Tobacco ConmU997i6il9261Scientifi qualit o origina researc article oe - PDF document

Tobacco ConmU997i6il9~261Scientifi qualit o origina researc article oenvironmenta tobacc smokDebora E Barnes Lis A BerInstitut fo HealtPolic Studies Schoo Medicine oCalifornia SaFrancisco CaliforniaR BarneDepartmen o ClinicaPharmacy Schoo oPharmacy Universit California SaFrancisco CalifornialABerCorrespondenc toO Lis A Bero Institut foHaslt PoJic Studies 138gutte SEteet 11t floor, SaFrancisco Californi 94109SA eiuaiibcro@cardio.ucsf.edAbstracObjective—To evaluat th scientifi qualit o origina researc article o thhealt effect o environmenta tobacc 2Barnes, BeroThes include researc sponsorshi (thsource o fundin acknowledged),-articl conclusio (tha environmenta tobacc smok iharmfu o no harmful) articl topic anstud designPreviou studie hav suggeste tha botresearc sponsorshi an articl conclusionhav influence th overal conten o scientifireports Fo example severa studie havshow tha researc sponsore b intereste tend t produc result tha arfavourabl t th sponsors.1*" Similarly thexistenc o publicatio bia suggest tha threport ma contai a preponderanc o studies.18" However th relatiobetwee thes factor an th qualit o th ha no bee studiedW hypothesise a prior tha poo articlqualit woul b associate wit publicatio i proceedings wit tobacc industrsponsorship an wit a conclusio thaenvironmenta tobacc smok i no harmfu thealth Articl topi and stud desig(experimenta versu observational wer a potentia confoundin variablesMethodSELECTIO O PEE REVIEWE AN SYMPOSIUARTICLEPee reviewe article o th healt effect oenvironmenta tobacc smok wer identifieb searchin Medlin from 1 Januar 198 t1 Marc 199 usin th ke word "environmenta tobacc smoke, "tobacc smokpollution, "passiv smok#, "involuntarsmok#, o "sidestrea smoke,* retrievin atota o 218 citations Th searc result wernarrowe electronicall b excludin non languag publication ( = 393) b studie no relate t huma (n = 215) an b excludin article b Medlin a reviews new items commentaries o letter t th ( = 504) Th remainin 107 article evaluate fo inclusionArticle wer include i the describe thstud o th healt effect o environmentatobacc smok i humans wer originaresearc studies an wer publishe i pee journals a define below O th107 article considered 38 (36% satisfieou inclusion/exclusio criteriaStudies of environmental tobacco smokeArticle wer include i environmentatobacc smok wa th focu o th study o ii wa on o severa risk factor examinedStudie o th effect o materna smokin o fetu wer exclude becaus th fetu iexpose t mainstrea smok rathe tha tobacc smokeStudies of health outcomesArticle wer include i the studie anhealt outcom relate t environmentatobacc smoke rangin from lun cance t irritation Article wer exclude i thediscusse polic issue relate t environmenta tobacc smoke measure exposur withouassessin th effects o discusse theoretica relate t stud desig o analysisStudies conducted in hitman subjectsOnl studie conducte i huma subjectwer included I vitro cellular an anima wer excludedStudies that presented original research findingsOrigina researc article wer define a article tha presente dat base o researc con b th authors Article wer include ithe incorporate a shor revie o publishereport followe b a presentatio o origina findings. Article wer als include i presente reanalyse o anothe investigator' data becaus th reanalysi wa t b origina work Meta-analyses woul hav require a differen fo qualit assessment wer bu ar bein assesse i a differenStudies published in peer reviewed journalsArticle wer include onl i the werpublishe i pee reviewe journals A journa considere pee reviewe i i explicitl tha it article wer pee reviewed i ipublishe a lis o pee reviewers o i i submissio o multipl copie omanuscript fo revie befor publicationArticle publishe i symposiu proceeding a pee reviewe journa wer considerefo inclusio a symposiu articles a belowSymposiu article o environmentatobacc smok wer identifie throug thresources (1 Fo a previou study, w identifie1 symposi tha primaril contai article oenvironmenta tobacc smok b searchinMedline Catalog Dialog Conferenc PaperIndex Toxline an Internationa Guid t usin th ke word "environmen tobacc smoke, tobacc smok pollution, s

