-29.9kg/m;Severeoverweight:BMI 30.0-40.0 2 7 kg/m .Thereference categoryisBMI20.0-24.9kg/m Source:dataobtained fromCentralBureauofStatisticsHealth Interview Surveys1981-83 (n=17166) -46Compared tonon-overweightwomen,severelyoverweightwomen reported that they suffered fromhypertension,diabetes,varicose veins,asthma/bronchitisand haemorrhoids more often,andconsulted their general practitioner about cardiovascular complaintsorroutinematters,andmedical specialistsforall reasonscombined.Theyhadalsobeenhospitalized moreoftenintheyear precedingtheinterview. Severely overweightmenreportedhavinghypertension moreoften thandid non-overweight men,andtheyalsohadconsulted theirgeneralpractitioneror medical specialistmore oftenabout complaintsofthetrunkorextremities.In addition,theywerealsomore likelytohave consultedamedical specialist about joint,back,ormuscle complaints.Levelofeducationappeared tomodify thislatter association:Table5showsthatthisassociationwasmost evident inmenwithanintermediate levelofeducation. Age appeared tomodifytherelationbetweenoverweightandhypertensionfor men;themagnitudeandstrengthofthisassociationdecreaseswithageandis notevidentatage65andover. Table 5.Odds ratiosinstrataofeffect-modifyingvariables (interactiontermsofthevariablewithBMI included intheregressionmodel Stratum-specific OddsRatios (95% Confidence Intervals) Moderately Severely Interaction termin Stratumeffect Overweight Overweight RegressionModel Modifier OR 95%CI OR 95% CI BMIxAge 20- 34yrs 4.17 1.94-8.95 6.94 1.85-26.09 35-49yrs 2.23 1.30-3.83 5.48 2.67-11.28 50-64yrs 1.15 0.79-1.66 1.75 0.89- 3.42 65+yrs 1.67 0.99-2.83 1.09 0.32- 3.72 Low 1.38 0.92-2.08 1.30 0.55- 3.07 Specialist for Middle 1.12 0.50-2.52 7.28 3.03-17.51 back/joint/muscles High 2.16 0.82-5.73 3.97 0.57-28.87 Dependent variable Hypertension (men) ConsultingMedical BMIxEducation (men) 2 Moderately overweight:BMI25.0- 29.9kgkg/m ;severelyoverweight:BMI30.0-40.0kg/m; the reference category wasBMI 20.0- 24.9kg/m 2 -47Table6.odds ratios,adjusted foreffectsofageandeducational level,fortheuseofprescribed andunprescribed drugsaccording tosexindifferent categoriesofBodyMass Index,in14dayspreceding interview Typeofmedicines Number ofcases Odds ratios (OR)and95%confidence intervals (CI) (%population) For stomach and intestinal 17 (0.6%) Moderately overweight Severely overweight OR CI OR 2.92 1.73-4.92 CI 2.96 1.00-8.85 complaints (unprescribed) Diuretics (prescribed) For cardiovascular/circulation 76 (2.7%) 2.00 1.52-2.65 4.57 3.30-6.34 147 (5.3%) 1.78 1.46-2.15 2.69 2.07-3.50 problems (including hypertension; (prescribed) Sleepingpill,sedatives, 125 (4.5%) tranquilizers (prescribed) For stomach and intestinal complaints; forwomenunder age50 ** (prescribed) 29 (1.1%) nouseforseverely overweight women 2 Moderatele overweight:BMI25.0-29.0kg/m ;severely overweight:BMI 30.0-40.0kg/m; Referencecategory:BMI 20.0- 24.9kg/m Agemodified theassociationwith overweight 2 -48Formoderateoverweighttheassociationsobservedbetweensevereoverweightand thevariouscomplaintswereweakerorabsent (table 4). Compared to non-overwightwomen,moderatelyoverweightwomen,reportedhypertensionand varicoseveinsmoreoftenand consultationswiththeirgeneralpractitioner aboutcardiovascular complaintsand routinemattersslightlymoreoften. Moderatelyoverweightmenunder 50yearsofage reportedmorehypertension than non-overweightmen, (table5)andconsultedamedical specialistmoreoften for complaintsoftrunkandextremities.Table6showstheodds ratiosforseverely andmoderatelyoverweightcompared tonon-overweightmenandwomen.Severely andmoderatelyoverweightmen reportedmoreuseofnon-prescriptionmedicines for stomachand intestinalcomplaintsthannon-overweightmen.Moderatelyand severelyoverweightwomen reportedmoreuseofprescribeddiuretics,sleeping pillsorsedatives,andmedicinesusedtotreatcardiovascular and circulation disordersthannon-overweightwomen.Moderatelyoverweightwomenunderage 50 usedmedicines forstomachand intestinalcomplaintsmoreoftenthan non-overweightwomen. DISCUSSION Basedontheself-reporteddata fromthecross-sectional surveys,itcanbe concluded thatsevereoverweight,especially inwomen,isaccompanied byan excessofchronicdisordersand increaseduseofmedical care.Moderate overweight isfarmorecommonthansevereoverweightand,although itisnota strong risk factoronanindividual level,itmayhave substantial impacton theprevalenceofsomechronicdisordersandthedemandonhealthcarefacilities.Theassociationsbetweenbodyfatnessandhypertensionanddiabetes mentionedarewelldocumented innumerousstudies. ' Data fromlarge prospective studies,liketheFraminghamstudy,donotconfirmour observation thatobesityandhypertensionareassociated onlyinmenunder 50yearsof 20 Varicoseveinsaremorecommoninoverweightwomenthaninthosewhoare 2122 notoverweight. ' Results froma recentstudyonfatdistribution indicate age. thatwhilehypertension,diabetes,andarthrosisweremoreprevalent inobese subjectswithanabdominal typeoffatdistribution,varicoseveinsweremore commoninobesewomenwhosefatwaspredominantly storedaroundhipsand thighs. 23 Theassociationsofsevereoverweightwithhaemmorhoidsandasthma/bronchitis inwomenarelessclearlyestablished intheliterature,although itiswell knownthat respiratory function isoften impaired inobese subjects. ~ -49Reducedphysicalactivity,resulting fromchronic illness,mighthave contributed totheexcessofweight.Therefore,associationsfound inthis studydonotnecessarily implycausalrelations. Comparing self-reporteddataofthehealthstatusofoverweightandnotoverweightsubjectsmaybebiased.Undiagnosed illnessmightbemorecommonin 27 thosewhoarenotoverweight ;whileontheotherhand,takingmedication (e.g.,anti-hypertensive)mightleadtounderestimationof risk.Although it hasbeenfound thatpeopletendtoreporttheirheightandweight inthe directionofculturallydesirablevalues (i.e.,slimwomen,tallmen), investigatorsgenerallyagreethatforlargegroups,self-reporteddataare sufficientlyaccurateanddonotsubstantiallyaffectconclusions. 28—32 Any bias introducedbytheunderreportingofweightwouldprobablybetoward the nullhypothesis.Thetrueprevalenceofmoderateand severeoverweight inThe Netherlandsmightbesomewhatunderestimated inthisstudy. Overweight,inthisstudy,wasalsofoundtobeassociatedwithmedical care notnecessarily resulting fromtheassociationswithchronicdisorders describedabove. More than70percentofDutchadultsarepublicly insured formedical careand 34 arenotpermitted tocontactaspecialistdirectly. The findingthatseverely andmoderatelyoverweightmenmoreoftenconsultedamedical specialistfor somecombinationsofcomplaints (Table4)impliesthatthese complaints generallywereconsidered seriousenoughbytheirgeneralpractitionar,who usuallyprovidesprimarymedical care,for referraltoa specialist.Forwomen, itwasshownthatsevereoverweightwasaccompaniedbyhigheruseofmedical careatall levels.Itisnotsurprising thatoverweightwomenwere found to usemorediureticsandmedicinesused totreatcardiovascular disorders (includinghypertension)sincethesedrugsareoftenprescribed forwomen in TheNetherlands.The findingthatseverelyoverweightwomenuse sleeping pills or sedativesmoreoftenthanwomenwhoarenotoverweightmaybethe consequenceofeither thesocialstigmathatobesitycarriesinaffluent societies, or thatstressfactors =ssfac mayhaveleadtoeatinghabitsthat,in 37,38 turn,causedweightgain. IntheNetherlands,itwas recently recommended thattreatment isindicated in everyonewho isseverelyoverweightand inthosewhoaremoderately overweight accompanied byincreased levelson risk factors,orwho suffer fromchronic disorders. The findingsofthisstudysupportthese recommendations.Once severeoverweight isestablished,however,itisverydifficult totreat successfully inthelongterm.Thepreventionofweightgain,especially in -50- youngadults,seemsamatterdeservingattention.Thegeneralpractitionar,who playsanimportantroleasafamilyphysicianandwhoprovidescontinuous primarycaretopatients,mightbeinanexcellentpositiontoidentifyand treatthosewhoareatriskofbecomingseverelyoverweight.Prospective studiesshouldbecarriedouttoinvestigatethepossiblebenefitsofsuchan approach. ACKNOWLEDGEMENTS ThisworkwasconductedunderacooperativeagreementbetweentheDepartmentof HumanNutritioninWageningenandtheCentralBureauofStatisticsinVoorburg, supportedbyagrantfromtheDutchPraeventiefonds.Theauthorswouldliketo thankJ.BuremaandS.Wardforhelpfulcomments.Partsofthispaperwere presentedattheannualmeetingoftheDutchworking-groupofepidemiologic research (WEON),May2-3,1985. REFERENCES 1.KannelWB:Healthandobesity:anoverview.In:ConnHL,DefeliceEA,Kuo PT (eds):Healthandobesity.NewYork:RavenPress,1983. 2.SimopoulosAP;vanItallieTB:Bodyweight.HealthandLongevity.Arm InternMed1984;100:285-295. 3.AhrensAH:Obesityandcoronaryheartdisease-newdimensions(editorial). Arteriosclerosis1984;4:177-179. 4.AndresR:Effectofobesityontotalmortality.IntJObesity1980;4; 381-385. 5.BergerM,BerchtoldP,GriesA,ZimmermanH.Indicationsforthe treatmentofobesity.In:BjorntorpP,CairellaM,HowardAN (eds):Recent AdvancesinObesityResearchIII.London:JohnLibbeyandCoLtd,1981. 6.RoyalCollegeofPhysiciansofLondon;Obesity.JRoyCollPhys1983;17: 1-58. 7.Gezondheidsraad.Adviesinzakeadipositas.'sGravenhage.Staatsuitgeverij, 1984. 8.GarrowJS:Treatobesityseriously-aclnicalmanual.Edinburghetc. ChurchillLivingstone,1981. 9.KokFJ,MatroosAW,vandenBanAW,HautvastJGAJ:Characteristicsof individualswithmultiplebehavioralriskfactorsforcoronaryheart disease:TheNetherlands.AmJPublicHealth1982;72:986-991. il- 10.KromhoutD:Obesitasencoronairehartziekten.HartBull1979;10:119-124. 11.GarrisonRJ,FeinleibM,CastelliWP,McNamaraPM:Cigarette smokingasa confounder ofthe relationshipbetween relativeweightand long-term mortality.TheFraminghamStudy.JAmMedAss1983;249:2199-2203. 12.G a m SM,HawthorneVM,PilkingtonJJ,Pesick SD:Fatnessandmortality in theWestofScotland.AmJClinNutr1983;38:313-319. 13. Sonsbeek JLAvan:"Nederlandersgemetenengewogen.<TheDutch,byheight andweight> MndberGezondheid (CBS)1985;6:5-18. 14. CentralBureauofStatistics.Thecontinuoushealth interviewsurvey;some introductoryaspects.MndberGezondheid1982;1:5-9. 15.AppelboomWJMJ.Quarterlydata,health interviewsurveys.Mndber Gezondheid 1983;2:40-45. 16.Sonsbeek JLAvan,StronkhorstLH.Statistische onderzoekingen.Methodische aspectenvandegezondheidsenquete.statistical investigations. Methodological aspectsofthehealthinterviewsurvey>. 'sGravenhage: Staatsuitgeverij,1983. 17.MantelN,HaenzelW. Statistical aspectsoftheanalysisofdata from retrospective studiesofdisease.JNatCancer Inst1959;22:719-748. 18.DixonWJ.BMDPStatistical software.BerkelyCA:UniversityofCalfornia Press,1981. 19. LemeshowS,HosmerDW.Estimatingoddsratioswithcategorically scaled covariates inmultiple logistic regressionanalysis.AmJEpidemiol1984; 119:147-151. 20.Ashley FW,KannelWB.Relationofweightchange tochangesinatherogenic traits.The Framingham Study.JChronDis1974;27:103-114. 21.VonMarshallM,UhligHE,HessH.UbergewichtundperiphereErkrankungen. Zeitschrift furAllgemeinmedizin1977;53:659-664. 22.Dall'Antonia F,OstuzziR,Bosello0.Obesityandvaricoseveinsofthe lowerlimbs.In:Mancini,LewisB,ContaldoM (eds):Medical complications ofobesity.London:AcademicPress:271-273,1979. 23. Seidell JC,BakxJC,deBoerE,Deurenberg P,HautvastJGAJ.Fat distributionofoverweightpersons inrelationtomorbidityand subjective health. Int J Obesity1985;9:363-374. 24. SharpJT,HenryJP,SwearySKet al;The totalwork ofbreathing innormal and obesemen.JClin Invest1964;43:728-739. 25.Gries FA,Berchtold P,BergerM:Adipositas-Pathophysiologie, Kliniek und Therapie.Heidelberg etc:SpringerVerlag,1976. -52- 26.BaeJ,TingEY,GioffridaJG.Theeffectofchangesinthebodycomposition ofobesepatientsonpulmonaryvolumeandventilatoryfunction.BullN.Y. AcadSci1976;52:830-837. 27.GarrowJS.Howtotreatandwhentotreat.In:MunroJF (ed.):The treatmentofobesity.Lancaster:MPTPressLtd:1-17,1979. 28.BiroG:Validityofself-reportedweightsandheightsinself-selected subjects.CommHealthStudies1980;4:46-47. 29.PaltaM,PrineasRJ,BermanR,HarmanP.Comparisonofself-reportedand measuredheightandweight.AmJEpidemiol1982;115:223-230. 30.PirieP,JacobsD,JefferyR,HarmanP.Distortioninselfreportedheight andweightinwomenandmen.IntJObesity1981;5:67-76. 31.StewartAL.Thereliabilityandvalidityofselfreportedweightand height.JChronDis1982;35:295-309. 32.StunkardAJ,AlbaumJM.Theaccuracyofself-reportedweight.AmJClin Nutr1981;34:1593-1599. 33.TuomilehtoJ,JalkanenL,TanskanenA,PuskaP.Selfestimatedbodyweight comparedtoactuallymeasuredbodyweightinapopulationsurvey (abstract).4thEuropeanNutrConference,Amsterdam:24-27,May1983. 34.RuttenFFH,GaagJvander.Referralsandthedemandforspecialist1carein theNetherlands.HealthServicesRes1977;12:233-250. 35.AllonN:Thestigmaofoverweightineverydaylife.In:WolmanBB (ed)Psychologicalaspectsofobesity:ahandbook.NewYorketc:Van NostrandReinholdCompany,1982. 36.FreedJ:Physicfactorsinthedevelopmentandtreatmentofobesity.JAm MedAss1947;133:369-378. 37.LupinskyL:Lifechangeeventsascorrelatesofweightgain.In:HowardA (ed):RecentAdvancesinObesityResearchI.London:NewmanPublishing Co:210-212,1975. 38.HuygenFJA:FamilyMedicine.Nijmegen:DekkerenVandeVegt,1978. 53AppendixI ListofchronicalillnessessorhandicapsshowntosubjectsintheC.B.S. HealthInterviewSurvey.Subjectscouldstatethepresenceofmaximal4 differentitems Descriptionofcomplaints 1.Asthma,chronicalbronchitisornasalcavityinflammation 2.Heartdiseaseorfunctionaldisturbancesoftheheart 3.Highbloodpressure 4.Stroke 5.Varicoseveins 6.Haemmorrhoids 7.Pepticulcerorotherchronicstomachcomplaints 8.Chronicintestinalcomplaints 9.Gallstones,gallbladderorliverdisease 10.Rupture 11.Renalcalculus,chronicalinflammationofthebladder 12.Prostatecomplaints (men),prolapse(women) 13.Diabetesmellitus 14.Thyroiddysfunction 15.Rheumatism,jointdisease,slippeddisc 16.Epilepsy,migraine,skindisease 17.Canceroustumoursorgrowth 18.Consequencesofaccident 19.Otherchronicaldisorders 34- Appendix II Classificationsof illnesses,complaintsoractionswhichwere characteristic fora reported consultationofageneralpractitionar ormedical specialist in 3monthspreceeding the intervieworhospitalization in1yearpreceedingthe interview Groupdescription Abbreviation (thesewere used inthetext) 1 Childhooddiseases Childhooddisease 2 Commoncold,flue,infalammationsofthroat,tonsils, Commoninfections/ cavityof jaw,frontal sinus,nasalcavity 3 Chronicalbronchitis,asthma,respiratoryillness inflammations Respiratory thighnessofchest,coughing,hay-fever,pneumonia, pulmonaryaffection 4 Heartdisease,stroke,hypertension,hypotension, Cardiovascular anemia,varicoseveins,heammorroids,soreleg 5 Toothache Toothache 6 Pepticulcer,duodenumulcer,stomachcomplaints Internalorgans/ indigestion,intestinal complaints,diarrhea, digestive tract constipation,gastroenteritis,liverandgallbladder disease,appendicitis 7 Renalcalculus,nephritis,cystitis,menstrual Urogenital abnormalities,fleshygrowths,menopausal complaints 8 Pregnancy,miscarriage,delivery,anticonceptiva, sterilization Reproduction/ fertility Disordersofthethyroidgland,diabetes,overweight Metabolicdisorders 10 Cancer Cancer 11 Complaintsordiseasesconcerningears,eys,speech 9 Ears/eyes 12 Allergy,eczemaor skindisease Allergy/skin 13 Backacheforvarious reasons,myelitis,rheumatoid Back/joint/muscles 14 Fracturse,injuries,concussionof thebrain, complaints,arthrosis,sprain,achingmuscles Accidents contusions,burns,intoxications 15 Insomnia,fatigue,vertigo,headaches,migraine, nerves Nervous/stress -5516 Complaintsofneck,shoulder,extremities,chest, Trunk/extremities ribs,hips 17 Painsnotincategoriesabove,itchings,fever 18 Injections,bloodpressurecontrol,recipies, Pain Routine generalexaminations,referrals 19 Operations Operations AppendixIII CategoriesofmedicinesusedintheDutchHealthInterviewSurveys(prescribed, andnotprescribed) a.Painrelieversandmedicinesusedtotreatfever(e.g.aspirin) b.Medicinesusedtotreatcoughs,commoncold,flue,anginaetc c.Medicineslikevitamins,mineralsandtonics d.Medicinesusedtotreatdisordesoftheheart,circulationandraisedblood pressure e.Diuretics f.Medicinesusedtotreatconstipation g.Medicinesusedtotreatstomach-andintestinalcomplaintsandindigestion h.Sleepingdrugs,sedativesandtranquilizers i.Antibiotics (likepencillin)andsulphonamides j.Medicinesusedtotreatskindisorders(includingeczema,itchings, dandruff) k.Medicinesusedtotreatrheumatismandarthritis 1.Medicinesusedtotreatallergy m.Medicinesusedtotreatbronchialastma n.Medicinesnotdescribedabove -56- CHAPTER3 UTILIZATIONOFPRIMARYHEALTHCAREOFOVERWEIGHTANDNON-OVERWEIGHTSUBJECTS -ASIXYEARFOLLOW-UPSTUDY J.C.Seidell,J.C.Bakx,F.J.A.Huygen,H.J.M.vandenHoogen,P.Deurenberg, J.G.A.J.Hautvast SocSciMed(submittedforpublication) ABSTRACT Inthisstudy,315adultmenand562adultwomenfromfourgeneralpractisesin TheNetherlandswerefollowedforsixyearsinacontinuousmorbidity registration.Follow-upbeganintheyeartheywereregisteredasoverweightby theirgeneralpractitioner.Theincidenceofnewspellsofillnessinthis overweightgroupwascomparedwiththatinacontrolgroupof438menand618 womenwho,foreachcalendaryearofstartoffollow-upwerematchedtothe overweightgroupaccordingtothegeneralpractisetheywerein,andtheirsex andage.Itwasshownthatnewspellsofillnessthatarenotlife-threatening orlongterm,weremorefrequentlyregisteredintheoverweightgroup.This differencewasalsoapparentafterstratificationforsex,age,andsocial class.Specificafflictionsthatwereregisteredmoreoftenintheoverweight groupwerecommoncoldandinfluenza,psychoneuroticcomplaints,skinproblems, myalgiaandlumbago,distortions(inmenandwomen),menstrualabnormalities(in women),andminorinjuries(inmen). Comparingscreeningdatafrom1978withtheoverweightandhypertensionstatus asgiveninthemorbidityregister,justpriortothescreening,indicatedthat thesensitivityoftheregistrationofoverweightandofhypertensionincreases withage(withtheexceptionofoverweightinwomen).Usingthescreeningdata, itcanbeinferredthatoverweightindicatesariskofhypertension,andthis riskisgreatertheyoungerthatobesityisconfirmed.Oddsratioscalculated theregistrationdataindicatedthatthesedatamaybebiased:however,no consistentpatternofthestrengthanddirectionofthisbiaswasseenin differentageandsexgroups.Itwasconcludedthatoverweightpatientsin generalpractisesdoshowmorenewspellsofillnessthanthosewhoarenot knowntobeoverweightbythegeneralpractitioners. INTRODUCTION The relationshipbetweenoverweight and the incidence ofvarious chronic disorders iswell known from reviewstudies [1]andhas recentlybeen confirmed inTheNetherlands [2].Theassociation ofoverweightwith less severe andmore commonafflictions islesswell known.Self-reported data from overweight persons indicated that theyvisited general practitioners andmedical specialistsmore frequently for specific reasons [3,4]. Self-reported datamight suffer frommisclassificationbiasand subjects canonlybeasked about previous medical care that coversashortperiod of time,because ithasbeen shown they soon forget [5], Inorder tostudymedical careandoverweightmore objectively we studied the incidence ofafflictions ina retrospective cohort study comparing registered overweight subjectswith subjects thathadnever been registered asoverweight. Ithasbeensuggested thatsuchdatamightalsobebiased by general practitioners whohaveatendency to register asoverweight those subjects who have ahigher morbidity. Inaddition, inoverweight subjectswho receive medical caremore oftenandmore specifically itmaybemore likely that certain disorderswillbedetected than innon-overweight subjects [6].Therefore,we paid special attention tothevalidityof the registration ofhypertension and overweight and tothepossibility ofdifferential misclassification of these conditions by thegeneral practitioners. Our studyalso indicates theextraworkload forgeneral practitioners that can beattributed totheiroverweightpatients. POPULATIONAND METHODS Morbidity registration Thecontinuous morbidity registration of theDepartment ofGeneral Practise of theUniversity ofNijmegenwas started in1967 intwopractises,andwas augmented by twoothers in1971.These generalpractices comprise about 12,000 patients andare located inabig industrial town,asmall town,urbanized countryside anda rural area. Inthese fourpractises,alldiagnoses and referralsare registered daily,according totheE-list,which isan adapted version of the International Classification ofDiseases [7]used inThe Netherlands.A distinction ismadebetweennewspellsof illnessand diagnoses and contactswithpatientsbecauseofan illness thathaspreviously been -58- recorded.Thisenabledustorestrictourstudytonewspellsofillnessonly. Inthispapertheresultsofanalysingalldefinitenewdiagnosesoveraperiod of6yearsretrospectivefollow-uparepresented.Totalmorbiditywasdivided intothreegroupsofseverityasclassifiedbytheDepartmentofGeneral Practise[8]: highestlevelofseverity=potentiallylife-threatening,or long-termillness;middlelevelofseverity=temporarilythreatening;lower levelofseverity=generallynotthreatening.Thegroupsofillnessstudiedin thispaperaredescribedinAppendicesIandII. Population Theoverweightgroupcomprisedallpatientswho,atanagebetween20and50 years,hadbeendiagnosedbytheirgeneralpractitionerasbeingoverweight.The criterionforoverweightwasaBodyMassindex(BMI=weightdividedbyheight 2 2 squared)exceedingabout26kg/m (forwomen)or27kg/m (formen).Weightand heightwasmeasuredbythegeneralpractitioner (eitheratroutineexamination orattheinitiativeofthepatientorthegeneralpractitioner).Theseweights andheightswerenotenteredinthemorbidityregistration. Foreachcalendaryearofstartoffollow-up,acontrolgroupthathadnever beenregisteredasoverweightintheperiod1967-1984wasselectedfromthe totalpopulation.Thesepersonswererandomlyassignedtostartingyearsof follow-upintheperiod1967-1978,followedbyafrequencymatchingprocedure doneaccordingtogeneralpractise,sexandageatstartoffollow-up(in5-year agegroups).Somecharacteristicsofthestudypopulationareshownintable1. Estimateofinformationbias