Re:ManuscriptIDBMJ.2014.022376entitled"Arandomized,controlledtrialoftheefficacyandharmsofparoxetineandimipramineinthetreatmentofadolescentmajordepression:RestoringStudy329"

Responsetothereviewers andcommitteeproviding,pointbypoint, repliestothecomments madebythereviewers andtheeditors,andexplaininghowwehavedealtwiththeminthepaper.

A = authorship

H = housekeeping amendment or request

B = COI and bias

P = presentation issue

Q = question seeking info or clarification

R = recommended change to methodology

Reviewercomment / commenttoeditor / changemade
Loderandcommittee
1.Thefull text onlineversionofyourarticle,ifacceptedafterrevision,willbetheindexedcitableversion(fulldetailsareat theprintandiPadBMJwillcarryan abridgedversionofyourarticle, usuallyafew weeksafterwards.Thisabridgedversionof thearticleisessentiallyanevidenceabstractcalledBMJpico,whichwewouldlikeyoutowriteusingatemplateandthenemailitto (therearemoredetailsbelowon howtowritethisusingatemplate).Publication of research onbmj.comisdefinitiveandis notsimplyinterim"epublication aheadofprint", soifyoudonotwish toabridge yourarticleusingBMJpico, you willbeabletoopt for onlineonlypublication.Pleaselet usknowifyouwouldpreferthisoption. / Wewould liketoabridge / -
2.Aswellas submitting your revisedmanuscript,wealso requirea copy ofthemanuscriptwithchangeshighlighted.Pleaseuploadthisasasupplemental filewithfiledesignation‘RevisedManuscriptMarkedcopy’. / done
3.didyou registerthestudyinan approvedtrialregistry? / wehavenotdoneatrial; wehaveprovidedthetrialregistration detailsinourpaper / Nonerequired
4.howmany versions oftheprotocolarethere,and iftherewas morethanone, how didyouchoosewhichonetofollow? / Toourknowledge,therewere 2fullversions:1993 (signedbyPrincipalInvestigator2June),andApril1994(amended24March;approved17April).The1994versionthatwefollowedincludedAmendment1,whichspecifiedsubstitutionofK-SADS-Linsteadof K-SADS-P;optionalexternalreviewofdiagnosis;additionalsafetymeasures;anda replacementSBMedical Monitor. TherewasalsoAmendment2in1996(28October; detailed intheCSR,pp.000027-000028)reducingthesample sizetoapproximately275patients,butotherwiseunchanged. / Nonerequired
5.*Weagreewithseveralofthereviewersthattheproblemof potential biasandconflictofinterestneedsmoreattention. Wewould liketohearyour thoughtsaboutthesemattersandwethinksomecommentinthepaperitselfmightbenecessary. / One pointbehind aRIAT articleis bymaking thedataavailablewehopetoallowotherstomakejudgementsaboutthepossibleinfluenceconflictsofinterestthatauthorsmightnotthemselvessee.
WithstandardCOIdeclarations, readersthen havetoguesswhetheritemised conflictshavehadaninfluenceornot.
Weaim tomake itpossibleforreaders todotheirownanalysisandifthisanalysisdiffers fromours theremay beapatterntothedifferencethatisindicativeofsomekindofbias.
WeexpectGSKtore-analyzethedataandproduceanalternatereading butwewelcomethis.One of ourmainpoints behindthearticleistoshowthat thereisnoprivilegedinterpretationthatsomeonewhohasnoconflictsofinterestcanarriveat. / COIsnowincorporatedintodocumenttitlepage,+seenewbox1
6.Wedidnotagree,forexample,thatitmakes sensetomovesymptomssuchasdizziness andheadacheout ofthenervoussystemcluster. / Thepublishedpaper putpsychiatriceventsintoaclustercalledNervousSystem.MedDRAseparatesPsychiatricfromnervoussystem.
Wearguethatitis inappropriatetocodedizziness andheadacheasneurologicalissues. Themostlikelycauseofdizzinessgiventhedrugsinvolvediscardiovascular.
Headaches mostcommonly stemfrommusclesandbloodvesselstothescalp–not partoftheCNS. / Therelevantsection(nowshiftedtoDiscussion)nowreads:
‘TheCSRandCRFfiguresalsodiffersubstantiallyfromotherfiguresquotedin Kelleretal,becausewedidnot code ‘dizziness’and‘headache’underNervousSystem,sincetheformer ismorelikely to
Theimportantpointbehindourcodingis notwherewe puttheseitemsbutrather drawingattentiontothefactthatwhereveritemslikethisareputcansignificantlyaffecttheinterpretation. / be attributable to‘cardiovascular’whileheadachesmostcommonlystemfrommusclesandbloodvessels tothe scalp.’
7.Weagreewith reviewersthatcodingofadverseeventsneedstoberedonebypeoplewhoareindependentofyourgroup. / 1.100%oftheinitialcoding was doneblind–alloftheextracodingfrom theCRFswas doneblind –asthedrug namewas notonthelist
Itwas only fortheelevenSAEs where itwas notpossibletobeblindandnot all ofthesegave extracodes –so99.995%of thecodingwasblind
2.JLNwasrecruitedtothegroupspecificallybecauseofherexpertisetocarryouttheseanalyses.Neither shenorJN,whodidtheanalysisindependentlyandagreedonall ratings,has COI.
3.JJ,whohas providedexpertopinionin aclass action,did notanalysetheefficacyortheadverseeventdata.
4.Weknowof no precedentforanalysestobecarriedoutoutsidearesearchgroup. / Added:
'Allof theinitialcoding fromthetheclinical descriptionsintheCSRwas done blind,as was coding fromtheCRFs.Only forsixeventsfromtheelevenseriousadverseeventnarratives wasitnotpossibletobeblind.Thiswas0.005%of events.'