moking, o "involuntar smokingand "symposia,* "proceedings, o "conferencpaper. (2 Additiona article wer identifie searchin Medlin a describe abov fo reviewe articles excep tha onl sympo article wer considere fo inclusion (3 w identifie symposi o indoo ai bsearchin Catalo usin th titl word "indooair an eithe "symposi#,M "proceedings,"conference, "meeting, o *congressM� anw the examine thei content fo article oenvironmenta tobacc smoke 30 articlewer identifie throug thes thre source an considere fo inclusion Article w# i the studie th healt effect o tobacc smok i anwer origina researc studies a define aborfo pee reviewe articles an i the w«publishe i a symposium "Symposium v-"3define a a publicatio tha contain a colletio o article base o a conferenc proceedingO th 30 symposiu article identified,(30% satisfie ou inclusion/exclusio crft*ria However 2 (24% o thes wer excludbecaus the wer duplicat publications *a f article on environmental tobaco2is the discusse th sam outcom dat froth sam grou o researc subject a anothesymposiu articl i th study I on case thsam articl ha bee publishe i fou differen symposia.3 " W identifie duplicat publication withi th symposiu articl databasb comparin article tha ha a leas onautho wit th sam las name Th articltha appeare t presen th mos completdiscussio o th dat wa selecte fo ( simila techniqu wa used t duplicat publication withi th peereviewe articl database non weridentified.Ou final sampl consiste o 6 articlefro 2 symposia 3 article ha bee i ou previou stud o symposi o tobacc smoke whil 3article wer fro ne symposi identifiethroug Medlin o Catalog Th numbe o pe symposiu range fro 1 t 11wit a media o 2 (interquartil range 1 t 4)Non o th symposi wer pee reviewedSixt eigh pee reviewe journa articlewer randoml selecte fo compariso wit 6 symposiu articles Becaus th qualityo th scientifi literatur ha improve ovetime/4� 2( w wante t ensur tha th publicatio date o th pee reviewe an symposiuarticle di no diffe significantly W determine th numbe o sympo article publishe fro 198 t 198 (n =17) 198 t 198 ( - 32) an 199 t 199(n =19 an used a rando numbe generatot selec th sam numbe o pee reviewe article withi eac tim block Th 6pee reviewe article selecte ha bee i 4 journals Th numbe oarticle pe Journa range fro 1 t 6 wit amedia o 1 (interquartil range 1 t 2)A minimu sampl siz o 3 article ieac compariso grou (symposiu v peereviewed wa necessar t detec a differenci quality score o 0.1 (o a scal o 0 t 1 aa tw taile a = 0.05 p = 0.0 (95 power)an standar deviatio = 0.12.2QUALIT ASSESSMENW hire tw reviewer t asses th quality o article independently usin a instrumen ha previousl bee teste fo validit an2 Bot reviewer wer postdoctorafellow wh ha receive doctorat degree i psycholog a wel a MP (Master i Health degree i epidemiology wer traine t us th instrumen wer give a detaile se o instruction fo durin th study In orde t minimis biases,3 3 reviewer wer unawar oou stud hypotheses di no hav extensivknowledg o th literatur o environmenta smoke state tha the ha neve bee wit eithe pro-tobacc o tobacc groups an state tha the di nohav an financia conflict o interestArticle wer maske s tha identifyincharacteristics—suc a autho name anaffiliations journa an boo titles date opublication source o funding anacknowledgments—ha bee completel removed Article wer sen t reviewer i a ran orde usin a rando numbe generato a computerOu qualit assessmen instrumen idesigne t measur th qualit o design analysis an interpretatio o origi researc articles."8 Thi instrumen waspecificall designe t b applicabl t a widrang o stud designs regardles o articltopic Compare t othe publishe qualityassessmen instruments our ha higreliability.3 Th instrumen consist o a 2ite questionnair tha ask reviewer t determin whethe certai item associate wit quality ar presen i a articl (item i tabl 2) Reviewer ma answe tha ite i "present" "partiall present" "no o "no applicable fo th article ar als aske t classif th stud o th articl int on o 1 categoriesA overal quality scor i calculate base oth stud desig an th numbe o itempresen usin a scorin syste ha bee previousl described.2