In1977/1978patientswerescreenedforcardiovascular riskfactorsinthreeof thefourpracticesthatparticipateinthemorbidityregistration.Eighty percentofallpatientsagedbetween20and50yearswereweighed(withoutshoes andwearingindoorclothingonly)andtheirheightandbloodpressurewere measured.Bloodpressurewasmeasuredusingasphygmomanometer,ascommonlyused ingeneralpractices[9]withthesubjectsitting,after15minutes'rest. Diastolicpressurewasreadatthedisappearanceofsounds(KorotkovphaseV ) . Theaverageoftworeadings,taken15minutesapart,wasusedforclassification ofhypertension(cut-offpoint100mmHg,whichcorrespondstothecriterion usedinthemorbidityregistration).Detailsofthescreeninghavebeengiven elsewhere[10J. Subjectswhosediastolicpressurewasbelowthecut-offpoint forhypertension,butwhoserecordsshowedtheywerereceivinganti-hypertension medicationwereclassifiedashypertensivesinthescreeningdata.Atthe -59- screeningexaminationBMIwasused toidentifyoverweight subjects (cut-off 2 2 points:26kg/m forwomen,27kg/m formen). Statisticalmethods Category-specific andadjusted risk ratiosandodds ratios (comparing overweight withnon-overweight subjects)werecalculated andtheirconfidence intervals estimatedusingTaylor seriesapproximation forvariance,asdescribed elsewhere [11].Thenumbersofafflictionswerecomparedusing theWilcoxon rank-order testofunpaired samples [12].Pvalues (two-sided)weretakentobe significant whensmaller than0.05. RESULTS Table 1showsthedistributionofsubjectsintheoverweightand thecontrol groupovercategoriesofsocialclassandofage.Intheoverweightgroupthe distribution isshifted towardsthelower socialclasses. Table 1.Distributionofthe studypopulationovercategoriesofsocial class (according toprofessionoftheheadofthehousehold)andover categoriesofage (atstartof follow-up) Social class Women Men OverweightNon-overweight Overweight Non-overweight n(%) n[V) ni%J n(%) low middle high 359(63.9) 174(31.0) 29( 5.1) 296(47.9) 255(41.3) 67(10.8) 562( 100) 618( 100) 202(64.1) 208(47.5) 100(31.8) 177(40.4) 13( 4.1) 53(12.1) 315( 100) 438( 100) 68(21.2) 113(35.2) 140(43.6) 91(20.8) 160(36.5) 187(42.7) 321( 100) 438( 100) Age 20- 29 30- 39 40- 50 140(24.9) 183(32.6) 239(42.5) 562( 100) 163(26.4) 208(33.6) 247(40.0) 618( 100) -60- Tabl« 2.Relative risks,adjusted forageandsocialclass,foratleast onenewspellofillnessbeing registered; overweight persons versusnot overweight persons Women Men Ilnness RR(95% confidence Registeredas RR(95%confidence Registered as (codes) interval) exhibiting illness interval) exhibiting illness Overweight % Control Overweight % % Common cold& influenza 1.16(1.09 - 1.24) 1.16(1.06 - 1.26) (240-3, 245) Anemia (111) Psychoneurosis 0.80(0.52 -1.22) 1.09(1.01 -1.18) 7.7 66.2 1.40(1.22 - 1.61) 1.44(0.94 - 2.22) 8.3 1.52(0.84 -2.76) 1.17(0.94 -1.45) 25.0 1.83(0.88 - 3.82) (135) Otitis externa (182) Cystitis acuta (313) Menstrual 1.36(1.13 -1.63) 31.7 disorders (325, 327-9) Skinproblems 1.33(1.12 -1.59) 1.27(1.01 - 1.59) (370,372,377,381) Myalgia, lumbago 1.13(1.01 - 1.27) 1.39(1.22 - 1.57) (407, 408) Distortions 1.54(1.08 - 2.21) 1.51(1.02 - 2.24) (480) Minor injuries (4951 p <0.05 1.14(0.95 - 1.37) 1.27(1.08 -1.481 Control % -61- Intable2therelativerisksfortheincidenceofoneormorediagnosesofnew spellsofillnessduringthesixyearsoffollow-upareshownfortheoverweight groupvis-a-visthenon-overweightgroup.Althoughformostofthediseases showninthetabletherisksareonlymoderatelyelevated,thehighratesof incidencedoindicateconsiderablymorecasesofsuchillnessesinthe overweightgroup. Thepercentagesofpersonswhomanifestedatleastonenewillnessduringthe sixyearsoffollow-up,asshownintable3,showthatalmosteveryoneinthe studypopulationwasseenbythegeneralpractitioneratleastonce.This tendencywasmorepronouncedintheoverweightgroup(especiallyinthelower andmiddlesocialclasses). Table3.Percentageofpersonsexperiencingatleastoneillnessinthecourse ofsixyearsfollow-upinamorbidityregistration,stratified accordingtosocialclass.Adjustedforage.Illnessesclassifiedby severityasdescribedinAppendixII) Men Women Severityof Social Overweight Non overweight illness class low middle high % a "6 Overweight Non % overweight % low 97.4 90.3 92.9* ** 83.5 97.5 middle 95.0 92.8 ** 81.3 high 80.5 85.1 84.6 75.5 low 95.6 92.5 88.5* ** 83.9 92.3 middle 95.0 87.5* ** 76.8 high 72.4 80.6 84.6 71.7 low 17.0 15.6 15.3 17.8 * 10.7 15.1 middle 16.7 12.6 20.0 high 20.5 14.9 7.7 * p<0.10 (Nonoverweightcomparedtooverweight) **p<0.05 (Nonoverweightcomparedtooverweight) -G2- The totalnumberofcomplaints inthesix-yearperiod,as reflected inthe averagenumber ofafflictions reportedperperson,washigher fornewspellsof illnessintheoverweight groupthaninthenot-overweightgroup (exceptforthe most severe illnesses). Illnessesof lowerandmiddle levelofseverity showa higheraveragenumber ofafflictions inwomenthaninmenandadecrease in incidencewith increasing socialstatus. Table 4.Averagenumber ofafflictionsperpersonduringsixyearsof follow-up inamorbidity registration,stratifiedbysocialclass.Weighted averagesofagegroups.Illnessesclassifiedbyseverity (see Appendix II) Women Severityof Social illness low middle high Men OverweightNonoverweight OverweightNonoverweight class low 8.3 7.0 6.9 5.3 middle 7.8 5.7* 4.4 3.8 high 6.4 4.6 3.7 3.0 low 6.4 4.6 middle 4.6 3.7 high 3.4 5.9 4.4 4.2 3.0 3.6 3.6 2.0 * low 0.2 0.2 0.2 0.2 middle 0.2 0.1 0.2 0.2 high 0.3 0.2 0.1 0.1 * p <0.10 (Nonoverweightcompared tooverweight) **p <0.05 (Nonoverweight compared tooverweight) Note Average istotalnumberof registrations inasex/age/social classcategory, dividedbythenumberofsubjectsinthatcategory.Differencesbetween groups were testedusingtheWilcoxon rank test -63- Table5.Sensitivityandspecificityoftheregistrationofoverweightand hypertensioninthreegeneralpractises(Datafromascreeningin1978 andfromthemorbidityregisterjustpriortothescreening) Disorder Sex n Age %withhypertension atscreening Hypertension Men 16.7 98.5 96.8 13.0 66.0 93.7 50.6 96.4 349 350 397 7.4 4.3 3.7 33.3 99.4 76.9 96.4 40- 50 29 39 50 17.6 89.9 95.4 20- 50 1086 9.0 74.5 97.3 29 343 30- 39 333 40- 50 363 12.9 59.1 95.3 27.9 68.8 94.2 31.4 77.2 88.0 20- 50 1039 24.2 70.9 92.6 20- 9.17 93.8 89.9 18.00 92.1 87.7 40- 29 349 39 350 50 397 39.04 87.7 85.5 20- 50 1086 23.02 89.6 87.9 202030- Men Women % 29.2 40- Overweight Specificity % 1.8 7.2 2030- Women Sensitivity 29 343 39 333 50 363 1039 20- 30- Note(forTable5): Sensitivity=percentageofthosewiththedisorderatthescreeningthathad beenregisteredashavingthatdisorder. Specificity=percentageofthosewithoutthedisorderatthescreeningandwho hadnotbeenregisteredashavingthatdisorder. Personswhoatthetimeofthescreeningwereundergoingtreatmentforhypertensionwereregardedashypertensiveregardlessoftheirbloodpressure -64- Resultsfroma screening inthreepractises in1978showthat theprevalenceof bothhypertensionandoverweight increasewithage.Thepercentageof subjects ashypertensive atthescreeningwhohadbeen registered inthemorbidity registerbecauseofoverweight increasedwithage.Theaccuracyofthe registrationofhypertensionappeared tobemuchbetter forwomenthanformen. The samewas trueforthe registrationofoverweight, registeredwasgceater thanthepercentageofcorrectly registeredhypertensives (table5 ) . Calculationsof theodds ratios forthepresenceofhypertension inoverweight subjectscomparedwithnon-overweight subjects (using thescreeningdata) indicated thatthestrengthofthecorrelationbetweenoverweightand hypertensiondecreaseswithage.Theodds ratioscalculated fromthe registrationdatashowedthatthistrendwaslessclearinmen,whileinwomen, an increase inodds ratioswithagewasobserved.Our resultsshowthatusing the registrationdatatocalculatethecorrelationbetweenoverweightand hypertensionwould resultsinunderestimate forwomenaged20- 29yearsand an overestimate formenaged 40- 50years. Table6.Odds ratiosforpresenceorabsenceofhypertension inoverweight subjects comparedwithnon-overweight subjects.Calculationsusing data frommorbidity register anddata froma screening in1978 Men Women Age Registration Screening Registration Screening Odds 95%CI Odds 95%CI Odds 95%CI Odds 95%CI ratio ratio ratio ratio 29 1.88 0.36-9.92 39 2.34 0.94-5.84 40- 50 3.94 2.27-6.82 20- 7.90 2.61-23 30- 1.82 0.55-5 20- 50 3.28 2.09-5.16 2.93 1.71-4 4.52 2.96-6.90 91 99 98 8,05 1.57-41 2.36 1.16-8 2.73 1.50-4 37 7.3 1.76-30 40 4.94 2-08-11 99 1.28 0.67-2 3.11 1.90-5.08 2.89 1.81- 4.64 45 73 49 DISCUSSIONAND CONCLUSIONS Inthisstudy itwas shownthatpatientsfromfourgeneralpractiseswhowere registered asbeingoverweightbytheir general practitioner displayed more new spellsof illnessduring sixyearsof follow-up thanpatientswhowere not registered asoverweight.Thesedifferenceswere alsoapparentwithinageand social classgroups.The registration inthese fourpractises is continually monitored andevaluated and isdoneaccording tostandardized criteria. Itis thereforevery likely thatthesedifferences are real.Nevertheless,itisnot possible toinfer acausal relationship betweenoverweightand the incidence of common illness.Personal characteristics notmeasured inthisstudycouldbe responsible forbothoverweight and illness. Some of theassociations between overweight and illness thatwereobserved in thisstudy (Table2)havebeenobservedbefore.Theassociation between overweight andmusculo-skeletaldisorders [3,13,14],skinafflictions[15], distortions andminor injuries [16,17]andmenstrual abnormalities [18,19]have been reported.Theassociation betweenoverweight andpsychoneurotic complaints isoften citedbut isdifficult tointerpret. Itisknownthatbeing overweight may causemuchdistress,especially inwomen [20,21,22].Thishasbeen confirmed intheNetherlands,where ithasbeen shownthatadultoverweightwomen reported that theyused sleepingpills,sedatives or tranquilizersmore often than non-overweightwomendid [3].Youngoverweight adultwomen reported having headaches andnervousbreakdownsmoreoftenthannon-overweightwomen [4].On theother hand, itisknownthatmood and certainmental statesmay contribute totheonset ofoverweight [23,24].Our finding thatthis relationshipwasmore pronounced inmen than inwomen isnotconfirmed intheliterature. Itmust be noted,however,thatmost researchonoverweight in relation to psychosocial problemshasbeencarried out inwomen.Huygenandhis colleagues [25] demonstrated that subjectswith recurrent psychosocial complaintsvisited their general practitioner more oftenbutwerealsomoreoftenoverweight.Overweight, psychoneurotic complaintsand other illnessesmay thusbe interrelated and it is notpossible to interpret theassociations between these intermsofcausality . Franksandhis colleagues [6]havepointed outaproblem thatmayarise from the use ofdataderived fromdiagnoses.They found that frequency ofvisits tothe general practitioner andpercentagewithpsychosocial complaints differed between obesepatients according towhether their obesityhadbeendiagnosed. In their study itwasclear thatthe registered obeseand thenot registered obese groupsdiffered inage (menandwomen)and social class (men). Inourdata,age -66andsocialclassvrereshowntobe importantassociatesof registered morbidity andmight thereforeexplain,atleastpartly,the findingsofFranksandhis colleagues.A relatedargumentwasputforwardbyGarrow [26],whoproposed that practitionersmayexamineoverweightsubjectsmorethoroughly fordisorders they suspecttobe related tooverweightthantheyexaminesubjectswhoarenot overweight.Onedisordermostlikelytobeaffectedbysuch informationbias is hypertension.Fromourcomparisonofdata fromascreeninganddata fr<3mthe morbidity register justprior tothescreening itwasevident thattherewas misclassificationofoverweightandhypertension,butnoconsistentpattern through theageand sexgroupscouldbe found.Theaccuracyofthe registration ofhypertensionandoverweight increasedwithage,andtherebywith the prevalenceofhypertensionandoverweight,andwasmoreaccurate inwomenthan inmen.This findingprobably reflectsthatwith increasingagepatients receive more routineexaminations (includingbloodpressuremeasurements)andcontact theirphysicianmoreoften.Thesameistrueforwomencompared tomen. Comparingtheodds ratioscalculated fromthescreeningdata tothose calculated fromthe registerdata revealed thatinsomeagegroupstheassociationbetween overweightandhypertension issomewhatunderestimated and inothers somewhat overestimated.Conclusionsbasedontheregistrationdatawoulddiffer fromthe screeningdata inwomenaged 20-29yearsandinmenaged 40-50years.Inthe other sexandagegroupstheassociationswerequite similar.Informationbias due todifferentialmisclassification ofhypertension inoverweightand non-overweight groups ispresentonlytoalimitedextent.Of coursethese comparisonsareonlyacrudewaytodetectaconsistentbiasandarebased on datameasuredononeparticular occasion.Itisverylikelythat,duetothis screening in1977-1978 theaccuracyofthe registrationofbothhypertension and overweight inthemorbidity registrationhasimproved considerably. Inaddition, hypertension isoneofthedisordersthat ismostlikelytosuffer frombias, and itisevenlesslikelythatasimilarbiasispresent inother associations observed inthisstudysuchastheassociationbetweenoverweightand common coldand influenza,anassociation that isprobablynotanticipatedbygeneral practitioners. It remainstobeestablishedwhether theassociationsobserved inthisstudyare causedbyoverweightorwhetheroverweightsubjectsaremerelymore hypochondriacal thannon-overweight subjects.Furtherevidence foracausal relationshipmaybeobtained fromstudiesinwhichtheeffectsofpreventionand treatmentofoverweight ingeneralpractisesontheuseofmedical careare studied.Suchevidencemaycontribute toaneventual reductionofutilizationof primaryhealthcare inpractises inwhichphysicianspaymuchattentiontothe reductionoftheprevalenceofoverweightamongtheirpatients. ACKNOWLEDGEMENTS Theauthorswould liketothank J.Burema forstatisticaladvice,P.Hoppener for computerassistance,J.Burrough forassistance inthepreparationofthe manuscript,personnel fromtheDept.ofGeneralPractiseoftheUniversityof Nijmegenforcollecting thedata,andtheDutchPraeventiefonds for financial support. REFERENCES 1.National InstitutesofHealthconsensusdevelopment conference statement.Healthimplicationsofobesity.Ann InternMed1985,-103:1073. 2. SeidellJC,BakxJC,DeurenbergP,HoogenHJMvanden,HautvastJGAJ, StijnenT.Overweightandchronic illness-a retrospectivecohortstudy withafollow-upof6-17 yearsinmenandwomen initially20-50yearsof age.JChronDis1986,inpress. 3.SeidellJC,DeGrootCPGM,VanSonsbeekJLA,Deurenberg P,Hautvast JGAJ.Associationsofmoderateand severeoverweightwith self-reported illnessandmedical care indutchadults.AmJPublicHealth1986;76:264. 4.DeurenbergP,vanPoppelG,HautvastJGAJ.Morbiditeitensubjectief welbevindenbijjongevolwassenenmetovergewicht.NedTijdschrGeneeskd 1984;128:940. 5.VandenBergJ.Memorybiasinreportingthenumberofconsultations withaGP.MndberGezondheid (CBS)1983;5:24. 6. FranksP,Culpepper L,DickinsonJ. Psychosocialbias inthediagnosis ofobesity.J FamPract1982;14:745. 7.Morbidity statistics fromgeneralpractice.Secondnational study 1970-71.London.HerMajesty'sStationaryOffice,1974. 8.NijmeegsUniversitairHuisartsen Instituut.Epidemiologic inde huisartsenpraktijk.Gewone ziekten.Eenaantalmorbiditeitsgegevensuit eenviertal huisartsenpraktijken.Nijmegen:NUHI,1980. 9.Hunyor SN,FlynnJM,CochineasC.Comparisonofperformanceofvarious sphygmomanometerwith intra-arterialbloodpressure readings.BritMedJ 1978;2:159. 10.VanReeJW.HetNijmeegsInterventieproject (thesis)Nijmegen,1981. ! -6811.KleinbaumDG,KupperLL,MorgensternH.EpidemiologicResearch Priciplesandquantitativemethods.LifetimelearningPublications, BelmontCA,1982. 12.SnedecorGW,CochranWG.StatisticalMethods.TheIowaStateUniversity Press,Ames,1980. 13.AroS,LemoP.Overweightandmusculoskeletalmorbidity:atenyear follow-up.IntJObes1985;9:267. 14.SilberbergR.Obesityandosteoarthritis.InMedicalcomplicationsof obesity(editedbyManciniMetal):301.AcadPress,London,1979. 15.HanzlickovaL,KrizekV,StepanekP.Dermatologischebefundebei fettleibigkeit.MunchMedWschr1967,-199:586. 16.SeidellJC,BakxJC,DeurenbergP,BuremaJ,HautvastJGAJ,HuygenFJA. Therelationbetweenoverweightandsubjectivehealthindutchadults. AmJPublicHealth1968,inpress. 17.HenschelA.Obesityasanoccupationalhazard.CanadJPublic Health1967;58:491. 18.HartzAJ,BarboriakPN,WangA,KatayamaKP,RimmAA.Theassociation ofobesitywithinfertilityandrelatedmenstrualabnormalitiesinwomen. IntJObes1979;3:57. 19.FriedlKE,PlymateSR.Effectofobesityonreproductioninthefemale. JObesityandWeightRegulation1985;4:129. 20.StewartA,BrookRH.Effecrtsofbeingoverweight.AmJPublicHealth 1983;73:171. 21.RodinJ,SilbersteinL,Striegel-MooreR.Womenandweight:anormative discontent.NebraskaSymposiumonMotivation.LincolnNE,Universityof NebraskaPress,1985. 22.AllonN.Thestigmaofoverweightineverydaylife.In:Psychological aspectsofobesity:ahandbook (editedbyWolmanBBetal).VanNostrand ReinholdCompany,1982. 23.LoweMR,FisherEB.Emotionalreactivity,emotionaleatingandobesity: anaturalisticstudy.JBehavioralMedicine1983;6:135. 24.VanStrienT.Eatingbehaviour,personalitytraitsandbodymass.Swets NorthAmericaInc.Berwin,1986. 25.HuygenFJA,HoogenHJMvanden,LogtATHvande,SmitsAJA.Nerveus functioneleklachtenindehuisartsenpraktijk.NedTijdschrGeneeskd 1984,-128:1321. 26.GarrowJS.Howtotreatandwhentotreat.In:Thetreatmentofobesity (editedbyMunroJF).MTPPress,London,1979. -69- Appendix I.Descriptionofcodesandgroupsofcodesused inanalysis Code Description 111 Irondeficiency anemiaandotherhypochronic anemia,excl.anaemia in 135 Emotional disorderswithorwithout somatic symptoms (or pregnancy. reactions),('Nervous-functional complaints'),incl.nervous reactions, psychogenic reactions,functional reactions,hypochondriasis, neurasthenia,overwork. 182 Otitisexterna,excl.thatasa resultofperforation. 240 Coryza,commoncold,incl.rhinitis,sore throatwithout temerature, 241 Febrile commoncoldand influenza-like illness,influenza,pharyngitis. 242 Tonsillitis,including lymphadenitis colly,tonsillar abcess,excl. 243 Sinusitis,acute. excl.allergic rhinitisandglueear. acute lymphadenitis. 245 Epidemic influenza. 313 Urinary infection,acute (acute cystitis),excl.pyelitisand acute 325 Dysmenorrhoea,excluding premenstrual symptoms. 327 Irregular menstruation, incl.amenorrhoeanotduringpregnancy and less 328 Menorraghia. 329 Menopausal and climacterialsymptoms. 381 Dermatitis,incl.diaper rash,intertrigo,allergic eczema,contact pyelonephritis. than 4months,excl.amenorrhoea more than 4months. dermatitis,varicose eczema,nickeleczema, rhagades,seborrhoeic dermatitis inadults,dermatitisdue tothe sun.Excl.dermatitisdue to occupation ordrugs,urticaria,insectbites,pigmented naevus,scabies, erysipelas,dystrichotic eczema,strophulus,sunburn,allergic sun reaction,herpeticdermatitis. 370 Furuncle,carbuncle (boil). 377 Skin infections:pyoderma,folliculitis,secondary impetigoof theskin, 372 Cellulitisandabscesseswithout lynphadenitis, incl.sebaceouscyst, excl.dermatophytosis,pilonidalsinus. pompholyx,sweatglanddisease,para-anal abscess,erysipeloid,excl. cellulitis of fingerand toeandnon-infected cellulitis. 407 Acute lumbago,excl.prolapsed intervertebral d i s c , sciatica. 408 Fibrositis,incl.myositis,myalgia,tenditisaround shoulder,stiff neck,lowback pain,overstrained muscle;excl.tenosynovitis. 480 Sprainsand strains,incl.effusionafter injury. 