8.WealsoagreewithseveralofthereviewersthatextrapolationofAEs from thenon-randomsampleofCRFs isunwise. Thisanalysis shouldberemovedfromthepaper.(Table6) / Agreed,willremove / Deletedfromvarioustables
9.*PleasepresentatrueITTanalysis (in otherwords, analyzeallsubjects inthegroupstowhichtheywererandomised,regardlessofwhethertheyreceivedthestudy drugornot).Ourstatisticiansuggeststhatyouconsiderhaving severalcolumns inyour results table. Thefirst wouldpresent anITTanalysis usingLOCF,thesecondusingimputationand correcting forstrata(12centres).Thethirdcolumncould showtheperprotocol or completecaseanalysisusingLOCFandthe fourth theperprotocol or / TheProtocolcalledfor evaluationofthe OC[Observed Case]dataandthe LOCF[LastObservationCarried Forward]datasetwiththelatterbeingdefinitive. TheLOCFmethodforcorrectingmissing valueswasthestandardatthetimethestudywas conducted. Itcontinues tobewidely used,thoughnewermodelssuchasMultipleImputationorMixedModelsarenowfrequentlypreferred.WechosetosticktotheProtocolandusetheLOCFmethodratherthanintroduceaposthocanalytictool. / Seenewfigure 2,andcommentaryinnew box 1
completecaseanalysisusing imputation.Thiswouldallowreaders tojudge forthemselvestheeffects,ifany,of usingmoremodern methods ofanalysis,whilestillshowingtheoriginallyintendedefficacyanalysis.
10. *Wewouldalsoliketoseethe results ofpairwisecomparisons. / Weconductedtheprotocol-specified omnibusanalyses,whicharenegativeasshown.Nevertheless,thepairwiseresultswereconfirmedasnon-significantas reportedbyKelleretal. Theseare tabled in theappendix2. / Figure1&2Appendix2–Tablei
11. *Canyouplease alsoincludeatablethatcontrasts allof yourfindings withthoseoftheoriginalpaper?Youdothis forAEsbutnotfortheefficacyoutcomes.Manyeditorscommentedthatitwasdifficultto understandhowandwherethereanalysesdifferfrom theoriginalones. / Thecontrastof relevanceiswiththeCSR ratherthanthepublishedpaper,butthereisnosignificantdiscrepancybetweenourresultsand GSK’s / ‘Therewerenodiscrepanciesbetweenany ofouranalysesandthosecontainedintheCSR.’addedtoefficacyresults
12. *we were disappointedthat you didnotexaminetheCRFs forallsubjects. This seemsaseriousproblem.Itis,weunderstand,amajorundertakingtoreviewallofthesedocuments,butseemsnecessarytosettherecordstraight.Afterall,thetrial itselfwas amajor effort onthepartoftheoriginalinvestigators. / Seealsoletter toDrLoder.
Wearenolongermakingevententativeextrapolationsfrom theaudit(and wenolongerusethatterm),sotheprimaryjustificationforcompletingitno longerexists.Furthermore:
1.Completingtheauditwouldtake about2000 hoursbecause GSK’smethodofpermittingaccess tothedataissoburdensomeandthis would essentiallyall havetobedoneby oneperson.(Thedifficulties facedbytheRIATteamareseveralordersofmagnitudegreaterthantheGSKteamwhodidtheoriginalwriteup.GSKcouldhavemade ita lot easierfor ustodotheauditexpeditiously andsafely but chosenottodoso.)
2.Itwouldgive only an illusionofcompletenessaswehavealreadyfound 1000missingpagessothat therearelikely3000+missingpages.
3.Becauseoftheenormousburdeningainingaccessto / allextrapolationsfrom the ‘audit’removed.
Alsohavechangedthe way wereportCRFfindings(see,e.g.,table
5)–nolongeras part ofatotal,butthenumberofadditionalcasesidentified,whichwethinkismoreinformative–no oneactuallyknowswhatthetotalis.
andauditingCRFs, nootherteam(apartfromGSK)islikely tohaveresources tocheckouraudit.Wewouldthereforepreferamodelthatseespublicationofthisversionofthepaperand thenhasBMJinaneditorialcallingonGSKand othercompaniestomake thedataavailabletoresearchersin auserfriendlyformatsothattheauditcanbereadily audited by others.
13. *Webelievetheoriginal investigators inthetrialshouldbeacknowledgedinthepaper. / theroles ofthevarious investigators,authorship andrelated issuesarethoroughly discussedin reference3. / AddedtoBackground:
'Weacknowledgetheworkoftheoriginalinvestigators.'
14. Youmentionthat insomecases itwas notclearwhathappenedintheoriginalstudy, forexamplewhysomesecondary outcomeswerechanged.Didyoumakeany attempttoasktheoriginalinvestigators?Ifnot,whynot? / Wehavehad considerablecorrespondencewithGSKhaspublished inaseries of rapidresponses intheBMJ.InparticularGSKhas not beenabletoproduceacopyoftheputative ‘analytic plan’ / Nonerequired
15. Didyouhaveany funding for thisreanalysis? / No / AddedtoAbstract:
‘Nofundingwasobtainedtosupport this restoration’
16. * Theabstract containsnonumericalfindings.Pleasepresentthe figures fortheprincipalstudyoutcomes intheabstract. / Done / ThefollowingsentencehasbeenaddedtotheAbstract'sResultssection:
'HAM-Dscoresdecreasedby10.73,
8.95and9.08points,respectively,fortheparoxetine, imipramine andplacebo groups(p=0.204).'
17. *Wethoughtthat information abouttheallegedproblemswiththeoriginalstudycouldbedealtwith inasingleparagraph intheintroduction. / Wealreadywrite:Kelleretal.,which waslargelyghostwritten,[3]claimedefficacyandsafetyforparoxetineatodds withthedata,[4]Thisisproblematicbecausethearticlehasbeeninfluentialin theliteraturesupportingtheuseofantidepressantsinadolescents.[5] / Nonerequired
18. Pleasebe carefulnottoincludeadhominemremarks. / Ifyoucan identify any such remarks,wewould behappytoremovethem / Nonefoundby us
19. Hasthe previous paper been retracted?Ifnot,howwill readers ofthatpaperknowaboutthisone? / Thepreviouspaper has notbeen retracted; perhapsthepublicationof thispaper will providemoreincentive forJAACAPtodoso. / Nonerequired
20. *Wethought youshouldcommentonthematterofdropouts.Theseseemedhigherin theplacebogroup. / Theyarenothigher in theplacebo group. Wealreadydiscussdropouts. / Nonerequired
21. Wealsowonderedwhether 8weeks istoosoontosee any possiblebenefitof anantidepressant.Severaleditorswhoarepracticingphysicians andusethesedrugsthoughtthat8weeks mightbetoosoontoexpectthedrugstodivergefromplacebo.Couldyoucommentonthis? / Tedeschinietal.'s(2011)pooled analysisof104clinicaltrialsrevealedthat'Fourweeksistheminimum adequatelengthofatrialinordertoreliablydetect drug versusplacebodifferences'.
TedeschiniE,FavaM,PapakostasGI.Placebo-controlled,antidepressantclinicaltrialscannotbeshortenedtolessthan4weeks'duration:apooledanalysisofrandomizedclinicaltrialsemployinga diagnosticoddsratio-basedapproach.JClinPsychiatry.2011Jan;72(1):98-113.
