Medicares Search for Effective Obesity Treatments Diets Are Not the Answer Traci - PDF document

Medicare’sSearchforEffectiveObesityTreatments DietsAreNottheAnswer TraciMann,A.JanetTomiyama,ErikaWestling,Ann-MarieLew,BarbraSamuels,andJasonChatmanUniversityofCalifornia,LosAngeles Theprevalenceofobesityanditsassociatedhealthprob-lemshaveincreasedsharplyinthepast2decades.NewrevisionstoMedicarepolicywillallowfundingforobesitytreatmentsofprovenefÞcacy.Theauthorsreviewstudiesofthelong-termoutcomesofcalorie-restrictingdietstoas-sesswhetherdietingisaneffectivetreatmentforobesity. critical.Third,theoverallqualityofevidenceacrossthesecriticaloutcomesisassessed.Fourth,thebalanceofbene-tsandharmsisconsidered.InassessingwhetherMedicareshouldfundcalorie-restrictingdietsforthetreatmentofobesity,weshallat-tempttoanswerthequestionsposedbytheGRADEsys-tem.Webeginbyevaluatingthequalityofevidenceforthemostcommonobesitytreatment—thesevererestrictionofcalorieintake(whichwewillrefertoasingtooneprimaryoutcomemeasure:sustainedweightloss.Wefocusonlong-termweightlossbecauseshort-termweightlossisnotacureforobesity.Population-levelchangesinobesity(andpresumablyitsconcomitanthealthproblems)willonlyoccuriflossesaremaintained.Directindicatorsofhealth,suchasbloodpressure,cholesterollevels,diseaseincidence,andevenmortality,arealsoim-portantoutcomesofobesitytreatments,andwediscusstheseoutcomesinthefewlong-termstudiesthatincludethem.Finally,webrieydiscussthebalanceofbenetsandharmsassociatedwithusingdietsasatreatmentforobe-AccordingtotheGRADEsystem,therstfactortoconsiderinevaluatingthequalityofscienticevidenceisthestudydesign.Thesystemassignsahigherstartinggradetorandomizedtrialsandalowerstartinggradetoobserva-tionalstudies(GRADEWorkingGroup,2004).Thestart-inggradeisthenadjustedonthebasisofthequalityofthestudymethodsandexecution.Werstdiscusslong-termrandomizedtrialsofdiets(higherstartinggrade)andthentwodifferenttypesofobservationalstudiesofdiets(lowerstartinggrades).Ineachsection,weassessthequalityofthestudymethodsandexecutionaswellastheconsistencyanddirectnessofthendings.Long-TermEffectsofDietingReviewsofthescienticliteratureondieting(e.g.,Garner&Wooley,1991;Jefferyetal.,2000;Perri&Fuller,1995)generallydrawtwoconclusionsaboutdiets.First,dietsdoleadtoshort-termweightloss.Onesummaryofdietstudiesfromthe1970stothemid-1990sfoundthattheseweightlossprogramsconsistentlyresultedinparticipantslosinganaverageof5%–10%oftheirweight(Perri&Fuller,1995).Second,theselossesarenotmaintained.Asnotedinonereview,“Itisonlytherateofweightregain,notthefactofweightregain,thatappearsopentodebate”(Garner&Wooley,1991,p.740).Themoretimethatelapsesbetweentheendofadietandthefollow-up,themoreweightisregained.Forexam-ple,inastudyinwhichobesepatientswerestarvedinthehospitalforanaverageof38days,patientswerefollowedforvaryinglengthsoftimeafterthestarvationperiod.Amongpatientswhowerefollowedforundertwoyears,23%gainedbackmoreweightthantheyhadlost.Amongpatientswhowerefollowedfortwoormoreyears,83%gainedbackmoreweightthantheylost(Swanson&Dinello,1970).Eveninthestudieswiththelongestfol-low-uptimes(offourorveyearspostdiet),theweightregaintrajectoriesdidnottypicallyappeartoleveloff(e.g.,Hensrud,Weinsier,Darnell,&Hunter,1994;Kramer,Jef-fery,Forster,&Snell,1989),suggestingthatifparticipantswerefollowedforevenlonger,theirweightwouldcontinuetoincrease.Itisimportantforpolicymakerstorememberthatweightregaindoesnotnecessarilyendwhenresearch-ersstopfollowingstudyparticipants.Long-TermRandomizedStudiesThemostrigorousdesignsinstudiesoflong-termweightlossmaintenancearethosethatrandomlyassignindividu-alstoadietconditionortoano-dietconditionandthenfollowthemovertime.Suchstudiesallowcausalconclu-sionstobedrawnabouttheeffectsofthedietonweight,andtheyareparticularlyusefulbecauseindividualswhodonotgoondietsareoftenfoundtoslowlygainweightovertime(e.g.,Burkeetal.,1996;Shah,Hannan,&Jeffery,1991).Veryfewsuchstudiesincludelong-termfollow-upsthatallowforclearcomparisonsbetweentheweightofdietersandtheweightofcontrolparticipants,presumablybecauseitisdifculttorequireobeseindividualswithaninterestindietingtoremaininwait-listcontrolgroupsforextendedperiodsoftime.Byconductingsearchesofonlinedatabases,inspect-ingstudieslistedinthereferencesectionsofseverallargereviewsofdietoutcomestudies(Astrup&Rossner,2000;Black,Gleser,&Kooyers,1990;Foreyt,Goodrick,&Gotto,1981;Jefferyetal.,2000;Leon,1976;Perri,1998;Perri&Fuller,1995;Saris,2001),andexaminingthereferencesectionsofthedietoutcomestudiesthemselves, Thetermhasbeenusedtorefertoawiderangeofbehav-iors,butweuseitsolelytorefertothespecicbehaviorofseverelyrestrictingone’scalorieintakeinordertoloseweight. TraciMannApril2007AmericanPsychologist wewereonlyabletolocatesevenstudiesthatrandomlyassignedparticipantstoadietorawait-listcontrolgroup,followthemforatleasttwoyears,andthenreportonweightoutcomes.Thedietsinseveralofthesestudiesarecombinedwithotherlifestyleinterventions,includingex-ercise,butweincludethesestudiesbecausetheyarethemostrigoroustypeofdietstudiesandbecausetheyhavecontrolgroupsthatallowforusefulcomparisons.Infact,onlyoneofthesesevenstudiesisaformaldietstudywithoutadditionalinterventions.Oneisanobesitypreven-tionprogram,andveareclinicaltrialsoflarge-scaleinterventionsdesignedtoreduceriskforheartdiseaseordiabetesortopreventorcontrolhypertension.Theformaldietstudyfollowedparticipantsfor2.5years(Jeffery&Wing,1995).Participantswererandomlyassignedtooneoffourtypesofdietsortoawait-listcontrolgroup.Attheendofthe2.5-yearfollow-up,par-ticipantsassignedtothewait-listcontrolgroupdidnotshowastatisticallysignicantweightgain