495 Small superficial injuriesorcontusions,blisters,excl.non-traumatic epistaxis. * Illnesseslooked atbut toofewforanalysis:infectiousdiseases,allergies, thyroiddisfunction,metabolicdisturbances,diseasesofblood and bloodforming organs,psychiatric diseases,addiction,illnessesofnervous systemand senses,digestive tract complaints **Chronicdiseases studiedwith longer follow-up:neoplasmata,diabetes mellitus,gout,cardiovascular disease,arthrosis -71- Appendix II.Morbidity subdivided intolevelsofseverity Levelof severity Complaint Codesof complaints inE-list lower skininfections anemia 021 111 psychoneurotic complaints 135 otitisexterna 102 varicoseveins 224 hemorrhoids commoncoldwithout fever gastroenteritis 225 240 289 cystitisacuta 313 menstrualdisorders 325-9 eczema middle 379-81 acnevulgaris 385 minor injuries 495 herpessimplex diabetesmellitis 014 091 gout commoncoldwith fever,influenza 093 241-5 bronchitis 247-8 prolapseduterus 324 skindisorders 370 psoriasis skeletalproblems 382 406-8 distortions higher allmalignanttumors cerebrovascular accident coronaryheartdisease vascular complaints 480 050-69 155 211-3 221-3 all fractures (excludingwristandcollarbone) 410-77 -72- CHAPTER4 OVERWEIGHTANDCHRONIC ILLNESS-A RETROSPECTIVECOHORTSTUDY,WITHA FOLLOW-UP OF 6-17YEARS,INMENANDWOMENOF INITIALLY20- 50YEARSOFAGE J.C.Seidell,K.C.Bakx,P.Deurenberg,H.J.M.vandeHoogen,J.G.A.J.Hautvast, T.Stijnen J ChronDis1986 (inpress) ABSTRACT A retrospective cohort-studywitha follow-upof6-17yearswas carriedoutin fourgeneralpractices inTheNetherlands intheperiod 1967-1983.Intotal 317overweightmenand 565overweightwomenwere followed ina continuous morbidity registration,startingintheyear theywerediagnosed asoverweight (atage20- 50years). Incidenceof illnessesinthisgroupwascompared to that inacontrol group (444menand627womenwithout registered overweight), matched onsex,ageandcalendar-year atstartoffollow-up. The incidenceof registeredmorbidity intheoverweightgroupwashigher for diabetes,gout,arterioscleroticdisease,arthrosis formenandwomen,andalso forvaricoseveins forwomen.IncreasingBMIatstartof follow-upwas associatedwith increased risk formost illnessesunder study.Forgoutand arterioscleroticdisease inmen,overweightappeared tobea risk factorat lower levelsofBMI thaninwomen. INTRODUCTION Recent reportsofexpertcommitteeshave stressed thatoverweightandobesity are importanthealthhazardsinaffluent societies [1,2,3].Manyprospective studieshave shownthattheoverweightandtheobese tendtohave shorter life spans [3],andanexcess risk formorbidity.Fromthe reportsor reviewsonthe risksofoverweight itisclear thatobesity isconsidered tobea risk factor formany illnesses,butthatthese riskshave rarelybeenstudied intermsof 2 BodyMass Index (BMIinkg/m )inafollow-upsituationoflongduration.In thisstudya retrospective cohortanalysiswascarriedout,usinga continuous morbidity registrationwith standardized criteria fordiagnosesofdiseases in fourgeneralpractices.The incidence insubjects,registered asoverweight -73- duringafollow-upperiodof6-17years,wascomparedtothatinacontrolgroup withoutregisteredoverweight.Withtheuseofweightandheightdataof overweightsubjects,measuredbythegeneralpractitioners,animpressionofthe risksindifferentcategoriesofBMIcouldbeevaluated. METHODS Forthisstudy,informationwasobtainedfromtheContinuousMorbidity RegistrationoftheDepartmentofGeneralPracticeoftheUniversityof Nijmegen.Infourpractices,alldiagnosesareregisteredonadaytodaybasis. Thesemorbiditydatacanbelinkedtoacomputerizedcontinuousregistrationof thepatientfiles(whichincludesup-to-dateinformationondemographic variables).Twopracticesstartedtheregistrationin1967,twoothersjoinedin 1971.Thetotalpopulationofthefourpracticestogethercomprisesabout12,000 patientsandisconsideredtoberepresentativeoftheDutchpopulation,with respecttothedistributionofageandsex.Onepracticeissituatedinan industrialtown(approximately180,000inhabitants),oneinamiddlesizedmixed industrial/ruraltown(approximately10,000inhabitants)andthetwoother practicesaresituatedinruralareas.Diagnosesareregisteredaccordingtothe E-list [4],whichisaversionofthe'InternationalStatisticalClassification ofDiseasesandCausesofDeath'adaptedforuseinTheNetherlands.A descriptionofthediagnosticcriteriafordiseasesstudiedinthispapercanbe foundintheAppendix.Twogroupsofpatients,anoverweightgroupandanonoverweightgroup,wereretrospectivelyfollowedup.Theoverweightgroup comprisedofallpatientswhowereregisteredasoverweightbytheirgeneral practitionersintheperiod1967-1978.Thethresholdcriteriafordiagnosisof overweight(relativeweight>115%)correspondstoaBMIbetween25.0-26.0 2 2 kg/m (forwomen)orbetween26.0-27.0kg/m (formen).Overweightwasonly registeredwheninitialdiagnosedoverweightpersistedinthesubsequentyear. Overweightsubjectswerefollowed-upinthemorbidityregistration,startingin theyeartheiroverweightstatusappearedforthefasttimeintheregistration (t0forthesepersons). Theoverweightpopulationwasrestrictedtopatientsforwhomoverweightwas registeredforthefirsttimebetweenages20-50years,andwhocouldbe followedforatleast6yearsintheregistration. Dataonheightandweightwereobtainedfromindividualpatientrecords.Dueto omissionsofgeneralpractitionersontheirpatientrecordsorrecordsmissing wasperformedwithregardtoageatstartoffollow-up(in5yearagegroups), -74sexandgeneralpractice.The ratioof thecell frequencies inthe576 (6*2*4*12)cellsofthecontrolgroupcompared tothecell frequenciesof the overweightgroupwasalwaysbetween1.0 and 1.5 (average1.3).Toobtainavalid control group,informationonnon-overweightstatusinthemorbidity registration throughoutthecompleteobservationperiod 1967- 1983wasused. Therefore,theoratically,thecontrolgroupcouldnotbecontaminatedwithnon overweight subjectswhosubsequentlybecameoverweight. Boththeoverweightandthecontrolgroupwere restricted,beforematching,to persons stillpresent inthegeneralpractices.Theywere senta questionnaire (response:70%;detailsaredescribedelsewhere) [7a].Thosewhowere candidates forthestudypopulation,butdiedbefore theendoftheobservationperiodwere firstanalysed separatelyandwerelateradded tothestudypopulation.Social classwasdefinedastheprofessionoftheheadofthehousehold and ratedas low (unskilledand skilledworkers);middle (lowerandmiddleemployees)orhigh (highemployees). Statistical analysis Since lengthof follow-updifferedbetweenindividuals,statistical techniques foranalysing survivaldatawereapplied tostudytheincidenceof registrations ofchronicdiseases fortheoverweightand thecontrolgroup [8].Using these techniques itwaspossible touse informationonthecompleteobservationperiod ofupto17years (1967- 1983).Endpoints intheanalysiswereeither the first occurrenceoftheparticulardiseaseunder studyor theendof theobservation period (1/1/1984).Onlypersonswhowere,according tothe registration,freeof thediseaseunder studyatthetimethefollow-upstarted,wereused for follow-up. Analysiswasperformed formenandwomen separately,and foragegroups (based onageatstartoffollow-up)separatelyaswellasforallagestogether. Kaplan-Meier estimatesofthe 'disease-free'proportionof thepopulationwere calculated foreachyearof follow-up,usingtheSAS-programPHGLM [9]. The logrank statistic [10]wasused totestdifferences,onayear-to-year basis,betweencumulative incidencesofthediseasesunder studyofthe overweightand thecontrolgroups.TheproportionalhazardsmodelofCox [11] wasapplied toestimatethe riskofoverweightand tocontrol for potentially confoundingeffectsofsex,age,socialclassandpractice,usingdummy variables forcategorically scaledvariables.Relative riskswere calculated as the ratioof the incidencedensities (incidencedividedbynumber ofperson yearsobserved)intheoverweightand thecontrolgroup.Itappeared that32men -75- (11.9%)and23women(5.9%)belongingtothecontrolgroupcouldnowbe consideredoverweight.Analyseswererepeateddeletingthese'falsenegative' subjectsfromthecontrolgroup.Endpointswerefirstregistrationsofthe followingchronicdisorders:anginapectorisormyocardialinfarction;gout, diabetesmellitus,varicoseveins;arthrosis;stroke;chronicalbronchitisand myodegeneratiocordis.Norelationbetweenoverweightandthelatterthree diseaseswasfound.Resultsoftheanalysesofthesediseasesareomitted. RESULTS InTable1somecharacteristicsofthestudypopulationareshown.The differencesbetweentheoverweightandthecontrolgroupindistributionover thesocialclassesreflectaninverserelationshipbetweenlevelofprofession andbodyfatness. Tablela.Ageandnumberofyearsoffollow-upofthestudypopulation Men Overweight group 37.6 8.0 Overweight group mean SD Ageatstartof Women Control group mean SD 37.4 8.1 Control group mean SD 36.8 8.6 mean SD 36.1 8.7 follow-up Yearsoffollow-up 10.6 3.1 10.4 3.0 12.0 3.1 11.3 3.1 Tablelb.Distributionofpersonsinthestudypopulationoversocialclasses Socialclass n % n low 203 64.0 210 47.3 361 63.9 297 47.7 middle 101 31.9 179 40.3 175 31.0 260 41.5 high 13 4.1 317 100.0 55 12.4 29 5.1 70 11.2 444 100.0 565 100.0 627 100.0 IT -76- BMI'satstartoffollow-upcouldbecalculated for83.3%ofoverweightmenand 2 88.3%ofoverweightwomen.MeanBMIofoverweightmen (28.7kg/m ;SD2.0)was 2 lower (p<0.05) thanthemeanBMIofoverweightwomen (29.3kg/m ;SD3.4). BMI'softheoverweight incidentcaseswerecalculated.Thedistributionsover categoriesofBMI fortheoverweight casesofvariousdiseasesand ofall overweightpersonsareshowninTable2. Table 2indicatesthat,forwomen,there isanincrease inrisk foralldiseases shown,except forgout,withhigher levelsofbody fatness.Theproportionof female casesisclearlyelevated,compared tothecontrolgroup,andcompared to 2 lower levelsofBMI,whentheBMI ishigher than30.0kg/m .A similartrend is seen fordiabetesandarthrosis inmen.The incidenceofgoutand arteriosclerotic disease,however,wasparticularlyhighforoverweightmen in theBMIcategory27.0- 29.9kg/m2. InTable 3thelogrank statistics fordifferences inthecumulative proportions ofpersonswith thediseaseunder studybetweentheoverweightand control groupsaregiven (foragegroupsseparatelyandallagescombined).Differences weregreater foragegroupsabove 30years. Adding thepersonswhodied intheperiod 1967- 1983tothestudypopulation in the retrospective cohortanalysisdidnotalteranyoftheconclusionsdrawn fromtheanalysisexcludingdeaths.Cox's regressionmodelwasused tocontrol for thepotential confoundingeffectsofsocialclassorgeneralpractice. Neither social classnordifferences inpracticenoraddingthesmallgroupof deceasedpersons tothestudypopulation,significantly influenced the relationshipbetweenoverweightandtheincidenceof registereddiseases.To giveanimpressionoftheBMI'softhesubjectsintheoverweightandthe control groupstheBMI'sbased onself-reported heightandweightsattheendof theobservationperiod (spring1984)areshown (Table4 ) . -77- tx> T r-i m o> <& m o m rH eo o eo i— co in ^o o \o •H <N -tr .-* vo H IN H «) m 01 m r-J m m o -H i£> vo I m m o c • 5 o o r- o o T c o I I -73- Table 3.RelativeRisks andlogrank statistics fortestingdifferences inincidencebetweenoverweight and control groupduring17yearsoffollow-upinacontinuousmorbidity registration ageatt disease freepopulation numberofincidentcases Relative p-value overweight:control overweight:control Risk forlogre observed statistic expected 0.24 20-29 64 : 93 0.4 0.6 30-39 113 :161 2.5 3.3 7.1 0.036 40-50 139 :187 3.4 4.6 4.0 0.063* 20-50 316 6.2 8.7 5.6 0.003** 20-29 140 :168 0.9 1.1 1.2 0.91 30-39 182 :210 2.0 2.0 3.5 0.32 40-50 237 :247 7.1 6.9 13.5 0.002** 20-50 559 :625 6.3 0.001** 441 ** Arteriosclerotic 20 29 64 : 93 disease 30 39 112 :162 0 4 2.1 3.0 5.8 0.076* 40 50 136 :185 15 8.3 11.7 4.1 0.002** 4.5 14.7 312 :440 20 29 140 168 30 39 183 210 5: 0 2.4 2.7 0.017** 40 50 240 248 16 : 4 10.1 9.9 0.008** 0: 0 21 : 4 men women Varicose veins women 29 64 93 5: 3 3.3 : 4.7 (2.4) 0.23 30- 39 113 160 11 : 5 6.6 : 9.5 (3.1) 0.024** 40- 50 139 187 25 :23 19.9 :28.5 1.5 20- 50 316 440 41 p <0.10 31 29.7 :42.3 0.14 1.8 0.007** 20- 29 140 168 2: 3 2.4 2.7 0.8 0.76 30- 39 183 210 19 : 9 13.8 14.2 2.4 0.049** 40- 50 238 248 52 :44 47.7 48.3 1.2 0.38 73 :56 62.6 66.4 20-29 64 : 93 1.4 1.6 30-39 113 :163 0.4 0.6 40-50 149 :187 1.7 2.3 50 4.6 1.5 0.066* 0.057* 0.23 3.8 9.5 0.19 0.010** 317 444 20 - -29 138 168 0 30 - -39 183 210 1.0 1.0 0.15 40 - -50 241 248 1.0 1.0 0.17 20 - -50 562 626 20 - -29 135 167 17 17 15.5 18.5 1.2 0.60 30 - -39 176 202 30 24 25.0 29.0 1.2 0.18 40 - -50 230 257 47 35 40.4 41.6 1.5 0.15 80.2 89.8 1.4 0.034* 626 p <0.05 0.0005** 20- 20 women 0.0004** 0 0.046* -79Table4.DistributionovercategoriesofBMIbasedonself-reportedweightand heightin1984* BMI Overweight group Controlgroup category men n (%) men n (%) (kg/m2) <25.0 25.0-26.9 27.0-29.9 >,30.0 27 48 64 43 (14.8) (26.4) (35.2) (23.6) women n 55 96 117 102 (%) (14.8) (25.9) (31.6) (27.6) 181 56 30 2 women n 313 53 17 6 (67.3) (20.8) (11.2) ( 0.7) (%) (80.5) (13.6) ( 4.4) ( 1.5) 182 (100.0) 370 (100.0) 269 (100.0) 389 (100.0) Unknown 135 192 175 238 Total 317 562 444 627 *Actualmeasuredweight(withoutclothes)andheight(withoutshoes)were requested.70%returnedthequestionnaire,ofthese95.2%gavebothweightand height Table4showsthatabout15%oftheoverweightpopulationcannolongerbe 2 consideredoverweight(BMI<25.0kg/m).Thismayhavebeentheresultof weightreducingeffortsresultingin,atleasttemporarily,normalweightorthe consequenceofchronicillness.Aneffectofageingorofunderreportingof weightcannotbeexcluded.Inthecontrolgroup,12%ofthemenand6%ofthe 2 womenhaveaBMI>27kg/m.Thesepersonscanbeconsideredasfalsenegative controlsubjects,whohaveneverbeenregisteredbytheirgeneralpractitioner asoverweightinthemorbidityregistration.Astheseimpropercontrolsubjects mightinfluenceestimatesofrelativeriskaspresentedintable2,analysis wererepeatedafterdeletingthesepersonsfromthecontrolgroup.Table5shows therelativerisksandp-valuesofthelogrankstatisticsafterremovalofthese subjects. -80Table5.Relativeriskscalculatedastheratioofincidencedensitiesofsome chronicillnessesintheoverweightgroupandinthecontrolgroup, afterexclusionofsubjectsinthecontrolgroupwho,accordingto 2 selfreporteddata,hadaBMI>27.0kg/m inthespringof1984 Illness Sex Relative Risk p-value for logrank statistic DiabetesMellitis men 5.2 5.6 0.004 4.2 5.1 0.0009 1.9 1.3 0.006 0.013 women 9.0 _ 0.05 women 1.4 0.04 women Arteriosclerotic men Disease women Arthrosis men women Gout Varicoseveins men 0.002 0.0008 0.10 DISCUSSION Thisstudydemonstratesthatoverweightpatientsingeneralpracticeareat higherriskfordevelopingarterioscleroticdisease,diabetesmellitus, arthrosis,goutandvaricoseveins(thelatteronlyforwomen).Itshouldbe notedthattheoverweightgroupandthecontrolgroupdifferedinfrequencyof illnessatthebeginningofthestudy,becausepersonsintheoverweightgroup, butnotinthecontrolgroup,wereselectedatthestartoffollow-uponthe basisofavisittothegeneralpractitioner (onwhichoccasiontheir overweightstatuswasdiagnosed. Manyinvestigatorshavestudiedtheassociationsbetweenoverweightand disease.Mostofthesestudies,however,havecross-sectionalorcase-control designs.Causalinferencefromthesestudiesisoftenproblematic,especially whendiseasessuchasarthrosis,gout,varicoseveinsandnon-fatal -81arterioscleroticdiseaseareunder study,becausethesediseasesmight contribute toweightchange.Thiscouldleadtomisleadingassociations.The prospective studiesthathavebeenundertakenhavegenerallyconsidered causes ofdeath.Onlyveryfewinvestigatorshave studied theincidenceofdisease in relationtooverweight [12].Inthisstudy,casesofparticulardiseasesunder studyatthestartoffollow-upwereexcluded,sotime-relationships between theonsetofoverweightandtheonsetofdiseasesarelikely.Foroverweight cases,theweightchangebetweenthestartofthefollow-upandtheendofthe observation-periodwereanalyzed.There isnoindicationthatanyofthe diseasesunder studycaused significantweightgainorweight loss. As ineveryotherobservational studyinwhichoverweightisthe independent variable,thetimebetweenthestartoffollow-upandtheoccurrenceofdisease shouldbe interpretedwithcaution,becausetheexact 'exposure time',i.e.the ageofonsetofoverweight israrelyknown.Inthisstudy,theuseofa continuousmorbidity registrationimpliesthat,especially inthefirstyears of the registration,manyoverweightpersonshadbeenoverweight fora considerablenumberofyearsbeforethestartoffollow-up.Thisexposure-time, which isnotaccounted for intheanalysis,probably increaseswith increasing ageatthestartof follow-up.Thismightexplain,notonlythe increasing incidenceofchronical illnesswithadvancingage,butalsothe stronger associationsof illnesswithoverweight intheolderagegroups.Thefinding in Table 4thatmorementhanwomenfromthecontrolgroupcanbe considered overweight,probably reflectstheselectiveattentionofgeneral practitioners with regard tooverweight inwomen.Comparingthe resultsfromTable 3with those inTable 5suggeststhattheinclusionofthefalsenegative controls introduceda slightbiasawayfromthenull.Thiscouldmeanthatundiagnosed overweightsubjectshave relativelylower incidenceofillnessescompared to diagnosedoverweight subjects.Suchselectionbiashasbeen reported ingeneral practices intheUnited States [13].The relationshipswithchronic illnesses demonstrated inthisstudyarediscussedbelow. Arterioscleroticdisease Bothmyocardial infarctionandanginapectorisarecommonmanifestationsof arteriosclerotic disease,themostcommoncauseofdeath inTheNetherlands [14].Bothhaveoneunderlying functionalabnormality: ischemiaor insufficient oxygen supplytothetissuesoftheheart.Inlong-term follow-upstudies, overweighthasbeenassociatedwitharterioscleroticdisease inmultivariate analysisasan independentpredictor [15].A largeproportionofcaseswith clinicalmanifestationsofarteriosclerotic diseasedosurvive,probablywith severe impairmentoftheirqualityoflife.Thisstudyshowsthat arteriosclerotic disease,includingandexcluding fatalcases,isclearly related tooverweight.An important finding isthatapronounced increase in risk inmenwasobservedatlowerlevelsofBMI inmenthaninwomen.Thismay bedue tosexdifferences infatdistribution,whichare related to hypertension [16,17]andcardiovascularmortality [18].Itmustbenoted that bodyweight isclearly related tomajor risk factors forcardiovascular disease likeserumcholesterol levelandbloodpressure.Adjustments forthese factors arelikely toyieldlowerestimatesofthe relative risksassociatedwith overweight,ashasbeendemonstrated intheFraminghamHeartStudy [12].No baselinedataofthesepossibleconfounding factorswerepresent inthe morbidity registrationonwhichwebasedouranalysis,norweremeasuresof fat distribution.Theselimitationsshouldbekeptinmindintheinterpretationof ourresults. Diabetesmellitus Manystudies,prospective [19,20]aswellascross-sectional [21,22,23],have demonstrated theassociationbetweenoverweightanddiabetes (TypeII).This studyconfirmsthatthere isanincrease inriskwithhigher levelsofBMI. Gout IntheFramingham study itwasshownthatthe risk forgoutinmen increased appreciablyonlywhenweightsexceeded 130%oftheappropriateweight for height [24].Althoughgoutwasnotacommondisorder inthisstudy,itis striking thatvirtuallyallcasesbelonged totheoverweightpopulation.Most of thecasesofgouthadaBMI,atthestartoffollow-up inthecategory 27.0 2 - 29.9kg/m ,which isclearlylower than130%oftheappropriateweight,as found intheFramingham study. Arthrosis TheRoyal CollegeofPhysicianshasemphasized thepossibility thatthe associationbetweenoverweightandarthrosismightbe the resultof reduced physicalactivity,leadingtoweightgain [1].Comparing theBMI'sofcasesof arthrosisatthestartoffollow-upwithBMI'sbasedonweightsmeasured bythe generalpractitioner anaverageof11yearslater,revealed thatonly 18%were inahigher categoryofBMIthanatthestartof follow-up.FromTable?4and 5 itcanbe seenthatanincrease inrisk forarthrosisbecomesapparentwhen the -83- BMI is27.0orhigherandthat30- 39yearoldoverweightwomenwere particularlyat risk fordevelopingarthrosis.The individualpatient records were studied fortheaffected sitesofarthrosis.Inabout35%ofthecasesin both theoverweightandthecontrolgroup,theaffected siteswere weight-bearing joints (kneesand hips).Also forother sites,nodifferencewas foundbetweentheoverweightandthecontrolgroup.Thefactthatarthrosis associatedwithoverweight isnot restricted toweightbearing jointshasbeen observed inprevious studies[25]. Varicoseveins Varicoseveinsweremuchmorecommoninwomenthaninmen.Ithasbeenobserved thatwomenwithvaricoseveinsareheavier thancontrolsofthesameage [26,27].Results fromtheBasle study [28]demonstrated thatparity isan important risk factorforvaricoseveinsinwomen. Inourpopulation,parity intheoverweightgroup (nulliparous:6%, one pregnancy:10%,twopregnancies:33%,threepregnancies:23%,fourormore pregnancies:28%)wasvery similar toparity inthecontrolgroup (nulliparous: 7%, onepregnancy:9%),(twopregnancies:31%,threepregnancies25%,fouror morepregnancies:28%).Comparing incidenceofvaricoseveinsbetweenthe overweightandcontrolgroupincategoriesofparity showedno confounding effectofparityontheestimatesof relativerisk. Intheanalysisnoadjustmentscouldbemade forthepossible confounding effectofsmoking,underweightortypeof fatdistribution.Since smokersare generallylessobese thannon-smokers,andalsohavehighermorbidity,taking smoking intoaccountwouldprobablyyieldhigherestimatesof relative risk. Fatdistribution,wasonlyassessed inasubpopulationoftheoverweight subjects intwogeneralpractices.Thewaist:thighcircumference ratiowas shown tobe related toarthrosis inmen,hypertension inmenandwomen, diabetesorgoutinwomen,andvaricoseveinsinwomen.Theseassociationswere independentofthedegreeofoverweight[17]. CONCLUSIONS Overweight leadstoanincreased frequencyofchronic illnesses,registered in generalpractice.Formen,thiselevated riskwasapparentatlower levelsof BMIthan forwomen.Thisisinagreementwith findingsofstudiesonfat distribution,whichhave shownthat,forthesamedegreeofoverweight,menare -84- athigher riskofvariousdisordersthanwomen[29]. Atpresent,overweight ingeneralpractice isacondition forwhich relatively morewomenare treated thanmen.Howeveroverweightasahealthhazard inmen shouldnotbeneglected.Thisstudydoesnotallowfordetectinga dose-response relationship,butitisapparent fromTable2thatoverweight in men isassociatedwithchronic illnessatlower levelsofBMIthan inwomen. Thediseasesthatwere foundtobeassociatedwithoverweightmightleadto important functional limitationsanddiminished subjectivehealth.ItShouldbe emphasized thatselectionbiasmighthave influenced the resultsofthe analysisand thattheelevated risksassociatedwithoverweightmightatleast partlybeattributedtoconfoundingeffectsoffactorslikeserumcholesterol andbloodpressure.Levelsofthese risk factorsforarterioscleroticdisease usuallytendtonormalizeandconditionslikediabetesmellitus,goutand arthrosisarelikelytoimprovewhenasubstantialweight reduction isachieved andmaintained [30]. Thus, treatmentofthosewhoarealreadyoverweightand,perhapsmore importantly,preventionofoverweight inthosewhoareat risk,deserves attention ingeneralpractice.Thegeneralpractitioner isable tomonitor weight changesona regularbasis.Whenpersons,especiallyyoungadults,who 2 arealreadymoderatelyoverweight (BMI>25kg/m )gainweight,attention shouldbepaid totheirnutritionalhabitsandotheraspectsoftheir lifestyles,inordertopreventthemfrombecomingseverelyoverweight,which would increase theirchancesofdevelopingachronicdisease. ACKNOWLEDGEMENTS Theauthorsaregrateful toF.J.A.Huygenand staffofthedepartmentof General Practice inNijmegenforgenerouslyproviding themorbiditydata;J. BuremaandS.Ward forhelpfulcomments;P.Hoppener forcomputer assistance andA.Zijlmans fortypingthefinaldraftofthemanuscript. REFERENCES 1.ReportoftheRoyalCollegeofPhysicians;Obesity.JournaloftheRoyal CollegeofPhysiciansofLondon1983,17:1-58. 2.Gezondheidsraad.Adviesinzakeadipositas.StaatsuitgeverijDenHaag,1984. 3.SimopoulosAP;vanItallieTB:Bodyweight,HealthandLongevity.Ann InternMed,1984;100:285-295. 4.RoyalCollegeofPractioners.InternationalStatisticalClassificationof Diseasesandcauseofdeath,1963. 5.GarrowJS.Treatobesityseriously,aclinicalmanual.London[etc] ChurchillLivingstone,1981. 6.KeysAC,AravanisH,BlackburnH,etal.Coronaryheartdisease:overweight andobesityasriskfactors.AnnIntMed1972;77:15-27. 7.StewartAL,BrookRH.Effectofbeingoverweight.AmJPublHealth1983;73: 171-178. 7a.SeidellJC,BakxJC,HoogenHJMvanden,DeurenbergP.Overgewichtin relatietomorbiditeitensubjectiefwelbevindeninvierhuisartspraktijken. In:BoermaWGW,HingstmanLed.Deeerstelijnonderzocht.Deventer:Van LoghumSlaterus,1985:225-236. 8.KaplanEL,MeierP.Nonparametricestimationfromincompleteobservations. JAmStatAss1958;53:457-481. 