Keller etal. commentedthat8weeksmightnotbesufficienttoachieveafull clinical response(p.770).SimilarlyitmightnotbesufficientforADRstoemerge.
22. * The methodssectionshouldgivemoreinformationabouthowsubjectswererecruited,numberofcentersinvolvedinthestudyandhowtheywerechosen.Whodidtheinterviews?Howwerethey trained? Yousaythatchildrensignedan informedconsentform, butshouldthisnotbe"assent?" / Asspecified in themanuscript, therewere12study centers(10 intheUnitedStates and2 inCanada).Thisis nowstatedintheabstractaswellasthe Methodssection. / Nowreads:
'Anundisclosednumberofpatientsidentifiedbytelephonescreeningaspotentialparticipantsweresubsequentlyevaluatedatthestudysitebyaseniorclinician(psychiatristorpsychologist).'
Thefollowingsentencehasbeenadded:
'Multiple meetingsandteleconferenceswereheldbythesponsoringcompany withsitestudyinvestigatorstoensurestandardizationacrosssites.'
Wehaveadded: ''The centerswereaffiliatedwith eitherauniversityorahospitalpsychiatrydepartmentandhadexperiencewithadolescentpatients.Theinvestigatorswereselectedfortheirinterestin the study and theirabilityto recruitstudypatients.'Therewasnoassentform. Wehaveadded:
‘thestudyinformedconsentformwassignedbybothpatientandparent;thereisnomentionofaseparateassentformintheprotocol orintheclinicalstudyreport.’
23. Pleaseexplainhowthedecisionwas madetoreducethenumberofsubjectsfrom 300to275. / Wealready explain this:
Theprotocolcalled for300subjects,butthiswasreducedto
275.Recruitmentwasslowerthanexpectedand,reportedlybecauseoflimited medicationsupplies(mainlyplacebo)duetoexpiry,amidcourseevaluationof189patientswascarriedout,withoutbreakingtheblind,revealinglessvariabilityinHAM-Dscores (SD8)thananticipated.Thereforethe recruitmenttargetwasreduced onthegrounds thatitwould havenonegativeimpacton theestimated80%powerrequiredtodetecta four-pointdifference betweenplaceboandactivedruggroups. / SeealsoNaudet,query 8,andDoshi,query6. Nowundersamplesizeas:
'Recruitmentwasslowerthanexpected,andreportedlymedicationsupplies(mainlyplacebo)werelimited duetoexpiry.Thereforeamidcourseevaluationof189 patients wascarriedout,withoutbreakingtheblind,revealinglessvariabilityinHAM-Dscores (SD8) thananticipated.Thereforetherecruitmenttargetwasreducedto275onthegroundsthatitwouldhavenonegativeimpactontheestimated80% powerrequiredto
detectafour-pointdifferencebetweenplaceboandactivedruggroups.'
24. Indescribingtheintervention,pleaseclarifythedefinitionof"nonresponder." / Thereisnoexplicitdefinitionfornon-response,justimplicitone,consideringthedefinitionofresponse.
AccordingtotheCSR,section5.2.4SustainedResponse,page000078, "Survivalanalysiswasperformedfortimeuntilsustainedresponse,definedasresponselasting untilendpointoftheacutephase. Responsewas definedasaHAM-Dtotalscoreless thanorequalto8 oradecreasefrombaselineinHAM-Dtotal score of 50%or greater.
Patientswereclassifiedasbeing a responder ornon-responder." / Revisedtoread
‘Non-responders (thosefailingtoreachrespondercriteria)...’
25. Althoughsubjects couldbe titratedupto60mgparoxetineor300 mgimipramine,how manyactuallydidachievethesedoses?Canyouprovideinformation aboutthemeanfinaldoseineachgroup andtherange? / Wehavealreadyreportedmeanfinaldoseand range.Wehave added number reaching highest dosefor imipramineandparoxetine. / Nowreads:
'Theparoxetinegroup wastitratedtoadoseof20mg/daybyweek4,with55%movingtoa higher dose(mean28.0mg/day, SD8.4mg)byweek8. Theimipraminegroupwastitratedto200mg/daybyweek4,with40%goinghigher(mean205.8
mg/day,SD63.9mg)byweek8.28patientsreachedthehighestpermissibledoseof40mgofparoxetine,and20patientsweretitratedtothemaximum300mgofimipramine.'
26. *Howmany subjects were screenedforthestudy?PleaseshowthisinFigure1. / Wehave notbeenableto findthisinformation. / Added:
‘Anundisclosednumberofpatients…’
27.Figure1alsoneedstoshowthenumberanalyzedfor thecompletecaseoutcomeat8weeks. / Displayed inDataTable / Figure 1
Reviewer:1(FlorianNaudet)
  1. commentsinthemethodsectionand intheresultssectionwhicharegenerallynottheplacetodiscuss choices and results.Pleasesee forexample:-intheintroduction :“Consequently,wehavereanalysedStudy329accordingtotheRIATstatement..Tothisend,wehaveusedtheClinicalStudyReport(CSR;GSK's'FinalClinicalReport')availableontheGSKwebsite,[7]otherpublicallyavailabledocuments,[8]andthedataaccesssystemSASSolutionsOnDemand,[9]onwhichGSKhas postedsomeStudy 329documents(availableonlytousersapprovedbyGSK).Followingnegotiation,[10]GSKpostedde-identified individualcasereportforms (CRFs)onthatsite. Atable ofsourcesofdataconsultedinpreparing eachpartofthispaperisavailableasAppendix1.” Thisshouldappearinthemethodsection;
/ Agreed / MovedtoMethodssection
29.-inthemethodsection,authors state“These imipramine dosesarehighforadolescents.Inthesixcomparatorstudiessubmitted bySKBaspartoftheir1991 ApprovalNDAforparoxetineinadults,themeanimipraminedoseoverall was140mg,withameanendpointdose of170mg” / Agreed / MovedtoResults belowtable 4
30.-inthe methodsectionwe canread “(weacknowledgediffering opinionsaboutthisissueinthe statisticalliterature).” This commenthasnoreference. / Referenceadded / Kline RB.Beyond SignificanceTesting.Statistics Reform intheBehavioralSciences,2013,p81.