Quaiit score ma rangfro 0 (lowes quality t 1 (highes quality)Ou analyse wer base o th mea qualityscor fo eac articl (tha is th averag o thtw reviewers scores) I th reviewers score b mor tha on standar deviationth maske articl wa discusse b bot unti consensu wa achieved an consensu scor wa used Th interrate betwee th tw reviewer wa usin Kendall' coefficien oconcordanc wit adjustmen fo tie rank(W an th intraclas correlatio coefficien(R) Initia interrate reliabilit wa goo ( =0.85 R - 0.64) Reyiewe score differe bmor tha on standar deviatio for 2article (si pee reviewed 1 symposia) Afte ha reache consensu o thes interrate reliabilit wa hig ( = 0.92 = 0.87)CLASSIFICATIO O OTHE VARIABLESponsorshipResearc sponsorshi wa determine base statement i th article an wa classifie eithe "tobacc industry" "government" o "non acknowledged. Articlewer classifie a tobacc industr sponsore ia tobacc company th Tobacc Institute thCounci fo Tobacc Research th Cente foIndoo Ai Research th Smokeles Tobacc Council o a combinatio o thesewer th sol source o fundin acknowledged Article classifie a governmen i on o mor governmen agencies a th Nationa Institute o Health werth sol sourc o fundin acknowledged Article tha acknowledge othe source ofunding suc a privat foundations o whicacknowledge fundin fro multipl source(fo example governmen an privat foundations wer classifie i th "other categoryArticle conclusionArticl conclusion wer code a "positive(environmenta tobacc smok harmful i tharticl conclude tha environmenta tobacc 2Barnes, Be?/)smok ha a advers effec o on o mor oth outcome measured "negative" (environ tobacc smok no harmful i th arti conclude tha environmenta tobacc di no hav a advers effec o an o outcome measured o "n conclusion ith articl controlle for environmentatobacc smok a a confoundin variabl an no dra an conclusion abou it healtStudy designOu scorin syste weight experimenta stud highe tha observationa studies.'5 T contro fo stud desig a a potentia confoundin w classifie eac articl a eitheobservational (fo example th effect oenvironmenta tobacc smok wer deter b askin peopl abou thei histor oexposure o experimental (fo example the:effect o environmenta tobacc smok werdetermine b exposin subject t i i a chamber)Article topicA preliminar examinatio o th datindicate tha pee reviewe an symposiu tende t stud differen topics T a articl topi a a potentiaconfoundin variable th topi o eac articl code a eithe "lun cancer" "respiratoreffects" "othe chroni disease" "biochemica example urinar rnutagenidty) buildin syndrome (fo exampleheadach an ey irritatio i th workplace)o "other.DUPLICAT PUBLICATIONA describe abov i th sectio o selectioo pee reviewe an symposiu articles weliminate duplicat publication tha occurre withi th symposiu articl databas(symposium/symposium duplicates anwithi th pee reviewe articl databas (peer duplicates) However w wer intereste comparin th qualit o duplicat tha wer publishe i bot a an a pee reviewe journa (peersymposiu duplicates) W identifie thesduplicat publication b searchin ou entir o 38 pee reviewe article usinth las name o th author fro th 6 symposiu article include i th study Articletha ha a leas on autho wit th sam lasnam wer compared Article wer considere b duplicate i the discusse th sam out dat fro th sam subjec population pair o peer/symposiu duplicate identified I fou cases ther wa onl overla betwee th article (fo example on articl discusse lun cancerrespirator disease an hear diseas i a o subjects wherea th othe articl discusse th dat relate t lun cancer)Al o th symposiu article an five o thpee reviewe article ha bee include i thmai study Th quality o th 1 remaininpee reviewe article wa assesse b mixin i randoml wit th othe articlesANALYSEW used %2 test t compar pee reviewe ansymposiu article i term o qualitativ T examin th relatio ou outcom variable mea qualitscore an th variou predicto variables wfirs conducte ursivariat analyse usin onwa analysi o varianc (ANOVA) W the conducte multivariat regressio analyse iwhic tw model wer developed I th firsmode (ful model) al o th variable werinclude regardles o whethe the wersignificantl associate wit q

uality scores variable wer represente usin variables T develop ou second simple model w bega b includin al variable wer significantl associate ( 0.05wit quality score i th ful model W the additiona variable on a a time focus o thos tha wer associate wit poo quality Variable wer retaine i the associate wit poo quality scores o i change th associatio betwee qualit an th othe variable i th model publication wer compare usin paire t test A test wer conducte usina tw taile a - 0.05 Confidenc intervalwer calculate base o th t distributionAssumption checkingBot ANOV an multivariat regressio analyse us th F test whic i base o thassumptio tha th outcom variabl (qualitscores i approximatel normall distributedwit constan varianc a al value o th inde variables I ou sample peereviewe article ha normall distribute bu th distributio o symposiu arti wa slightl skewe (Shapiro-Wil test P ~0.03) However th media symposiu scor(0.375 di no diffe greatl fro th measymposiu scor (0.365) I addition th deviation fo pee reviewe article(0.08 an symposiu article (0.13 di nodiffe greatly Furthermore th F tes i robus whe sampl size ar tha 3 an roughl balance i eacfgroup A plo o residua value from th fufmultivariat regressio mode di no reveaan obvious violation o th assumptions V?therefor believ tha ou analyse provid estimate o th relation betwee articl pee revie status an th othe variable examinedResultCHARACTERISTIC O PEE REVIEWE ANtSYMPOSIU ARTICLE jTabl 1 show tha pee reviewearticle an non-peer-reviewearticle differe fro eac othe i term ^ source acknowledged articl com stud design an topic Symposiumst'cle wer mor likel tha pee revie t fai t acknowledg thei sources t conclud tha environin smok i no harmful to stud designs an t stud £ ?relate t sic buildin syndromereviewe articles o th othe hand ***** Qualityof• articles oner,•voironmental tobacco smoke2Table IanidesCharacteristics of peer reviewed and symposium.Sourc o fundingTobacc industrGovernmenOtheNon acknowledgeArticl conclusion.ET harmfuET no harmfuN conclusioStud designExperimentaObservationaTopic1Lun canceRespirator diseasOthe chroni diseasBiochemica effectSic buildin syndromMiscellaneouPer csm of articlesPeer reviewedarticles2.27-23.45.76.20.2.98.91.10,51.56.10.4.7.Symposiumctnicks(h=68)4.7.14.73.45.41.3 3.25.75.13.35.1.10.27.11.ETS environmenta tobacc smokebXJffi14.4jdf=2jP«0.00].likel tha symposiu article t acknowledggovernmen sponsorship t conclud tha tobacc smok i harmful t observationa stud designs an t stud disorder an othe chroni Give th underlyin difference pee reviewe an symposiuarticles i i particularl importan t conducrnultivariat analyse i orde t ensur thaan difference observe i quality score arno actuall du t thes othe factorsUNIVARIAT ANALYSEI th univariat analyses articl qualit wahighl associate with pee revie status PeeTable 2 hem 6 item analysis of quality $coresreviewe journa article ha mea quality o 0.4 (95 confidenc interva (CI) t 0.47) wherea non-peer-reviewe article ha mea quality score o (95 CI 0.3 t 0.40 ( = 21,15 d = IP 0.0001) Th score o pee reviewe arti range fro 0.2 t 0.61 whil th score symposiu article range fro 0.0 tTabl 2 present a ite b ite analysi oth difference betwee pee reviewe an articles Pee reviewe article highe tha symposiu article o mos th item i ou quality assessmen Fo example pee reviewe wer mor likel tha symposiu arti t hav a wel describe stud question t a appropriat stud design t specif an exclusio criteria t stud subjec populations t reporfindings completely an r hav conclusiontha wer consisten wit thei results Th cat i whic pee reviewe and symposiu di no diffe tende t b thos i bot group score poorly Fo example symposiu no pee reviewe articleroutinel reporte blindin o investigator o Symposiu article di no receiv highe score tha pee reviewe fo an o th criteri measuredTh univariat analyse als reveale thamea quality score wer associate wit fund sourc acknowledge ( - 2.87 d = 3 P =0.039 an wit articl topi ( - 9.33 d = 5P 0.0001) Pos ho testin usin th Scheff suggeste tha article tha faile t thei fundin source ha poore tha thos tha acknowledge an fundin source althoug th diffe