9.SASSupplementalLibraryUsersGuide.SASInstitute,1980:83-102and 119-131. 10.PetoR,PikeMC,ArmitageP,etal.Designandanalysisofrandomized clinicaltrialsrequiringprolongedobservationofeachpatient.BrJCancer 1977;35:1-39. 11.BreslowNE.Covarianceanalysisofcensoredsurvivaldata.Biometrics1974; 30:89-99. 12.KannelWB.Healthandobesity:anoverview.In:ConnHL,DefeliceEA,KuoPT eds.Healthandobesity.NewYork:RavenPress,1983. 13.FranksP,CulpepperL,DickinsonJ.Psychosocialbiasinthediagnosisof obesity.JFamPract1982;24:745-750. 14.CentraalBureauvoordeStatistiek.Hart-envaatziekten-eenstatistische verkenning.CBS-publicationVoorburg,1983. 15.HubertHB,FeinleibM,McNamaraPM,CastelliWP.Obesityasanindependent riskfactorforcardiovasculardisease:a26-yearfollow-upofparticipants intheFraminghamHeartStudy.Circulation1983;67:968-977. -36- 16.BlairD,HabichtJ,SimsEAH,SylvesterD,AbrahamS.Evidenceforan increasedriskforhypertensionwithcentrallylocatedbodyfatandthe effectofraceandsexonthisrisk.AmJEpidemiol1984;119:526-540. 17.SeidellJC,BakxJC,BoerEde,DeurenbergP,HautvastJGAJ.Fat distributionofoverweightpersonsinrelationtomorbidityandsubjective health.IntJObes1985;9:363-374. 18.LarssonB,SvardsuddK,WelinL,WilhelmsenL,BjorntorpP,TibblinG. Abdominaladiposetissuedistribution,obesity,andriskofcardiovascular diseaseanddeath:13yearfollowupofparticipantsinthestudyofmen bornin1913.BrMedJ1984;288:1401-1404. 19.KannelWB,GordonT,CastelliWP.Obesity,lipidsandglucoseintolerance. TheFraminghamStudy.AmJClinNutr1979;32:1238-1245. 20.WestlundK,NicolaysenR.Ten-yearmortalityandmorbidityrelatedtoserum cholesterol.Afollow-upof3751menage40-49.ScandJClinLabInvest 1972,-Suppl127:30. 21.RimmAA,WernerLH,vanYserlooB,BernsteinRA.Relationshipofobesityand diseasein75,532weight-consciouswomen.PublicHealthReports1975;90: 44-51. 22.SeidellJC,deGrootCPGM,vanSonsbeekJLA,DeurenbergP,HautvastJGAJ. Associationsofmoderateandsevereoverweightwithselfreportedillness andmedicalconsumptioninDutchadults.AmJPublicHealth1986;76:264-269. 23.BonhamGS,BrookDB.Therelationshipofdiabeteswithrace,sexand obesity.AmJClinNutr1985;41:776-783. 24.RimmAA,WhitePL.Obesity:itsrisksandhazards.In:BrayGAed.Obesity inAmericaWashington:NIHPublicationNo.79;1979:103-124. 25.SilberbergR.Obesityandosteoarthrosis.In:ManciniL,LewisB,Contaldo M,Medicalcomplicationsofobesity.London:Acad.Press1979:301-315. 26.VonMarshallM,UhligHE,HessH.UbergewichtundperiphereErkrankungen. ZeitschriftfurAllgemeinmedizin1977;53:659-664. 27.Dall'AntoniaF,OstuzziR,Bosello0.Obesityandvaricoseveinsofthe lowerlimbs.In:ManciniL,LewisB,ContaldoMeds:Medicalcomplications ofobesity.London:AcademicPress:271-273,1979. 28.WidmerLK,KachK,MadarG,KamberV.Riskfactorsofvaricosity.In:Widmer LKed.PeripheralVenousDisorders,prevalenceandsocio-medicalimportance. Observationsin4529apparentlyhealthypersons.BaslestudyIII,Bern, Stuttgart,Vienna:HansHuberPublishers,1978:58-66. 29.BjorntorpP.Hazardsinsubgroupsofhumanobesity.EurJClinInvest1984; 14:239-241. -87- i30.BlackburnGL,ReadJL.Benefitsof reducing-revisited.PostgradMedJ1984; 60 (suppl3):13-18. ! APPEM)IX Diagnosticcriteria forchronicdiseasesinthecontinuousmorbidity registration. Diabetesmellitus A singleblood-glucoselevel >10mmol/1wasusedasthecriterion.Indoubtful cases,thiswasconfirmedbyanabnormalOralGlucoseToleranceTestaccording toWHOcriteria. Gout Goutwas registeredwhentheusualclinical symptomswereobserved,ifnecessary confirmedbylaboratoryinvestigation. Arthrosis Arthrosisdeformanswas registeredwhentherewere clear clinical signsof arthrosis (e.g.,pain,stiffnessand inflammation),confirmedbyanX-ray. Varicoseveins Varicoseveinsof lower limbswere registeredwhenclearlyvisiblevaricose veinswerepresented,observedeitherat routineexaminationor following specific complaints fromthepatient. Myocardial infarction Myocardial infarctionwas registeredwhenaclinical suspectionofamyocardial infarctioncouldbeconfirmedbyanabnormalECGand/orbymeasurementsof the concentrationsoftheenzymesCPK,SGOT,LDH (exceptionsweremade forpatients whodiedbefore such investigationswere possible). Anginapectoris Anginapectoriswasdiagnosedwhenaclinicalexaminationprovided sufficient evidence foranginapectoris (notnecessarilyconfirmedbylaboratory-orECGtest). CHAPTER5 THERELATIONBETWEENOVERWEIGHTANDSUBJECTIVEHEALTH-WITHATTENTIONTOTHE EFFECTSOFAGE,SOCIALCLASS,SLIMMINGBEHAVIORANDSMOKINGHABITSONTHIS RELATION J.C.Seidell,K.C.Bakx,P.Deurenberg,J.Burema,J.G.A.J.Hautvast, F.J.A.Huygen AmJPublicHealth1986;76:(inpress) ABSTRACT Subjectivehealthstatuswasassessed,inrelationtooverweight,by administeringalistof51healthcomplaintstoadultmenandwomenwhowere eitherchronicallyoverweightasdefinedbyBodyMassIndex(BMI)ornot overweight,inacontinuousmorbidityregistrationinfourgeneralpractices duringtheperiod1967-1983.Responseswerereceivedfrom455men(182 overweight)and790women(386overweight),age26-66years.Responserate (71%)andagedistribution(meanage48)wereequalinoverweightand non-overweightgroupsofbothsexes.BMIwascorrelatedwiththetotalnumberof complaintsinwomen(r=0.15)butnotinmen(r=0.07).Multipleregression analysisrevealed,however,thatagewasaneffectmodifierinthisrelation, therebeinganegativeassociationbetweenBMIandsubjectivehealthinyounger menandapositiveassociationinoldermen,whereas,inwomen,theassociation betweenBMIandsubjectivehealthwasmuchmorepronouncedatyoungeragesthan atolderages.Inaddition,currentsmokinghabitsandsocialclass(inmenand women)andreportedslimmingbehavior (inwomen)hadanindependentrelationto thetotalnumberofhealthcomplaints.BMIwasalsorelatedtospecific complaintsandgroupsofcomplaints,particularlyinwomen. INTRODUCTION Overweightisariskfactorforthedevelopmentofvariousdiseases' andis 12 verycommoninaffluentcountriesincludingTheNetherlands.Wehaveshownthat overweightisaccompaniedbyanexcessofbothobjectivelyregisteredmorbidity andsubjectivereportedillness,increaseduseofmedicalcareforparticular reasons,andtheuseofcertaindrugs.Thesubjectivehealthstatusof -89overweightpersonscouldbemediatedthroughtheassociationbetweenoverweight andmorbidity.Ontheotherhand,intheabsenceofillness,overweightperse mayberelatedtoadiminishedfeelingofwellbeing.Deviationsfromacultural 45 idealofslimmnessmaybeinvolvedinsucharelation' aswellasfunctional limitationsduetoexcessweight. Theassociationofoverweightwithstatesofmentalhealthhasbeenthesubject ofmanystudies,buttheassociationwithsubjectivehealthcomplaintshasbeen asomewhatneglectedsubject.ResearchintheU.S.A. indicatedthatoverweight wasrelatedtofunctionallimitations,pain,worryandrestrictedactivities.In Sweden,associationsbetweenoverweightandindicesofhealthwerefoundtobe 78 lesspronouncedinmiddleagedmenthaninwomen'. InthisstudyweinvestigatedtheassociationbetweentheBodyMassIndex(BMI= 2 weight/height)andresponsestoalistof51subjectivehealthcomplaints.The relationsofBMIwiththetotalnumberofcomplaintsaswellaswithgroupsof complaintswerestudied,adjustingforthepossibleconfoundingeffectsof slimmingbehavior,smokinghabits,age,andsocialclass. METHODS Samplecharacteristics Thedatapresentedinthispaperwereobtainedfromaquestionnairethatwas administeredinthespringof1984tomembersofastudypopulationthathad q beenfollowed-upinacontinuousmorbidityregistration .Fortheoverweight group,personswereselectedonthebasisofdiagnosisofoverweightatage2050yearsbygeneralpractitionersfromfourgeneralpracticesinamixed rural/urbanarea.Thecontrolgroupcomprisedarandomsampleofadultswhohad neverbeendiagnosedasoverweight;thisgroupwasmatchedtotheoverweight groupinage,sex,practiceandcalendar-yearatstartoffollow-up.Detailsof q thefollow-upstudyandsamplingmethodshavebeendescribedelsewhere. Thetotalpopulationinthefourgeneralpracticescomprisesabout12.000 individuals,andisrepresentativefortheDutchpopulationwithrespecttothe distributionofageandsex.AllpersonssampledwereCaucasianandofDutch nationality. Personswhoseaddresseswereknown(93%)weresentaquestionnaire,whichwas returnedby1241persons(71%response).Theresponsewasnotselectivewith regardtooverweightstatus,sex,ageorsocialclass.Table1showssome characteristicsofthestudygroup. -90- Descriptionofmeasures 1.Overweightwasdiagnosedintheperiod1967-1978bygeneralpractitioners accordingtocriteriathatcorrespondtoaBodyMassIndex(BMI= 2 2 2 weight/height)ofatleast26.0kg/m (forwomen)or27.0kg/m (formen). Heightandweightweremeasuredbythegeneralpractionersatthetimeofthe diagnosisofoverweight.Onaverage,thiswas12yearspriortothe administrationofthequestionnaire (range6 - 1 7 years).Aspartofthe follow-upquestionnairein1984,subjectswereaskedtoweighthemselvesin themorningwithoutshoesandclothesandmeasuretheiractualheightwithout shoes.Virtuallyallsubjectsindicated,whenaskedinthequestionnaire, thattheyhadbeenabletofollowtheseinstructionsforweighingthemselves andmeasuringheight.Inaddition,self-reportedheightsandweightswere foundtobeaccuratewhencomparedtomeasuredheightsandweightsina 25 sampleof305oftheoverweightsubjects .Unlessotherwisespecified,BMI whenusedinthispaperreferstocurrentBMI. 2.Socialclasswasdefinedaslower(unskilledandskilledmanualworkers), middle(loweremployees)orupper (higheremployees),accordingtothe professionoftheheadofthehousehold. 3.Smokingbehaviorwasclassifiedfromquestionnaireintofourcategories (neversmoked,stoppedsmoking,smokinglessthan10cigarettesperday,and 10cigarettesormoreperday).Inmultipleregressionanalysesdummy variablesforthecategories"stoppedsmoking",and"smoking"wereused, while"neversmoked"wasthereferencecategory. 4.Slimmingbehaviorwasclassifiedintofivecategories("never","seldom", "regularly","often",or"(almost)always",tryingtoreduceweight).Inthe multipleregressionanalyses,"regularly"and"often"slimmingwerecombined intoonecategory.Dummyvariableswereused,with"neverslimming"asthe referencecategory. 5.Subjectivehealthwasmeasuredbyratedhealthona7-pointscaleandbythe answerstoalistof53healthcomplaints .ThislistisoftenusedinThe Netherlandstoassesssubjectivehealthstatus .AppendixIgivesashort descriptionofsomesingleitemsand,itemsandclustersidentified:using 12 factoranalysis .Allitemsweredichotomous;twoitems("do youfeeltoo thin?"and"doyoufeeloverweight?")wereexcludedfromtheanalysissince theseitemscouldseriouslyconfoundtherelationbetweenoverweightandthe totalnumberofcomplaints.Thesquarerootofthetotalnumberofcomplaints (fromtheremaining51items)hadacloseapproximationtothenormal distributionandshowedtheclearestlinearrelationshipwithBMIandother -91- variables;itwasthereforeused intheanalysisasanindex forgeneral subjectivehealth.Inlogistic regression,scoresonsubscaleswere consideredaswellassomesingle items.These itemswere selectedwhen,in analysisofcontingencytables,afteradjustments forsex,socialclassin theperiod 1967-1978andage,theanswersonthese itemswerediffered betweentheoverweightandthecontrolgroupataprobabilitylevelof 0.05. 6. Statisticalmethods.Multiplelinearandlogistic regressionanalyseswere performedwiththeuseofthestatisticalpackageprogramsBMDP2Rand BMDPLR .Presenceof interactionwasassessedbyaddingcross-product terms ofvariablestoamodel thatcontainedallthemaineffects.Whenthe F-to-enter ofan interactiontermhadaprobabilityvalue lessthan0.10,the cross-product termwas included inthestraight regressionanalysiswhose resultsarepresented inthetables.Thedependentvariable inmultiple linear regression (table4)wasthesquare rootofthetotalnumberof complaints. IndependentvariableswereBMI,age (andacross-product termof BMIandage,reflectingstatistical interactionofBMIandageinthe analysis), socialclass,smokinghabitsandslimmingbehavior.For categories of smokinghabitsandslimmingbehavior,dummyvariableswereused (using never-smokingandnever-slimmingas reference categories). Inmultiple logistic regression (table6)thesame independentvariableswereused.The dependentvariablewasdichotomous:presenceorabsenceofa complaint (single items)orpresenceorabsenceofatleastonecomplaint (groupsof items). RESULTS Table 1showsthedistributionofsubjectsinBMI categoriesandcategoriesof possible confoundingvariables.At thetimeofthestudy,about27%ofthe originally registeredoverweightwomenand41%of theoriginally registered overweightmencouldnotbeconsidered overweightanylonger,whileabout6% of thewomenand 12%ofthemen inthecontrolgroupwouldbeconsidered overweight,according tothesamecriteriausedearlierbytheirgeneral practitioners. InTable 1,currentoverweight isinversely related tosocial class;ageand smokingmorethanwomen;subjects intheoverweightgroups reported slimmingmoreoftenthanthoseinthecontrolgroup,thisbeingmore pronounced inwomen thaninmen.Subjectivehealth,ratedona7-pointscale, washighlycorrelatedwith thesquare rootofthenumberofcomplaints (men:r= 0.61,95%CI= 0.55 -0.66;women:r= 0.65,95%CI=0.61 - 0.69) -92- Table1.BodyMass Index,socialclass,age,smokinghabits,andslimming behaviourofthose respondingtothequestionnaire,classified accordingtooverweight status (registeredin1967-1978)andsex Men Women Overweight n A. BodyMassIndex Control n % % Overweight Control n % n % (kg/m2) 181 67.3 <25.0 55 14.8 313 80.5 27 14.8 25.0-26.9 96 25.9 53 13.6 48 26.4 56 27.0-29.9 117 31.6 17 4.4 64 35.2 30 11.2 >30.0 102 27.6 6 1.5 43 23.6 2 0.7 16 - 15 0 - 4 - Lower 248 64.2 186 46.1 117 64.3 124 45.4 Middle 118 30.6 177 43.8 59 32.4 117 42.9 Upper 20 5.2 41 10.1 6 3.3 26-39 82 21.2 35 19.2 62 22.7 40-49 98 25.4 131 32.4 64 35.2 78 28.6 50-66 205 53.1 184 45.5 74 40.7 130 47.6 163 42.2 156 38.6 22 12.1 40 14.7 85 22.0 99 24.5 59 32.4 79 smoking<10cigs/day 43 11.1 59 14.6 41 22.5 smoking>10cigs/day 95 24.6 90 22.3 60 32.0 108 39.6 52 13.6 207 55 31.1 190 69.6 108 55 20.1 16 5.9 unknown - 20.8 B. SocialClass 32 11.7 C. Age(years) 87 21.5 D. Smokinghabits neversmoked stoppedsmoking 28.9 46 16.8 E. Slimmingbehaviour never 51.6 seldom 83 21.8 26.9 59 33.3 regular 119 31.2 56 14.0 34 19.2 often 48 12.6 12 3.0 15 8.5 0 0.0 (almost)always 79 20.7 18 4.5 14 7.9 12 4.4 unknown Total 5 386 z 100 3 404 100 5 - 0 _z_ 182 100 273 100 -93- InwomenBMIwascorrelatedwiththenumberofcomplaints (r= 0.15;95%CI: 0.08 - 0.21),butinmenthecorrelationwasvery slight (r= 0.07;95%CI: -0.03- 0.17). Intable2theaverageofthesquare rootofthetotalnumber of complaints invariouscategoriesofBMIareshown.Itcanbe seenthat inall categoriesofBMImore complaintswere reported intheoverweight groupthan in thecontrol group.Overweight statusinthistablewasbasedonthe classificationbythegeneralpractitioners.Insubsequentanalysiswedidnot use thegeneralpractitioner classificationasthiswouldprobablywould have led toanoverestimationoftheassociationbetweencurrentoverweightand current subjectivehealth. Table2.Square rootofthetotalnumberofcomplaintspercategoryofcurrent BodyMass Indexbyoverweight statusbasedonthediagnosisoftheir general practitioners 6-17 yearsearlier Women A. BodyMass Index (kg/m) <25.0 25.0-26.9 27.0-29.9 >,30.0 Men Overweight Control Overweight Control mean mean mean mean 3.3 3.2 3.4 3.5 3.0 2.9 2.8 3.0 3.1 3.3 3.2 3.2 2.8 2.5 2.8 3.0 2 Theaveragenumberofcomplaints incategoriesofothervariablesthatwereused inthisstudyare shown inTable 3.Thenumberofcomplaintsgenerally increased withage,withheavier smoking (inmenandwomen)andwithmore frequent slimming inwomen.A clear inverse relationbetweensocialclassand thenumber of complaintswas found inmenandwomen. As thevariables intable3areknowntobe related tobothoverweightand to subjectivehealthweperformed amultiple regressionanalysis inwhichall these variableswere takenintoaccount.Intheevaluationof thedataaclear interactionbetweenageandBMIwasobserved inbothmenandwomen.Therefore this interactionwas included inthemultiple regressionmodel.The resultsare presented intable4. -94- Table3.SquarerootofthetotalnumberofcomplaintsaccordingtocurrentBMI, socialclass,age,smokinghabits,andslimmingbehaviorformenand women Women n=790 BodyMassIndex(kg/m) <25.0 25.0-26.9 27.0-29.9 >30.0 2 Men n=451 mean mean 3.0 3.1 3.3 3.5 2.8 3.3 3.0 2.7 3.1 2.9 3.1 3.1 Socialclass Low Middle High 2.6 2.4 Age(years) 26-39 40-49 50-66 3.0 3.2 3.3 2.7 3.0 3.2 3.1 3.4 2.7 2.9 3.0 Smokinghabits neversmoked stoppedsmoking smoking<10cigs/day smoking>10cigs/day 2.6 3.0 3.1 Slimmingbehavior Never Seldom Regular Often (Almost)always 2.9 3.0 3.3 3.4 3.5 2.9 2.8 2.8 3.9 2.9 -95- Table4.RelationofBMI,age,socialclass,smokinghabits,andslimming behavior tothesquare rootofthetotalnumberofcomplaints(51 items).Resultsofmultiplelinear regressionanalysis Sex Variable Regression StandardError coefficient ofregr.coeff. Women BMI(kg/m2) Age (years) BMIx (Age-48) + Social class 0.022 0.014 0.078* 0.032 -0.002* 0.001 -0.272* 0.081 Smoking* 0.326* 0.117 Ex-smoking* 0.278* 0.131 Slimming (seldomP 0.233* 0.118 Slimming (often/regular)* 0.352 0.220 Slimming (almostalways)* 0.438* 0.179 0.020 0.021 -0.092* 0.051 0.004* 0.002 -0.344* 0.099 0.376* 0.189 2 TotalR *100=7.6 Men BMI(kg/m2) Age (years) BMIx (Age-48)+ Social class Smoking" Ex-smoking' 1 Slimming (seldom)" ' -0.199 0.207 -0.003 0.145 Slimming (often/regular)? 0.832* 0.422 Slimming (almostalways)^ 0.228 0.284 TotalR 2*100=10 Notes:*p<0.05 + Cross-product termofBMIandageindicating statistical interaction betweenBMIandageintheirassociationwith subjectivehealth(48 yearsistheaverageageinmenandwomen) + Dummyvariablesforcategoriesofsmokinghabits ("neversmoked"is the reference category) vDummyvariablesforcategoriesofslimminghabits ("neverslimming"is the reference category The resultsinTable4areinagreementwiththesimpleanalysisinthat social classwasfoundtobenegatively relatedtosubjectivehealth,that smokershad -96- morehealthcomplaintsthanthosewhoreportedtheyneversmoked.Menwho stoppedsmokinghadlesscomplaintsandwomenwhostoppedhadmorecomplaints thanthosewhoneversmoked.Inmostcategoriesofslimmingfrequencythenumber ofcomplaintswashigherthaninthecategoryof"neverslimming"alsoafter adjustmentforBMIandtheothervariablesandthiswasmorepronouncedinwomen thaninmen.Thepresenceofaninteractionterminmultipleregressionintable 4complicatestheinterpretationoftheregressioncoefficientsofBMIandage. Forinterpretationpurposeswecalculatedthepredictedaveragenumberof complaintsaccordingtotheregressionmodelsintable4fordifferentvaluesof BMIandage(correctedforthetransformationofthepredictedsquareroot valuestoactualnumberofcomplaints). TheresultsareshowninTable5.Inmen,ahigherBMIwasassociatedwithmore healthcomplaintsattheageof55butwithlesscomplaintsat35.Inwomena higherBMIwasassociatedwithmorecomplaintsattheageof35butnotatthe ageof55.Anincreaseinnumberofcomplaintswithagewasclearerin overweightmenthaninleanmen,andclearerinleanwomenthaninoverweight women. Table5.Averagenumberofcomplaints,predictedbyregressioncoefficientsof theregressionmodelfromtable4,attwodifferentagesandfortwo differentlevelsofBodyMassIndex Sex Age BodyMass 2 23kg/m Women Men 35years Index 28kg/m2 9.6 11.1 55years 12.3 12.4 35years 9.2 9.8 11.3 55years 8.3 Note:estimatescorrectedfortransformingthepredictedsquarerootof complaintstotheaveragenumberofcomplaints(byaddingtheresidual varianceofthenumberofcomplaintsinthesquarerootscale) AstheBMImightberelatedtospecificcomplaintsbutnottoothers,westudied therelationofBMItoabsenceorpresenceofspecificcomplaintsandabsences orpresenceofatleastonecomplaintoutofgroupsofcomplaints.The -97- complaintsthatwere foundtobe relatedtoBMI,independentofage,social class,and smokinghabitsareshowninTable 6. Table6.Groupsofcomplaints (clusters)and singlecomplaints related tothe BodyMass Index,adjusting forage,socialclass,slimmingbehavior, andsmoking habits.Resultsfrommultiplelogistic regression analysis Sex Description ofcomplaints i Women Stomach/digestion Skeleto/muscular Nervousness Tiredness Nose/throat Shortnessofbreath Painchest/heart Oftenhavingmishaps Tight inchest Men Partial logistic regression ternsin % reporting 1 complaints cluster fromcluster b SE(b) No.of Stomach/digestion Skeleto/muscular Shortnessofbreath coefficientofBMI 5 30.6 0.0422 0.0194 5 4 5 3 1 1 1 1 67.2 0.0703 0.0244 48.3 0.0570 0.0196 50.0 0.0432 0.0194 22.6 0.0371 0.0208 18.0 0.115 0.0228 13.8 0.0390 0.0230 0.115 0.0309 12.4 0.0584 0.0250 27.5 0.0760 0.0344 59.3 0.0767 0.0359 12.8 0.0898 5 5 1 6.8 0.0368 Note:all regressioncoefficientssignificantlydifferent fromzero (p< 0.05). Thedependentvariable intheanalysiswasdichotomous:presenceor absenceofaparticular complaint (singleitems)orpresenceorabsence ofatleastonecomplaintoutofagroup (cluster)ofcomplaints Complaintsofthedigestive tract,theskeleto-muscularsystemand shortnessof breathwereassociatedwithBMIinmenandwomen.Inaddition,inwomenan associationwas foundbetweenBMIandcomplaintsofnervousness,tiredness,nose and throat,pain inthechestandheart region,oftenhavingmishaps,andoften feeling tightinthechest.Slimmingbehaviorwasindependently relatedto complaintsoftirednessandoftenhavingmishaps (notshowninthe tables). -98DISCUSSION Ourdata showthat increasedBodyMass Indexisassociatedwithmore subjective healthcomplaintsbutthatageisaneffectmodifier inthisassociation.Inmen a positiveassociationcouldbedemonstrated inolderagesandanegative association inyounger ages. Inwomen,ontheotherhand,thepositive correlationbetweenBMIandnumberofcomplaintsdiminishedwithage.Inmost studiesthathave reportedanassociationbetweenoverweightand subjective healthapossible interactionbetweenageandBMIwasnottakeninto account 6,7 ' 8 . Theeffectmodificationofage inmencannotbeeasilyexplained.Itmaybe that a longperiodofoverweight isneededbeforeaneffectofoverweighton subjectivehealthbecomesapparent inmen.The reasonwhy inyoungerwomena muchstronger relationbetweenoverweightandcomplaintswas foundthaninolder womenmaybethatyoungerwomenworrymoreandaremoreconcernedaboutbeing overweight,hence,exhibitmorehealthcomplaints.Someofthespecifichealth complaintsassociatedwithBMIarepsychosomatic (e.g.nervousness,titedness, oftenhavingmishaps). Inthese relationsnosignificant interactionwas observedbutitmustbenoted that inlogisticanalysis,interaction is evaluated onamultiplicative scalewhereas inmultiple