31.-intheresultsection“(withadifferenceof4points beingpre-specified asclinicallysignificant)”:it isinalreadyinthemethod / agreed / Deleted
sectionandshouldnot appear intheresultswhicharedescriptive;
32. -intheresultsection‘(ScoresontheHAM-D can varyfromzerotoa maximum of52)’thatshould appearinthemethodsection. / agreed / MovedtoMethods
33.-intheresultsection“theprotocolalsolistedtherelapse rateinthecontinuationphasefor respondersasasecondaryoutcomevariable.Ourcalculation differedfrom theCSRcalculationbecauseweincludedthosewhoseHAM-D scoresroseabovethe ‘response’rangeandthose whointentionallyoverdosed.” / Wethinkthis needstostay whereitisin ordertomakewhatfollowsintelligible / Notchanged
34.-intheresults sectionauthors states that“alternative treatmentsofthedata couldgivedifferentresults.”It must be inthe discussionsectionandnotintheresults. / Nolonger relevantasestimates fromauditnowexcluded / deleted
35.-I alsothinkthat,forclarity purpose,theinformationaboutchanges insamplesizecanbepresentedafter thesamplesizecalculationforclaritypurposes. / Movedandedited forclarity, seeLoder,query23
36.thereweretwopre-specifiedoutcomevariables,withthreegroups.Wasthere acorrection for multiplecomparisonsmentionedintheprotocol?Thesepoints mustbe detailed. / Nocorrection / Seenew box 1
37. IfIunderstand,therewas alsoachangeofprimaryoutcomecriteriawhichwasdoneaposterioriandafter breakingtheblind. / Yes,but thisis discussedindetail inaprevious publication,andwethinkitneednotberehearsedhere / Nochange
38. Canauthors givethedate of:-Breakingtheblind;Changesmadeintheoutcomes criteria / thisis discussedindetailinaprevious publication,andwethinkitneednotberehearsedhere / Nochange
39. It wouldbe also helpfulto listandcomparealltheoutcomesreportedinthepublishedpaperbyKelleret al. / we think it bettertofollowDoshi’srecommendedapproachand restrictdiscussionofKellertotheintroductionanddiscussion / SeeDoshisection forchanges
40. Inthe sentence:“Globalimpressionscale?” / Agreeconfusing,wewerebeingobsessionalabout / changeto'ClinicalGlobal
pleasesuppressthe “?” andexplainthat it istheCGI(as reportedinthetable). / accuracy,buthavechangedforclarity / Impression(CGI)'
41. Theprimaryefficacyvariablereported inthestatisticalmethodsandin theprimaryoutcomevariablesarenotthesame.Pleaseexplainorcorrect. / Wehaverewrittenthissection,whichweagreewasconfusing / Nowreads:
'Oneofthe twoprimaryefficacyvariables,proportionof responders(response),andonesecondaryefficacyvariable,proportionofpatientsrelapsing,weretreatedascategoricalvariables.Thesecondprimaryefficacyvariable,changeintotalHAM-Dscoreovertheacutephase,andtheremainingsecondaryefficacyvariablesweretreatedascontinuousvariables.'
42. In Table1:pleaselegend(mean[SD]). / Assumethisrefers totable 3 / Done
43. Figuresarerepresentedfor OC analysis,pleaseprovidethedataforW8 (endpoint)ITTanalysiswithLOCFwhich was definedas theprincipalpopulationofanalysis. / Under “PatientsValid For TheEfficacyAnalysis”,the
Protocolstates,“Allpatientsrandomizedtostudy treatmentandforwhomatleastonevalidpost-treatmentefficacyevaluationis availablewill bevalidfor inclusion in an
'intent-to-treat'analysis.”
44.Pleasealsoindicatethenumberofpatientineachgroup underthefigureforeachtimepoint. / This dataistoocumbersomeformainpaper,sohaveaddedasanappendix / SeeTablexivinAppendix2.
45. Iunderstandthatitis timeconsuming anddifficult, but I thinkthattheanalysis ofCRFshould becompletetoavoidanymisinterpretation.Itis indeed importantsincethisauditprocess gave risetoadditionalAEs.Indeed,sincethis analysisis notcomplete,andsinceitwas not at random, itis amajorlimitationsand one can beverycriticonthispoint. / Seeabove
46. Intables wheretheCRFestimates are / Weagree / deleted
presented,Ithinkthatthisestimatesarehighlyspeculativeandthatthedatacannotbeanalysedinthisway.I suggesttodelete thiscolumnandtoanalysealltheCRF.
47. SAEhavea specificdefinitioninMEDRA. I’mnotsurethatitisstrictlyoverlappingwiththenotion ofseverity. Thus thecomparisonwithKeller’set al.paperisverydifficultas stated bytheauthors. ForMEDRA,a SAEisseriouswhen itresultsindeath,life-threatening,hospitalization (initialorprolonged),a disabilityorPermanentDamage,inacongenitalAnomaly/BirthDefect,itrequired InterventiontoPreventPermanentImpairment,andfor otherSerious(ImportantMedicalEvents).
This last category is acrucialpointanditisprobablynotstriclyoverlappingwiththenotionofsevereAE(usedby theauthors) : itiswhentheeventmayjeopardizethepatientandmayrequire medicalorsurgicalintervention(treatment)topreventoneoftheotheroutcomes.Examplesincludeallergicbrochospasm(aseriousproblemwith breathing)requiringtreatmentinanemergencyroom,seriousblooddyscrasias (blood disorders)orseizures/convulsions thatdo notresultinhospitalization.Thedevelopmentof drugdependenceor drugabusewouldalsobeexamplesofimportantmedicalevents. / TheproblemwithSAE as usedby Kelleretal is thatacomponentofthesestemsfromthejudgementofthedoctor–wecannotreplicatethis.
Lodgingthedatawith BMJmeansthatanyonewillbeabletogointoourspreadsheetsandseeexactly whatwascodedandhow andwill beabletocomeup withalternatecodings.
Nomatterwhodoesthecoding,itwill bepossible forothergroups tomakeacasethatinbetween1–5%ofcases thattheywouldhavedonethingsslightly differently.
This is simplythenatureofthebeast. Coding is notsomethingyoucangetright –itisinherentlycollaborative. / Seeresponse toLoder, query7
48. Whenauthorsstate that “The majorityofpatientsstoppedatthispointweredesignatedbyGSKaslackofefficacy(seeTable11).Investigatorsinfourcentresreported lackof / Weprovidethedata forotherstomaketheirowninterpretation.Others arequitewelcometocodetheseasGSK have done.GSK simplydon’tprovide uswitha basisforgoingalong withwhatthey havedone.Ourapproach / Nochange
efficacyasareasonforstopping sixplacebopatientseventhoughtheHAM-Dscorewasintheresponderrange andas low as2 or 3pointsinsomeinstances.”I wouldliketosee moredetails. Additionaly,IthinkthatthechangeofcodingbetweenLackofEfficacyandAdverseEventisdifficultandcouldbemisleading. Manytimes,discontinuation occurs forbothlackofefficacyandadverseevents,sinceonecaneasilyconsiderthatadverseeventslikedrymouthcanbemoreacceptableinthecaseoftreatmentefficacy. This pointcould beaddressedin thediscussionandI’mnot sure that aaposterioriinterpretationof theCRFcangiveaperfectinformationabouttheindividualpatientexperience(evenifitisverybetterthanaggregateddata ofcourse…). Moroever,I alsothinkthatalackof efficacycanbeconsideredforpatientsevenifthey areresponder upontheHDRS. Patients are notjust a score onascale.
The authors’aposterioriproposalforrecodingthiscanbethuserroneous. / has more facevalidity – but couldas hesaysbe wrong.We’renotafraid tobewrong.
49. Pleaseexplain,inthediscussion,for readersthattheinterpretationofqualitative informationinCRF isverysubjectiveandpronetoaninterpretation bias(including forthefirstmanuscriptandforthisone).Pleaseexplainwhyitis notpossibletocollectAE inanotherway(orexplainhowtheyshouldbecollected) andtheinterestofMEDRA. / Firstinterpreting thedataontheCRF was essentiallyblind.Therewasnoindicationon thedocumentthatindicated whichdrugwasinvolved.
Insofarascoding isanactof interpretation,thenyestherewas interpretationandtheriskofbias.Thiswassomethingthatcouldnotbeovercomeowingtothelimitations GSKimposedonus. Wecouldnotprint offthematerialandsubmitit topanels ofcoders inanefforttoreducebiasandwecouldnotconvenepanelsofcoders aroundtheperiscope.
Thecollectionof AEwasdone16-20 years ago-notby
us.Itwas done intheusual ways itisdonein drugtrialsthenandnow.This is avery poorsystem.Itwouldbepossibletodesignmuch bettersystemsifyouwereinterestedtodiscoveradverseeventsbutthisisadifferenttopicand wearestuckwith whatGSKinfactdid
50. Table5canbedeletedsinceitpresentsresultsthatarealsopresentedintable6. / Table5movedtoAppendix2.
51. Legendoftable6is missing(SOC*). / fixed
52. Intable11,please legendwhat is“RIATproposed”? / fixed
53. It isstated that “Roughly1000pagesweremissingfromthe CRFsaudited”. Canauthorsprecisewhy? / InsomecasesGSKstate theseare missing,insomecasesthey aresimplymissingwithoutnote;wecoulddetectnopatterntothis / Added:
'withnodiscerniblepatterntomissinginformation'
54. Inthe box Patient00039,pleasedetailwetheritwasAEorSAE. / This was severeAE– butnotseriousSAE / Patient00039,whohad asevere(butnot serious)AE
55. Inthediscussion section,whenauthorsstate that “TheRIATapproach[…]outcomevariables.”It must be recalledthat the messageisvery differentsincetheKeller’s reportstateinthe abstract that “Paroxetinedemonstratedsignificantlygreaterimprovementcomparedwithplacebo inHAM-Dtotal scoreor=8,HAM-D depressedmooditem”. / We can’t see that anythingisbeingrequestedhere
56. Whenthey state “Inour opinion,statisticallysignificantornot,allrelevantprimaryandsecondaryoutcomes,andharmsoutcomes,shouldbe explicitlyreported”.I’mnot sure thatitis only theopinionofthis
paper’authors.RCTsare oftenunderpoweredfordetectingthesechanges. / We can’t see that anythingisbeingrequestedhere
57. TheURL notexactly thegoodURL…Please donot test…andcorrect… / This URLisaplaceholderuntilwefindoutwherethedocumentswillbe housed / Pendingconfirmationdocswillbehoused onBMJwebsite,and ona
dedicatedstudy329site.
58. Wheretheystate:“Theyrevealevidenceconsistentwithdependence onandwithdrawalfromparoxetine.”I wouldnuance,“withpossible dependence”. / We disagree.‘Consistent with’isalready aqualifier,soadding‘possible’wouldbetautological
Reviewer:2(PeterDoshi)
59Organizational issue.Ithinkthatin generaltheauthorsdonotneedtomentiontheKelleretal.publication intheMethodsor ResultssectionsofthisRIATmanuscript.Themisreportingof study 329intheKellermanuscripthasbeenwelldocumentedbytheauthorselsewhere.Theprimarypurposeofthismanuscript,asIseeit,is on presentinganhonestandaccuratereportofthestudy329resultsthanitistofurther documentmisreportingof Kelleretal.If additionalaspectsofmisreportingin Kelleretal.werediscoveredintheprocess ofRIATing study 329, thisisimportantandIthinktheauthorscanincludethisinformation,butIthinkit would be besttokeepthistotheIntroductionand Discussionsections. / Agreed. / Kellerreferences removedfromresults;modifiedversionincludedinpara4-6ofDiscussion. Oldtables68 incorporated intonew tablexinDiscussion.
60.Auditofnon-random sampleofAEs. TheRIATauthors carriedoutanauditoftheadverseeventsectionofcasereportforms (CRFs) foranon-randomsampleof93 ofthetotal275 trialparticipants.Theauthorsareveryclearthroughoutthemanuscripttoindicatethatthiswasa non-randomsample.It wouldhavebeenbetterofcourseif100%of CRFswereaudited,butgiventhenumberofhoursittooktoaudit93(approx. 1000hours they sayinthetext),a / Addressedabove.Notethisreviewerrecognisedtheimpracticalityofauditingallcases:
Itwouldhavebeenbetter ofcourseif100%ofCRFswereaudited, butgiventhenumberof hours ittookto audit93(approx. 1000hours theysay inthetext),afullauditlikelyonly willhappen ifanothergrouppicksupthebaton. / Changes as detailedabove.
fullauditlikely onlywillhappenifanother group picks upthebaton.I think theauthorsarecorrectto includeanalysesandtables thatshowthe pre-auditand post-audittallies ofAEs.HoweverIdonot thinkitwise fortheauthorstoextrapolateandpresentestimates,basedonfindingsfromtheirnon-randomsample, ofthenumberofadditionalAEs they wouldhavediscoveredhadtheybeenabletoauditall275CRFs.(This mightbeOKif it were arandomsamplebut itis not.)Buthereinparticular,Idonotthinkitwisebecausemy impression ofthenon-randomsample–of allparticipantsthatwithdrewfromthestudy (85)plus8childrenknowntohavebecomesuicidal–is thatitisasamplemorelikelytohaveproblems inthetransferofinformationfromCRFtoCSR.