rence wer no significant I addition articlerelate t sic buildin syndrom ha lowe score tha article o al topics. base o th pos ho ScheffItemPercent (number) of anklessmamningitemPeer retn&wed articles Symposium articles Fischer's exact P value1 Stud questio wei describe 96 (65/68 75 (50/67 0.00062 Stud desig appropriat 84 (57/68 55 (36/65 0.00063 Inclusion/exclusio criteri specifie 56 (37/66 15 (9/59 4 Fo cas repor only patien characteristic reporte 0 (0/0 100 (I/I n/5 Subjec populatio appropriat 88 (60/&8 59 (36/61 0.00026 Conrro populatio appropriat 64 (1*22 28 (7/25 0.07 Subject randoml selecte 14 (9/66 9 (5/58 0.48 Metho o rando selectio describe 0 (0/9 0 (0/6 a/9 Metho o rando allocatio describe 0 (0/0 0 (0/2 n/1 Investigator blinde 4 (2/53 7 (3/46 O.S1 Subject blinde 7 (3/45 3 (1/39 0.61 Measuremen bia accounte for 46 (3I/6S 27 (17/62 0.01 Confounder accounte fo b stud desig 43 (18/42 16 (8/50 0.001 Confounder accounte for b analysi 70 (42/60 50 (29/5S 0.01 A prior sampl siz justificatio 3 (2/67 0 (0/61 0.5% 6 Post-ho powe calculation for non-significanresult 36 (22/61 18 (9/51 0.01 Statistica analyse appropriat 83 (54/65 6S (36/53 0,01 Statistica test state 95 (62/65 80 (40/50 0.019 Exac P value o confidenc interval reporte 47 (30/64 32 (16/50 0.12 Attritio o subject discusse 67 (6/9 23 (3/13 0.02 Result completel reporte 97 (66/68 75 (46/61 0.00042 Finding suppor conclusion 96 (64/67 73 (48/66 0.0003Th dat ar presente a th percentag o article fo whic on o bot reviewer rate th criterio a "present" Thdenominato i eac cei varie becaus th articl wa exclude i on o bot reviewer fel tha th criterion wa "noapplicable (n/a fo th article*Usin a Bonfemm adjustmen for multipl comparisons onl P 0.002 sho«! b considere statisticall significan 2Table 3 Muhivariau regression analysis: full modelPee revie statuPee reviewe v symposiuSourc o fundinTobacc industr v otheGovernmen v otheNon v otheArticl conclusioET harmfu v n conclusioET no harmfu� t n conclusioStud desigExperimenta v observationaTopiLun cance v miscellaneou disease v miscellaneou chroni disease v miscellaneou studie v miscellaneou buildin syndrom v miscellaneouCoefficient0.04-0,060.01-0.020.020.020.080.070.060.080.01-0.0795% Cl0.00 t 0.08-0.16 t 0.04-0.04 t 0.07-0.07 t 0.01-0.04 t 0.08-0.04 r 0.090.02 t 0.14-0.0 t 0,16-0.00 t 0.14-0.00 t 0,17-0,06 t 0.09-0.15 t 0.00t2.24-1.180.58-1,290.680.682.851,731,850,26-1.84Barnes, BeroP value0.020.20.560,190.40.40.000.080.060.060.780.06CI confidenc intervaltest Articl conclusio wa marginallassociate wit quaiity score ( = 2.99 df 2P = 0.054) bu stud desig wa no ( - 0.75MULTJVAKIAT DEGRESSIO ANALYSITh result o ou multivariat analysi usin ful mode ar presente i tabl 3 Th importan predictor o articl quality pee revie statu ( = 0.027 an studdesig ( ~ 0.005) Th coefficient ma b a th chang i mea quality scor wit a give variabl whilcontrollin fo al othe variable i th modelThu whe controllin fo al othe factor pee reviewe article ha meaqualit score tha wer 0,04 point highe(o a scal o 0 t 1 tha symposiu articlesSimilarly experimenta studie ha mea quaiit score tha wer 0.08 point highe thaobservationa studie whe controllin fo alothe variables Neithe fundin sourc no conclusio wa significantl associatewit articl quality i th multivariat modelRoughl 37 o th variabilit i qualit score explaine whe al o th variable wer (R2 = 0.372)Th result o th final simple mode argive i tabl 4 W include bot pee reviestatu an stud design becaus the wer significan i th ful model I addition w th mos significan effect whe thfundin variabl wa dichotomise a an acknowledge versus n fundin an th topi variabl wa a sic buildin syndrom versus othe topics I th simple model thTable 4 Multivariate regression analyses; final modelprimar predictor o articl qualit wer peerevie statu ( = 0.005 an sic buildinsyndrom ( 0.0001) Roughl 33 o thvariabilit i quaiity score wa explaine usinth simple mode (R2 = 0.33)DUPLICAT PUBLICATIONOu analysi o duplicat publication revealetha th article publishe i pee reviewe ha a mea quality scor o 0.4 (95CI 0.4 t 0.54) wherea th sam studiepublishe i symposi ha a mea qualit o 0.4 (95