,andtheirweightchangewasonlymarginallydifferentthanthatofthedieters.Thedietershadkeptoffanaverageofonly1.7kg(3.7lb).Thismethodologicallyrigorousstudyshowsthatinthelongterm,dietersdonotfarenotablybetterthanAnotherstudycomparedindividualsassignedtoeitheroftwoobesitypreventionprogramswiththoseassignedtoawait-listcontrolgroup(Jeffery&French,1999).Thefocusoftheobesitypreventionprogramswastoencourageparticipantstopayattentiontotheirweightandtomakesmallchangestotheirdietandexercisehabits.Afterthreeyears,therewerenosignicantdifferencesintheweightchangebetweenpreventionparticipantsandwait-listpar-ticipants.Participantsgainedanaverageof1.6kg(3.5lb),andtheresearchersconcludedthattheobesitypreventionprogramhadnotsucceeded.Fivedifferentlarge-scaleclinicaltrialsrandomlyas-signedindividualstointerventionswithmanycompo-nents,inwhichonecomponentwasafocusonlosingweightthroughdieting.Theseinterventionswereaimedatlargegroupsofparticipants,andtheytendedtolastseveralyears,sotherewerenotalwaysrigorouscontrolsonwhetherparticularindividualsactuallyparticipatedintheintervention(andwhethercontrolparticipantstrulydidnotdiet).Still,thesestudiesprovideinformationonthelong-termweighttrajectoriesofdieterscomparedwithnon-Thestudywiththemostpromisingresultsintermsofhealthoutcomesrandomlyassignedoverweightorobeseindividualsatriskfordiabetestoalifestyleinterventioninvolvingdietandexerciseortooneoftwocontrolgroups(DiabetesPreventionProgramResearchGroup,2002).Atathree-yearfollow-up,participantsinthelifestyleinterven-tionlostanaverageofabout4kg(8.8lb),whilepartici-pantsintheplacebocontrolgrouphadgainedabout0.5kg(1.1lb).Inthissametimeframe,however,thelifestyleinterventionreducedtheincidenceofdiabetesby58%comparedwiththeplacebocontrolgroup.Theseresultsmaynotdirectlybeduetothedietpartoftheintervention,andinfactparticipantsinthelifestyleinterventionengagedinlargeamountsofphysicalactivity(averaging227min-utesperweek),andthismaybethepotentfactor.Aswediscussindetailbelow,exerciseseemstobeanimportantfactorinweightandhealthoutcomes.Inaddition,thendingsmaynotapplytoallobesepeople,butratherjusttothespeciallychosensubsetofobesepeopleincludedinthestudy—obesepeople(andoverweightpeople)withelevatedplasmaglucoseconcentrations,ariskfactorforThelargestweightlossoccurred,notsurprisingly,inthestudywiththeshortestfollow-uptime.Inthatstudyofanonpharmacologicinterventionforolderindividualswithhypertension,participantsinthedietconditionsmaintainedanaverageweightlossof4.7kg(10.4lb)atthe2.5-yearfollow-up,whereascontrolparticipantsmaintainedalossof0.9kg(1.9lb;Wheltonetal.,1998).Thisstudyalsoincludedhealthoutcomesandfoundnosignicantdiffer-encebetweenparticipantsinthedietconditioncomparedwiththecontrolconditioninsystolicordiastolicbloodpressureorinthenumberofcardiovascularevents(e.g.,strokes,myocardialinfarction)experienced.However,par-ticipantsinthedietconditiondidappeartohaveareducedneedforantihypertensivedrugs. Severalstudiesthatareoftencitedassupportforthelong-termsuccessofdietsfollowedindividualsforlessthanoneyear(Foreyt&Kennedy,1971),sixmonths(e.g.,Klem,Wing,Simkin-Silverman,&Kuller,1997),orevenlesstime(Dahlkoetter,Callahan,&Linton,1979;Presnell&Stice,2003).Studiesthatrandomlyassignedentirecommunitiestointerventionsandcontrolgroupsarenotincludedherebecauseofthelowpowerinherentinsuchdesigns(Fortmann,Williams,Hulley,Haskell,&Far-quhar,1981;Jefferyetal.,1995;Tayloretal.,1991).Thesenumberswerenotreportedinthearticlebutwereestimatedfromagraphdepictedinit. A.JanetApril2007AmericanPsychologist Astudydesignedtopreventhypertension(Hyperten-sionPreventionTrialResearchGroup,1990)andastudydesignedtocontrolhypertension(Stamleretal.,1987)followedparticipantsforthreeandfouryears,respectively,andfoundsimilarresultsintermsofweightlossbutdif-feredonthehealthoutcomes.Inbothstudies,participantsinthediet-typeinterventionsmaintainedanaverageweightlossof1.8kg(3.9lb),whereascorrespondingcontrolparticipantsgainedanaverageof1.8kg.Inthehyperten-sionpreventionstudy(HypertensionPreventionTrialRe-searchGroup,1990),participantsonthedietdidshowsignicantlygreaterreductionsindiastolicandsystolicbloodpressureoverthethreeyearsofthestudythandidcontrolparticipants,thoughthesedifferenceswereclini-callysmall(1.8millimetersofmercuryand2.4millimetersofmercury,respectively).Dietandcontrolparticipantsdidnotdifferinwhethertheyneededdrugtreatmentforhy-pertension,nordidtheydifferinperiodsofhospitalizationordeaths.Inthehypertensioncontrolstudy(Stamleretal.,1987),hypertensiveparticipantswereassignedtoacontrolgrouportoaninterventionthatcombinedadiet,sodiumrestriction,andalcoholreduction.Theywerethenremovedfromtheirhypertensionmedication.Themostimportantoutcomeofthestudywaswhetherparticipantswereabletokeeptheirbloodpressurelowenoughtoremainofftheirhypertensionmedication.Onthisoutcome,39%ofpartic-ipantsinthedietinterventionsucceededafterfouryears,whereasonly5%ofcontrolparticipantssucceeded.Al-thoughanimportantandclinicallysignicantnding,itlikelyfailsthetestofdirectnessrecommendedbytheGRADEsystem.First,theinterventionwasnotsolelyadietbutalsorequiredsodiumandalcoholreduction,andthebenecialeffectsofitmaybeduetothosecomponents.Second,theinterventionisaneffectivetreatmentforhy-pertension,notobesity.Assuch,Medicaremightfunditforindividualswithhypertension,regardlessofwhethertheyareobese,andthenewpolicyontreatingobesityperseisnotnecessaryforthistooccur.Participantsintheremainingstudywereindividualsatriskforcardiovasculardisease,accordingtotheirdiastolicbloodpressure,cholesterollevel,andcigaretteuse.Abouttwothirdsoftheparticipantswerehypertensive.Partici-pantswereassignedtoaninterventiondesignedtolowercardiovascularriskfactorsortoausual-carecontrolgroupandthenwerefollowedforsixyears(Grimm,Cohen,Smith,Falvo-Gerard,&Neaton,1985).Theinterventionincludedalow-cholesteroldietandinstructioninsmokingcessation,amongothercomponents.Atthe6-yearfollow-up,theaverageweightlossmaintainedforindividualsintheinterventionconditionwasabout0.5kg(1.1lb),whereascontrolparticipantsgainedanaverageof0.5kg.Bloodpressurelevelsdecreasedforparticipantsinbothconditions,butthedecreasesweresignicantlylargerforparticipantsintheinterventionthanforparticipantsinthecontrolgroup.However,thisdifferenceislikelyattribut-abletothefactthatsignicantlymoreinterventionpartic-ipantsweretakingantihypertensivemedicationthancon-trolparticipantsthroughoutthestudy.Mortalityratesdidnotdifferbetweenparticipantsinthetwogroupsatthe6-yearfollow-up,butmortalityrateswerelowerforinter-ventionparticipantsata10.5-yearfollow-up(MultipleRiskFactorInterventionTrialResearchGroup,1990).Thesemortalitychangesmayhavebeenduetosmokingchangesratherthandietaryones(ortotheuseoftheantihypertensivemedications).Thedirectnessofthisstudytoobesitytreatmentmustbequestionedaswell,astheparticipantswerenotnecessarilyobese,andinfactpartic-ipantswhoweighedmorethan50%overtheiridealweight(accordingtoheightandweightcharts)wereexcluded.Insum,acrossthesestudies,thereisnotstrongevi-dencefortheefcacyofdietsinleadingtolong-termweightloss.Intwoofthestudies,therewasnotasigni-cantdifferencebetweentheamountofweightlossmain-tainedbyparticipantsassignedtothedietconditionsandthoseassignedtothecontrolconditions.Inthethreestudiesthatdidndsignicantdifferences,thedifferenceswerequitesmall.Theamountofweightlossmaintainedinthedietconditionsofthesestudiesaveraged1.1kg(2.4lb),rangingfroma4.7-kg(10.4-lb)losstoa1.6-kg(3.5-lb)gain.Participantsinthecontrolgroupsaveragedweightgainsof0.6kg(1.3lb;rangingfromlossesof0.9kg[1.9lb]togainsofupto1.8kg[3.9lb]).Themostpositiveconclusionisthatdietingslowstheslightweightgainthatoccurswithageamongtheaveragenondieter,andeventhisslightweightgainwasnotseeninalloftheno-dietcontrolgroups.Itishardtocalltheseobesitytreatmentseffectivewhenparticipantsmaintainsuchasmallweightloss.Clearly,theseparticipantsremainobese. Twostudiesdidnotreportsufcientinformationtoassesswhetherdifferencesweresignicant. April2007AmericanPsychologist Thereissomeevidencefortheeffectivenessofdietsinleadingtootherbenecialhealthoutcomes,particularlyinhelpingpeoplestayoffantihypertensivedrugsandpre-ventingdiabetes,butthisevidenceisnotconsistentacrossthestudies.Inaddition,itisnotpossibletodetectwhetherthedietcomponentsoftheseinterventionswerepotent,astheinterventionsallcontainedothercomponentsthatmayhavereducedhypertensionorpreventeddiabetes(e.g.,in-creasesinphysicalactivity,reductionsinsmoking,alcoholuse,andsodium).Inaddition,theeffectsonbloodpressureseemtobespecictoindividualswhohavehypertension,ratherthantoindividualswhoareoverweight,andhencethoseinterventionsaremoreaptlylabeledobesitytreatments—andmightbeofferedtohypertensiveindividualsregardlessoftheirweight.Sim-ilarly,onlyindividualswithaspecicriskfactorfordia-beteswereincludedinthediabetespreventionstudy,soitsgeneralizabilitytoobeseindividualsingeneralisnotknown.Finally,althoughallofthestudiesinthissectionusedstrongresearchdesigns,instudiesthattrackedpartic-ipantsforsuchlongperiodsoftime,itisdifculttoretainallparticipantsinfollow-upassessments.Itisnotentirelyclearfromtheresearchreportsexactlywhichparticipantswereincludedateachassessmentpoint,aproblemthatwediscussinmoredetailbelow.ObservationalStudies:Type1.Long-TermFollow-UpsWithoutControlGroupsMostconclusionsintheliteratureonthelong-termout-comesofdietingcomefromstudiesinwhichparticipantsinstructuredweightlossprogramsarerecontactedfromtwotoveyearsafterthedietends.Thesestudiesdonotincludecontrolgroupsofnondietingcomparisonpartici-pants,sonocausalconclusionsabouttheeffectsofthedietscanbedrawnfromthem.Despitethisproblem,thesestudiesdogiveausefulsenseoftheweighttrajectoriesofindividualswhogoondiets.Bysearchingonlinedatabasesandreviewsofdietstudies(Anderson,Konz,Frederich,&Wood,2001;As-trup&Rossner,2000;Black,Gleser,&Kooyers,1990;Foreytetal.,1981;Jefferyetal.,2000;Leon,1976;Perri,1998;Perri&Fuller,1995;Saris,2001),wewereabletolocate14studiesthatfollowedparticipantsforatleastfouryearsafteradiet(seeTable1forthefeaturesofthesestudies).Theaverageweightlossonthesedietswas14kg(30.8lb),andbythelong-termfollow-up,participantshadgainedbackallbut3ofthosekilograms(6.6lb).Eightofthestudiesreported(ormadeitpossibletocompute)thepercentageofparticipantswhoweighedmoreatfollow-upthanbeforetheywentonthediet.Theseratesaveraged41%andrangedfrom29%(Pekkarinen&Mus-tajoki,1997)to64%(Wadden,Sternberg,Letizia,Stunkard,&Foster,1989),includingonestudythatfoundthat50%oftheparticipantsweighedmorethan5kg(11lb)abovetheirstartingweightbyveyearsafterthediet(Foster,Kendall,Wadden,Stunkard,&Vogt,1996).Ofnote,studiesalwaysreportthepercentageofparticipantswhomanagetokeepoffsomepercentageofthelostweight,butonlyasubsetreportedonparticipantsforwhomthedietwascounterproductive,eventhoughthispercent-ageistypicallylargerthanthepercentagewhokeptoffsubstantialweight.Althoughthendingsreportedsofargiveableakpictureoftheoutcomesofdiets,therearefourreasonswhytheactualeffectivenessofdietsisevenworse.First,thestudieshaveverylowfollow-uprates,andthisisespeciallytrueforthelongertermfollow-ups.Second,manyoftheparticipantsinthesestudiesself-reportedtheirweightoverthephoneorbymail.Third,mostofthestudiesconfoundeffectsofthedietwitheffectsofexercise.Fourth,asub-stantialpercentageofparticipantsinthesestudieshavebeenonotherdietssincethestudieddietended.Eachofthesemethodologicalproblemsbiasesthestudiestowardshowingmoreeffectivemaintenanceofthelostweight.Wewillbrieyconsidereachoftheseproblemsforthe14studiesinwhichindividualswerefollowedforfourormoreyearsafterthedietended.Follow-uprates.Thefollow-upratesinthesestudieswerequitelow(seeTable1).Overall,33%oftheoriginalparticipantsinthesedietstudiesreturnedfortheirlong-termfollow-up.Notsurprisingly,thefourstudieswiththehighestfollow-uprates(from81%to88%)wereamongthevestudieswiththesmallestsamplesizes(36par-ticipants).Eightofthestudieshadfollow-uprateslowerthan50%.Itisgenerallybelievedthatlowfollow-upratesbiastheresultsofdietstudies,makingthedietsappeartobemoreeffectivethantheywere,becauseindividualswhogainbacklargeamountsoftheirweightareparticularlyunlikelytoshowupforfollow-uptests.Inonestudy,researchersexaminedtheweighttrajectoriesofparticipantsonthebasisofhowmanyfollow-upappointmentstheyattended(Hovelletal.,1988).Theinitialamountofweight April2007AmericanPsychologist lostdidnotdifferamongtheseparticipants,butparticipantswhoshowedupforonlyonefollow-upappointmenthadthesteepestrateofweightregainoverthenext30months,regaining82%oftheweighttheyhadlost.Participantswhoshowedupfortwotofourfollow-upappointmentsregained59%ofthelostweight.Inaddition,asurveyofindividualswhohadparticipatedinacommercialweightlossprogramfoundthatindividualswhorespondedtothesurveyhadlostsignicantlymoreweightthanthosewhodidnotrespond(Grodstein,Levine,Spencer,Colditz,&Stampfer,1996).Onestudywithahighfollow-uprateenticedpartici-pantsbacktothestudybyofferingfurtherdietservices(Graham,Taylor,Hovell,&Siegel,1983).Only23%ofparticipantsrespondedtorequeststoreturnforfurtherassessmentswhenthoserequestsweremadewithastan-dardletterandphonecall.Butwhenresearchersofferedaninexpensivefollow-upclass,anadditional60%ofpartici-pantsagreedtoreturn,and17%moreagreedtoreturnafterbeingofferedhomeweightsandafreeclass.Thispatternofreturnsuggeststhattherecruitmentpracticesusedbynearlyallotherstudies(aletterandaphonecall)mayinadvertentlyloseparticipantswhofeelthattheystillneedfurtherweightlossservices.Givingparticipantsthesensethatitisnetoreturnforadditionalfollow-upseveniftheyregainedsignicantamountsofweightmayhelptoin-creasefollow-upratesandreducebias.Itisdifculttocomparestudiesbasedontheirfol-low-uprates,becausestudiesreportedandcomputedtheseratesindifferentways.Ideally,theserateswouldbecom-putedbydividingthenumberofparticipantsinthefol-low-upbythetotalnumberofparticipantsintheoriginaldiet(asreportedinTable1).Few,ifany,studies,however,computetheirfollow-upratethatway.Moststudiessub-tractvarioussubgroupsofparticipantsfromthetotalnum-berofparticipantsintheoriginalstudy,therebyreducingthedenominatorofthisequationandincreasingthere-portedfollow-uprate.Thesubsetofparticipantstypicallysubtractedfromtheoveralltotalincludesparticipantswhohavediedsincethestudyendedandparticipantsforwhomresearcherscouldnotlocateavalidphonenumberoraddress.Attritionforthosereasonsisconsideredrandombymostresearch-ers,butitmaynotbe.Becausebeingobeseislinkedtohealthoutcomes,participantswhohavediedmayhavebeenmorelikelythanotherparticipantstohaveregainedtheirlostweight.Instudiesinwhichparticipantsknowtherearelong-termfollow-ups,beingdifculttolocateisnotnecessarilyarandomevent.Participantswithpooroutcomesmaynotbothercontinuingtoupdatestudyper-sonnelontheirwhereabouts.Evenmoreproblematicisthatmostresearchersalsosubtractadditionalcategoriesofparticipantsfromtheiroriginalstudytotals(andtheirdataanalyses),andexclud-ingthesecategoriesdoesnotjustmakethefollow-uprateappearhigherthanitactuallywas,butitalsotypicallymakesthedietappearmoresuccessfulthanitwas.Theseadditionalcategoriesofexclusionsincludeparticipantswhodidnotlosesufcientamountsofweightduringthestudy,participantswholeftsubstantialportionsofques-tionnairesblank,participantswhohadparticipatedinasimilardietbefore,participantswhorefusedtoparticipateinearlierfollow-upsforthestudy,participantswhodidnotreturncalls,andparticipantswhohadgastricbypasssur-geries(orothertypesofsurgery)toinduceweightlosssubsequenttothestudy.Onestudyreportedexcludingtwoparticipantsfromanalysesbecause“inclusionofthetwopatientsstronglyskewedtheresultsagainstweightlossmaintenance”(Walsh&Flynn,1995,p.232).Thesetypesofexclusionscanleadtofollow-upreportsonfracturedsamples.Forexample,onestudyenrolled426participantsinadietprogram(Anderson,Vichitbandra,Qian,&Kryscio,1999)andthenexcludedallbut154participantsfromanalysisforavarietyofthereasonsreportedabove.Researchersobtainedfollow-upweightsfor112ofthose154participantsattheirrsttimepointandthereforereportedafollow-uprateof73%,eventhoughonly26%oftheoriginalparticipantswereincludedinthefollow-up.Inaddition,thearticleisalong-termfollow-upstudy,butthenalfourfollow-uppointsonlyincludefrom15to42participantseach,atinyfractionoftheoriginalsamplesize.Whichcategoriesofparticipantsitisappropriatetoexcludefromstudytotalsis,ofcourse,ajudgmentcall,butitisimportanttorememberthatalloftheseexclusionslikelymakeweightlossmaintenanceappearmoresuccess-fulthanitis.Itisworthnotingthatalthoughstudiesndnumerousreasonstoexcludeparticipantswhomightmakethedietlookineffective,itdoesnotappearthatanystudiesexcludeparticipantswhomightinappropriatelymakethedietlookeffective.Forexample,wecouldnotlocateanystudiesthatexcludedparticipantswhohadahistoryofstableweightbeforeasinglerecentepisodeofweightgain,perhapsafterapregnancyoranepisodeofclinicaldepres- April2007AmericanPsychologist