61.I didn’t seeaCOIstatementfor the authorsinany ofthemanuscriptandappendixfiles? / Wedidsubmit them,but theydidn’t get intothe PDFforsomereason / Seeabove;have addedCOIstatementstomainmanuscript
62.Methods.Cantheauthors explainwhytheychosetofollowthe1994protocol instead ofthe1993 or1996versions?WhichversionoftheprotocolwasthelastversionbeforeparticipantrecruitmentbeganinApril1994?Whichversions dotheauthorshave thefulltextfor? / SeeLoder, query 4
63.Methods.“Whererelevant,we havereferredtothese variations.” What does thismean? / Agreethisisconfusingand haveclarified / Nowreads:
'Furthermore,theCSRreportedsomeproceduresthatvariedfromthosespecified intheprotocol,andwehavenotedvariationswherevertheywereconsideredsignificant.'
64.Methods/Participants.“Theprotocolcalledfor300subjects,but thiswasreducedto275.” / SeeLoder,query23,Naudet,query8
Canthisbeclarified?Sothe1993protocolcalledfor300subjects butthiswasrevisedto275inthe1994protocol?
65.Methods/sourcesofharmsdata.“Roughly1000 pagesweremissing fromtheCRFsaudited.”Canthe authorsexplainhowtheyknewpages weremissingandcanconcludethis?(e.g. numberedpagesindicatingmissingpages etc.)Wereall93participantswhoseCRFswereauditedmissingthesamepages/sections?Also,did they alertGSKtothis andif sowhatwasGSK’sresponse? / SeeLoder,query 25.
66.Methods/coding ofAEs.Intheparagraphbeginning,“Classifying aproblem…” cantheauthors clarify ifMedDRAputs‘sore through’inthecentralnervoussystembucket? / MedDRAhasparticularproblemswithsorethroat–as anycodingsystemwould.
Thereareoptions forittogo intotheinfectious,gastric,respiratoryandnervoussystemSOC.
Wehave looked atall instances blind tothe studydrug andallocated ittonervoussystem,respiratoryand infectiousrespectively andtothesurpriseofatleast one of us (DH)theresults didnot panoutasexpected –aclearpreponderanceofnervoussystemproblemsonparoxetine.
67.Box1.“Mostrecodingissueswere clear-cut.”What ismeant by‘clear-cut’? / MEDDRAisamorestraightforwardprocess less opentobiasthanusing ADECS.
inalmostall instances theclinicaldescriptionsweresufficientlyclearthatmostcoderswouldcometothesameMEDDRAcode / Nowreads
'Mostrecodingwasstraightforward.'
68.Competinginterests statementappearsmissing.The authorssay“asattached” but Icouldnot findtheattachment. / Clarifiedabove
69. Methods/analysisofharms data.TheauthorschosetoanalyzeMedDRASOCclassespsychiatric,cardiovascular,gastrointestinal,respiratoryandplaceall otherAEs in“other”. / Thesecategorieswerespecificallychosentocorrespondwiththe Kellerpaper ‘Table3’,inordertohelpwithanycomparisons.They presenteddatausingthecategories:‘Cardio-vascular’,‘Digestive’,‘Nervous’,‘Respiratory’and
Afterlookingattheresults tables,theselooklikereasonablechoices tome,butcantheauthorsincludeasentencethatexplains howtheymadethischoice? / ‘Other’.
70. Methods/patientwithdrawal.Intheparagraphbeginning“TheCSRstatesthat theprimaryreason…” it mentions “CSRAppendixG”.Cantheauthors sayherebrieflywhatAppendixGcontains? / 329DEPAppendixGCaseReportFormTabulations byPatientIntent-to-TreatPopulation[2073pages]:demographics,presentingconditions,concomitantmedication,adverseexperiences,vitalsigns,laboratorydata / Wethinkitwouldaddtoo manywords fornotenoughgaintofullyexplain whateachappendixwerefertocontains
71. Methods/blinding.Couldtheauthors alsomentionwhethertheyreviewedtheCertificatesofAnalysisforthestudy medications todouble-checkwhethertheyappeared tohavebeencorrectlyformulatedtoensure blinding? / WehavereviewedtheCertificatesofAnalysis forthestudy.Thestudypillsthemselves differed, thoughallwereprovidedasover-encapsulatedbluishgreentablets.Noinformationwasavailableregardingblindingsuccess.Asdescribedinourmanuscript:“Paroxetine wassuppliedasfilm-coated,capsule-shapedyellow(10mg)andpink(20mg)tablets.Imipramine(50mg)wasboughtcommerciallyandsuppliedasgreenfilm-coated round 50mgtablets.‘Paroxetine placebos’matchedthe paroxetine 20mgtablets,and‘imipramineplacebos’matchedtheimipraminetablets.Alltabletswereover-encapsulatedinbluish-greencapsules topreserve blinding.”
72. Methods/statisticalmethods.Theauthorswrite,“Wefollowedthe methodology ofthe apriori1994studyprotocol.” Whyisthe 1994protocollabeled“apriori”? Wasit the lastversionpriortoparticipantenrollment? / Correct
73. Methods/statisticalmethods.Intheparagraphbeginning“Theprimaryefficacyvariable”,there are twosentenceswiththephrase “primaryefficacy variable”.I supposethisis a reflectionofthetrialhaving twooutcomesprespecifiedas “primary”? / As notedabove,wephrasedthispoorlyand havecorrectedit / SeeNaudet,query14
74. Discussion. Does thefollowingtextreferto / Thisrefers primarilytotheCSR,which deviatedfromthe / Nowreads:
Kelleret al.orthe CSR:“Theauthors/sponsorsdepartedfromprotocolbyperformingpairwisecomparisonsoftwoofthethreegroups whentheomnibusANOVA showed nosignificanceineitherthecontinuousordichotomousvariables.”This shouldbe clarified.Ifthis referstotheCSR,thentosomeextentthereisadiscoveryamongtheRIATauthorsthattheyhave found reportingbias within theCSRitself,andIthink this is an importantfinding whichthey shouldhighlightas such. / protocol.ThiswasuncriticallyacceptedbyKeller etal. / 'Theauthors/sponsorsdepartedfrom their study protocol intheCSRitselfbyperformingpairwisecomparisonsoftwoofthethreegroups whentheomnibusANOVAshowednosignificanceineitherthecontinuousordichotomousvariables.'