CI 0.4 t 0.46 (paire t P ~ 0.01) Qualit score wer highe fo reviewe article tha fo symposiu article i 1 o th 1 pairsDiscussioOu finding confir ou primar hypothesi article publishe i symposiu proceeding ar associate wit poo quality W fountha symposiu article ha significanti mea quality score tha pee reviewe eve afte controllin fo th effect ofundin sourc acknowledged articl conclusio regardin th healt effect o tobacc smoke typ o stud used an articl topicTh criterio b criterio analysi showetha pee reviewe article wer superio t article i term o stud desigreporting an interpretation Fo example reviewe article wer mor likel t usappropriat stud design an subjecpopulations t repor thei stud objectivesinclusion/exclusio process an resultCoefficient95% ClP valuePee revie statuPee reviewe v symposiuSourc o fundinNon acknowledge v an acknowledgeStud desigExperimenta v observationaTopiSic buildin syndrom v ai othe topic0.050.030.120.01 t 0.08-0.06 t 0.00-0,00 t 0.08-0.17 to-0.082.86-1.821.87-5.390.00O.O70.06CI confidenc interval Quality ofarticles on environmental tobacco smoke2adequately ari t dra conclusion tha werconsisten wit thei findings.Ou tw othe a prior hypothese wer noconfirmed W ha originall propose thatobacc industr sponsorshi woul b associate wit poo quality However becaus onlfiv article i ou stud (tw pee reviewedthre symposium acknowledge tobaccIndustr sponsorship w di no hav enougstatistica powe t evaluat thi hypothesis ou findings sugges tha th fail t acknowledg fundin source i margin associate wit poo quality I i possibltha article tha fai t acknowledg fundinsource ma ten t hav poo reportin i an thi ma explai wh thei quality ten t b lower Thi findin shoul binvestigate i futur studiesOu thir a prior hypothesi wa thaarticle wit negativ conclusion (environmenta tobacc smok i no harmfu t woul b associate wit poo quality hypothesi wa no confirmed Articlconclusio wa no associate wit articl whe controllin fo factor suc apee review W hav previousl foun thapee reviewe article ar mor likel tha article t conclud tha environmenta tobacc smok i harmful.5 Thtobacc industr ha suggeste tha thi ma du t a bia i th pee reviewe journal publishin negativ studie o environ tobacc smoke.3 Ou finding suggesthowever tha negativ studie o environmenta tobacc smok ar publishe i th pee journal whe the ar o higOu findings als sugges tha article relatet sic buildin syndrom ma b o poore tha article o othe topics W di noestablis thi hypothesi a priori; an thi find- shoul b furthe evaluate b othe studiesOu analysi o duplicat publicationshowe tha pee reviewe article ar highe i eve whe the discus tri sam dat symposiu articles Becaus thes pair o b definition ha th sam studdesig characteristics th difference observebetwee pee reviewe an symposiu article mos likel t hav bee du t factorsuc a poo reportin an poo interpretatioo findingsTh findings fro thi stud complemenou previou wor o th conten o symposi environmenta tobacc smoke W hav foun tha symposiu article o tobacc smok ar associate a lac o balanc an tha the ten t th tobacc industr positio thaenvironmenta tobacc smok i no harmful.2Ou findings reporte i thi articl sugges symposiu article o environmentatobacc smok ar als associate wit pooquality Take together thes findings suggestha symposiu article ar no reliabl sourceo informatio abou environmenta tobaccOu findings ar als consisten witpreviou researc o th quality o symposiuarticle i th pharmaceutica literature Fo Rochon3 ha foun tha randomisecontrolle dru trial ar o poore qualitwhe publishe in journa supplement tha i journals O th othe hand Ch anBero'J di no find a differenc i th quality oarticle o drug publishe i symposi anpee reviewe journals however the note thei powe t detec a differenc whecontrollin fo th effec o stud desig waLIMITATIONTher ar severa limitation t ou findings a mentione above w di no havenoug statistica powe t asses th relatiobetwee fundin sourc an articl quality ou findin tha articl qualit wa associate wit sponsorshi shoul binterprete cautiouslyAnothe potentia sourc o criticis lie iou selectio o pee reviewe an symposiu Ou symposiu article wer drawfro 2 symposi whil ou pee