linear regression 14 analysis interaction isassessedonanadditive scale Inour study itwasnotpossibletodeterminewhetherpsychosomatic complaints were thecauseor the resultofoverweight ' .Theeffectsofthestigmaof 45 obesity,especially inwomen,havebeen reviewed recently ' and ithasbeen suggested thatsociety'sdiscriminationagainsttheobeseandthepressuresfor thinnessareenoughtoaccount formanypsychological complaints inthe 17 overweight .Ontheotherhand,ithasbeenshownthatpsychosocial problems 18 maycontribute toweightgain and thatweightchangesareaccompanied by 1920 changes inmood ' .Inwomen,ahigher frequencyof slimmingbehaviorwas related tomorehealthcomplaints,independentlyofthedegreeofoverweight; only 15%ofthewomen,however,reported thathealth reasonswere importantas motives forslimming,whereas the reasons 'problemswithclothes'and 'my figure'wereeachendorsedby65%ofthewomenwhohadever tried tolose weight.Therefore,itisnotlikelythatthe relationbetween slimmingand subjectivehealthcomplaintscanbeexplainedbythefactthatmostwomentryto loseweightbecauseofhealthproblems.Perhapsslimmingbehavior initself is associatedwith increaseddistress.Whenwe stratified theoverweightgroup (as classifiedby thegeneralpractitioners)intopersonswhoshifted outofthe -99overweightcategory (basedcurrentBMI)and thosewho remained inthe overweight category itwas surprising tofindthat thesetwogroupshadanequal average number of complaintsandbothgroupshadmore complaints thanthegroup thathad remained non-overweight.Thus, successful slimming cannotaccount for improved subjectivehealth status.Only 10%or lessofthevariance inthenumber of complaints couldbeexplained byallthevariables inthe regression analysis (Table 4 ) .Thus theremustbeotherunmeasured determinants,thataccount for theunexplained partof thevariance. Although subjects received instructions formeasuring heightsandweights,the useof self-reported measurements has itslimitations.Forexample,evenwhen all subjectswould have followed the instructions carefully, insystematicerrors mayhaveoccurred becauseof thepoorqualityof somehomeweighing-scales. Wemay conclude that theassociation ofoverweightandpoorer subjective health statusdiffersbetweenmenandwomenand thatthisassociation isdependent on age. Inyoungerwomen,thisassociation isprobably related toan increased concernof thesewomenabout theirweightor figure.Inthisparticular age group itmaybe important toinvestigatewhether the 'legitimization'of being overweight (assuggested byStewartandher colleagues )mayhaveapositive effect onsubjective health. Incaseof increased health risk it remains necessary totreat thosewhoareoverweight.Publichealthmeasures should be directed towards theprevention ofoverweightand increasing awarenessabout the health risksassociatedwithoverweightwhenaccompanied byanabdominal fat 22 distribution ,a familyhistoryofcardiovascular disease ordiabetes,elevated 23 levelsof risk factors forcardiovascular disease and inthose caseswhen subjects suffer fromafflictions thatmaybenefit fromweight reduction .In our opinion,public healthmeasures should,besides informing thepublic about thehealthhazardsofoverweight,payattention totheunwarranted social pressures forthinnessespecially inthose inwhomoverweight ispredominantly a 24 cosmeticproblem . ACKNOWLEDGEMENTS Theauthors like tothank thePraeventiefonds inTheNetherlands for financial support for the studyand theheadand staffof theDepartment ofGeneral Practice of theUniversity ofNijmegenfor their cooperation and generously providing of the information fromtheir ContinuousMorbidity Registration. Theyparticularly like tothank P.Hoppener,H.vandenHoogen,S.Ward ,W. Knuiman-HijlandA.M. Zijlmansfor technical assistancewith the study. 21 -100- REFERENCES 1.SimopoulosAP,vanvanItallieTB.Bodyweight,HealthandLongevity.Ann InternMed1984;100:285-295. 2.ReportoftheRoyalCollegeofPhysicians;Obesity.JournaloftheRoyal CollegeofPhysiciansofLondon1983;17:1-58. 3.SeidellJC,deGrootCPGM,vanSonsbeekJLAetal.Associationsofmoderate andsevereoverweightwithself-reportedillnessandmedicalconsumptionin Dutchadults.AmJPublicHealth1986;76:264-269. 4.AllonN.Thestigmaofoverweightineverydaylife.In:WolmanBB,ed: Psychologicalaspectsofobesity:ahandbook.NewYorketc:VanNostrand ReinholdCompany,1982: 5.TobiasAL,GordonJB.Socialconsequencesofobesity.JAmDietAss1980; 76:338-342. 6.StewartAL,BrookRH.Effectsofbeingoverweight.AmJPublicHealth1983; 73:171-178. 7.LarssonB,BjorntorpP,TibblinG.Thehealthconsequencesofmoderate obesity.IntJObes1981;3:97-116. 8.BengtssonC,TibblinE.Epidemiologiskasynpunkkerpaovervikthaskvinnor. Lakartidningen1974;71:4189-4191. 9.SeidellJC,BakxJC,HoogenHJMvanden,DeurenbergP.Overgewichtin relatietotmorbiditeitensubjectiefwelbevindeninvier huisartspraktijken.In:BoermaWGW,HingstmanL.ed.Deeerstelijn onderzocht.Deventer:VanLoghumSlaterus,1985:225-236. 10.DirkenJM.Hetmetenvan'stress'inindustrielesituaties.Groningen: Wolters,1967:130-138. 11.VisserAP.DebetekenisvandeVOEG:enkelegegevensoverde begripsvallditeit.GezondheidenSamenleving1983;4:177-188. 12.PhilipsenH,JoostenJ,ReekJvan.Eersteuitweidingoverfactorstructuur ensubschalenvandeVOEG.Maastricht:RULimburg,1984. 13.DixonWJ.BMDPStatisticalsoftware.BerkelyCA:UniversityofCalfornia Press,1981. 14.KleinbaumDG,KupperLL,MorgensternH.EpidemiologicResearch-principles andquantitativemethods.LifetimeLearningPublications,Belmont,1982. 15.Rodin,J.Psychologicalfactorsinobesity.In:BjorntorpP,Cairella, HowardedsRecentadvancesinobesityresearchIII.London:JohnLibbeyand Co.Ltd.,1980:106-123. -101- 16.AsherWL.Theenigmaofobesity:anoverview.InAsherWLed.Treatingthe obese.NewYork:MedicalPress,1974:1-16. 17.JordanHA.Weightregulationinman:physiologicandpsychologicalfactors. InAsherWLed.Treatingtheobese.NewYork:MedicalPress,1974:17-27. 18.VanStrienT.Eatingbehaviour,PersonalitytraitsandBodyMass(thesis). Lisse:Swets&Zeitlinger,1986. 19.WeighillVE,BuglassD.Weightchangeandpsychologicalstateinobese women.Appetite1984;5:95-102. 20.Linet01,MetzlerCM.Emotionalstatusduringaweightreductionprogram.J ClinPsychiatry1981;42:228-231. 21.BlackburnGL,ReadJL.Benefitsofreducing-revisited.PostgradMedJ1984; 60(suppl3):13-18. 22.BjorntorpP.Obesityandtheriskofcardiovasculardisease.AnnClinRes 1985;17:3-9. 23.JamesWPT.Dietaryaspectsofobesity.PostgradMedJ1984;60(suppl. 3): 50-55. 24.PolivyJ,HermanCP.Breakingthediethabit:thenaturalweight alternative.NewYork:Basicbooks,1983. 25.MantelN,HaenszelW.Statisticalaspectsoftheanalysisofdatafrom retrospectivestudiesofdisease.JBNatCancerInst1959;22:719-748. 26.SeidellJC,BakxJC,deBoerE,DeurenbergP,HautvastJGAJ.Fat distributionofoverweightpersonsinrelationtomorbidityandsubjective health.IntJObesity1985;9:363-374. -102- AppendixI A.Descriptionofclustersofcomplaints (identifiedusingfactoranalysis (Philipsenetal.1983) Nameofcluster Item Women Men Stomach/digestion 1.Doesyourstomachoftenfeelfulland bloated? + + + + + 2.Doyouoftenhavepainsinthestomach region? 3.Doyouoftenhaveanupsetstomach? 4.Doyouoftenhavestomachcomplaints? 5.Doyouoftenhaveindigestion? -Skeleto-muscular 1.Doyouoftenhaveunpleasantlycoldfingers, handsorfeet? 2.Doyourbonesormusclesoftenache? 3.Areyoubotheredbyweakorachingfeet? 4.Areyouoftentroubledbybackache? + + + 5.Doyourarmsandlegsoftengodeador tingle? -Nervousness 1.Doyouoftengetupset? 2.Doyouoftenfeelnervous? + 3.Areyouoftenirritable? + 4.Doyourhandsoftenshake? -Tiredness 1.Doyouoftenfeeltired 2.Doyouoftenfeelsleepishorsluggish? 3.Doyouoftenfeellistless? + + + 4.Doyougenerallygetupfeelingtiredand notrestedinthemorning? 5.Doyoufeeltiredsoonerthanyouthinkis normal? + -103- -Noseandthroat 1.Areyouregularlybotheredbycoughing? + 2.Doyouoftenhavetoclearyourthroat? + 3.Isyournoseoftenblocked? + B.Othercomplaints,forwhichtheanswerswerefoundtodifferbetweenthe overweightandcontrolgroups(p<0.05,afteradjustmentforsocialclass). Onlycomplaintsthatarenotincludedintheclustersofcomplaintslisted abovearepresented. -Complaint 1.Doyougetshortofbreatheasily? + + 2.Doyoufallasleepeasilyanddoyousleep well? + 3.Doyousometimessweatheavilyevenwhenit isnothot? + 4.Doyouoftenhavelittlemishaps? 5.Doyouoftenfeeltightinthechest? + + +=Stratifiedanalysis(withadjustmentforsocialclass)revealeda differencebetweenoverweightandcontrolgroup(p<0.05)(usingthe 14 chi-squaredtestofMantelHaenszel ) Note:clusters(factors)identifiedwhen,afterrotation,thefactorloadingof theitemsononlyoneofthefactoreswasatleast0.40.Moreinformation abouttheclusteranalysisisavailableonrequesttotheauthors. ..-. . -104- I ,T ' CHAPTER6 FATDISTRIBUTIONOFOVERWEIGHTPERSONSINRELATIONTOMORBIDITYAND SUBJECTIVE HEALTH J.C.Seidell,J.C.Bakx,E.deBoer,P.Deurenberg,J.G.A.J.Hautvast IntJObesity 1985;9:363-374 ABSTRACT Theassociationbetween fatdistribution,morbidityand subjectivehealthwas studied in95overweightadultmenand210overweightadultwomen.Retrospective morbiditydataweretakenfromacontinuousmorbidity registrationmadeby generalpractitionersoveraperiodofmaximally 17years.Inaddition informationabout subjectivehealthandweighthistorywasobtained froma selfadministeredquestionnaire.Anthropometricmeasurementswere takenand,onthe basisofwaist-hipandwaist-thighcircumference ratios,subjectswere classified intoupperbodysegmentobesity,intermediateobesity,and lower body segmentobesity. Itwas found that,adjusted forageandbodymass index,ahighwaist-thigh circumference ratiowasa risk factor forhypertensionand forgoutor diabetes inwomenandarthrosis inmen.A lowwaist-thigh ratiowasassociatedwith a highprevalenceofvaricoseveins inwomen. Theassociationsofwaist-hipscircumference ratiowithmorbiditywere less pronounced,with theexceptionofhypertension formen.Information frontthe questionnaire revealedthatpersonswithupperbody segmentobesity (especially men)feltlesshealthyandhadmorehealth complaints.These findingsweremore pronounced for subjectsyounger than50yearsofage thanfor thoseof 50years andolder. Theweighthistoriessuggestthatwomenwith lowerbodysegmentobesityhad a longerhistoryofobesitythanwomenwithupperbodysegmentobesity.Thiswas not found inmen.Itisconcluded thatclassificationofobesityonthebasisof circumference ratiosisuseful fortheevaluationofhealthhazardsof overweightsubjects. -105- INTRODUCTION Thedistributionof fatover thebodyisimportant intheassociationofobesity and somemetabolicdisturbancesinlipidandcarbohydratemetabolism. A relativepredominanceof fatintheupperpartofthebody,or,more specifically,intheabdominal region,isrelatedtometabolicdisturbancesas wellastotheprevalenceofvariousdisorders,suchasdiabetes ,,.. 9,15,16,18,21,29,31, . . 3,16,18,21 16 ,and nlK1 ,, ,. mellitus • > • • • > hypertension ' ' ' gallbladder disease menstrualdisorders .Fatdistributionhasalsobeenshowntobean important predictor ofdeathduetocardiovasculardiseaseor stroke,andofdeathsfrom IT j 21a all causesin.men 22 andwomen Usefulmeasuresfortheclassificationofdifferenttypesof fatdistributionin epidemiological studies includethecircumference ratioofwaistand i iR 1ft y \ y? 1 hips > • • i an( jthe ratioofwaistandthighdiameters. Bothwaist-hipsandwaist-thighcircumference ratioswereused inthisstudyto describe the fatdistribution inagroupofoverweightDutchmenandwomen.The aimofthestudywastodeterminewhetherassociationsoffatdistributionwith diseasecouldbeconfirmedorestablished,andwhether fatdistributionwas relatedtosubjectivewell-beingandweighthistory.The resultsofthestudy confirmthatfatdistribution isanimportantindicator forboth registered disordersand subjectivehealth status (thelatterespecially inmen). SUBJECTSANDMETHODS Subjects Subjectswerepatientsoftwogeneralpracticestakingpart intheContinuous MorbidityRegistrationattheDepartmentofGeneralPracticeoftheUniversity ofNijmegen.Personswhohadbeendiagnosed,bythegeneralpractitioner as 'obese' (criterium:Broca Index >115%)atanagebetween20and 50years,and were stillobeseatthetimeofthisinvestigation,receivedamailed questionnaire.Thosewhocompleted thequestionnaire,fromwhich information about subjectivehealthandweighthistorywasobtained,wereaskedto participate inananthropometric study.548Personswere sentaquestionnaire, 401 (73%)returned thequestionnaireand,ofthese,305 (56%;95men,210women) agreed toparticipate intheanthropometricstudy. Measurements The subjectsweremeasuredeitherathomeor inthedoctor'sofficebyatrained -106- assistant.Height instandingposition (tothenearest0.1 cm)andweight (to thenearest0.5 kg,onacalibratedbalance)weremeasuredofsubjectswithout shoes,wearingonlyunderwear orlight indoorclothing.Measurementsofthe circumferencesofsubjectswere takeninastandingposition,breathing normally.Theminimal circumference ofthewaist,themaximalhips circumferencesandthehighesthorizontal circumferenceoftheleftupper thigh weremeasured (tothenearest0.1 cm)usingametaltapemeasure. Calculations The ratiosofwaisttothighcircumference ratio (waist-thigh ratio)andwaist tohipscircumference (waist-hips ratio)werecalculated.Waist-hips ratiohas been showntobeausefulmeasure forclassificationof fatdistribution in 9 "71R 1ft 91 99 relationtometabolicassociatesofobesity ' < > < < mTherewere indications fromtheliterature thatthewaist-thigh ratiocouldbeatleastan equallygood indicatorof fatdistribution .Thereforebothcircumference ratios wereused intheanalysis.As inother studies.UpperBodySegmentObesity (UBSO)wasdefinedastheupper tertileofthecircumference ratios,when these wereusedascategorically scaledvariables,andLowerBodySegmentObesity 9 22 (LBSO)asthelower tertileofthe ratios' 9 BodyMass Index (BMIinkg/m )wascalculatedusingweightandheightas measuredbythegeneralpractitionersatthetimeofwhichobesitywasdiagnosed (t Q )andusingweightandheightattimeoftheanthropometric studyin1984.In analysis,withcategorically scaledvariables,twoclassesofobesitywere distinguished,according totheclassificationofGarrow :BMI25.0-29.9 (GradeIormoderateobesity)andBMI >30.0 (GradeIIorsevere obesity). Morbidityanalysis Forallpersonsthattookpartinthestudy,wewere informed aboutthe diagnosesmadebythegeneralpractitionersduring theprevious17years(i.e. since thestartoftheContinuousMorbidityRegistration)ofanyofthe following chronicdisorders:diabetesmellitus,gout,anginapectoris, hypertension,varicoseveins,non-fatal ischaemicheartdisease,arthrosis, eczemaandprolapsuteri. Thediagnosesweremadeusingaclassification systembasedonthe International ClassificationofDiseases (ICDversion6)using standardized procedures shortdescriptionofdiagnostic criteria isincluded inAppendix I.The a 24 -107- prevalence of somepossibly relevantdisorders for thisstudywas toolow for analysise.g. cancer,stroke,gallstones,myodegeneratiocordisand periferal arterial heartdisease. Combinations of somedisorderswerealso studied (e.g.diabetesand/or gout,and a combination of seriousdisorders:diabetes,gout,anginapectoris, ischaemic heartdisease and arthrosis). Questionnaire The subjectswere asked to ratetheir subjective health statusona7-point scale (1-verygood,7«very bad). Thequestionnaire included alistof 53 items (subjective health complaints)oftenused tomeasure subjective health in 4 epidemiologic surveys inTheNetherlands .For thispaper,a 39-item abbreviationwasused,inanalysis since these itemshadbeenpublished in 26 English .The totalnumber ofcomplaints onthis39-item listwas calculated foreach subject.Thedistribution of thesecumulative scoreswas markedly skewed tothe left.The square rootwasapproximately normallydistributed and wasthereforeused inanalysis.Subscoresofsomeclustersofcomplaintswere calculated.These clustershadbeen indentified using factoranalysisondata 25 from largeepidemiologic surveys . Information aboutweight history,about smokinghabitsandabout slimming frequencywasalsoobtained fromthequestionnaire.Age and socio-economic status,based ontheprofession ofthehead ofthehousehold,were registered in 24 themorbidity registration . Statistical methodsmorbidity analysis Ina forward stepwise multiple logistic regression analysis,using the computer package programme BHPLR ,the following continuousvariableswere candidates for inclusion inthelogistic modelasindependentvariables:age,BMI, circumference ratiosand,categorically scaled, socio-economic status (lower, 24 middle,upper class) Analysisusing age (<50;> 10 years in 1984),BMI (<30;>30kg/in in 1984) and circumference values (lower,middleandupper tertile)as categorically scaledvariables produced comparable results.Thelatter procedure yielded multiple regression coefficients intervals oftheodds ratiosaccording to 23 LemeshowandHosmen (1984) capaing theupper tertileof the circumference ratios tothe lower tertile.Analysiswereperformed separately formenand -108- womenandforthetwocircumferenceratios. Questionnaireanalysis Fortheanalysisoftheinformationobtainedfromthequestionnaire,t^testsor chi-squaredtestswereperformedtocomparesubjectsintheuppertertile(upper bodysegmentobesity)withthoseinthelowertertile(lowerbodysegment obesity)ofthewaist-thighratio.Multiplelinearregressionanalyseswere performedusingtheprogramBMDPLR. RESULTS SomecharacteristicsofthestudypopulationaregiveninTable1. Table1.Meanandstandarddeviationofageandsomeanthropometric measurementsofthestudypopulation Men (n=9S) Women (n=210) Mean Age years Weight kg cm Height 2 BMI kg/m Waist-circumference cm cm cm Hips-circumference Thigh-circumference S.D. Mean S.D. 49.2 9.2 11.9 90.3 11.8 89.6 6.1 4.1 9.7 177.4 109.9 10.0 105.0 60.1 5.0 59.2 6.1 3.1 8.8 5.8 4.0 49.9 9.7 78.7 163.4 29.5 28.7 101.6 Waist-hipsratio 0.82 0.05 0.97 0.06 Waist-thighratio 1.42 0.16 1.72 0.15 InTable2theclassificationofmenandwomenintypesoffatdistribution (upperbodysegmentobesity,intermediateobesityorlowerbodysegment obesity)isshown.Whenclassifyingaccordingtobothwaist-hipsratiosand waist-thighratiosnoexactlyequalgroupscouldbeformed(especiallyinthe caseofthewaist-hipsratiodistribution).Thereforeonlyapproximationsof thetertilesoftheratioswereused.Theratioswerehighlycorrelated (r= 0.76;p<0.001)andonlyafewsubjectsappearindifferentextremetertiles ofthedifferentratios. -109- Table2.Numberofsubjectsinvariousclassesofwaist-hipscircumference ratioandwaist-thighcircumferenceratio* Women Waist-thigh ratio Waist-hipsratio <0.78 (LBSO)** <1.34 (LBSO)** 1.35-1.49 (Intermediate) > 1.50 (UBSO) allwomen 0.79-0.84 (Intermediate) >0.85 all women (UBSO) 45 9 3 24 44 19 2 14 49 71 67 71 57 87 65 209 Men Waist-thigh ratio Waist-hipsratio <0.94 <1.65 (LBSO)** 1.66-1.78 (Intermediate) >1.79 allmen (UBSO) 0.95-0.99 >1.00 allmen (LBSO)** (Intermediate) 26 4 1 4 15 8 2 13 22 32 31 27 37 95 (UBSO) Cut-offvaluesweresettoobtainapproximatetertilesofthe distributionofthecircumferenceratios.Becauseofthedistributionof thecircumferenceratiosexactlyequalgroupscouldnotbeformed ** LBSO(LowerBodySegmentObesity)wasdefinedasthecategorywithlow valuesofthecircumferenceratios.Intermediatewasdefinedasthemiddle categoryofthecircumferenceratios. UBSO(Upperbodysegmentobesity)wasdefinedasthecategorywithhigh valuesofthecircumferenceratios FromTable3itcanbeseenthatwomenwithupperbodysegmentobesitywere heavieratthetimeofinvestigation(1984)thanattimeofdiagnosisofobesity (t Q ).Upperbodysegmentobesewomengained2.7+7.6kg(mean+standard deviation).Therewasnosignificantchangeofweightinsubjectsinanyofthe othercategoriesofwaist-thighratio. 32 31 -noTable3.AgeandBodyMassIndex(BMI)ofthestudypopulationattimeof investigation(1984)andattimeofdiagnosis(t0)indifferentclasses offatdistribution Number LBSO** ofsubjects Mean SD Intermediate Mean UBSO** SD Mean SD Women 71 Numberofsubjects 71 67 Ageatt Q 34.03 8.91 38.27 9.08 41.04 8.41 Agein1984 46.24 9.65 50.49 9.83 53.15 8.51 BMIatt Q 28.15 2.04 29.15 3.56 29.74 3.23*** BMIin1984 27.84 2.99 29.20 3.75 31.05 4.34*** Men 32 Numberofsubjects 32 31 Ageatt„ 35.52 8.69 37.83 7.54 41.04 6.95 Agein1984 45.34 9.65 48.59 7.74 53.65 8.32 BMIatt Q 28.43 1.54 28.76 1.38 29.60 2.70 BMIin1984 27.74 2.31 28.77 2.49 29.46 4.02 *t Q=1967-1983.Subjectswereselectedonthegroundsthattheyhadbeen diagnosedbytheirgeneralpractitionaratage20-50years ** LBSO(LowerBodySegmentObesity),IntermediateandUBSO(UpperBody SegmentObesity)weredefinedasthelower,middleanduppertertileof waist-thighratiorespectively (seeTable1) *** DifferenceinBMIbetweentimetQand1984statisticallysignificant, (p<0.05) Theindependentvariablesincludedinthestepwiselogisticregressionanalysis, themultiplelogisticregressioncoefficientsandtheirstandarderrors:,using age,BMIandcircumferenceratiosascontinuousvariablesareshowninTable4. Thenumberofcasesforeachdisorderorcombinationofdisordersisalpogiven. TheresultsshowninTable4wereessentiallythesamewhencategoricallyscaled variablesforage,BMIandcircumferenceratioswereused. Theodds'ratios(showninTable5,withtheirconfidenceintervals)are estimatesofrelativeriskforhavingadisorderforsubjectswithupperbody segmentobesitycomparedtopersonswithlowerbodysegmentobesity.Theodds -111- able 4.Multiple regression coefficients obtained instepwisemultivariate logistic regressionanalysis,using disorders asdependent variables andBMI,ageandcircumferences ratios asindependent continuous variables ender Dependent variable (Disorder) onen Hypertension Gout ordiabetes Circum- Number Regression coefficients ference circum- SES ratio ference high/low middle/low wr 9.45 0.10 WH 8.28 0.10 wr 9.43 4.58 WH 15.80 11.50 ofcases SES - 0.07 0.12 62 62 13 13 Arthrosis wr/WH* 8.10 Varicose veins wr 1.19 2.70 WH 4.26 6.44 myocardial infarction WT/WH* 8.88 0.12 - 13 Prolapsuteri WT/WH* 8.64 0.13 0.97 22 0.12 0.03 36 55 55 Angina pectorisor Serious disorders combined** Hypertension WT - 8.17 3.96 - 0.87 0.97 16 WH - 12.60 19.10 0.26 0.51 0.39 16 WT - 8.86 7.49 0.25 - - 12 WH - 7.50 - - 1.56 - 1.25 12 myocardial infarction WT/WH - 8.11 - Serious disorders WT - 3.54 5.86 26 combined** WH - 4.79 - 26 Serious disorderor WT - 3.06 4.67 32 hypertension WH - 3.21 - 32 Arthrosis Angina pectorisor NeitherWTnorWHwere included inthestepwise regression analysisandthereforeproduced thesame Oneormore ofthe following disorders:gout,diabetes,arthrosis,anginapectoris,myocardial infarction, prolapsuter LIregression coefficients were statistically significant from zero (b/SE(b)greaterthan 2) -112ratiosarederived from regressioncoefficientsofthecategorically scaled circumference ratiosandareadjusted forothervariables inthemodel.The resultsindicateahigherprevalenceofhypertension,goutordiabetes inwomen intheupper tertileofthewaist-thigh ratiotheninwomen inthelower tertile,andalowerprevalenceofvaricoseveins inwomen intheupper;tertile ofbothcircumferenceratios. Table 5.Odds ratiosand95%Confidence intervalsofOdds ratiosfordisorders, comparing theupper tertiletothelower tertileofthecircumference ratiosofwaist-hipsandwaist-thigh Circum- 95% Confidence Interval of theOdds Ratio Disorder Odds ratio Hypertension 2.44 1.09 - 5.56 ference Ratio Women WT WH Men WT Goutordiabetes 4.68 2.10 - 10.44 VaricoseVeins 0.37 0.27 - 0.60 Goutordiabetes 2.66 0.68 - 10.38 VaricoseVeins 0.40 0.16 - 0.86 Hypertension 3.69 1.74 - 5.64 Arthrosis Seriousdisorders WH Hypertension 11.88 1.38 - 35.40 5.58 2.86 - 10.89 54.98 4.02 - 752.00* Note: Seriousdisorders:oneormoreofthefollowingdisorders:anginapectoris, Ischemicheartdisease,gout,diabetes,arthrosis - Oddsadjusted forothervariables inthelogistic regressionmodel -WT=waist-thigh ratio WH=waist-hips ratio * Themagnitudeof theodds ratioand itswide confidence interval canbe explainedby thelargenumberofvariables inthemodel (4)while therewere only 16malecasesofhypertension Formen,ahighprevalenceofarthrosis,orofat leastone seriousdisorder, was found intheupper tertileof thewaist-thigh ratio.Using thewaist-hips ratioasindicator of fatdistribution,hypertensionwasmore frequent in -113- subjects intheupper tertile ofboth ratios. Upper body segmentobesemenandwomen,asdefined as theupper tertile of the waist-thigh ratio, reported a lower subjective health status than lower body segment obesemenandwomen (p<0.05).When stratified intotwoage groups( < 50years; >,50years), thisdifference disappeared inthegroup ofwomen of 50 years and older. Themeans and standard deviation of the square rootof thenumber of complaints are given inTable 6.Without taking age intoaccount,a statistically significantdifferencewas foundbetween the results forupper body segment obesemenand lower body segment obesemen.When subjectswere stratified into twoage groups,a significantdifference (p<0.05)was found only for men younger than 50yearsofage. Table 6. Crude and age-specific estimates of the square root*ofthe total number ofhealth complaints ona listof 39complaints (meanand standard deviations) inmenandwomenwithUBSOand LBSO** LBSO** age n mean score* LBSO** n st.dev. Women< 50yrs > 50yrs all ages Men <50yrs >,50yrs all ages 42 19 61 2.88 - 0.83 17 11 28 2.25 - 1.32 2.94 - 1.26 2.90 - 0.97 2.48 - 1.01 2.34 - 1.19 mean score p-value for difference st.dev. in mean score 17 44 61 3.37 - 1.05 0.1314 3.02 - 1.39 0.8873 3.11 - 1.31 0.2384 6 16 22 3.63 - 0.80 0.0214 3.08 -1.13 0.2184 3.23 - 1.06 0.0124 * Square rootwas takenbecause itsdistributionwas approximately normal (ahigher value indicatesmorehealth complaints) ** LBSO and UBSO Lower andUpper Body SegmentObesity,defined as the lower and upper tertile of thewaist-thigh distribution, respectively Multiple linear regressionwasperformed,with the square rootof the total number of complaints asa continuousdependentvariable and the waist-thigh ratio,age,BMI in1984 (all three continuous), socio-economic status, slimming frequency,and smokinghabits (allthreeascategorical variables) as independent variables. -114- Formen,12.0%ofthevariancewasexplainedbythefullmodel,withthe 2 waist-thighratioastheonlysignificantindependentvariable(r =6.7%).For women10.2%ofthevarianceinthenumberofcomplaintswasexplainedbyallsix variablesinthemodel,withslimmingfrequencyastheonlysignificant 2 independentvariable (r =5.6%). Whenclustersoforgan-specificcomplaintswereanalysed,upperbodysegment obesemenwerefoundtosuffermorefromheartandchestpains,complaintsof bonesandextremitiesandshortnessofbreaththanlowerbodysegmentobesemen. Thesedifferenceswerestillsignificant(p<0.05)alsowhenadjustedforage. Formenyoungerthan50years,stomachcomplaintsandindigestionwerealso significantlymorefrequentinupperbodysegmentobesesubjects(p<0.05)(see AppendixIIforalistingoftheitemsoftheclustersofcomplaints). Amultiplelineairregressionwasperformedwithwaist-thighratioasa continuousdependentvariableandsixindependentvariables:age,BMIatt Q ,BMI in1984ascontinuousvariables,andsmokinghabits,slimmingfrequencyand socio-economicstatusascategoricallyscaledvariables.Formenthefullmodel explained30.1%ofthevariationinthewaist-thighratio.Thesinglevariables 2 withsignificantpredictivepowerwereage(r =16.5%)andsocio-economic 2 status(r =3.3%). Forwomenthefullmodelexplained22.0%ofthevariationin thewaist-thighratio,withtheBMIin1984(r =15.9%)andage(r =9.0%)as significantpredictors.Ahighwaist-thighratio(upperbodysegmentobesity)is thusassociatedwitholderageinmenandwitholderageandhigherdegreeof obesityinwomen. Womenwithupperbodysegmentobesityreportedlessoftenthattheyhadbeen "toofat"atbirth(p<0.05),inchildhoodorinadolescence (p<0.10)than lowerbodysegmentobesewomen.Mendidnotconsiderthemselvestohavebeen overweightbefore20yearsofage.Whenthesubjectswereaskedaboutthe maximalweighttheyrememberedfromactualmeasurementsduringsixageintervals 1 5 - 1 9 , 2 0 - 2 9 , 3 0 - 3 9 , 4 0 - 4 9 , 5 0 - 5 9 , 60yearsorolder)theaverageBMI inallageintervals,except15-19and20-29years,wasfoundtobehigher forwomenwithupperbodysegmentobesitythanforwomenwithlowerbodysegment obesity.TheaverageBMIofmenwithupperbodysegmentobesitywas significantlylowerthanthatofmenwithlowerbodysegmentobesityintheage interval20-29years.Inotherageintervalstherewasnodifferencebetween theaverageBMIofupperandlowerbodysegmentobesemen.Nodifferenceswere foundbetweenupperbodysegmentobesesubjectsandlowerbodysegmentobese subjectsinslimmingfrequency,smokinghabitsorsocio-economicstatus. -115DISCUSSION The studypopulationwasagroupofobese subjects intwogeneral practices in twomedium sizedmixed rural/industrial towns inTheNetherlands (near Nijmegen).Thepopulationwasprobablynot representative fortheentire obese population inTheNetherlandsbutwithin thepopulation ofobese subjects of thepractices the studypopulationwasnotselected onthebasisofBMIand morbidity.Nevertheless,extrapolation toother overweight populations should be donewithcaution. Obesity hadbeendiagnosed by thegeneral practitionarswithin the last 5 - 1 7 yearsbefore thisstudy.From theweight histories inthequestionnaire itwas clear that the subjects included inthe studypopulationhadbeenobeseat least since thetimeofdiagnosisof obesity. There isevidence that fatdistribution isa relatively permanent charactistic of subjects,evenwhen therearemajorweight changes '' ' ' . I t was considered appropriate therefore,tolinkpresent fatdistribution tothe retrospective morbiditydata. Upper bodysegmentobesityandlowerbodysegmentobesitywere arbitrarily defined astheupper tertileand lower tertileof thecircumference ratios,an 9 22 approachused alsobyother investigators ' .Both circumference ratios probablymeasure approximately thesameconcept:the fatstored inthe abdominal region relative tothat inthegluteal (hips)and femoral (thighs) regions.Both gluteal and femoral regionsprobablyhave similar physiological characteristics . The results from themorbidity analysis confirm that fatdistribution, as measured bycircumference ratios,isassociated with theprevalence of some chronic disorders inseriously ormoderately obese subjects.Theassociations of a relatively central deposition of fatwithdiabetes,glucose tolerance and hyperinsulinemiaarewell k n o w n 9 ' 1 0 ' 1 1 ' 1 5 ' 1 6 ' 1 7 ' 1 8 - 2 5 ' 2 9 ' 3 1 .T h e possible mechanismsof this relationshiphave recentlybeen studiedbyEvanset al. '' .Diabeteswas significantly related tothewaist-thigh ratio forwomen inunvariate analysisbutnotafter adjustments forageandBMI.Both diabetes and goutare known tobe the resultofmetabolic aberrations associated with 1928 obesity ' andVague reported thatall ofhisgout-afflicted subjectswere 29 hyperandroid . Diabetes (11 female cases)andgout (2female cases)werenotvery common inthe studypopulation but the combinationwas independently related to fat distribution. -116- We founda strongassociationbetweenhypertensionandboththewaist-hips ratio formen,andamoremoderateassociationbetweenhypertensionandthe waist-thigh ratioforbothmenandwomen.Other studieshavealsofoundthatfat distribution,independentlyofdegreeofobesity,isrelated to 3 1Rfi 1 1ft?1?Q hypertension > > ' • ' , sincehypertension isa risk factor for cardiovascular diseaseand stroke,thisobserved relationmaypartlyexplain the 21a22 findings fromGothenburg ' inwhich itwasdemonstrated that,formen,the waist-hips ratioisaprognostic factor fordeath fromcardiovascular disease or strokeand fordeathofallcauses. Wearenotawareofanystudy reportinganassociationbetweenfatdistribution andarthrosis.Itistemptingtosuggestthattheassociationobservedhere indicates thatmetabolicdisturbancemayplayagreater rolethanweight induced wearand tear intheetiologyofarthrosis.Infact,afteradjustment forthe waistthighratio,anegativemultiple regressioncoefficient forBMIwasfound 27 (Table 4). Silberberg hasreviewednumerousstudiesontheassociationof obesityandarthrosis,andhasconcluded thatthere isno reasontoassumea direct causal relationshipbetweenjointdiseaseandthemechanical overload causedbyobesity.An injuriousmetaboliceffectoffatorcomponentsof faton weightbearingaswell asnon-weightbearing jointswashypothesized .Itcould 27 be importanttonotethatanassociationbetweenfatdistributionandblood 8 921 lipidslevelshasbeenobserved '' .Univariateanalysis revealed thatfat distribution inwomenwasassociatedwitharthrosis (resultsnotshown)but this associationdisappearedwhenadjustments forBMIandageweremade.The reason for thissexdifference inrisk isnotclear,although itshouldbenoted that ingeneralmenhavehigher circumference ratiosthanwomen (Table 2). Thenegativeassociationbetweencircumference ratiosandvaricoseveinshasnot been reported inanyother studieson fatdistributionanddisease,although it hasbeensuggestedbyVague thatinsufficientvenouscirculation isoneofthe 29 characteristicsofgynoidobesity .Upperbody segmentobesityandandroid obesityare thoughttooverlapconsiderablyasarelowerbody segment Obesity 1 21 andgynoidobesity' Theanalysisof theinformationobtained fromthequestionnaire revealedthat, formenalsootherdisorders,notmeasured inthemorbidity registration,were associatedwith fatdistribution.A highwaist-thigh ratiowasassociatedwith stomachcomplaintsand indigestion,heartandchestpains,skeletalcomplaints, problems intheextremities,and shortnessofbreath. Forwomena relativepredominanceof fatstorage intheabdominal region (upper bodysegmentobesity)wasassociatedwithdegreeofobesity,andprobably also -117- witharelativelylateonsetofobesity.Thisisinagreementwiththeresults ofstudiesdemonstratingtherelationbetweenadultonsetofobesityoradult weightgainandvariousmetabolicassociatesofobesity.Theserelationsare essentiallythesameasthosefoundbetweenfatdistributionandmetabolic aberrations. Wemayconcludethatintheevaluationoftheriskprofileofmoderatelyand seriouslyobesesubjectsitisimportanttoassesstheirfatdistribution. Circumferenceratiosmaybeusefulforthispurpose.Theyareeasytomeasure andcanbeappliedinepidemiologicresearchaswellasingeneralpractice.The waist-thighratiocouldbeconsidered,aswellasthewaist-hipsratio,because theformerratiomightbeatleastequallysensitiveasanindicatorofchronic disordersasthelatterratio.Furtherresearch,preferablylarge-prospective studies,shouldrelatetheincidenceofdisordersnotonlytobaselinefat distributionandBMIbutalsotochangesinthesecharacteristics,totakeinto accountindividuals'weighthistoryandfamilyhistoryofobesityandmorbidity. TheeffectsofweightreductionOnbothfatdistributionandhealthshouldalso becarefullystudied. Untilmoreprecisecriteriaaredeveloped,thetertilesofthewaist-thighratio andwaist-hipsratiocanbeusedasaroughmeasureforclassificationinto extremetypesoffatdistribution. ACKNOWLEDGEMENTS TheauthorsliketothankthePraeventiefondsinTheNetherlandsforfinancial supportforthestudyandtheheadandstaffoftheDepartmentofGeneral PracticeoftheUniversityofNijmegen(Director:prof.F.J.A.Huygen)fortheir cooperationandgenerouslyprovidingoftheinformationfromtheirContinuous MorbidityRegistration.TheyalsoliketothankW.vandenBosch,P.Hoppener, H.vandenHoogen,D.Kaayk,T.Kempener,S.WardandA.M.Zijlmansfor technicalassistancewiththestudy. REFERENCES 1.AshwellM,ChinnS,StalleyS,GarrowJS.Femalefatdistribution:a photographicandcellularitystudy.IntJObesity1978;2:289-302. 2.BjorntorpP.Morphologicalclassificationsofobesity:whattheytellus, whattheydon't.IntJObesity1984;8:525-533. -118- 3.BlairD,HabichtJ,SimsEAH,SylvesterD,AbrahamS.Evidenceforan increasedriskforhypertensionwithcentrallylocatedbodyfatandthe effectofraceandsexonthisrisk.AmJEpidemiol1984;119:526-540. 4.DirkenJM.Arbeidenstress.Hetvaststellenvanaanpassingsproblemenin werksituaties.Groningen:JBWolters,1967. 5.DixonWJ.BMPDStatisticalSoftware.Berkeley,CA:UniversityofCalifornia Press,1983. 6.EdwardsDAW.Observationsonthedistributionofsubcutaneousfat.ClinSci 1950;9:259-270. 7.EvansD,HoffmannR,WilsonRK,KalkhoffR,KissebahAH.Relationshipof regionalbodyfatdistributiontoglucosetoleranceandresponsetodietin premenopausalwomen.Proc.NorthAm.Assoc.StudyObesity,Poughkeepsie,New York (Abstr),1982. 8.EvansDJ,HoffmannRG,KalkhoffRK,KissebahAH.Relationshipofandrogenic activitytobodyfattopography,fatcellmorphologyandmetabolic aberrationsinpremenopausalwomen.JClinEndocrinolMetab1983;59: 304-310. 9.EvansDJ,HoffmanRG,KalkhoffRK,KissebahAH.Relationofbodyfat topographytoinsulinsensitivityandmetabolicprofilesinpremenopausal women.Metabolism1984;59:68-75. 10.EvansDJ,MurrayR,KissebahAH.Relationshipbetweenskeletalmuscle insulinresistance,insulin-mediatedglucosedisposalandinsulinbinding: effectsofobesityandbodyfattopography.JClinInvest1984;74:1515-1525. 11.FeldmanR,SenderAJ,SiegelaubAB,OaklandMS.Differenceindiabeticand non-diabeticfatdistributionpatternsbyskinfoldmeasurements.Diabetes 1969;18:478-486. 12.G a m SM.Relativefatpatterning:anindividualcharacteristic.HumBiol 1955;26:75-89. 13.GarrowJS.Treatobesityseriously.Aclinicalmanual.Churchill Livingstone,London,1981. 14.GraigLS,BayerLM.Androgenicphenotypesinobesewomen.AmJPhysiol. Anthropol1967;26:23-33. 15.HartzAJ,RupleyDC,KalkhoffRD,RimmAA.Relationshipbetweenobesityand diabetes:influenceofobesitylevelandbodyfatdistribution.PrevMed 1983;12:351-357. 16.HartzAJ,RupleyDC,RimmAA.Theassociationofgirthmeasurementswith diseasein32.856women.AmJEpidemiol1984;119:71-80. -11917.JoosSK,MuellerWH,HanisCL,SchullWJ.Diabetesalertstudy:weight historyandupperbodyobesityindiabeticandnon-diabeticmexican-american adults.AnnHumBiol1984;11:167-171. 18.KalkhoffRK,HartzAJ,RupleyDC,KissebahAH,KelberS.Relationshipof bodyfatdistributiontobloodpressure,carbohydratetoleranceandplasma lipidsinhealthyobesewomen.JLabClinMed1983;102:621-627. 19.KannelWB.Healthandobesity:anoverview.In:Healthandobesityed. Conn LH,deFeliceEA, KuoP.RavenPress,NewYork. 20.KissebahAH,VydelingumN,MurrayR,EvansDJ,HartzAJ,KalkhoffRK,Adams RW.Relationofbodyfatdistributiontometaboliccomplicationsofobesity. JClinEndocrinolMetab1982;54:254-260. 21.KrotkiewskiM,BjorntorpP,SjostromL,SmithU.Impactofobesityon metabolisminmenandwomen.JClinInvest1983;72:1150-1162. 21a.LapidusL,BengtssonC,LarssonB,PennertK,RyboE,SjostromL Distributionofadiposetissueandriskofcardiovasculardiseaseanddeath: a12yearfollowupofparticipantsinthepopulationstudyofwomenin Gothenburg,Sweden.BritMedJ1984;289:1257-1261. 22.LarssonB,SvardsuddK,WelinL,WilhelmsenL,BjorntorpP,TibbllnG. Abdominaladiposetissuedistribution,obesityandriskofcardiovascular diseaseanddeath:13yearfollow-upofparticipantsinthestudyofmen bornin1913.BritMedJ1984;288:1401-1404. 23.LemeshowS,HosmerDW.Estimatingoddsratioswithcategoricallyscaled covariatesinmultiplelogisticregressionanalysis.AmJEpidemiol 1984;119:147-151. 24.NijmeegsUniversitairHuisartsenInstituut.GewoneZiekten.Nijmegen,1980 25.PhilipsenH,JoostenJ,ReekJvan.Eersteuitweidingoverfactorstructuur ensubschalenvandeVOEG.Maastricht:RULimburg,1984. 26.vanReekJ,DiederiksJ,PhilipsenH,vanZutphenW,SeelenT.Subjective complaintsandbloodpressure.JPsychosomRes1982;26:155-165. 27.RoyalCollegeofPhysicians.Obesity.JRoyCollPhysLondon1983;17:1-58. 28.SilberbergR.Obesityandosteoarthrosis.In:Medicalcomplicationsof obesity,ManciniM,LewisB,ContaldoFeds.:301-315,AcadPress,London, 1979. 29.VagueJ.Thedegreeofmasculinedifferentiationofobesity;afactor determiningpredispositiontodiabetes,atherosclerosis,goutanduric calculus.Am J ClinNutr1956;4:20-34. -12030.VagueJ,RubinP,JubelinJ,LamGvan,AubertF,WassermanAM.Regulation ofadiposemass:histometricandanthropometric aspects.In:The regulation ofadipose tissuemass,VagueJ,BoyerJeds. :296-310.Exerptamedica, AmericanElsevier,Amsterdam,NewYork,1974. 31.VagueJ, CombesR,TranciniM,Angeletti S,RubinP.Thediabetogenic adiposemass. In:Medical complicationsofobesity,ManciniM,LewisB, Contaldo Feds.:45-51,AcadPress,London,1979. -121- AppendixI Diagnostic criteria forchronicdiseasesinthecontinuousmorbidity registration. Hypertension Hypertensionwas registratedwhendiastolicbloodpressure (measured inlying position)washigher than100mmHgatatleastthreedifferentoccasions. Diabetesmellitus A singlebloodglucose-level >10mmol/1wasusedasthecriterion.Indoubt (6-10mmol/1)confirmedbyglucosetolerancetest. Gout Goutwas registeredwhen severalclinical symptomsclearly indicatedgout (e.g. painand swelling inalargetoe,unilateral arthritis infeetor toes, hyperuricarmia,manifestationorsuspectionoftophus,presenceofurale crystals injointsetc.) Arthrosis Arthrosisdeformanswas registeredwhentherewere clearclinical signsof arthrosis (likepain,stiffnessand inflammation)were found,inmost cases confirmedbyanX-ray. Varicoseveins Varicoseveinsoflower limbswere registeredwhenclearlyvisiblevaricose veinswerepresented,observedeitherat routineexaminationor following specific complaintsofthepatient. Myocardial infarction Myocardial infarctionwas registeredwhenaclinical suspectionofamyocardial infarctioncouldbeverifiedbyabnormal ECGand/orofconcentrationsof enzymesCPK,SGOT,LDH (exceptionsweremade forpatientswhodiedbefore such investigationswere possible). Anginapectoris Anginapectoriswasdiagnosedwhenaclinicalexaminationprovided sufficient evidence foranginapectoris (notnecessarilyconfirmedbylaboratory-orECG - test). -122- AppendixII Clustersofcomplaintstakenfromalistof53healthcomplaints' .subjects wereaskedtocircle'yes'or'no'foreachitem -Digestivetractcomplaints -Doesyourstomachfeelfullandbloated? -Doyouoftenhavepainsinthestomachregion? -Doyouoftenhavestomachtrouble? -Doyouhaveunspecificstomachcomplaints? -Doyouoftenhaveindigestion? -Skeletalcomplaints -Doyouregularlyhaveunpleasantcoldfingers,handsorfeet? -Doyourbonesormusclesoftenache? -Areyoubotheredbyweakorachingfeet? -Areyouoftentroubledbybackache? -Doyourarmsandlegsoftengodeadortingle? -Heartandchestcomplaints -Doyouoftenhavepainsinthechestorheartregion? -Doyouoftenhavepalpitationsoftheheartorthrobbinginyourheart region? -Doyouhavefeelthightinthechest? -Shortofbreath -Doyougetshortofbreatheasily? 