75. Box3.“The inability toaccessallCRFsmayhave introducedsome error.” Notsurewhat ismeantbythis.Aretheauthorstalkingabouttheirinability due totime/resources toauditeverything?Isthisa reference tothedifficulttouseportalforaccessing thestudy data?Or isthisareferencetotheapproximately1000pagesthatweremissingfromthe CRFsthatGSKmadeavailablethrough theirportal? / Agreed / Nowreads:
'Timeand resourcespreventedaccesstoallCRFsbecauseofthedifficulties inusingtheportalforaccessingthestudydataandbecausesignificantdataweremissing.'
76. RIATAR.Why are someitems so long?Forexample,somany sourcesare givenforFunding(#25). / Becausethereare somanypotentialambiguitiesandcontradictions wethoughtitimportanttodiscloseallpossiblesources of data;bettertobethoroughthanreadable
77. Abtract/Results.Suggestchanging,ifappropriate,“foranymeasure” to“foranyprimaryorsecondary[efficacy] outcome.” / Agreed,Changed / Nowreads:
'Theresponses toparoxetineandimipraminewerenotstatisticallyorclinicallysignificantlydifferentfromplaceboforanypre-specifiedprimaryorsecondaryefficacyoutcome.'
78. Background. “RIAT publicationofStudy 329whichwasfundedby…”Changeto “RIAT / Agreed / Rewordedassuggested
publicationofStudy329.Theoriginalstudywasfundedby…”
79. Background.“On14June2013,theRIATresearchersnotifiedGSKthatKelleretal.appeared… Study329.”Thisrefers toaletterIsentGSK.Wedidnotspecifically mentionstudy329 inthisemail.Inorder tomakethesentenceaccurate,I suggest rewording:“On14June2013,theRIATresearchersaskedGSKwhetherithadanyintention torestore any ofthetrialsitsponsored.” / Agreed / Rewordedassuggested
80. Similarly,change “GSK didnot signalanyintenttopublishacorrected versionofthearticle.”to“GSKdid not signalany intent topublishacorrectedversionofanyofitstrials.” / Agreed / Rewordedassuggested
81. Methods/SecondaryEfficacyVariables. “We couldnot findanydocument that providedanyscientific rationaleforthesepost-hocchanges…”Didyoufindany“non-scientific”rationale? Ifnot,perhapsdelete “scientific”. / Westickbytheuseofthetermscientific,becausealthoughitis outsidethescopeofthispaper, thestory of 329isrepletewithnonscientific(mostlymarketing-based)rationales
82. Methods/Outcomes.Theheadings1.Principal Endpoints for Efficacyand 2.PrincipalEndpoints for Harms.Ithink thisis slightlyconfusingwiththelanguage of“primary” and“secondary” efficacy variables.How about justlabelingthe sections“EfficacyEndpoints” and“HarmsEndpoints”? / Agreed / Rewordedassuggested
83. Methods/Harms.I think the “(p.18)”at theendofthequotedparagraphis atypo as itisalsostatedabove. / Agreed / Rewordedassuggested
84. Box1.“At the week6visit … GSK…”DotheauthorsmeanSKB? / Yes / WehavealteredallreferencesfromGSKtoSKB whereappropriate
85. Avarietyoftermsareusedtorepresenttheprovenance ofAEdata e.g. “CSR recoded”and“CRFaudit”from table7,“AEsinAppendix D”fromtable9,and“AEsreported (CSRcheck)”intable12.Iwonder if better terms canbeusedtomakethemeaningmoretransparent.Perhapssome variant of“SKB/GSKcoded”,
“RIAT recoded”,and “RIATrecodedplus CRFaudit”?Anotherthought is touse termslikeADECSandMedDRA e.g. “SKB/GSKADECS
coded”,“RIAT MedDRArecoded”,and “RIATMedDRArecodedplusCRFauditdiscoveredadditionalAEs”.Irealize that some ofmyproposedtitlesarelongandwon’t fit willinthespaceofatighttable, but my suggestiontoremovethe Keller columns aswellasthe“CRFestimated” i.e.extrapolatedAEcount columnsfrom theResults sectionwillhopefullyfreeupsomespace. / Theproposals aregood / Haveadoptedthisterminologyintables
86. Results/Discontinuations.“Consort” shouldbe “CONSORT”. / Corrected / CONSORT
87. Results/Discontinuations.“GSK regardedthesepatientsas participantsinthecontinuationphase…” Shouldthis be SKB? / Asabove
88. Box 2/section8. “…becauseitbecame clear that the blindhadbeenbroken…”Canyoujustbeclearwhoseblindyouaretalkingabout?I.e.I think thisisSKB’sblind,but I’mnot 100% sureaspartoftheRIATersrecoding happenedblindwhileother parts didnot. / This hasbeenclarified / Nowreads:
'becauseitbecameclearthattheblindhadbeenbroken in severalcasesbeforerelatednesswasadjudicatedbytheoriginalinvestigators'
89. Discussionsection/twoparagraphs beforeConclusion.“…analysisofadverse eventsrequiresaccesstoindividualpatientleveldata / Agreed / Nowreads:
'analysisofadverseevents requiresaccesstoindividualpatientlevel
(CRFs).”I wouldrewordthe ending to“…requiresaccesstoindividualpatientleveldata intheformof CRFs.” / dataintheformofCRFs.'
90. Conclusion.“Study 329showednoadvantage… onanyofthe specifiedparameters.”Wouldusingtheword“pre-specified” bebetterthan“specified”? / Agreed / Added‘pre-‘
91. Methods/Interventions.“Theseimipraminedoses arehigh foradolescents.Inthesixcomparatorstudiessubmitted bySKBaspartoftheir1991ApprovalNDAforparoxetine
inadults,themeanimipraminedoseoverall was140mg,withameanendpointdoseof170mg.[14]”I think this shouldgototheDiscussionsectionunlessitwaspartoftheoriginalmethods. / Agreewithmove / InaccordancewithNaudet,query2,movedtoResults below Table4
92. Methods/Sourceofharmsdata.Suggestmovingthe following toResults: “Of the elevenparoxetinepatientswithAEsdesignatedasserious,ninediscontinuedbecauseof AEs.AlargenumberofotherpatientsdiscontinuedbecauseofAEsthatwerenotregarded asserious,orforlackofefficacyor protocolviolations(seeFigure1).Noneoftheselatterdiscontinuationsledtopatient narratives.” / This isinfact amethodologicalissueas itpertainstoavailabilityofdata.Haveclarifiedthatbyrewording: / Nowreads:
‘Additionalinformationwasavailable fromthesummarynarratives inthebody oftheCSRforpatientswhohad AEsthatweredesignatedasseriousorledtowithdrawal.(Ofthe elevenparoxetinepatientswithAEsdesignatedasserious, ninediscontinuedbecauseofAEs.)However,thelargenumberofotherpatientsdiscontinuedbecauseofAEsthatwerenotregardedasserious,or for lackofefficacyorprotocolviolations(seeFigure 1),didnotgeneratepatientnarratives.’