reviewe article wer draw fro 4 journals I a singl o symposiu containe extremelhig o lo qualit articles the i coul theoreticall hav shifte th mea qualit scor oth entir group W d no believ tha thi becaus n singl journa o dominate ou samples Thmaximu numbe o article pe symposiu 1 (16%) whil th maximu numbe oarticle pe pee reviewe journa wa si (9%) qualit o article fro thes source dino diffe fro th overal qualit o article ith groupsAnothe potentia sourc o selectio bia itha th pee reviewe article i ou stud identifie exclusivel throug Medlinewherea th symposiu article wer identifiethroug a variet o electroni databases W Medlin t identif pee reviewe journa o environmenta tobacc smok i i th databas mos commonl use th Unite State t gathe informatio healt relate research However severa reviewe journal tha publis article t environmenta tobacc smoke Tobacco Control, wer no indexe bMedlin durin th perio o ou study.3 3Futur studie o th quality o th scientifi shoul therefor conside drawin sample fro multiple sources suc a a variet o electroni databases searchin journals an checkinAnothe potentia criticis o ou stud itha th differenc i mea quality scoreobserve betwee pee reviewe an articles althoug statisticall wa no particularl large W fee tabl 2 provide qualitativ insigh int th difference betwee th groups tabl show tha pee reviewe article ar likel tha symposiu article t b weldesigned t repor result completel anaccurate an t dra appropriat conclusionsWhe thes qualitativ an quantitativfindings ar take together the provid stron 2Barnes, Beroevidenc tha pee reviewe article ten t bo highe quality tha symposiu articlesFinally i i no clea whethe ou findingsma b generalise t symposi o othetopics Ou stud primaril containe articlefro symposi o environmenta tobaccsmok an symposi o indoo air an wfoun tha articl qualit score wer similarlpoo i bot group (dat no shown) Othestudie hav foun tha th qualit osymposiu article i th pharmaceuticaliteratur tend t b poor.3 Take togetherthes findings sugges tha symposi i generama b poo i quality However thi shoulb confirme b futur studiesCONCLUSIONI summary ou stud provide stronevidenc tha pee reviewe journa article oth healt effect o environmenta tobaccsmok ar superio t symposiu articlesThes findings suppor th decisio o th USuprem Court1 tha exper scientifitestimon shoul b "scientificall valid antha "[a pertinen consideratio i whethe ththeor o techniqu ha bee subjecte t peerevie an publication.However pu findings als sugges tha peerevie i no a guarante o hig quality Thpee reviewe article i ou sampl ha amea quality scor o 0.4 o a scal o 0 t 1suggestin tha ther i a grea dea o roo foimprovement Becaus decision shoul bbase o th highes quality evidenc availablew propos tha researchers polic makers judge shoul evaluat th qualit o allscientifi researc befor usin th findings iresearch policy o lega settings A lac o peerevie shoul b see a a flag tha a articlha a highe probabilit o bein poo i qualityO th basi o ou findings, w believ thajourna an boo editor shoul developstandard fo labellin scientifi material I iofte difficul t determin whethe a articlha hsen publishe i a pee reviewe journao a non-peer-reviewe symposiu becausthi informatio i no printe directl o tharticles W therefor recommen tha th firstpag o al scientifi researc article shoul bclearl labelle t indicat whethe th articlwa pee reviewe an whethe i i bein publishe a par o a symposium Thi wil allopolic makers researchers th press an thpubli t identif thos article tha ar moslikel t b associate wit hig qualityThi stud wa rande b th Rober Woo JohnsoFoundatio (awar N 024783) an b th Cigarett anTobacc Surta Fun o th Stat o California throug thtobacc relate diseas researc progra o th Universit o (awar N 4RT0035)W wis K than ou qualit assessors Valeri HoffmanPhD MP an Judit Moskowi PhD MPH th WritinSemina a th Institut fo Healt Polic Studies whicprovide invaiuab insigh durin th preparatio o Ehi Philli LolSa fo administrativ assistance anStev Seivi Ph fo statistica adviceTh result o thi stud wer presente i par a th 199meetin o th America Publi Healt Associatio i SaDiegO California1 Ber LAS Galbrait A Rermi D