425 -123- CHAPTER7 ASSESSMENTOF INTRA-ABDOMINALANDSUBCUTANEOUSABDOMINAL FAT-RELATION ANTHROPOMETRYANDCOMPUTEDTOMOGRAPHY J.C.Seidell,A.Oosterlee,M.A.O.Thijssen,J.Burema,P.Deurenberg, J.G.A.J.Hautvast,J.H.J.Ruijs AmJClinNutr (inpress) ABSTRACT Theabilitytodistinguishbetween intra-abdominal andsubcutaneousabdominal fatmaybeimportant inepidemiologicandclinical research.Inthisstudy anthropometricmeasurementsweretakenfrom71menand 34womenpresenting for routinecomputed tomography (CT).Areasofabdominal fatwerecalculated fromCT scansmadeattheleveloftheL4vertebra.Theamountsofintra-abdominal and subcutaneousabdominal fatcouldbeaccuratelypredicted fromseveral 2 circumferences,skinfoldmeasurements,bodymass indexandage (R ranged from 0.79 to 0.84). Inaddition,itwas foundthattheareaof intra-abdominal faton theCTscanwas related tothewaist/hipcircumference ratio (r=0.75 inmen,r =0.55 inwomen)andtothewaist/thigh circumference ratio (r= 0.55 inmen,r • 0.70 inwomen). Thecorrelationsofthecircumference ratioswith theareasof subcutaneous fatwere invariablylower. INTRODUCTION Itisnowgenerally recognized thatfatdistribution,asascertained fromthe ratioofwaist circumference tohipcircumference,isan importantprognostic indicatorof theoccurrenceofmetabolicabnormalities,diabetesmellitus (type II), hypertension,cardiovasculardisease,strokeanddeath inmenandwomen (1). This relationshipmightdependontheincreasedaccumulationof intra-abdominal fat (2).Ithasbeenshownthatsubcutaneous fat isless responsivethanomentaladiposetissue tothelipolyticeffectsofepinephrine andnor-epinephrine (3)andtotheantilipolyticeffectof insulin (4).Free fattyacidsare therefore releasedata rapid rateanddraindirectly intothe portalvein.Long-termexposureofthelivertohighconcentrationsof fatty acidsmay resultinmetabolicderangements (5,6). -124- Ashwelletal(7)recentlycomparedinformationfromcomputedtomographyscans madeattheleveloftheumbilicuswithcircumferenceratiosof28adultwomen andshowedthatthesecircumferenceratioswererelatedtotheamountof intra-abdominalfatandtheratioofintra-abdominalfattosubcutaneous abdominalfat. Kissebahandhiscollegues(8)showedthatinwomen,thelatterratiocorrelates strongly(r»0.8approximately)withincreasedglucoseresponsetooralglucose challengeandtoinsulinlevels. Inourstudyoftheserelationshipsweincludedbothmenandwomen.Furthermore, wemeasuredcircumferencesatmorelevelsandincludedskinfoldmeasurements takenattheleveloftheumbilicus.Wedevelopedmultipleregressionmodelsto predicttheamountandproportionsofintra-abdominalfatandsubcutaneous abdominalfatfromthesesimpledataonanthropometryandage.Although extrapolationoftheresultingregressionequationstootherpopulationsmustbe donewithgreatcautiontheyprovideinsightintotheindependentcontribution oftheindependentvariablestothevariationinfatareasontheCTscans. PATIENTSANDMETHODS Patients Westudiedpatientswho,duringaperiodofthreemonths,werereferredtothe RadiodiagnosticInstituteoftheUniversityHospitalinNijmegenforaCTscan oftheabdomen.Theycomprisedin-patientsaswellasout-patientsand^inmost cases,theycamefordiagnosisortherapeuticevaluationofcancerorother malignancies.Weexcludedpatientsundertheageof19andallthosewhoseCT scansrevealedabnormalfeaturesthatcouldinfluencethemeasurementsoffator anthropometricmeasurements.Thus,patientswithlargelocalabnormalities, enlargedliverorspleen,andthosewithastomaatthelevelatwhichthe measurementsweretakenwereexcluded.Menopausalstatuswasassessed,andin onecasewherethiswasuncertainbecauseofillnessandtherapy,awomanwas excluded.Thefinalpopulationcomprised71malesand34females.Theprocedures followedwereinaccordwiththeethicalstandardsoftheCommitteeonHuman ExperimentationoftheInstituteofRadiodiagnostics. ComputedTomography TheCTscanningwasdonewithaSiemensSomatomDR3(SiemensUBMed, Henkestrasse127,8520ErlangenGFR).Radiographicfactorswere125KVand350 mAs.Withalateralcomputedradiographindexedtothescannertableposition, -125- theCT-scanthroughthethefourthlumbarvertebra(L4wasobtained.The resultingscanwasofalayer8mmthick .TheL4levelcorrespondstothe umbilicuslevel,atwhichwaistcircumferenceisusuallymeasured (7,9).As describedinotherstudies(10)theamountoffatattheumbilicuslevelis representativeofthetotalabdominalfatandithasbeensuggestedthatyoucan bestdistinguishthesexesonthebasisoffatattheumbilicus(11).Thedata fromthescanswereanalysedusingahistogram-basedvolumetricanalysis techniquebywhichtissueareaonaCTscanimagecanbemeasuredusinga certainrangeofHounsfieldUnits(10,12).Wedeterminedtherangefor measurementsoffattobe-150to-50Hounsfieldunits,whichisinaccordance withrangesdescribedinsimilarstudies (7,9,10). Weusedarangeof-700to3071Hounsfieldunits(therebyexcludingsmall amountsofairpresentonthescan)forthetotalareaofthecross-section.The areaofintra-abdominalfatwasdiscernedfromtheareaofsubcutaneousfatby tracingalinewithalightpencursorthroughthem.rectusabdominus,m. obliquusinternus,m.quadratuslumbarumandthelongbackmuscles.Thetotal areaoffatwithinthislinewasdefinedastheintra-abdominalfat.Theamount ofsubcutaneousfatwasascertainedasthedifferencebetweenthetotalareaof fatshownonthescanandtheareaofintra-abdominalfat.Thismethodhasbeen provedtobeveryaccurateandreproducibleashasbeenreviewedbyGraueret al(12).Theratioofintra-abdominalfattosubcutaneousabdominalfatwas calculated.Fatpercentagewascalculatedastheratiooftotalareaoffatto totalareaofthecross-section. Anthropometricmeasurements Allpatientswereaskedaseriesofquestions.Theirresponsestogetherwiththe patientfilesandtheevaluationoftheradiologists,providedinformationabout currenthealthstatusandpastillness,typeoftreatmentoruseofdrugs, menopausalstatus,long-termbedrest,weightchangesintheyearprecedingthe examination,andage. Wemeasuredtheheighttothenearest0.1cmandbodyweighttothenearest0.5 kgonacalibratedweighingscaleofthesubjectswhentheywerenotwearing shoesandwerecladonlyinunderwearorlightindoorclothing.BodyMassIndex (BMIinkg/m2)wascalculatedasweightdividedbyheightsquared. Circumferencesweremeasuredatthefollowinglevels,usingaplastictape measure:smallestcircumferenceatthewaist,waistattheumbilicuslevel, widesthipcircumference,hipsattheleveloftheanteriorsuperioriliacspine oftheiliaccrest,andthighcircumferenceatthehighestlevel. -126Most resultspresented inthispaperwereobtainedusingthe ratioofwaist circumference:hipcircumference (WHR)andthe ratioofthewaistcircumference: thighcircumference (WTR)basedonwaistattheumbilicuslevel,hipsatthe iliaccrestandthehighestthighcircumference. Twoskinfoldmeasurements (Harpender skinfold caliper,CMSWeighing Equipment Ltd,London)were takenonthehorizontal linefromtheumbilicus tothe mid-axillary line:onehorizontally,atone-thirdofthedistance fromthe umbilicus (para-umbilicalis),andtheotherobliquely,onthemid-axillary line itself (whichinmostcasescorrespondswiththesupra-iliac skinfold).All measurementsweretakenwiththesubjectstandinguprightandbreathinglightly. Statisticalanalysis The relationsbetweenanthropometricmeasurements,weightchangesinthe precedingyear,menopausal status,andage,andofvarious fatareasand proportionsof fatareasasobtained fromtheCTscanswereanalysed by calculatingproduct-moment correlationcoefficientsandbyusingmultiple linear regressionanalysis (13).Comparinguntransformedvariableswithlog transformationsand square roottransformationsofthevariables revealed that the square rootsofall theCTscanmeasurementsgaveamuchbetter approximation tothenormaldistributionandtolinear relationsinthe regressionanalysis.Therefore,inour statisticalanalysisweused thesquare rootsofthesevariables. RESULTS Table1showssomecharacteristicsofthe sample.Onaverage,inthisstudythe BodyMass Index (BMI)ofthewomenwashigher thanthatofthemenbut, nevertheless,inwomen the ratiosofwaistcircumference tohip circumference andwaist circumference tothighcircumferencewere lower.When theWHRbasedon minimumwaistgirthandmaximumhipgirthwasused,thedifference betweenmen andwomenwasevenmorepronounced (average ratioformen:0.93,forwomen: 0.87).CT-scanmeasurementsattheleveloftheL4showthat,onaverage,women hadalarger totalareaofbody fatthanmen:thiscouldlargelybeattributed todifferences inamountsofsubcutaneous fatandwasalso reflected inthe proportionsof intra-abdominal fatinmenandwomen. -127- Table1.Details of71menand34womenattendingforcomputedtomography Men Women mean SD rang s Age (years) 51.5 17.2 19 BodyMass Index (kg/m) 23.4 3.1 17.8 - mean SD range 85 52.4 13.4 19 72 31.6 26.0 5.0 19.4 - 38.3 Waist:hip ratio 0.96 0.05 0.80 - 1.10 0.94 0.08 0.79 - 1.09 Waist:thigh ratio 1.75 0.18 1.41 - 2.10 1.68 0.22 1.28 - 2.08 CTscandata; Totalareaoffat (cm) Area ofintra-abdominal fat (cm) 2 Areaofsubcutaneous fat (cm} Ratioof intra-abdominal tosubcutaneous fat 20.2 112.9 24.2 - 487.1 347.9 169.5 90.8 - 691.0 89.9 53.0 12.7 - 247.0 91.8 52.7 18.8 - 221.0 30.3 70.9 11.4 - 319.1 256.2 129.9 71.5 - 567.5 0.77 0.39 0.11- 2.42 0.38 0.19 InTables2and3thecorrelationsbetween someoftherelevantvariablesare presented.InmenandinwomentheWHRcorrelated positivelywiththetotal area of fat,areaofintra-abdominalfatand areaofsubcutaneousfat.Inwomen,but notinmen,theWHRiscorrelatedwiththeratioofintra-abdominal fatto subcutaneousfat.Inwomen,thecorrelationsbetweentheWTRandareasoffat are consistentlyhigher thanthosebetweentheWHRandareasoffat.Inmen,the reverseistrue,thecorrelationsoftheWHRwithareasoffatbeinghigher than thoseoftheWTRwiththeexceptionofthecorrelationswith theratioof intra-abdominal fatandsubcutaneousfat. Inmostepidemiologic studiestheeffectsoffatdistributionareestimated after adjustingforBMIandage.Table4showsthesignificanceofthepartial correlationsofWHRandWTRwiththeareasoffat,adjustedforageandBMI.The partial correlationswithintra-abdominalfatwere invariablyhigher thanthose with subcutaneousfat.Thestatistical associationbetweenWHRinmenandareas offatwassimilartothatofWTRinwomenandareasoffat.Thesameholdsfor theWTRinmenandtheWHRinwomen. 0.08 - 1.04 -128- Table 2.Correlationmatrix ofvariables inmen 1.Age 2.BodyMassIndex 0.15 0.24 0.76 (weight/height ) 3.Waist:hipratio 4.Waist:thigh ratio 1.00 0.68 0.14 1.00 0.47 5.Total areaoffat 0.81** 0.33* (CTscan) 6.Totalareaofintra- 0.54 abdominal fat (CTscan) ** 0.75 0.76 ** 0.75 1.00 ** ** 0.35* 1.00 0.55** 0.90 0.16 0.95** ** 1.00 7.Areaof subcutaneous 0.76** 0.14 fat (CTscan) 0.68** 0.72 8.Ratioof intra-abdominal tosubcutaneous fat 0.53** 0.02 0.13 0.56** -0.03 0.37** -0.33 1.00 9.Ratiooftotal fattototal body area 0.29 0.67 0.66 0.29 0.96 0.83 0.93 -0.10 0.70** 0.59** 0.05 0.77** 0.57** 0.83** -0.33* 1.00 10.Sum skinfolds (para-umbilical andsupra-iliac) -0.09 0.73 1.00 Table 3.Correlationmatrixofvariables inwomen 1.Age 2.BodyMass Index 1.00 0.51 1.00 3.Waist:hip ratio 0.40 0.47 1.00 4.Waist:thigh ratio 0.60 0.55 0.77 2 (weight/height} 5.Total areaoffat (CTscan) 0.54 6.Area ofintra-abdominal 0.91 0.62 0.47 0.82 1.00 0.65 0.55 1.00 0.70 0.86 1.00 fat (CTscan) 7.Area ofsubcutaneous fat (CTscan) 0.46 0.87 0.39 0.57 0.98 0.73 0.42 0.16 0.34 0.42 0.14 0.61 0.47* 0.78** 0.47* 0.63** 0.95** 0.78** 0.56 0.89 0.73 1.00) 8.Ratioof intra-abdominalto subcutaneous fat -0.07 1.00 9.Ratiooftotalfatto totalbodyarea 0.94 0.08 1.00 10.Sumof skinfolds (para-umbilical and supra-iliac) <0.05; <0.001 0.43 0.83 0.73 0.88 0.05 0.36 1.00 -129Table4.T-valuesforthepartial correlation coefficientsofwaist:hip ratioand waist:thigh ratiowithfat areasontheumbilical CTscan;adjusted forageandBMI Men Dependent variable 4.33 Area oftotalfat Area of intra-abdominal Area of subcutaneous Women Waist:hip Eat fat Ratioof intra-abdominal 4.86 2.96 to 0.64 ** ** * Waist thigh Waist hip Waist:thigh 1.79 0.53 2.73* 3.38* 1.66 2.63 0.66 0.22 1.66 2.10* 1.41 1.48 * subcutaneous fat Percentage fat p-values: <0.05 ; <0.001 InTables5and6theresultsofstepwisemultiple regression analysis arepresented,usingtheareasoffatcalculated fromtheCTscansasdependent variablesandanthropometricdataandageasindependentvariables.Most (79.4% to90.5%)ofthevarianceintotalareaoffataswellasinareasofwiththe results fromtable4,theresultsinTables5and6showthat,inmen,theWHRis a better indexofcharacteristicsoftheareasoffatthantheWTR,whereasin womentheWTRisthemoreappropriate index.Note that,inwomen,menopausal statewasmorecloselyassociatedwiththeareaofintra-abdominal fatthanage. Menopausal statehadnorelationwithtotalfatorsubcutaneous fat.Tolearn moreaboutwhytheWHRandWTRbehaveddifferentlyinthestatistical analysis we repeatedtheanalysisusing separatecircumferencemeasurements rather than ratiosofcircumferences. inmen,thehipcircumference,afteradjustmentforalltheothervariables, correlatednegativelywiththeintra-abdominalandsubcutaneousareasoffaton theCTscan,whereaswaistandthighcircumferenceshadpositive partial correlationcoefficients.Inwomen,thethighcircumferencehadanegative partialcorrelationcoefficientwithbothfatareas,whereasthehip circumference correlatedpositivelywithtotalandsubcutaneousfatand negativelywith intra-abdominal fat. Waist,inwomen,remained positively correlatedwithbothareasoffat. TT -130- Table 5.Multiple regressionmodels linkingfatmeasurementstakenfromCTscansto anthropometric variables andage Men Dependent Independent Regression variable variable coe fficients of regr.coeff. Total areaoffat (cm BMI 0.404 Skinfolds 1.480 0.243 34.122 10.284 0.060 0.012 WHR Age Intercept:- Standarderror 0.108 39 953 Area of intra-abdominal BMI 0.350 0.084 fat (cm >:I Skinfolds 0.405 0.189 WHR 33.118 8.032 Age 0.068 0.009 0.108 Intercept:- 37 322 Areaof subcutaneous BMI 0.157 fat (cm ): s Skinfolds 1.630 0.227 29.159 11.397 Age 0.051 0.018 WTR -10.694 5.060 0.029 0.009 WHR Intercept:Ratio of I:S 18 798 BMI Skinfolds - 0.109 WTR 1.583 Intercept:- 0.021 0.272 1 323 Table 6.Multiple regressionmodelslinking fatmeasurementstakenfromCTscans (atthelevelofthe umbilicus)toanthropometricvariables andage Women Dependent Independent Regression variable variable coefficientsof St* ndarderror Total areaoffat (cm BMI 0.508 0.097 Skinfolds 1.044 0.297 WTR 9.958 3.871 regr.coeff. Intercept:-15.282 Area of intra-abdominal BMI 0.367 0.061 fat (cm ): I WTR 10.911 3.660 Menopausal state 0.994 0.497 (1=post;0=pre) Intercept:-15.083 Area of subcutaneous fat (cm 2 ): S BMI 0.373 0.105 Skinfolds 1.301 0.313 Intercept:-3.327 Ratioof I:S -0.036 0.017 WTR 0.870 0.369 Age 0.0026 0.0019 Skinfolds Intercept:-0.4105 Note (table 5&6)thesquare rootsweretakenofthedependent variablesandofwaist:hip ratio (WHR), waist:thigh ratio (WTR)andsumofskinfolds (Skinfolds);variableswereselectedbymeans ofstepwisemultiple regressionanalysis (pvalueforF-to-enter lessthan0.10);other candidates for regressionwere:height,weightchange intheyearpreceding theexamination andall cross-products ofvariablesofthe independent variables All other regressioncoefficients statistically significant different fromzero (p<0.05) -131- DISCUSSION WHRandWTRareoftenusedinepidemiologicandclinicalinvestigationsas indicesoffatdistribution.TheresultsofourstudyindicatethatWHRandWTR inmenandwomenareassociatedmorecloselywithintra-abdominalfatthanwith subcutaneousabdominalfatatthelevelofthefourthlumbarvertebra. Wealsodemonstrated (Tables5and6)thatamountsofintra-abdominaland subcutaneousfatcanbepredictedaccuratelybymeansofeasy-to-measure anthropometricvariablesandage.Thesepredictionsmaybeimportantinfuture researchbecausetheyenablethespecificeffectsofthesefatdepotson morbidityandmortalitytobeestimated.Recently,Sjostrometal.(11) presentedpreliminaryresultsthatindicatethatevenhighercorrelationsof anthropometricvariableswithtotalvolumeofvisceraladiposetissuemaybe obtainedbyassessingthetotalcross-sectionalbodyareabymeansoftransverse andsagittaldiametersofthebody(inmen,correctedforsubcutaneousfat). Althoughtheirapproachwasdifferent,theaimoftheirstudywassimilarto ours.Thereforeitwillbeimportanttocomparetheresultswhenalldetailsof theirstudybecomeavailable. AscanbeseenfromTable1,menhaveagreaterproportionofintra-abdominal fatthanwomen.Thismaypartlyexplainwhy,whenequallyobese,menhavea greaterriskforcardiovasculardiseasethanwomen.Thisdifferencebetweenmen andwomenagreeswithresultsfromotherstudies(9,12)andprobablyreflects theimportanceofhormonesand/orgeneticinfluencesonfatdistributioninthe abdomen. Borkanetal.(14)suggestedthatinmen,ageisanimportantdeterminantfor theproportionofintra-abdominalfat.Thisobservationisconfirmedbyour data,aswefoundthat,inmen,agecorrelatedwiththeamountof intra-abdominalfat(r=0.54)butnotwiththeamountofsubcutaneousfat(r= 0.14).Inwomenthiswaslessclear (Table3). OurresultsconfirmthefindingsreportedbyAshwelletal.(7)whosestudyof theCTscansof28womenshowedthatcircumferenceratioswererelatedmore stronglytointra-abdominalfatthantosubcutaneousfat.Wehavenowshownthat thisisalsotrueinmen.Wewereunabletoconfirm,however,thatafter adjustmentsforageandBMI,circumferenceratiosarecorrelatednegativelywith subcutaneousfat.AlthoughourstudywassimilartothestudydonebyAshwellet al,weusedWHRbasedonwaistgirthattheleveloftheumbilicusratherthan theminimumwaistcircumference,andthehipcircumferenceatthelevelofthe anterioriliaccrestratherthanthewidesthipcircumference.Werepeatedour -132- analysisusingthesame ratiosasAswellandhercolleguesbut,inconttastto their findingswe foundevenpoorer correlationswiththe ratioof intra-abdominal fatandsubcutaneous fat,bothinmenand inwomen.Not*that WHRscalculated fromthedifferentcircumferences correlated strongly ifimen (r = 0.90)butlesswell inwomen (r= 0.77).Thisindicatesthat,especially in women, itisimportant tostandardizethesemeasurements. The resultspresented inTables5and6indicate thattheWHRandWTRdiffer betweenmenandwomenasindicesof fatdistribution.WHR inmen resemblesWTR inwomen inthestatistical analysiswhereasWTR inmen resemblesWHR inwomen. Thenegativepartial correlationcoefficients forareaofabdominal fatand the hipcircumference inmenand thethighcircumference inwomen indicate that thesecircumferencesprovideadditional informationonnon-fattissue,i.e. that atagivenwaistandthighcircumferencealargerhipcircumferenceindicates, inmen,ahigherproportionofleanbodymassintheabdomen,whereas inwomen withagivenwaistandhipscircumference,alargerthigh circumference indicatesmore leanbodymassintheabdomen. Analysisof residuals indicated thattherewerenodeviations fromlinearity in themultiple regressionanalysespresented inTables5and6.Whenthedatawere stratified into10-year agegroupsandcategoriesofBMI,theanalysis revealed similarassociations throughout the rangeofageandBMI. Itshouldbenoted that,aswedescribed inthePatientsandMethodssection, thesampleofpeoplewe studied cannotbe regardedas representative ofithe totalpopulationofthesameage range.Wecarefullyexcludedallpatielntsin whom illnesscouldhavedistorted ourmeasurements.We cannotbecertain, however,that illnessortherapydidnot influence theamountordistribution of fat intheabdomen.Alteredmenopausal statusandweight lossarethemost likely causesforsuchchanges.We repeatedtheanalysis,includingweight change intheyearprecedingtheexaminationbutwe foundno independent relationofthisvariablewithanyofthedependentvariablesweused.Thus, although thesizeofareasof fatontheCTscanmayhavebeen influencedby factors thatwedidnotmeasure,itseemslikelythatanysuchfactorwould have produced corresponding changes incircumferences,skinfoldsandweight. Summarizingour results,wecanconfirmthat,inourstudypopulation,inboth menandwomen,theWHR andWTR are relatedmore strongly totheamountof intra-abdominal fatthantotheamountofsubcutaneousabdominal fat.This findingmightbe important forexplaining theconsistent relationsbetweenWHR andWTRandmetabolicaberrations,manifest illnessanddeath. The regressionmodelswepresent inthispaperarebasedonobservationsonmen -133- andwomencoveringawide rangeofageanddegreeofobesity.Giventhefairly restrictednumberof subjects,especiallyinwomen,theextrapolationofthe regressionmodelstootherpopulationsmustbedonewithgreatcaution.The multiple regressionanalysisshows,however,that,inthestudiedgroups,the variation inthe fatareasontheCTscansareexplained toahighdegreeby the simpleanthropometricvariables,age,andmenopausal state.Inaddition,the multiple regressionmodelsprovideinformationaboutthe independent contributionsofeachoftheindependentvariablestotheexplanationof the variance ofintra-abdominal and subcutaneousabdominal fatareasontheCTscans attheL4level. ACKNCWLEDGEMENTS Theauthorswould liketothank the radiologistsandtechnical staffof the DepartmentofRadiologyoftheUniversityHospital inNijmegenfor invaluable assistance intheCTmeasurements,M.Rookus forhelpfuldiscussionsand J. Burrough forhelpwiththepreparationofthemanuscript.The financial support of theDutchPraeventiefonds isgratefullyacknowledged. REFERENCES 1.BjorntorpP.Adiposetissueinobesity (Willendorflecture). In:Recent advances inobesity research IV:163-170.HirschJ,Van ItallieTB (eds). JohnLibbey,London,Paris,1985. 2. LarssonB,Svardsudd K,WelinL,WilhelmsenL,BjorntorpP,TibblinG. Abdominaladiposetissuedistribution,obesity,and riskof cardiovascular diseaseanddeath:a13year followupofparticipants inthe studyofmen born in1913.BritMedJ1984;288:1401-4. 3.Goldrick RB,McLouglinGM.Lipolysisandlipogenesis fromglucose inhuman fatcellsofdifferentsizes.JClinInvest1970;49:1213-1223. 4. Bolinder J, KagerL,OstmanJ,Arner P.Differencesatthe receptorand postreceptor levelsbetweenhumanomentaland subcutaneousadipose tissue in theactionof insulinonlipolysis.Diabetes1983;32:117-23. 5.SmithU.Regionaldifferences inadipocytemetabolismandpossible consequences invivo.In:HirschJ,VanitallieTB (eds).Recentadvances in obesity research IV:33-36.JohnLibbey,London,Paris1985. -134- 6.BjomtorpP.Obesityandtheriskofcardiovasculardisease.AnnClinRes 1985;17:3-9. 7.AshwellM,ColeTJ,DixonAK.Obesity:newinsightintotheanthropometric classificationoffatdistributionshownbycomputedtomography.BritMedJ 1985;290:1692-1694. 8.KissebahAH,EvansDJ,PeirisA,WilsonCR.Endocrinecharacteristicsin regionalobesities:roleofsexsteroids.In:VagueJetal.(eds) Metaboliccomplicationsofhumanobesities:115-130.ElsevierScience PublishersBV,Amsterdam,1985. 9.DixonAK.Abdominalfatassessedbycomputedtomography:sexdifferences indistribution.ClinRadiol1983;14:189-191. 10.BorkanGA,GerzofSG,RobbinsAH,HultsDE,SilbertCK,SilbertJE. Assessmentofabdominalfatcontentbycomputedtomography.AmJClinNutr 1982;36:172-177. 11.SjostromL,KvistH,TylenK.Methodologicalaspectsofmeasurementsof adiposetissuedistribution.InVagueJetal.(eds).Metabolic complicationsofhumanobesities:13-19.ElsevierSciencePublishersBV, Amsterdam,1985. 12.GrauerWO,MossAA,CannCE,GoldbergHI.Quantificationofbodyfat distributionintheabdomenusingcomputedtomography.AmJClinNutr 1984;39:631-7. 13.DixonWJ.BMDPstatisticalsoftware.BerkelyCA.UniversityofCalifornia Press,1983. 