sion.Theseindividuals,whoseweightsarehigherthantheirlong-standingtypicalweight,arethoughttohaveaneasiertimemaintainingweightlossafteradiet.Self-reportsofweight.Anotherreasonwhytheresultsfromthe14long-termdietstudiesarelikelybiasedtowardoverestimatingtheeffectivenessofthedietsisthatasubstantialpercentageoftheparticipantsinthesestudiesself-reportedtheirweight(bymailorphone)in-steadofbeingweighedatthelab(seeTable1).Researchersgenerallyacknowledgethatthemostaccuratewaytoassessindividuals’weightistomeasureitwithascaleinthelaboratory.Becausemanyparticipantsareunableorun-willingtocomebacktothelab,itisnotusuallypossibleforresearcherstomeasureeachparticipant’sweight.Inthesestudies,56%ofparticipantsself-reportedtheirweight.Thepercentageofself-reportedweightsvariessignicantlyacrossthestudies,withthelargerstudiestendingtohavehigherself-reportratesthanthesmallerstudies.Inthreeofthestudies,includingthreeofthefourlargeststudies,100%oftheparticipantsself-reportedtheirweight.Self-reportratesfortheremaining10studiesrangedfrom0%(in5studieswithsmallsamplesizes)to63%.Forsomeofthesestudies,researcherstriedtocorrectforself-reportingbiasbyadding2.3kg(5lb)toeachparticipant’sweight.ItisnotclearifthisisanadequateTherehasbeenaconsiderableamountofresearchcomparingindividuals’self-reportedweightstotheirscale-measuredweight,butmostofthatworkwasnotconductedwithobeseindividualsorwithindividualswhowerepar-ticipatinginweightlossprograms.Ameta-analysisof15studiescomparingindividuals’self-reportedweightwiththeirscale-measuredweightlocatedonlytwostudiesofparticipantsinweightlossprogramsthatreportedsufcientinformationtobeincludedinthemeta-analysis(Bowman&DeLucia,1992).Participantsinthe15studiessigni-cantlyunderestimatedtheirweightbyapproximately2.1kg(4.6lb).Moreextremebias(underestimatesof3.7kg[8.2lb])wasfoundwhenconsideringjustthestudiesofindi-vidualsinweightlossprograms,althoughtheseconclu-sionsshouldbeconsideredtentativeastherewereonlytwosuchstudies.Nevertheless,itisimportantforresearcherstousescale-measuredweightsasoftenaspossibleandtobeawarethattheuseofself-reportedweightswillbiasstudiestowardmakingdietslookmoreeffectivethantheyare.Confoundingdietandexercise.Effectsofdietingareconfoundedwitheffectsofexercisinginmanylong-termfollow-upsofdietstudies.Individualsondietsshouldexercise,butifthegoalistoevaluatetheeffectsofrestrictingcalorieintake—nottheeffectsofexercise—onsubsequentweightpatterns,theoccurrenceofexercisebe-comesamethodologicalproblem.Thisconfoundispartic-ularlysignicantbecausecorrelationalstudiesconsistentlyndthatindividualswhoreportedthemostexercisealsohadthebestweightlossmaintenance(reviewedinFo-gelholm&Kukkonen-Harjula,2000;Wing,1999).Exercisehasalsobeenshowntoimprovethemainte-nanceofweightlossintwometa-analysesofcontrolledtrialsinwhichparticipantswererandomlyassignedtodietandexerciseorjusttodiet(Fogelholm&Kukkonen-Harjula,2000;Wing,1999).Inonestudy,forexample,participantswererandomlyassignedtoadiet-only,exercise-only,ordiet-plus-exerciseinterventionforoneyear(Skenderetal.,1996).Allpartic-ipantslostsimilaramountsofweightduringtherstyear.Whenparticipantswerereassessedduringthesecondyear,thediet-onlyparticipantsaveragedafollow-upweightthatwasabout0.9kg(1.9lb)heavierthanbaseline,whereasthegroupsthatincludedexerciseremained2.5kg(5.5lb)belowbaseline.AccordingtotheNationalWeightControlRegistry,alongitudinalstudyofindividualswhomaintainedaweightlossof13.6kg(29.9lb)foratleastoneyear,90%ofitsparticipantsusedregularphysicalactivityasastrategytomaintaintheloss(Klem,Wing,McGuire,Seagle,&Hill,1997).Similarresultswerefoundinanothersampleofindividualswhomaintainedasubstantialweightloss(Kay-man,Bruvold,&Stern,1990)andinasurveyofsubscrib-erstoConsumerReports(“TheTruthAboutDieting,”2002).Infact,asurveyofparticipantsinacommercialdietprogramfoundthatexercisefrequencywasthestrongestpredictorofweightlossmaintenance(Grodsteinetal.,1996).Ifsubstantialpercentagesofparticipantsindietstudiesareexercising,thedietwillappearmoreeffectivethanitactuallyis.Nineofthe14studieswithlong-termfollow-upsdidnotreportinformationontheexercisehabitsofparticipants(seeTable1).Ofthe5studiesthatdidreportonexercise,3foundthatparticipantswhoexercisedregularlymain-tainedsignicantlygreaterweightlossthanparticipantswhodidnotexercise.Onestudydidnotndacorrelationbetweenayes–noquestionaboutexerciseandweightlossmaintenance,butithadasmallsamplesize(26)that April2007AmericanPsychologist mayhaveprecludedndingsuchacorrelation(Murphy,Bruce,&Williamson,1985).Theremainingstudyassignedparticipantstoweightlossprogramsthateitherincludedordidnotincludeanexercisecomponent(Stalonas,Perri,&Kerzner,1984).Nodifferencesintheamountoflostweightmaintainedwerefoundbetweenparticipantsassignedtothetwotypesofprograms,butitisimportanttonotethatresearchersdidnotcomparetheweightlossmaintenanceofparticipantswhoactuallyengagedinexercisewithpartic-ipantswhodidnot.Overall,itseemslikelythatlong-termfollow-upsofdietstudiesgiveoverlyoptimisticviewsofthesuccessofsuchdietsbecauseparticipantswhoexercisetendtoshowgreaterweightlossmaintenancethanpartic-ipantswhosimplydiet.Participationinadditionaldiets.factorthatobscurestheresultsoflong-termdietstudiesisthatstudyparticipantsoftenparticipateinadditionaldietsafterthedietbeingassessedinthestudyendsbutbeforethelong-termfollow-updataarecollected.Elevenofthe14studieswithlong-termfollow-upsreportedsomeinforma-tiononadditionaldieting(seeTable1).Sevenofthesestudiesreportedthepercentageofparticipantswhosaidtheyhadbeenonadditionaldiets,andtheseratesrangedfrom20%to65%ofparticipants.Threeadditionalstudiesaskedparticipantshowmanyotherdietstheyhadbeenonsincetheoriginaloneandreportedmeansofonetothreeotherdiets.Finally,onestudyreportedthat12%oftheparticipantslostmorethan10kg(22lb)onotherdiets(butdidnotreporttheprevalenceofsimplyparticipatinginotherdiets;Pekkarinen&Mustajoki,1997).Participatinginadditionaldietscanmaketheoriginaldietlookmoreeffectivethanitwas,becauseaconsistentoutcomeofdietsissignicantweightlossintheshortterm(which—aswehaveshown—willlaterberegained).onestudy,participantswereaskedtoreportwhattheyweighedwhentheystartedanadditionaldiet,andaftertakingthisweightintoaccount,theresearchersconcludedthat“failuretocorrectfortheeffectsofadditionaltherapywouldhaveresultedinsignicantoverestimationofthelongtermefcacyofthistherapy”(Waddenetal.,1989,p.42).Anotherstudyfoundthatparticipantshadlostanaverageof11.8morekg(26lb)onadditionaldietsandnotedthatparticipantsdidnotdietduringfollow-upuntiltheirweightexceededtheirbaselineweight(Fosteretal.,1996).Asurveyofparticipantsinacommercialdietpro-gramaskedparticipantstoreporttheirmaximumweightsincecompletingthediet.Itfoundthat60%ofparticipantsweighedmorethantheirstartingweightatsomepointinthethreeyearssincethedietended,eventhoughonly40%currentlyweighedmorethantheirstartingweight(Grod-steinetal.,1996).Insum,long-termdietstudieswithoutcontrolgroupsndlittlesupportfortheeffectivenessofdietinginleadingtosustainedweightloss.Fromonethirdtotwothirdsofparticipantsindietswillweighmorefourtoveyearsafterthedietendsthantheydidbeforethedietbegan.Thisconclusioncomesfromstudiesthatarebiasedtowardshowingsuccessfulweightlossmaintenancebythefourfactorsdescribedaboveandmustbeconsideredaconser-vativeestimateofthepercentageofindividualsforwhomdietingiscounterproductive.Thetruenumbermaywellbesignicantlyhigher.ObservationalStudies:Type2.ProspectiveStudiesWithoutRandomizationAtanyparticulartime,individualsaregaining,maintain-ing,orlosingweight(Klesges,Isbell,&Klesges,1992).Prospectivenonrandomizedstudiescanhelpdeterminetheeffectsofdietingonweightbytrackingweighttrajectoriesovertimeforbothdietersandnondieters.Inthesestudies,individuals’weightanddietstatusesareassessedatbase-lineandthentheirweightismeasuredatcertainfollow-uppoints.Thesestudiesdonotrandomlyassignindividualstodietornotdiet,socausalstatementscannotbemadefromthem,buttheydoallowforusefulcomparisonsbetweenpeoplewhochoosetodietandpeoplewhochoosenottodiet.Todate,however,fewstudieshaveusedsuchdesignsandcontrolledforpotentialconfoundingvariables.Ofthe10prospectivestudieswelocated,only1foundthatdietingatbaselineledtoweightlossovertime,2foundnorelationbetweendietingandweightchange,and7foundthatdietingledtoweightgain.Inthe1studythatshoweddietingtobeeffective,1,120adultswereweighedatbase-lineandagain4yearslater(French,Jeffery,&Murray,1999).Duringtheensuingyears,participantsreportedthenumberofweeksduringwhichtheyengagedin17differentweightlossbehaviors.Althoughthemajorityofthebehav-iors(includingparticipatinginweightlossgroups,eatinglow-caloriedietfoods,andeatingfewercarbohydrates)didnotsignicantlypredictweightchangeoverthe4-yearperiod,reducingcaloriesdid.Cumulativedurationofcal-oriereduction(inweeks)predictedweightloss4yearslater,aftercontrollingforbaselineweight.Twostudiesfoundnorelationshipbetweendietingandweight.Oneweighedagroupof24dietersandnon-dietersdailyforsixweeksandthenagainsixmonthslater(Heatherton,Polivy,&Herman,1991).Althoughneithergroupshowedsignicantweightgainorlossatsixmonths,dietingstatuswasasignicantlybetterpredictorofweightvariabilityoverthesixweeksofdailyweightrecordingsthanwasinitialbodyweight.Theshortdurationofthestudymayhavepreventedsignicantweightchangesfromoccurring,andthesmallsamplesizemayhavemadeanychangesdifculttodetect.Asimilarstudyusedalargersamplesize(305)andalongerfollow-uptime(2.5years),however,andshowednorelationshipbetweenre-straintstatusandweightchange(Klesges,Klem,Epkins,&Klesges,1991),althoughself-reportsofweightwereused.Theremainingsevenprospectivestudieseachfoundthatdietingwasrelatedtoweightgainovertime,aftercontrollingforbaselineweight.Inonemethodologicallyrigorousstudy,researchersattemptedtoexplorethelong-termeffectsofdietingonweightgainover6and15yearsinalargecohort(7,729)ofadultFinnishtwins Repeatedlylosingandthenregainingweight,calledweightcyclingisnotrecommendedasaweightlossstrategy.April2007AmericanPsychologist Rodin,1994;NationalTaskForceonthePreventionandTreatmentofObesity,1994).Thereisevidencefromlarge-scaleobservationalstudiesthatweightcyclingislinkedtoincreasedall-causemortality(Blair,Shaten,Brownell,Col-lins,&Lissner,1993;Lee&Paffenbarger,1992)andtoincreasedmortalityfromcardiovasculardisease(Hamm,Shekelle,&Stamler,1989).Inaddition,weightcyclingisassociatedwithincreasedriskformyocardialinfarction,stroke,anddiabetes(Frenchetal.,1997),increasedhigh-densitylipoproteincholesterol(Olsonetal.,2000),in-creasedsystolicanddiastolicbloodpressure(Kajioka,Tsu-zuku,Shimokata,&Sato,2002),andevensuppressedimmunefunction(Shadeetal.,2004).Ithasoftenbeensuggestedthattheharmfuleffectsofweightcyclingresultfromunintentionalweightloss(i.e.,fromsmokingorillness)ratherthanfromintentionaldiet-ing(French&Jeffery,1994;NationalTaskForceonthePreventionandTreatmentofObesity,1994).However,atleasttwolarge-scalestudiesthatcontrolledforuninten-tionalweightlossstillfoundthatintentionalweightlossislinkedtomortalityrisk(Andres,Muller,&Sorkin,1993;Pamuk,Williamson,Serdula,Madans,&Byers,1993),andthebalanceofevidencedoesseemtoimplicateintentionalweightlossinadversehealthoutcomes.Insum,thepotentialbenetsofdietingonlong-termweightoutcomesareminimal,thepotentialbenetsofdietingonlong-termhealthoutcomesarenotclearlyorconsistentlydemonstrated,andthepotentialharmsofweightcycling,althoughnotdenitivelydemonstrated,areaclearsourceofconcern.Thebenetsofdietingaresimplytoosmallandthepotentialharmsofdietingaretoolargeforittoberecommendedasasafeandeffectivetreatmentforobesity.ResearchAgendaFurtherstudyisneededinthreeprimaryareastoaddressthehealthproblemsassociatedwithobesity.Itisclearthatdietingdoesnotleadtosustainedweightlossinthema-jorityofindividuals,andadditionalstudiesoftheeffectsofdietingonweightarenotneeded.Acallformorerigorousdietstudiesseemsunwarrantedasithasbeennotedthatamongdietstudies,“greatermethodologicalrigorseemstobeassociatedwithpoorerresults”(Krameretal.,1989,p.126).Wedonotthinkfurtherstudyofexistingdietswillleadtoadifferentassessment,nordowethinkanewdietformulationwillappearthatleadstomorefavorableout-comes.Instead,werecommendthatresearchersconductlong-termrandomizedstudiesontheeffectsofdietingonhealthoutcomes,withaspecicfocusonwhethertheshort-termhealthbenetsofweightlosspersistaftertheweightisregained.Suchstudiesshouldmeasurehealthindicators,suchascholesterollevels,bloodpressure,andbloodglucose,aswellasillnessoutcomes,andshouldmakeanefforttoassesswhetherobesepeopleingeneralwillbenetorwhetheronlypeoplewithriskfactorsforparticularillnesseswillbenet.Thesestudiesmustmakeeveryefforttominimizethebiasesthatledtooverlyoptimisticreportsofthesuccessofdiets(i.e.,lowfol-low-uprates,self-reportedweights,participationinaddi-tionaldiets,andconfoundswithexercise).Second,althoughwedonotrecommendfurtherstudyoftheeffectsofdietingonweightloss,itisstillnecessarytoexplorethehealthconsequencesofweightregain.Be-causethemajorityofindividualswhoengageindietstendtoregainmostoftheirlostweight,nodietcanberecom-mendedwithoutconsideringthepotentialharmsofweightcycling.Researchmustcontinuetoexaminetheeffectsofweightcycling,andinparticularthisworkmustmakeafocusedefforttodistinguisheffectsofintentionalweightlossfromthoseofunintentionalweightloss.Inaddition,researchonweightcyclingmustattempttolocatemecha-nismsthatlinkweightcyclingtohealthoutcomes.Sofar,sucheffortshavehadmixedsuccess(Brownell&Rodin,1994;NationalTaskForceonthePreventionandTreat-mentofObesity,1994).Finally,asnotedearlier,exercisemayverywellbethepotentfactorleadingtosustainedweightloss,particularlyamongparticipantsintheNationalWeightControlRegis-try(Klem,Wing,McGuire,Seagle,&Hill,1997).Acom-prehensivereviewoftheeffectsofexerciseonhealthstatedthatinadditiontoitseffectsonweight,exercisealsohasbeenlinkedtopositivehealthoutcomes,includingde-creasedmortality,decreasedcardiovasculardisease,de-creasedType2diabetes,decreasedhypertension,andin-creasedmentalhealth(U.S.DepartmentofHealthandHumanServices,1996).Wethereforesuggestthatfutureresearchfocusonexerciseasatreatmentforobesity.Es-peciallyrecommendedarerandomized,controlledtrialsthatcompareexercise-onlyinterventionstobothdiet-onlyinterventionsandcombinedinterventionssothattheeffectsofexercisecanbedistinguishedfromtheeffectsofdieting.Studiestypicallyconfoundtheeffectsofdietandexercisebycomparingdiet-onlyinterventionswithcombineddiet-plus-exerciseinterventions(forareview,seeAvenelletal.,2004).Thesestudiescannotassesswhetherexercisealoneisasbenecial,orevenmorebenecial,thandietplusexercise.Inonestudythatwasabletoassesstheuniquecontributionofexercisebycomparingparticipantsas-signedtodiet-only,exercise-only,oracombinedinter-vention,theexercise-onlygroupshowedbetterlong-termweightlossmaintenancethanthecombinedgroup(Skenderetal.,1996).Morestudieswiththistypeofdesignareneeded.Inthestudiesreviewedhere,dieterswerenotabletomaintaintheirweightlossesinthelongterm,andtherewasnotconsistentevidencethatthedietsresultedinsignicantimprovementsintheirhealth.Inthefewcasesinwhichhealthbenetswereshown,itcouldnotbedemonstratedthattheyresultedfromdieting,ratherthanexercise,med-icationuse,orotherlifestylechanges.Itappearsthatdieterswhomanagetosustainaweightlossaretherareexception,ratherthantherule.Dieterswhogainbackmoreweightthantheylostmayverywellbethenorm,ratherthananunluckyminority.IfMedicareistofundanobesitytreat-ment,itmustleadtosustainedimprovementsinweightandhealthforthemajorityofindividuals.Itseemscleartousthatdietingdoesnot.April2007AmericanPsychologist forweightlossusingfoodprovisionandmonetaryincentives.ofConsultingandClinicalPsychology,63,Jordan,H.A.,Canavan,A.J.,&Steer,R.A.(1985).Patternsofweightchange:Theinterval6to10yearsafterinitialweightlossinacogni-tive–behavioraltreatmentprogram.PsychologicalReports,57,Kajioka,T.,Tsuzuku,S.,Shimokata,H.,&Sato,Y.(2002).Effectsofintentionalweightcyclingonnon-obeseyoungwomen.ClinicalandExperimental,51,149–154.Kayman,S.,Bruvold,W.,&Stern,J.S.(1990).Maintenanceandrelapseafterweightlossinwomen:Behavioralaspects.AmericanJournalofClinicalNutrition,52,800–807.Klem,M.L.,Wing,R.R.,McGuire,M.T.,Seagle,H.M.,&Hill,J.O.(1997).Adescriptivestudyofindividualssuccessfulatlong-termmaintenanceofsubstantialweightloss.AmericanJournalofClinicalNutrition,66,239–246.Klem,M.L.,Wing,R.R.,Simkin-Silverman,L.,&Kuller,L.H.(1997).Thepsychologicalconsequencesofweightgainpreventioninhealthy,premenopausalwomen.InternationalJournalofEatingDisorders,21,Klesges,R.C.,Isbell,T.R.,&Klesges,L.M.(1992).Relationshipbetweenrestraint,energyintake,physicalactivity,andbodyweight:Aprospectiveanalysis.JournalofAbnormalPsychology,101,668–674.Klesges,R.C.,Klem,M.,&Bene,C.(1989).Effectsofdietaryrestraint,obesity,andgenderonholidayeatingbehaviorandweightgain.nalofAbnormalPsychology,98,499–503.Klesges,R.C.,Klem,M.L.,Epkins,C.C.,&Klesges,L.M.(1991).Alongitudinalevaluationofdietaryrestraintanditsrelationshiptochangesinbodyweight.AddictiveBehaviors,16,Korkeila,M.,Rissanen,A.,Kapro,J.,Sorensen,T.I.A.,&Koskenvuo,M.(1999).Weight-lossattemptsandriskofmajorweightgain:AprospectivestudyinFinnishadults.AmericanJournalofClinicalNutrition,70,Kramer,F.M.,Jeffery,R.W.,Forster,J.L.,&Snell,M.K.(1989).Long-termfollow-upofbehavioraltreatmentforobesity:Patternsofweightregainamongmenandwomen.InternationalJournalofObe-sity,13,Lantz,H.,Peltonen,M.,Agren,L.,&Torgerson,J.S.(2003).Adietaryandbehaviouralprogrammeforthetreatmentofobesity.A4-yearclinicaltrialandalong-termposttreatmentfollow-up.JournalofIn-ternalMedicine,254,Lee,I.M.,&Paffenbarger,R.S.(1992).ChangesinbodyweightandJournaloftheAmericanMedicalAssociation,268,Leon,G.R.(1976).Currentdirectionsinthetreatmentofobesity.chologicalBulletin,83,Lissner,L.,Steen,S.N.,&Brownell,K.D.(1992).Weightreductiondietsandhealthpromotion.AmericanJournalofPreventiveMedicine,154–158.Medicareprogram:RevisedprocessformakingMedicarenationalcov-eragedeterminations,68Fed.Reg.55634(Sept.26,2003).Mokdad,A.H.,Marks,J.S.,Stroup,D.F.,&Gerberding,J.L.(2005).Correction:ActualcausesofdeathintheUnitedStates,2000.oftheAmericanMedicalAssociation,293,MultipleRiskFactorInterventionTrialResearchGroup.(1990).Mortalityratesafter10.5yearsforparticipantsintheMultipleRiskFactorInterventionTrial:Findingsrelatedtoapriorihypothesesofthetrial.JournaloftheAmericanMedicalAssociation,263,Murphy,J.K.,Bruce,B.K.,&Williamson,D.A.(1985).Acomparisonofmeasuredandself-reportedweightsina4-yearfollow-upofspouseinvolvementinobesitytreatment.BehaviorTherapy,16,524–530.NationalCenterforHealthStatistics.(2006).Health,UnitedStates,2006,withchartbookontrendsinthehealthofAmericans.Hyattsville,MD:NationalCenterforHealthStatistics.NationalTaskForceonthePreventionandTreatmentofObesity.(1994).Weightcycling.JournaloftheAmericanDieteticAssociation,272,1196–1202.Olson,M.B.,Kelsey,S.F.,Bittner,V.,Reis,S.E.,Reichek,N.,Hand-berg,E.M.,etal.(2000).Weightcyclingandhigh-densitylipoproteincholesterolinwomen:Evidenceofanadverseeffect.AreportfromtheNHLBI-sponsoredWISEStudy.JournaloftheAmericanCollegeofCardiology,36,Pamuk,E.R.,Williamson,D.F.,Serdula,M.K.,Madans,J.,&Byers,T.E.(1993).WeightlossandsubsequentdeathinacohortofU.S.AnnalsofInternalMedicine,119,744–748.Pekkarinen,T.,&Mustajoki,P.(1997).Comparisonofbehaviortherapywithandwithoutvery-low-energydietinthetreatmentofmorbidobesity:Ave-yearoutcome.ArchivesofInternalMedicine,157,Perri,M.G.(1998).Themaintenanceoftreatmenteffectsinlong-termmanagementofobesity.ClinicalPsychology:Science&Practice,5,526–543.Perri,M.G.,&Fuller,P.R.(1995).Successandfailureinthetreatmentofobesity:Wheredowegofromhere?Medicine,Exercise,Nutrition,andHealth,4,Presnell,K.,&Stice,E.(2003).Anexperimentaltestoftheeffectofweight-lossdietingonbulimicpathology:Tippingthescalesinadifferentdirection.JournalofAbnormalPsychology,112,166–170.Saris,W.H.(2001).Very-low-caloriedietsandsustainedweightloss.ObesityResearch,9S,295S–301S.Schotte,D.E.,&Stunkard,A.J.(1990).Theeffectsofweightreductiononbloodpressurein301obesepatients.ArchivesofInternalMedicine,Shade,E.D.,Ulrich,C.M.,Wener,M.H.,Wood,B.,Yasui,Y.,Lacroix,K.,etal.(2004).Frequentintentionalweightlossisassociatedwithlowernaturalkillercellcytotoxicityinpostmenopausalwomen:Possi-blelong-termimmuneeffects.JournaloftheAmericanDieteticAsso-ciation,104,Shah,M.,Hannan,P.J.,&Jeffery,R.W.(1991).Seculartrendinbodymassindexintheadultpopulationofthreecommunitiesfromtheuppermid-westernpartoftheUSA:TheMinnesotaHeartHealthProgram.InternationalJournalofObesity,15,499–503.Skender,M.L.,Goodrick,G.K.,DelJunco,D.J.,Reeves,R.S.,Darnell,L.,&Gotto,A.M.,etal.(1996).Comparisonof2-yearweightlosstrendsinbehavioraltreatmentsofobesity:Diet,exercise,andcombi-nationinterventions.JournaloftheAmericanDieteticAssociation,96,Stalonas,P.M.,Perri,M.G.,&Kerzner,A.B.(1984).Dobehavioraltreatmentsofobesitylast?Ave-yearfollow-upinvestigation.tiveBehaviors,9,Stamler,R.,Stamler,J.,Grimm,R.,Gosch,F.C.,Elmer,P.,Dyer,A.,etal.(1987).Nutritionaltherapyforhighbloodpressure:Finalreportofafour-yearrandomizedcontrolledtrial—TheHypertensionControlJournaloftheAmericanMedicalAssociation,257,1484–Stice,E.,Cameron,R.P.,Killen,J.D.,Hayward,C.,&Taylor,C.B.(1999).Naturalisticweightreductioneffortsprospectivelypredictgrowthinrelativeweightandonsetofobesityamongfemaleadoles-JournalofConsultingandClinicalPsychology,67,Stunkard,A.J.,&Penick,S.B.(1979).Behaviormodicationinthetreatmentofobesity.ArchivesofGeneralPsychiatry,36,801–806.Swanson,D.W.,&Dinello,F.A.(1970).Follow-upofpatientsstarvedforobesity.PsychosomaticMedicine,32,209–214.Taylor,C.B.,Fortmann,S.P.,Flora,J.,Kayman,S.,Barrett,D.C.,Jatulis,D.,etal.(1991).Effectsoflong-termcommunityhealthedu-cationonbodymassindex.AmericanJournalofEpidemiology,134,Torgerson,J.S.,Hauptman,J.,Boldrin,M.N.,&Sjostrom,L.(2004).XENicalinthePreventionofDiabetesinObeseParticipants(XEN-DOS)study:ArandomizedstudyoforlistatasanadjuncttolifestylechangesforthepreventionofType2diabetesinobesepatients.betesCare,27,Thetruthaboutdieting.(2002,June).ConsumerReports,67,26–32.U.S.DepartmentofHealthandHumanServices.(1996).Physicalactivityandhealth:AreportoftheSurgeonGeneral.Atlanta,GA:U.S.DepartmentofHealthandHumanServices,CentersforDiseaseControlandPrevention,NationalCenterforChronicDiseasePreventionandHealthPromotion.Wadden,T.A.,&Frey,D.L.(1997).Amulti-centerevaluationofaproprietaryweightlossprogramforthetreatmentofmarkedobesity:Ave-yearfollow-up.InternationalJournalofEatingDisorders,22,Wadden,T.A.,Sternberg,J.A.,Letizia,K.A.,Stunkard,A.J.,&Foster,G.D.(1989).Treatmentofobesitybyverylowcaloriediet,behaviorApril2007AmericanPsychologist