Th publicatio o sponsore symposium i medica journals N Engl J Med1992;32?:1135~402 Ber LA Galbrait A, Renn� E Sponsore symposi oenvironmenta tobacc smote JAMA 1994;27l;612--73 Rocho P Evaluatin th qualit o article publishe ijourna supplement compare wit th qualit o thos i fee paren journal JAMA 1994;272:IO8-~134 Ecobicho DJ Wt JM Bnvirormieniai tobacco smoke:proceed*ings of ihs International Symposium at McGill University,Lexington Massachusetts Lexingto Books 19905 Symposiu proceedings Assessin low-risf agent fo Huncancer methodologica aspects (Internationa Symposiu hel i Clearwater FL USA Octobe 10-14,1989.JmJEpidemiol I990;I (supp I)6 Worksho proceedings ETS—Environmenta tobaccsmoke (Worksho o effect an exposur levels hel ath Universit o Geneva Switzerland Marc 15-171983. EurJRespirDis 1984;65(supp 133)7 Worksho proceedings Environmenta tobacc smokeffect o th non-smoker (Workshop Bermuda 1974.Scand J Respir Dis 1974 supp 918 Perr R Kir P Indoor and ambum Mr quality. LondonSelper 19889 Symposiu proceedings Internationa Symposiu oMedica Perspective o Passiv Smoking PrevMed 1084131 Ber LA Giant SG Tobacc industr response to a risassessmen o environmenta tobacc smoke Tobacco Con-;1 R Reynold Tobacc Compan advertisement "Passivsmoking a activ controversy. (Availabl fro authorupo request.1 Bveringhan R Woodwar S Tobacco Litigation. SydneyLega Books 19911 Ch MK Ber LA Th qualit o dru studie publishe vxsymposiu proceedings Ann Intern Med 1996;124485-93 4 Barne DE Ber LA Industry-ftind« researc an conflico interest a analysi o researc sponsore b th industr throug th Cente fo Indoo AiResearch Ji&a&f Politics Policy LAW 1996;21:515-421 Rocho PA Gurwit JH Simm RW et at. A stud ornanufecturer-supporEfi trial o nortsteroida anti drug i th treatmen o arthritis Archj1 Swae G, Meijer J Influenc o desig characteristic oth outcom o retrospectiv cohor studies Br J InditstM4198S456241 Davidso R Sourc o fundin an outcom o clinica vi~&\$.3 Intern Med 1986;l:155-83 Didcers K. Th existenc o publicatio bia an ris factor fo it occurrence JAMA 1990;263:1385-91 Dickersi K Mi Y-I Mdne CL Factor influencinpublicatio o researc results Follow-u o applicationsubmitte t tw institutiona revie boards JAMA 1992267:374-82 Webe A Acut effect o environmenta tobacc smokeBur J Respirm00; £ 1984;6S:9S~-1082 Webe A Annoyanc an irrigatio b passiv smokingm00; P?£iMed 1984;13:618~252 Webe A Irrigatin an annoyin effect o passiv smokingTskaiJExp Om Med 1985;10:341-52 Webe A Environmenta Tobacc smok exposure acutfeffects-—acceptanc level-—protectiv m«3stires In Bfifglun E Ljndval T Sundel J eds Indoor air. Volum 2.Radon, passive smoking, particutate and hottsirig epidemiology*Stockholm Lyba Tryc AB 1985 297-3012 Hemmmk E Qmlit o clinica trials— concer fo thredecades Meth Inform Med l§82;2l:8l~5.2 Andre E Eid H Fugleru P et al. Publication o clinica trial wit X-ra contras media difference i qualitbetwee journal an decades EurJ Radial 1990;!0:52-J2 Soni Js Joine J Th Qualit o clinica trial publishe iThe Journal of Family Practice, 1974-1993 } Fam Pf1994^39:225-552 Hulle SB Cumming SR Designing clinicaBaltimore Maryland W$km & WSikins !9882 Ch MK Ber LA. Instrumen fo assessin th quaiityodru studie publishe i th medica literature JAm 101-4 _2 McNut RA Evan AT Fletche RH Fletche SW Teffect o blindin o th qualit o pee review JAi^§Q3?9;2I3 Erns E Resc K~L Reviewe bias a blinde experimenstudy J Lab din Med 19M 124:178-82 „ „3 Mohe D Jada AR; NIcho G Penma M i JWals S Assessin th qualit o randomisecrisis a annotate bibliograph o scale andConzroikdClin Trials I995;16:62~733 Ber LA Glant SA Renni D j?ublkatior bia ^healt polic o environment tobacc smoke J1994;272U33~63 Flue-Cure Tobacc Cooperativan Th Counci fo Burle Tobacco In Tobacc Company In an Phili Morri Jrate an R J Reynold Tobacc Compan anVendin Compan vs Unite State Environing Agenc an Caro Browner 19933 Liguor A Hughe JR Wher i smokin res«published Tobacco Control I996;S:37-83 Davi RM Tobacco C&nirol\am MdTobacco Control 1996;5:993 Suprem Cour o di Unite States ^^^i^ 2MerreH D Pharmaceuticals, Inc (N 92-102