14.BorkanGA,HultsDE,GerzofSG,RobbinsAH,SilbertCK.Agechangesin bodycompositionrevealedbycomputedtomography.JGerontology1983; 38:673-677. i -135- GENERALDISCUSSION This thesis reportsontheassociationsbetweenoverweightandseveralaspects ofmorbidity.Thelasttwochaptersprovide someinsight inthe roleof fat distribution intheseassociations.Inthestudiesdescribed inthisthesis overweightandobesity (orsevereoverweight)aredefined intermsoftheBody 2 Mass Index (BMI=weight/height ).Although itisknown thatinindividuals the BMIcanbeanunreliableestimateofbodyfatness,inepidemiologic research it isconsidered tobe themost satisfactorymeasureofobesity (1,2,3).Themost seriousmisclassificationofobesity statusonthebasisoftheBMIoccursin theveryold,theverymuscularandthosewhohavenot reached theendof puberty (4).The treshold levelsofBMI forthedefinitionof (moderate) overweightandobesity (orsevereoverweight)arestillarbitraryalthough itis 2 nowgenerallyagreed thatwhen theBMIexceeds30kg/m aperson isobesewith clearlyassociated increased risksforprematuredeath,cardiovasculardisease, diabetesmellitusandseveralothermoreorlessseriousmetabolicabberations. Mostoftheseassociationswereconfirmedbytheresultsofthestudies 2 presented inthisthesis.BelowaBMIof 30kg/m increasedmorbidity risksare lessclearandprobablyconfinedtosubgroups:thosewithanabdominal typeof fatdistribution,thosewhowerealreadyoverweightasyoungadults,and those whohavea familyhistoryofchronicdisordersor increasedlevelson cardiovascular riskfactors. Fromcross-sectional studies (Chapters2and 5)aswellasfrom longitudinal studies (Chapters 3and4)itisnotdirectlypossible tomake inferencesabout thecausal relationshipsbetweenoverweightandmorbidity. Forsomeofthe observed associations (e.g.thoseofBMIwithhypertension,cardiovascular disease,diabetesmellitus,arthrosis,menstrualabnormalities)there isnow muchevidence fromtheliterature thatthedisordersarethe resultof increased body fatness.An importantaspectoftheevidence,with implications forthe treatmentofoverweight is,thatwhenoverweightpersonssucceed inlosing weight,bloodpressure,glucose toleranceandlevelsofblood lipidsgenerally tend tonormalize.Overweight subjectswho suffer fromjointdisorders, menstrual abnormalities,shortnessofbreathetc.normally find reliefoftheir symptomswhentheyloseweight(5). Insomeoverweight subjects,especially in women,mental state improveswhentheysucceed inmaintaining substantialweight loss.Although thebenefitsof reducingweightareclear,itmustbenoted that slimming initselfmaycauseunwanted sideeffects.For instance,ithasbeen argued that slimmingwill increasethelithogenityofthebile,thereby -136- increasingtheriskofgallstones.Unsuccessfulslimmingattemptsmaylftadto increaseddepression.Thus,slimmingshould,ideally,besupervisedbymedical professionalists. TheassociationofBMIandpsychosomaticcomplaintsisanexampleofth$complex interrelationsbetweenthetwovariables.Itmaybethatpsychosomatic complaintsaretheresultofbeingoverweightbutsomepsychologicalstudies haveshownthatpsycho-socialproblemsmaycontributetoweightgain(6K In addition,thepossibilitythatpersonalcharacteristicscauseoverweightaswell aspsycho-somaticcomplaintscannotbeexcluded.Inthepopulationfromfour generalpractices,describedinChapters3-6itwasfoundthatespecially womenweremuchconcernedabouttheirweightandthatinthosewhoconsidered themselvesoverweight,cosmeticproblemsdominatedandwerefarmoreimportant thanconcernsabouthealth(7).Whenlookingattheregistersofgeneral physiciansonewouldtendtoconcludethatoverweightispredominantlyaproblem inwomen.InChapter4,however,itwasshownthatinmenanincreasedriskfor cardiovasculardiseasebecomesapparentatlowerlevelsofBMIthaninwomen. Thus,asGarrowhasstated (8),manypeopledonotcorrectlyappreciatethe hazardsofobesity.Somepeople,whodonotneedto,worryabouttheirweight. Others,whodoneedtoworryabouttheirweight,donot. Thefindingthatmenseemtobemorevulnerabletotheeffectsofobesityon cardiovasculardiseaseriskfactorsthanwomencan,atleastpartly,be! explainedbysexdifferencesinfatdistribution (9).Thisbringsustothe topicoffatdistributionwhichisthesubjectofChapters6and7.Ithasbeen shownintheseandotherstudies(reviewedinChapter1)thatthewaistj/hip circumferenceratioandthewaist/thighcircumferenceratioareconvenient indicatorsoffatdistribution.TheresultspresentedinChapter6confirmthat thesecircumferenceratiosare,independentofthedegreeofoverweightandage, relatedtotheprevalenceofcertaindisorders.Thefindingthat,inwomen,a highwaist/thighratiowasassociatedwithareportedonsetofoverweightat lateragethaninthosewithalowwaist/thighratiomightgiveanindicationof oneaspectoffatdistributionthathasreceivedlittleattention (Fig).Levels ofsex-hormonesthatchangewithageingareprobablythemostlikelyexplanation forthisobservation. Hormonalfactorsareprobablyofprimaryimportanceindeterminingan individual'sfatdistribution.Thetypicalfemaleadiposetissueinthe femoral-glutealregionisthoughttohaveaspecificfemalefunction:energy storageforlactationpurposes(10). Theinfluenceofmenopausalstate,the administrationofsexhormonesandthelevelsofsex-hormonesinobesewomen -137- withdifferenttypesoffatdistribution(11)demonstratetheimportanceof hormonalinfluences.Thefindingthatanabdominaltypeoffatdistributionis associatedwithcomplicatingdisordersisshownunanimouslyinnumerousstudies. Yetitremainstobeproventhattheseassociationsareindependentofhormonal factors.Probably,sexhormoneshavebothadirecteffectonglucoseandlipid metabolismaswellasanindirecteffectviatheirrelationtofatdistribution (11).Theassociationofmenopausalstatewithintra-abdominalfatbutnotwith subcutaneousabdominalfat(describedinChapter7)indicatesthatahighwaist thighratioinpost-menopausalwomenmaybemoredeletorioustohealththanin pre-menopausalwomen.This,however,requiresfurtherinvestigation. Circumferencemeasurementsshouldbeincorporatedinthesetofanthropometric measurementsinanyepidemiologicandclinicalresearch.Studiesinwhichthe independenteffectsofbothintra-abdominalandsubcutaneousabdominalamounts offattometabolicdisordersandclinicalillnesscanbeestablishedare urgentlyneeded.Suchstudieswouldbeimportantforfurthercharacterizationof subgroupsofoverweightsubjectsthathaveincreasedhealthrisks.Theeffects oftreatmentinthesesubgroupsshouldbestudiedcarefullyinwellcontrolled interventionstudieswithlongfollow-up.Detailedclassification,onthebasis ofthefatdistributionoverandwithinthebodyofoverweightsubjectswillbe importantforthediagnosisandtreatmentofoverweightandobesity. 9 • l.b.o. u.b.o. FIGURE-Percentageofwomenwhoindicatedtohavebeen 'toofat'at differentstagesoflife. 1=atbirth;2=childhood «12yrs);3=adolescence (12-20yrs);4=young adult (20-30yrs);5=after30yearsofage. l.b.o.=lowertertileofwaist-thighcircumferenceratio u.b.o.=uppertertileofwaist-thighcircumferenceratio (Chapter6) -138- REFERENCES 1.Gezondheidsraad.Adviesinzakeadipositas.StaatsuitgeverijDenHaag,1984. 2.RoyalCollegeofPhysicians.Obesity.JRoyCollPhys1983;17:1-65. 3.BurtonBT,FosterWR,HirschJ,vanItaalieTB.Healthimplicationsiof obesity:AnNIHconsensusdevelopmentconference.IntJObesity 1985;9:155-159. 4.GarrowJS.Treatobesityseriously.Aclinicalmanual.Edinburg:Churchill Livingstone,1981. 5.BlackburnGL,ReadJL.Benefitsofreducingrevisited.PostgradMedJ 1984;60(suppl.3):13-18. 6.VanStrienT.Eatingbehaviour,personalitytraitsandbodymass(thesis), Wageningen,TheNetherlands,1986. 7.SeidellJC,BakxJC,vandenHoogenHJM,DeurenbergP.Overgewichtin relatietotmorbiditeltensubjectiefwelbevindeninvierhuisartsen praktijken.In:Deeerstelijnonderzocht(BoermaWGW,HingstmanLreds): 225-36.Deventer,VanLoghumSlaterus,1985. 8.GarrowJS.Whyarewefat?NutrFdSci1986;22:21-2. 9.Bjorntorpp.Obesityandtheriskofcardiovasculardisease.AnnClinRes 1985;17:3-9. 10.Rebuffe-ScriveM.Regionaldifferencesinadiposetissuemetabolismin relationtosexsteroidhormones(Thesis),Goteborg,Sweden,1986. 11.KissebahAH,EvansDJ,PeirisA,WilsonCR.Endocrinecharacteristicsin regionalobesities:roleofsexsteroids.In:Metaboliccomplicationsof humanobesities(VagueJetal.eds):115-30.Amsterdam,ElsevierScience Publishers,1985. -139SAMENVATTING Inditproefschriftwordenonderzoekenbeschreven naardeverbanden tussen overgewicht envetverdeling enerzijdsenaspectenvangezondheid anderzijds. De idealeproefopzet omdezeverbanden teonderzoeken ismetbehulpvan prospectieve cohort studieswaarin groteaantallen individuen, liefstvan enkele generaties,vanafhungeboorte tothundood zoudenmoetenworden gevolgd. Gedurende deobservatieperiode zouvandeonderzoekspopulatie zowelde gewichtsontwikkeling, devetverdeling alswelhetoptredenvan ziektenenhet subjectiefwelbevindenmoetenwordenbijgehouden. Inzo'n ideale onderzoeksopzet zoudenook sociale kenmerken,familiairekenmerkenenleefgewoonten moeten wordenbetrokken endeniveausvan risicofactoren voor chronische aandoeningen herhaaldelijk moetenworden bepaald. De inditproefschrift beschreven onderzoeken zijn,inhetlichtvande hierbovengeschetste onderzoeksopzet,noodzakelijkerwijszeerbeperkt en hierdoor kunnen slechtsenkeleaspectenvandeverbanden tussenovergewicht en vetverdeling endeobjectieve ensubjectieve gezondheidstoestand wordenbelicht. Er zijngoede redenenomdezeverbanden teonderzoeken.Nietalleenkomt overgewicht veelvoor inlandenzoalsNederland meteengeindustrialiseerde samenlevingmaarer zijn tevensveelaanwijzingen datdeprevalentie van overgewicht indie landen toeneemt.Bovendienmaakteengrootdeelvande Nederlandse volwassenen zich,omuiteenlopende redenen,zorgenover het lichaamsgewicht.Opdevraag ofdeze zorgen,vanuit gezondheidskundig oogpunt, terecht zijnkanditproefschrift geendirectantwoord geven.De beschreven onderzoeken geven,opzijnbest,een schattingvanhet relatieve risicodat personenmeteenbepaaldematevanovergewicht ofeenbepaalde vetverdeling hebben tenopzichte vanpersonen zonderovergewicht ofmeteen 'gunstige' vetverdeling.Het relatieve risicogeefteenindrukvandeverhoudingvan kansen ophethebbenofkrijgenvanbepaalde aandoeningen indegroependie worden vergeleken. Of eenindividu zich zorgen zoumoetenmakenover eenverhoogde kansophet krijgenvaneenaandoening iseenzuiverpersoonlijke aangelegenheid. Of een samenleving zichdaarover zorgenmaakt iseenpolitieke enethische kwestie. Inhoofdstuk 1wordteenoverzicht gegevenvandehuidige inzichten inde verbanden tussenovergewicht,vetverdeling enhetoptredenvanstoornissenin hetmetabolismeen,eventueel daaruitvoortkomende,aandoeningen. -140- Inhoofdstuk 2wordteenindruk gegevenvandeprevalentievanmatig (QI25- 30 2 2 kg/m )enernstig (QI>30kg/m )overgewichtbijNederlandsevolwassenen, gebaseerdopdegewichtenenlengteszoalsongeveer 19000Nederlanders,ouder dan20jaar,zeopgavenaanenquetricesvanhetC.B.S.Bepaalde aandoeningen werdenvakergenoemddoorpersonenmetovergewichtdandoorpersonen zonder overgewicht (o.m.hypertensie,suikerziekte,spataderen,asthma/brochitigen aambeienbijvrouwen,hypertensiebijmannen).Ookblekenpersonenmet ' overgewicht,omsommige redenen,vakerbijdehuisartsofeenspecialistte komenenwerdenbepaaldemedicijnenvakergebruikt. Dezeverschillenwarenvooralduidelijkwanneerpersonenmeternstigovergewicht werdenvergelekenmetpersonenzonderovergewicht;bijpersonenmetmatig overgewichtwarendeverschillenveelminder grootenmaar inenkelegevallen statistischsignificant. Hoofdstukken3totenmet6hebbenbetrekkingopeen retrospectief cohort onderzoek datwerdverrichtmetbehulpvandegegevensverzameld indeContinue MorbiditeitsRegistratievanhetNijmeegsUniversitairHuisartsen Instituut. AllepersonenbijwieovergewichtdoordehuisartsenwerdgeregistreerdVan20 tot 50-jarige leeftijd indeperiode 1967- 1978endienogbereikbaarwaren in 1983,werdenvergelekenmeteengroeppersonenbijwienooitovergewichtwas vastgesteld.Degroepenwaren,watbetreftleeftijd,geslachten huisartsenpraktijk, zogoedmogelijk aanelkaargelijk gemaakt. Allepersonenwaren6jaartevolgenindeContinueMorbiditeitsRegistratie. Hetbleekdat,gedurendedie6jaar,zoweldemannenalsvrouwenmetovergewicht meerverkoudheid/griep,nerveus-functionele klachten,huidproblemen,myalgieen lumbagoendistorsiesvoorkwamendanindegroepzonderovergewicht.Bijvrouwen wasdatbovendienooknoghetgevalvoormenstruatiestoornissen,enbijmannen voor kleineongevallen (Hoofdstuk 3 ) .De follow-upwerduitgebreid totniaximaal 17jaren (Hoofdstuk 4)ende incidentiesvanverschillende chronische aandoeningenwerdenbestudeerdenvergelekenmetbehulpvan 'survival-arialyse' technieken.De incidentievandiabetesmellitus, jicht,myocard infarct/angina pectorisenartrosewashoger indegroepmetovergewicht (zowelbijmannenals bijvrouwen).Bijdevrouwenwasdithetgevalvoor spataderen.De incidentie vanhypertensiewasookverhoogd indegroepmetovergewicht (nietbeschrevenin ditproefschrift). In1983werddepersonenuitdegevolgde cohorteneenvragenlijstgestuurd.Deze werd ingevuld teruggestuurd door71%. Enkeleanalysesvandezeenquete zijnvermeld inhoofdstuk 5.Veel andere o gegevenszijneldersgepubliceerd .Indevragenlijstwasondermeereenlijst -141- opgenomenvan 51klachten (dezgn.VOEG-lijstvanDirken)entevenswerd gevraagd naarhethuidige gewicht,delengte,rookgewoonten ende frequentie van vermageren.De somscorevande 51klachtenbleekgecorreleerd aandeQI bij vrouwenmaarnietbijmannen.Inmultipelelineaire regressiebleek echter dat leeftijd een interactievertoondemetdeQI innunrelatiemetdeklachtenscore. Bij jongemannenwasereennegatiefverband tussendeQIende klachtenscore, bijouderemanneneenpositiefverband.Bijvrouwenbleek deassociatie tussen deQI endeklachtenscore veelduidelijker bijjongevrouwendanbijoudere vrouwen. DeQIbleek, inmultipele logistische regressie analyse,ook gerelateerd aan afzonderlijke klachten ofgroepjesvanklachten.Bijmannenenvrouwenwaren deze klachten:klachtenaanhetmaagdarmkanaal,spier-engewrichtsklachten en kortademigheidenbijvrouwen:nervositeit,moeheid,neus-enkeelklachten, pijn indehart-enborststreek,benauwdheid opdeborstenhetvaak hebbenvan kleineongevalletjes. InHoofdstuk 6iseenonderzoek beschrevenwaarbijbij 305personenuitde groep metovergewicht devetverdelingwerdbepaald aandehandvandemiddel/heup omtrekverhouding (middel/heupratio)endemiddel/dijomtrek verhouding (middel/dij ratio).Demiddel/dij ratiobleek,ookna correctievoor leeftijd en QI, positief geassocieerd metdeprevalentie vanhypertensie bijmannen en vrouwen, jichtofdiabetesbijvrouwen,arthrose bijmannenen negatief geassocieerd metdeprevalentie van spataderenbijvrouwen.Deassociaties van demiddel/heupratiometdezeaandoeningen blekenminder sterk (met uitzondering vanhypertensie bijmannen).Deklachtenscore (ziehoofdstuk 5)bleek bijmannen positief gecorreleerd metdemiddel/dij ratio.Groepenvanklachten die geassocieerd warenmetdemiddel/dij ratiowaren:pijn indehart-en borststreek, gewrichtsklachten enkortademigheid.Vrouwenmeteenhoge middel/dij ratiogavenaanpasopoudere leeftijd 'tedik'tezijngeworden dan devrouwenmeteenlagemiddel/dij ratio.Ditwerdbevestigd doornun gerapporteerde gewichtsontwikkeling. Hoofdstuk 7beschrijft eenonderzoek waarbijmetbehulpvan 'Computed Tomography'de relatiewerd onderzocht tussendeomtrekverhoudingenende hoeveelheid endeverhouding van intra-abdominaalveten subcutaan abdominaal vet terhoogtevandeL4 (meestalopnavelhoogte).Hetbleek datde correlaties tussendeomtrekverhoudingen endehoeveelheid intra-abdominaal vet significant waren,ookna correctie voorde leeftijd endeQI.De correlaties tussende omtrekverhoudingen endehoeveelheid subcutaanvetwaren steeds lager.Er bleken opvallende verschillen tebestaan tussenmannenenvrouwen.Bijmannenwas -142- relatiefeenveelgroterdeelvanhetabdominalevet indebuikholteop^eslagen. Ditaandeelnamtoemetdeleeftijden,bijvrouwen,nademenopauze. Devoornaamsteconclusiesvandeonderzoeken zijndaternstigovergewicht,enin minderemateookmatigovergewicht,samengaanmetmeermorbiditeitenm£t verminderdsubjectiefwelbevinden.Vanuitgezondheidskundig oogpunt zouj rten bij deevaluatievanpersonenmetzowelmatigalsernstigovergewichtrekenirig moetenhoudenmetdevetverdeling. REFERENTIES 1.BakxJC,SeidellJC,DeurenbergP,vandenHoogenHJM.Thedevelopmentof hypertensionofoverweightpatients seeningeneralpractice.J Family Practice 1987 (in press). 2. SeidellJC,BakxJC,vandenHoogenHJM,DeurenbergP.Overgewicht ihrelatie totmorbiditeitensubjectiefwelbevinden invierhuisartsenpraktijken.In: BoermaWG*J,HingstmanL (red.). Deeerstelijnonderzocht.Deventer:Van LoghumSlaterus,1985:225-236. -143- SUHHARY Thisthesis reportsontheassociationbetweenoverweightand fatdistribution withaspectsofmorbidity inDutchadults. The idealstudydesign for investigating theseassociationswouldhavebeen to carryoutalargeprospectivecohortstudyinwhichalargenumberof individuals,preferably fromseveralgenerations,couldbe followedup from birth todeath.Throughout theobservationperiod thedevelopmentofweightand fatdistributionwouldhave tobe registeredaswellastheoccurrenceofall diseasesandlevelandchangeofwell-being.Suchanidealstudywould involve measuresoffamiliarity,elementsoflife-styleandthelevelsofknownrisk factorsfordisease. Itshouldbenoted thatthestudiesdescribed inthisthesisare,compared to thedescribed studydesign,severelylimitedanditmayverywellbethat associationsobserved inthestudiescanbe,atleastinpart,explainedby variablesthathavenotbeenmeasured inthestudies. Thereareseveralgood reasonsforstudyingtheassociationsbetweenoverweight andaspectsofhealth.Notonlyistheprevalenceofoverweight inindustrial societies,likeinTheNetherlands,considered tobehighbut,moreover,there aremany indications thattheprevalenceofoverweighthasbeen increasingover thepastdecades.Inaddition,alargeproportionofDutchadultsare,because ofvarious reasons,concernedabouttheirweight.Whetherornotthisconcernis justifiedonmedical groundsisnotthedirectsubjectofthisthesis.The investigationspresentedherecan,atbest,giveanestimationofthe relative risksforcertaindisorderscomparingseveraldegreesofoverweight.Whether or notan individualworriesaboutacertainincrease inhealth risksisapersonal matter.Societiesconcernaboutoverweightand itspossibleconsequences isa politicalandethicalmatter.Itisnotthetaskofepidemiologists totell individualsandgouvernmentshowtofeelandact. Chapter 1ofthisthesis isaliterature review,givinganoverviewof current insights intheassociationsbetweenoverweight,fatdistributionandthe occurrenceofmetabolicdisturbancesand,possibly resulting,complicating disorders. Chapter 2givestheprevalenceofmoderateoverweight (BMI25- 30kg/m )and 2 severeoverweight (BMI>30kg/m )in19,000Dutchadults (self-reported data). Itcouldbeshownthat,afteradjustmentsforageandeducational level,certain chronicdisorderswere reportedmoreoften inoverweightsubjectsthanin subjectswithoutoverweight (i.e.hypertension,diabetesmellitus,varicose 2 -144- veins,asthma/bronchitisandhemorroidsinwomen,hypertensioninmen). In addition,itappeared thatmoreoftheoverweightsubjectsconsulted their generalpractitioner oramedical specialist forparticular reasonsandjahigher use ofseveralmedicinsthanthosewhowerenotoverweight.These differences weremorepronouncedwhencomparing severelyoverweight subjectsto non-overwight subjectsand lessclearorabsentformoderately overweight subjects. i Chapters 3 - 6 consider retrospective cohortstudiesthatwereperformedondata collected inacontinuousmorbidity registration infourgeneral practices (source:DepartmentofGeneral Practice,UniversityofNijmegen).All subjects whowere registered asoverweightattheageof20- 50years intheperiod 1967 - 1978were followed inthe registration.Themorbiditydatawerecompared to those inacontrol groupfrequencymatchedonage,sex,generalpracticeand durationof follow-up.All subjectswere selectedonthebasisthattheycould be followed foratleast6years.During these sixyears (Chapter 3)itwas shownthatmoreoverweightmenandwomenwere seenbytheirGP'sforcommoncold & influenza,psychosocial problems,skinafflictions,myalgyandlumbago, distortions,menstrualabnormalities (womenonly),andsmallinjuries (menonly) thanmenandwomen inthenon-overweightgroup.The follow-upperiodwas extended tomaximal 17years (Chapter 4)and theincidencesof severalChronic diseaseswere comparedbetweentheoverweightandthenon-overweightgroupwith theuseof 'survivalanalysis'techniques.The incidenceofdiabetesmellitus, gout,arteriosclerotic disease,artrosis,varicoseveins (womenonly),ftnd hypertension (presented elsewhere )was increased intheoverweightgroup.In Chapter 5some resultsarepresentedofthe relationbetweenBMIand complaints listed inaquestionnaire thatwasadministered tothestudypopulation (spring 1984). The response ratewas71%.Manyother resultswere reportedelsewhere . Thequestionnaire included,besidesthelistof 51subjectivehealthcomplaints, questionsaboutcurrentweightandheight,smokinghabitsand slimming, frequency. Inmultiple linear regressionanalysisthesumofcomplaintswas correlatedwiththeBMIinwomen (especiallyyoungwomen),positively inolder menandnegatively inyoungermen.Inaddition,itcouldbe shownthat,in multiple logistic regressionanalysis,thepresence ofparticular complaintsor groupsofcomplaintswas related totheBMI (inmenandwomen:digestive tract complaints,muscleand jointcomplaints,shortnessofbreath;inwomenonly: nervousness,tiredness,nose-andthroatcomplaints,painintheheart-or chest region,tightnessatthechestandoftenhavinglittlemishaps). InChapter 6,a studyisdescribed relatingthefatdistribution (in310 2 -145- subjectsoftheoverweightstudygroup)toregisteredmorbidity.Thewaist/hip circumferenceratio(WHR)andwaist/thighcircumferenceratio(WTR)weretaken asmeasuresoffatdistribution.TheWTRwas,afteradjustmentsforageandBMI, positivelyrelatedtotheprevalenceofhypertension(inmenandwomen),goutor diabetes(inwomen),artrosis(inmen),andnegativelyrelatedtotheprevalence ofvaricoseveinsinwomen.TheassociationsoftheWHRwiththesedisorders wereequalorlessstrong(withtheexceptionofhypertensioninmen).Thetotal numberofsubjectivehealthcomplaints(seealsoChapter5)waspositively relatedwiththeWTRinmen.WomenwithahighWTRreportedtohavebeen'too fat'atolderagesthanwomenwithalowWTR.Thiswasconfirmedbytheirweight history.InChapter7,therelationofcircumferenceratiostofatareason ComputedTomographyscansattheL4levelwasstudied.Itwasfoundthatthe correlationofthecircumferenceratioswerehigherwiththeamountof intra-abdominalfatthanwithsubcutaneousabdominalfat.Therewerepronounced differencesintheabdominalfatdistributionbetweenmenandwomen.Inmen,a relativelylargerproportionoffatwasstoredintheabdominalcavity.This proportionincreasedwithageand,inwomen,aftermenopause. Itisconcludedthat,fromamedicalpointofview,itisnecessarytoinclude measuresoffatdistributionintheevaluationofoverweightandobesesubjects. REFERENCES 1.BakxJC,SeidellJC,DeurenbergP,vandenHoogenHJM.Thedevelopmentof hypertensionofoverweightpatientsseeningeneralpractice.JFamily Practice1987(inpress) 2.SeidellJC,BakxJC,vandenHoogenHJM,DeurenbergP.Overgewichtinrelatie totmorbiditeitensubjectiefwelbevindeninvierhuisartsenpraktijken.In: BoermaWGW,HingstmanL (red.).Deeerstelijnonderzocht:225-236.Deventer: VanLoghumSlaterus,1985. -146- CURRICULUMVITAE JacobCaesar Seidellwerdop 3november 1957geborenteWeert.HetAthejieum-B diplomawerdbehaald in1976aanhetEckartCollege teEindhoven.Inhetzelfde jaarwerd zijnstudieaandeLandbouwhogeschoolteWageningenaangevang*n. Tijdens zijnpraktijktijd deedhijonderzoek ophet 'DunnClinicalNutrition Centre' teCambridge,Engeland. Injuli1983slaagdehijvoorhetdoctoraial examen (metlof)metalshoofdvakkenHumaneVoedingenToxicologie.Per1 September 1983tradhijindienstbijdeVakgroepHumaneVoeding teWageningen alswetenschappelijkassistentalwaar,met financiele steunvanhet Praeventiefonds,hetinditproefschriftbeschrevenonderzoekwerdverricht.Per 1 januari 1987zalhij,metbehulpvaneenstipendiumvanZWO,gedurendeeen jaarverder onderzoek naaraspectenvanvetverdelingverrichtenaande UniversiteitvanGoteborg teZweden.