93. Box1 lookslikeitbelongs inResults,notMethods. / Similarlywethinkthisspeakstomethodologicaldifficulties
94. Table8is great,but perhaps shouldgointheDiscussion? / While wehadthoughtitmightbeinappropriatetohavetables inthediscussion,weareOKwiththis. / Modifiedto includecomparison oftotalpsychiatricAEs.
Reviewer:3(HildePA vanderAa)
95.Theauthors followedthemethodologyasstatedinthepre-specifiedprotocolof1994,inwhichproposedstatisticalapproachesorstatisticalassumptionswerenotjustified.Outdated techniqueswereusedtoanalysethedata,leading tomoreuncertainty.Iwouldrecommendauthorsto(also)includemoderntechniquesof data-analysisoratleastmentionthis ‘limitation’inthe discussionpart ofthepaper:
-One ofthelimitationsofthistrial isthelargenumberofdropouts. Therefore, alinearmixedmodelsapproachtoanalysethedata withamaximumlikelihoodassumptionisbettersuitedtoestimateeffectsthanthechosenANOVAandGLM.
-Ifauthors,however,dodecidetouseANOVAandGLMmultipleimputationwouldbea betterway tohandlemissingdatathanthecurrentlyusedLOCF,see forexamplethepaper byBeunckensC,MolenberghsG,KenwardMG.Directlikelihoodanalysisversussimpleforms ofimputationformissing datain randomizedclinicaltrials.ClinicalTrial,2005;2:379-86. / Seenew box 1
96.-Authorsdescribedthatthey didnotcorrectforattritionandnon-complianceinthesamplesizecalculation. Inaddition, theyalsodid notcorrectforthedifferentstratain theirsample(12 centresincluded).Thisshouldalsobereported. /
everythingwedidtakestheeffectofthesites intoaccount[LSMean,ANOVA,X²] –see above / inthenew box:
'TheProtocolcalledfor ANOVAtesting[GLM]forcontinuousvariables usingamodelthatincluded theeffectsofSITE,TREATMENT,andSITEx
TREATMENTinteraction,withthelatterdropped ifp>0.10.LogisticalRegression[chiSquare2x3]wasprescribedforcategoricalvariablesunderthesamemodel.'
97.Limitations of thecurrentstudy shouldbedescribed in moredetail. / SeeresponsetoHetrick,query13
98.Thelimitationsofthestatistical analysis (asmentionedabove)should bementioned. / Seenew box 1
99.the authorsstate that‘Theinability toaccessallCRFs mayhaveintroducedsomeerror.’(page27,line 25). Thisshouldbeexplainedinmoredetail. / Thisclausenowdeleted
100.Atthebeginning ofthediscussionauthorsstatetodrawminimalconclusionsregardingefficacyandharms,invitingotherstooffertheirownanalysis. Ithink this is ajustconclusionbasedonpreviouslymentioned limitations.However,thiscautiousapproachofinterpretingthe results oftheRIATstudy shouldalsobereflectedintheconclusionpartoftheabstract / Webelievethat ourconclusionsinboththeabstractandthediscussionarefully justifiedbythe datathatwehavepresented
andthediscussion.
101.Throughoutthewholepaperauthorsdescribe the‘new study’comparedtothe‘oldstudy’ofKelleret al.Thismakesitdifficulttoreadandtodistinguishthemethods usedintheRIATtrial. Thoughitis importantto reportthesedifferences,theymightfor instancebecollectedinboxesor reportedinitalicorcombinedinthemethodssectionofthepaper. / weaddressthisbyremoving mentionofKeller fromtheresultssectionandsimplifying tables
102.Theabstractdoes notfollowthestandardstyle of‘The BMJ’forresearcharticles:objectives,design,setting,participants,intervention,mainoutcomes,resultsandconclusion. / Agreed / SeerevisedAbstract,whichadherestothisformat
Reviewer:4(SarahHetrick)
103.It’shardtoknow exactlywhat shouldbe inthebackground,orindeedwhathe objectivesareorhow a paper like thisshouldbewrittenup.Onone handitis simply the descriptionof atrial,butontheotherhand ithas severalimportantother objectivesIthink:first,tocorrecterrorsofthepreviouswrite-up;second,tohighlighttheissueof reporting bias.Iam not100%surethatthesecondobjectivewasclearlyarticulatedorachieved,and perhapsthisistheobjectiveof RIATbut notnecessarilyofthispaperas such. My personal opinion is that morecould be madeofitinthis paper(and perhapsthiswouldaddress my concernsaboutoriginalitymadeabove)andthatthebackgroundappearstoindicatethatthatcorrectingerrorsandhighlightingtheissueofreporting biasiswhatthepaperisabout. / Wehavedeliberately downplayed criticismoftheKelleret alpaperin termsof itsreportingbias,partlybecauseasbeendealt withelsewhere,but alsobecause wedidn’t wanttodistractfromthe straightforwardre-presentationofStudy329accordingtotheRIATapproach / Nochangemade
104.ShouldthebackgroundincludesomethingaboutletterbyJon Jureidiniand Martin Kellersresponsein2003?Thissawthecorrectionoffindingstoacertainextent. / We don’t think so,forsimilarreasonsandbecauseKeller’sresponsein2003correctednothingbuttrivialitiesintheinitialreporting
105.Iwas interestedtoknowwhetherthereadershouldjustbelievethattheKeller 2001paperwasghostwritten orwhetherthereissomekindofproofofthis?Howdidtheauthorsfindthisout/know? / Wehavedocumentedelsewherethatthereisnodoubtthatthispaperwas ghostwritten.A referencetothispaperisincluded in ourintroductory section.
106.Inthefourthparagraph theauthorsrefertothe RIATstatement,but I wasn’t clearwhat thiswas? / Agreeunclear / have changed‘statement’to‘recommendation’
107.Inthefifth paragraph theauthorsoutlinethe objectivesofthe originalstudybut don’tstatewheretheseobjectiveswerederivedfrom?TheKeller paper,orfrom theSKBreports? / Wehavecorrected thistomakeitclear / Nowreads: